blob: c8ca0640a10ca65993ddc903b48031bc0b8844fb [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;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800218 struct rlimit *rlim;
219
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100220 acct_arg_size(bprm, size / PAGE_SIZE);
221
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800222 /*
223 * We've historically supported up to 32 pages (ARG_MAX)
224 * of argument strings even with small stacks
225 */
226 if (size <= ARG_MAX)
227 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700228
229 /*
230 * Limit to 1/4-th the stack size for the argv+env strings.
231 * This ensures that:
232 * - the remaining binfmt code will not run out of stack space,
233 * - the program will have a reasonable amount of stack left
234 * to work from.
235 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800236 rlim = current->signal->rlim;
Jiri Slabyd554ed892010-03-05 13:42:42 -0800237 if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700238 put_page(page);
239 return NULL;
240 }
241 }
242
243 return page;
244}
245
246static void put_arg_page(struct page *page)
247{
248 put_page(page);
249}
250
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700251static void free_arg_pages(struct linux_binprm *bprm)
252{
253}
254
255static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
256 struct page *page)
257{
258 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
259}
260
261static int __bprm_mm_init(struct linux_binprm *bprm)
262{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800263 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700264 struct vm_area_struct *vma = NULL;
265 struct mm_struct *mm = bprm->mm;
266
267 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
268 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800269 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700270
Michal Hockof268dfe2016-05-23 16:26:02 -0700271 if (down_write_killable(&mm->mmap_sem)) {
272 err = -EINTR;
273 goto err_free;
274 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700275 vma->vm_mm = mm;
276
277 /*
278 * Place the stack at the largest stack address the architecture
279 * supports. Later, we'll move this to an appropriate place. We don't
280 * use STACK_TOP because that can depend on attributes which aren't
281 * configured yet.
282 */
Michal Hockoaacb3d12011-07-26 16:08:40 -0700283 BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700284 vma->vm_end = STACK_TOP_MAX;
285 vma->vm_start = vma->vm_end - PAGE_SIZE;
Cyrill Gorcunovd9104d12013-09-11 14:22:24 -0700286 vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP;
Coly Li3ed75eb2007-10-18 23:39:15 -0700287 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Rik van Riel5beb4932010-03-05 13:42:07 -0800288 INIT_LIST_HEAD(&vma->anon_vma_chain);
Tavis Ormandy462e635e2010-12-09 15:29:42 +0100289
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700290 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800291 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700292 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700293
294 mm->stack_vm = mm->total_vm = 1;
Dave Hansenfe3d1972014-11-14 07:18:29 -0800295 arch_bprm_mm_init(mm, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700296 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700297 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700298 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700299err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800300 up_write(&mm->mmap_sem);
Michal Hockof268dfe2016-05-23 16:26:02 -0700301err_free:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800302 bprm->vma = NULL;
303 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700304 return err;
305}
306
307static bool valid_arg_len(struct linux_binprm *bprm, long len)
308{
309 return len <= MAX_ARG_STRLEN;
310}
311
312#else
313
Oleg Nesterov0e028462011-03-06 18:02:54 +0100314static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100315{
316}
317
Oleg Nesterov0e028462011-03-06 18:02:54 +0100318static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700319 int write)
320{
321 struct page *page;
322
323 page = bprm->page[pos / PAGE_SIZE];
324 if (!page && write) {
325 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
326 if (!page)
327 return NULL;
328 bprm->page[pos / PAGE_SIZE] = page;
329 }
330
331 return page;
332}
333
334static void put_arg_page(struct page *page)
335{
336}
337
338static void free_arg_page(struct linux_binprm *bprm, int i)
339{
340 if (bprm->page[i]) {
341 __free_page(bprm->page[i]);
342 bprm->page[i] = NULL;
343 }
344}
345
346static void free_arg_pages(struct linux_binprm *bprm)
347{
348 int i;
349
350 for (i = 0; i < MAX_ARG_PAGES; i++)
351 free_arg_page(bprm, i);
352}
353
354static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
355 struct page *page)
356{
357}
358
359static int __bprm_mm_init(struct linux_binprm *bprm)
360{
361 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
362 return 0;
363}
364
365static bool valid_arg_len(struct linux_binprm *bprm, long len)
366{
367 return len <= bprm->p;
368}
369
370#endif /* CONFIG_MMU */
371
372/*
373 * Create a new mm_struct and populate it with a temporary stack
374 * vm_area_struct. We don't have enough context at this point to set the stack
375 * flags, permissions, and offset, so we use temporary values. We'll update
376 * them later in setup_arg_pages().
377 */
Yuanhan Liu9cc64ce2013-02-20 13:16:01 +1100378static int bprm_mm_init(struct linux_binprm *bprm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700379{
380 int err;
381 struct mm_struct *mm = NULL;
382
383 bprm->mm = mm = mm_alloc();
384 err = -ENOMEM;
385 if (!mm)
386 goto err;
387
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700388 err = __bprm_mm_init(bprm);
389 if (err)
390 goto err;
391
392 return 0;
393
394err:
395 if (mm) {
396 bprm->mm = NULL;
397 mmdrop(mm);
398 }
399
400 return err;
401}
402
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100403struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100404#ifdef CONFIG_COMPAT
405 bool is_compat;
406#endif
407 union {
408 const char __user *const __user *native;
409#ifdef CONFIG_COMPAT
Al Viro38b983b2012-09-30 13:38:55 -0400410 const compat_uptr_t __user *compat;
Oleg Nesterov0e028462011-03-06 18:02:54 +0100411#endif
412 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100413};
414
415static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100416{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100417 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100418
Oleg Nesterov0e028462011-03-06 18:02:54 +0100419#ifdef CONFIG_COMPAT
420 if (unlikely(argv.is_compat)) {
421 compat_uptr_t compat;
422
423 if (get_user(compat, argv.ptr.compat + nr))
424 return ERR_PTR(-EFAULT);
425
426 return compat_ptr(compat);
427 }
428#endif
429
430 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100431 return ERR_PTR(-EFAULT);
432
Oleg Nesterov0e028462011-03-06 18:02:54 +0100433 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100434}
435
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436/*
437 * count() counts the number of strings in array ARGV.
438 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100439static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440{
441 int i = 0;
442
Oleg Nesterov0e028462011-03-06 18:02:54 +0100443 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100445 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 if (!p)
448 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100449
450 if (IS_ERR(p))
451 return -EFAULT;
452
Xi Wang6d92d4f2013-01-11 14:31:48 -0800453 if (i >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 return -E2BIG;
Xi Wang6d92d4f2013-01-11 14:31:48 -0800455 ++i;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700456
457 if (fatal_signal_pending(current))
458 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 cond_resched();
460 }
461 }
462 return i;
463}
464
465/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700466 * 'copy_strings()' copies argument/environment strings from the old
467 * processes's memory to the new process's stack. The call to get_user_pages()
468 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100470static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700471 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472{
473 struct page *kmapped_page = NULL;
474 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700475 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 int ret;
477
478 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100479 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 int len;
481 unsigned long pos;
482
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100483 ret = -EFAULT;
484 str = get_user_arg_ptr(argv, argc);
485 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100488 len = strnlen_user(str, MAX_ARG_STRLEN);
489 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100491
492 ret = -E2BIG;
493 if (!valid_arg_len(bprm, len))
494 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700496 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700498 str += len;
499 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500
501 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
Roland McGrath9aea5a62010-09-07 19:37:06 -0700504 if (fatal_signal_pending(current)) {
505 ret = -ERESTARTNOHAND;
506 goto out;
507 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700508 cond_resched();
509
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700511 if (offset == 0)
512 offset = PAGE_SIZE;
513
514 bytes_to_copy = offset;
515 if (bytes_to_copy > len)
516 bytes_to_copy = len;
517
518 offset -= bytes_to_copy;
519 pos -= bytes_to_copy;
520 str -= bytes_to_copy;
521 len -= bytes_to_copy;
522
523 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
524 struct page *page;
525
526 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700528 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 goto out;
530 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700532 if (kmapped_page) {
533 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700535 put_arg_page(kmapped_page);
536 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 kmapped_page = page;
538 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700539 kpos = pos & PAGE_MASK;
540 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700542 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 ret = -EFAULT;
544 goto out;
545 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 }
547 }
548 ret = 0;
549out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700550 if (kmapped_page) {
551 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700553 put_arg_page(kmapped_page);
554 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 return ret;
556}
557
558/*
559 * Like copy_strings, but get argv and its values from kernel memory.
560 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100561int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100562 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563{
564 int r;
565 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100566 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100567 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100568 };
569
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100571 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100573
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 return r;
575}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576EXPORT_SYMBOL(copy_strings_kernel);
577
578#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700579
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700581 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
582 * the binfmt code determines where the new stack should reside, we shift it to
583 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700585 * 1) Use shift to calculate the new vma endpoints.
586 * 2) Extend vma to cover both the old and new ranges. This ensures the
587 * arguments passed to subsequent functions are consistent.
588 * 3) Move vma's page tables to the new range.
589 * 4) Free up any cleared pgd range.
590 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700592static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593{
594 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700595 unsigned long old_start = vma->vm_start;
596 unsigned long old_end = vma->vm_end;
597 unsigned long length = old_end - old_start;
598 unsigned long new_start = old_start - shift;
599 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700600 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700602 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700604 /*
605 * ensure there are no vmas between where we want to go
606 * and where we are
607 */
608 if (vma != find_vma(mm, new_start))
609 return -EFAULT;
610
611 /*
612 * cover the whole range: [new_start, old_end)
613 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800614 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
615 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700616
617 /*
618 * move the page tables downwards, on failure we rely on
619 * process cleanup to remove whatever mess we made.
620 */
621 if (length != move_page_tables(vma, old_start,
Michel Lespinasse38a76012012-10-08 16:31:50 -0700622 vma, new_start, length, false))
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700623 return -ENOMEM;
624
625 lru_add_drain();
Linus Torvalds2b047252013-08-15 11:42:25 -0700626 tlb_gather_mmu(&tlb, mm, old_start, old_end);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700627 if (new_end > old_start) {
628 /*
629 * when the old and new regions overlap clear from new_end.
630 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700631 free_pgd_range(&tlb, new_end, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700632 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700633 } else {
634 /*
635 * otherwise, clean from old_start; this is done to not touch
636 * the address space in [new_end, old_start) some architectures
637 * have constraints on va-space that make this illegal (IA64) -
638 * for the others its just a little faster.
639 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700640 free_pgd_range(&tlb, old_start, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700641 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 }
Linus Torvalds2b047252013-08-15 11:42:25 -0700643 tlb_finish_mmu(&tlb, old_start, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700645 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800646 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700647 */
648 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
649
650 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651}
652
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700653/*
654 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
655 * the stack is optionally relocated, and some extra space is added.
656 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657int setup_arg_pages(struct linux_binprm *bprm,
658 unsigned long stack_top,
659 int executable_stack)
660{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700661 unsigned long ret;
662 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700664 struct vm_area_struct *vma = bprm->vma;
665 struct vm_area_struct *prev = NULL;
666 unsigned long vm_flags;
667 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800668 unsigned long stack_size;
669 unsigned long stack_expand;
670 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
672#ifdef CONFIG_STACK_GROWSUP
James Hogand71f2902014-05-13 23:58:24 +0100673 /* Limit stack size */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800674 stack_base = rlimit_max(RLIMIT_STACK);
James Hogand71f2902014-05-13 23:58:24 +0100675 if (stack_base > STACK_SIZE_MAX)
676 stack_base = STACK_SIZE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
Helge Dellerd045c772015-05-11 22:01:27 +0200678 /* Add space for stack randomization. */
679 stack_base += (STACK_RND_MASK << PAGE_SHIFT);
680
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700681 /* Make sure we didn't let the argument array grow too large. */
682 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 return -ENOMEM;
684
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700685 stack_base = PAGE_ALIGN(stack_top - stack_base);
686
687 stack_shift = vma->vm_start - stack_base;
688 mm->arg_start = bprm->p - stack_shift;
689 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700691 stack_top = arch_align_stack(stack_top);
692 stack_top = PAGE_ALIGN(stack_top);
Roland McGrath1b528182010-09-07 19:35:49 -0700693
694 if (unlikely(stack_top < mmap_min_addr) ||
695 unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
696 return -ENOMEM;
697
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700698 stack_shift = vma->vm_end - stack_top;
699
700 bprm->p -= stack_shift;
701 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700703
704 if (bprm->loader)
705 bprm->loader -= stack_shift;
706 bprm->exec -= stack_shift;
707
Michal Hockof268dfe2016-05-23 16:26:02 -0700708 if (down_write_killable(&mm->mmap_sem))
709 return -EINTR;
710
Hugh Dickins96a8e132008-07-10 21:19:20 +0100711 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700712
713 /*
714 * Adjust stack execute permissions; explicitly enable for
715 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
716 * (arch default) otherwise.
717 */
718 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
719 vm_flags |= VM_EXEC;
720 else if (executable_stack == EXSTACK_DISABLE_X)
721 vm_flags &= ~VM_EXEC;
722 vm_flags |= mm->def_flags;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700723 vm_flags |= VM_STACK_INCOMPLETE_SETUP;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700724
725 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
726 vm_flags);
727 if (ret)
728 goto out_unlock;
729 BUG_ON(prev != vma);
730
731 /* Move stack pages down in memory. */
732 if (stack_shift) {
733 ret = shift_arg_pages(vma, stack_shift);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800734 if (ret)
735 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 }
737
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700738 /* mprotect_fixup is overkill to remove the temporary stack flags */
739 vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
740
Michael Neuling5ef097d2010-03-05 13:42:57 -0800741 stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */
Michael Neuling803bf5e2010-02-10 13:56:42 -0800742 stack_size = vma->vm_end - vma->vm_start;
743 /*
744 * Align this down to a page boundary as expand_stack
745 * will align it up.
746 */
747 rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700748#ifdef CONFIG_STACK_GROWSUP
Michael Neuling803bf5e2010-02-10 13:56:42 -0800749 if (stack_size + stack_expand > rlim_stack)
750 stack_base = vma->vm_start + rlim_stack;
751 else
752 stack_base = vma->vm_end + stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700753#else
Michael Neuling803bf5e2010-02-10 13:56:42 -0800754 if (stack_size + stack_expand > rlim_stack)
755 stack_base = vma->vm_end - rlim_stack;
756 else
757 stack_base = vma->vm_start - stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700758#endif
Eric B Munson3af9e852010-05-18 15:30:49 +0100759 current->mm->start_stack = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700760 ret = expand_stack(vma, stack_base);
761 if (ret)
762 ret = -EFAULT;
763
764out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 up_write(&mm->mmap_sem);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800766 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768EXPORT_SYMBOL(setup_arg_pages);
769
Nicolas Pitre7e7ec6a2016-07-24 11:30:18 -0400770#else
771
772/*
773 * Transfer the program arguments and environment from the holding pages
774 * onto the stack. The provided stack pointer is adjusted accordingly.
775 */
776int transfer_args_to_stack(struct linux_binprm *bprm,
777 unsigned long *sp_location)
778{
779 unsigned long index, stop, sp;
780 int ret = 0;
781
782 stop = bprm->p >> PAGE_SHIFT;
783 sp = *sp_location;
784
785 for (index = MAX_ARG_PAGES - 1; index >= stop; index--) {
786 unsigned int offset = index == stop ? bprm->p & ~PAGE_MASK : 0;
787 char *src = kmap(bprm->page[index]) + offset;
788 sp -= PAGE_SIZE - offset;
789 if (copy_to_user((void *) sp, src, PAGE_SIZE - offset) != 0)
790 ret = -EFAULT;
791 kunmap(bprm->page[index]);
792 if (ret)
793 goto out;
794 }
795
796 *sp_location = sp;
797
798out:
799 return ret;
800}
801EXPORT_SYMBOL(transfer_args_to_stack);
802
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803#endif /* CONFIG_MMU */
804
David Drysdale51f39a12014-12-12 16:57:29 -0800805static struct file *do_open_execat(int fd, struct filename *name, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200808 int err;
David Drysdale51f39a12014-12-12 16:57:29 -0800809 struct open_flags open_exec_flags = {
Al Viro47c805d2011-02-23 17:44:09 -0500810 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
Al Viro62fb4a12015-12-26 22:33:24 -0500811 .acc_mode = MAY_EXEC,
Al Virof9652e12013-06-11 08:23:01 +0400812 .intent = LOOKUP_OPEN,
813 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500814 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815
David Drysdale51f39a12014-12-12 16:57:29 -0800816 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
817 return ERR_PTR(-EINVAL);
818 if (flags & AT_SYMLINK_NOFOLLOW)
819 open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW;
820 if (flags & AT_EMPTY_PATH)
821 open_exec_flags.lookup_flags |= LOOKUP_EMPTY;
822
823 file = do_filp_open(fd, name, &open_exec_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400824 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200825 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200827 err = -EACCES;
Al Viro496ad9a2013-01-23 17:07:38 -0500828 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400829 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200830
Eric W. Biederman90f85722015-06-29 14:42:03 -0500831 if (path_noexec(&file->f_path))
Al Viro6e8341a2009-04-06 11:16:22 -0400832 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200833
834 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400835 if (err)
836 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837
David Drysdale51f39a12014-12-12 16:57:29 -0800838 if (name->name[0] != '\0')
839 fsnotify_open(file);
840
Al Viro6e8341a2009-04-06 11:16:22 -0400841out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200842 return file;
843
Al Viro6e8341a2009-04-06 11:16:22 -0400844exit:
845 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200846 return ERR_PTR(err);
847}
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800848
849struct file *open_exec(const char *name)
850{
Paul Moore51689102015-01-22 00:00:03 -0500851 struct filename *filename = getname_kernel(name);
852 struct file *f = ERR_CAST(filename);
853
854 if (!IS_ERR(filename)) {
855 f = do_open_execat(AT_FDCWD, filename, 0);
856 putname(filename);
857 }
858 return f;
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800859}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860EXPORT_SYMBOL(open_exec);
861
Mimi Zohar6777d772009-08-21 14:32:48 -0400862int kernel_read(struct file *file, loff_t offset,
863 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864{
865 mm_segment_t old_fs;
866 loff_t pos = offset;
867 int result;
868
869 old_fs = get_fs();
870 set_fs(get_ds());
871 /* The cast to a user pointer is valid due to the set_fs() */
872 result = vfs_read(file, (void __user *)addr, count, &pos);
873 set_fs(old_fs);
874 return result;
875}
876
877EXPORT_SYMBOL(kernel_read);
878
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500879int kernel_read_file(struct file *file, void **buf, loff_t *size,
Mimi Zoharbc8ca5b2016-01-24 10:07:32 -0500880 loff_t max_size, enum kernel_read_file_id id)
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500881{
882 loff_t i_size, pos;
883 ssize_t bytes = 0;
884 int ret;
885
886 if (!S_ISREG(file_inode(file)->i_mode) || max_size < 0)
887 return -EINVAL;
888
Mimi Zohar39eeb4f2016-01-30 22:23:26 -0500889 ret = security_kernel_read_file(file, id);
890 if (ret)
891 return ret;
892
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200893 ret = deny_write_access(file);
894 if (ret)
895 return ret;
896
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500897 i_size = i_size_read(file_inode(file));
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200898 if (max_size > 0 && i_size > max_size) {
899 ret = -EFBIG;
900 goto out;
901 }
902 if (i_size <= 0) {
903 ret = -EINVAL;
904 goto out;
905 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500906
Stephen Boyda098ecd2016-08-02 14:04:28 -0700907 if (id != READING_FIRMWARE_PREALLOC_BUFFER)
908 *buf = vmalloc(i_size);
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200909 if (!*buf) {
910 ret = -ENOMEM;
911 goto out;
912 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500913
914 pos = 0;
915 while (pos < i_size) {
916 bytes = kernel_read(file, pos, (char *)(*buf) + pos,
917 i_size - pos);
918 if (bytes < 0) {
919 ret = bytes;
920 goto out;
921 }
922
923 if (bytes == 0)
924 break;
925 pos += bytes;
926 }
927
928 if (pos != i_size) {
929 ret = -EIO;
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200930 goto out_free;
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500931 }
932
Mimi Zoharbc8ca5b2016-01-24 10:07:32 -0500933 ret = security_kernel_post_read_file(file, *buf, i_size, id);
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500934 if (!ret)
935 *size = pos;
936
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200937out_free:
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500938 if (ret < 0) {
Stephen Boyda098ecd2016-08-02 14:04:28 -0700939 if (id != READING_FIRMWARE_PREALLOC_BUFFER) {
940 vfree(*buf);
941 *buf = NULL;
942 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500943 }
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200944
945out:
946 allow_write_access(file);
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500947 return ret;
948}
949EXPORT_SYMBOL_GPL(kernel_read_file);
950
Mimi Zohar09596b92015-11-19 12:39:22 -0500951int kernel_read_file_from_path(char *path, void **buf, loff_t *size,
952 loff_t max_size, enum kernel_read_file_id id)
953{
954 struct file *file;
955 int ret;
956
957 if (!path || !*path)
958 return -EINVAL;
959
960 file = filp_open(path, O_RDONLY, 0);
961 if (IS_ERR(file))
962 return PTR_ERR(file);
963
964 ret = kernel_read_file(file, buf, size, max_size, id);
965 fput(file);
966 return ret;
967}
968EXPORT_SYMBOL_GPL(kernel_read_file_from_path);
969
Mimi Zoharb844f0e2016-02-01 08:36:21 -0500970int kernel_read_file_from_fd(int fd, void **buf, loff_t *size, loff_t max_size,
971 enum kernel_read_file_id id)
972{
973 struct fd f = fdget(fd);
974 int ret = -EBADF;
975
976 if (!f.file)
977 goto out;
978
979 ret = kernel_read_file(f.file, buf, size, max_size, id);
980out:
981 fdput(f);
982 return ret;
983}
984EXPORT_SYMBOL_GPL(kernel_read_file_from_fd);
985
Al Viro3dc20cb2013-04-13 20:31:37 -0400986ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len)
987{
Al Viroec695572014-02-04 19:08:21 -0500988 ssize_t res = vfs_read(file, (void __user *)addr, len, &pos);
Al Viro3dc20cb2013-04-13 20:31:37 -0400989 if (res > 0)
990 flush_icache_range(addr, addr + len);
991 return res;
992}
993EXPORT_SYMBOL(read_code);
994
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995static int exec_mmap(struct mm_struct *mm)
996{
997 struct task_struct *tsk;
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700998 struct mm_struct *old_mm, *active_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
1000 /* Notify parent that we're no longer interested in the old VM */
1001 tsk = current;
1002 old_mm = current->mm;
1003 mm_release(tsk, old_mm);
1004
1005 if (old_mm) {
Konstantin Khlebnikov4fe7efd2012-06-20 12:53:01 -07001006 sync_mm_rss(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 /*
1008 * Make sure that if there is a core dump in progress
1009 * for the old mm, we get out and die instead of going
1010 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001011 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 */
1013 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001014 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 up_read(&old_mm->mmap_sem);
1016 return -EINTR;
1017 }
1018 }
1019 task_lock(tsk);
1020 active_mm = tsk->active_mm;
1021 tsk->mm = mm;
1022 tsk->active_mm = mm;
1023 activate_mm(active_mm, mm);
Davidlohr Bueso615d6e82014-04-07 15:37:25 -07001024 tsk->mm->vmacache_seqnum = 0;
1025 vmacache_flush(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 if (old_mm) {
1028 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +02001029 BUG_ON(active_mm != old_mm);
Oleg Nesterov701085b2012-03-19 17:04:01 +01001030 setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +01001031 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 mmput(old_mm);
1033 return 0;
1034 }
1035 mmdrop(active_mm);
1036 return 0;
1037}
1038
1039/*
1040 * This function makes sure the current process has its own signal table,
1041 * so that flush_signal_handlers can later reset the handlers without
1042 * disturbing other processes. (Other processes might share the signal
1043 * table via the CLONE_SIGHAND option to clone().)
1044 */
Arjan van de Ven858119e2006-01-14 13:20:43 -08001045static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046{
1047 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001048 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 spinlock_t *lock = &oldsighand->siglock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001051 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 goto no_thread_group;
1053
1054 /*
1055 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001058 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 /*
1060 * Another group action in progress, just
1061 * return so that the signal is processed.
1062 */
1063 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 return -EAGAIN;
1065 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066
Oleg Nesterovd3441932010-05-26 14:43:11 -07001067 sig->group_exit_task = tsk;
1068 sig->notify_count = zap_other_threads(tsk);
1069 if (!thread_group_leader(tsk))
1070 sig->notify_count--;
1071
1072 while (sig->notify_count) {
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001073 __set_current_state(TASK_KILLABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 spin_unlock_irq(lock);
1075 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001076 if (unlikely(__fatal_signal_pending(tsk)))
1077 goto killed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 spin_lock_irq(lock);
1079 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 spin_unlock_irq(lock);
1081
1082 /*
1083 * At this point all other threads have exited, all we have to
1084 * do is to wait for the thread group leader to become inactive,
1085 * and to assume its PID:
1086 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001087 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -08001088 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001089
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001090 for (;;) {
Oleg Nesterove56fb282013-04-30 15:28:20 -07001091 threadgroup_change_begin(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001092 write_lock_irq(&tasklist_lock);
Kirill Tkhaidfcce792015-04-16 12:48:01 -07001093 /*
1094 * Do this under tasklist_lock to ensure that
1095 * exit_notify() can't miss ->group_exit_task
1096 */
1097 sig->notify_count = -1;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001098 if (likely(leader->exit_state))
1099 break;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001100 __set_current_state(TASK_KILLABLE);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001101 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -07001102 threadgroup_change_end(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001103 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001104 if (unlikely(__fatal_signal_pending(tsk)))
1105 goto killed;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001106 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
Roland McGrathf5e90282006-04-10 22:54:16 -07001108 /*
1109 * The only record we have of the real-time age of a
1110 * process, regardless of execs it's done, is start_time.
1111 * All the past CPU time is accumulated in signal_struct
1112 * from sister threads now dead. But in this non-leader
1113 * exec, nothing survives from the original leader thread,
1114 * whose birth marks the true age of this process now.
1115 * When we take on its identity by switching to its PID, we
1116 * also take its birthdate (always earlier than our own).
1117 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001118 tsk->start_time = leader->start_time;
Oleg Nesterov266b7a02013-07-03 15:08:35 -07001119 tsk->real_start_time = leader->real_start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -07001120
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07001121 BUG_ON(!same_thread_group(leader, tsk));
1122 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 /*
1124 * An exec() starts a new thread group with the
1125 * TGID of the previous thread group. Rehash the
1126 * two threads with a switched PID, and release
1127 * the former thread group leader:
1128 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -08001129
1130 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -07001131 * The old leader becomes a thread of the this thread group.
1132 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -08001133 * is called. Odd but simple and correct.
1134 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001135 tsk->pid = leader->pid;
Oleg Nesterov3f418542013-07-03 15:08:25 -07001136 change_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001137 transfer_pid(leader, tsk, PIDTYPE_PGID);
1138 transfer_pid(leader, tsk, PIDTYPE_SID);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001139
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001140 list_replace_rcu(&leader->tasks, &tsk->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001141 list_replace_init(&leader->sibling, &tsk->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001143 tsk->group_leader = tsk;
1144 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -06001145
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001146 tsk->exit_signal = SIGCHLD;
Oleg Nesterov087806b2011-06-22 23:10:26 +02001147 leader->exit_signal = -1;
Oleg Nesterov962b5642005-11-23 13:37:43 -08001148
1149 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
1150 leader->exit_state = EXIT_DEAD;
Oleg Nesteroveac1b5e2011-07-21 20:00:43 +02001151
1152 /*
1153 * We are going to release_task()->ptrace_unlink() silently,
1154 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1155 * the tracer wont't block again waiting for this thread.
1156 */
1157 if (unlikely(leader->ptrace))
1158 __wake_up_parent(leader, leader->parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -07001160 threadgroup_change_end(tsk);
Oleg Nesterov81879262008-12-01 14:18:16 -08001161
1162 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001163 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001165 sig->group_exit_task = NULL;
1166 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167
1168no_thread_group:
Oleg Nesterove6368252012-03-19 17:03:22 +01001169 /* we have changed execution domain */
1170 tsk->exit_signal = SIGCHLD;
1171
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +04001173 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +03001174
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001175 if (atomic_read(&oldsighand->count) != 1) {
1176 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001178 * This ->sighand is shared with the CLONE_SIGHAND
1179 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001181 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1182 if (!newsighand)
1183 return -ENOMEM;
1184
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 atomic_set(&newsighand->count, 1);
1186 memcpy(newsighand->action, oldsighand->action,
1187 sizeof(newsighand->action));
1188
1189 write_lock_irq(&tasklist_lock);
1190 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001191 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 spin_unlock(&oldsighand->siglock);
1193 write_unlock_irq(&tasklist_lock);
1194
Davide Libenzifba2afa2007-05-10 22:23:13 -07001195 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 }
1197
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001198 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 return 0;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001200
1201killed:
1202 /* protects against exit_notify() and __exit_signal() */
1203 read_lock(&tasklist_lock);
1204 sig->group_exit_task = NULL;
1205 sig->notify_count = 0;
1206 read_unlock(&tasklist_lock);
1207 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208}
Oleg Nesterov0840a902007-10-16 23:27:22 -07001209
Andrew Morton59714d62008-02-04 22:27:21 -08001210char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211{
1212 /* buf must be at least sizeof(tsk->comm) in size */
1213 task_lock(tsk);
1214 strncpy(buf, tsk->comm, sizeof(tsk->comm));
1215 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -08001216 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217}
J Freyensee7d74f492011-05-06 16:56:47 -07001218EXPORT_SYMBOL_GPL(get_task_comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219
Al Viro6a6d27d2012-08-21 09:56:33 -04001220/*
1221 * These functions flushes out all traces of the currently running executable
1222 * so that a new one can be started
1223 */
1224
Adrian Hunter82b89772014-05-28 11:45:04 +03001225void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226{
1227 task_lock(tsk);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001228 trace_task_rename(tsk, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1230 task_unlock(tsk);
Adrian Hunter82b89772014-05-28 11:45:04 +03001231 perf_event_comm(tsk, exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232}
1233
1234int flush_old_exec(struct linux_binprm * bprm)
1235{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001236 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237
1238 /*
1239 * Make sure we have a private signal table and that
1240 * we are unassociated from the previous thread group.
1241 */
1242 retval = de_thread(current);
1243 if (retval)
1244 goto out;
1245
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001246 /*
1247 * Must be called _before_ exec_mmap() as bprm->mm is
1248 * not visibile until then. This also enables the update
1249 * to be lockless.
1250 */
Matt Helsley925d1c42008-04-29 01:01:36 -07001251 set_mm_exe_file(bprm->mm, bprm->file);
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001252
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 * Release all of the old mmap stuff
1255 */
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001256 acct_arg_size(bprm, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 retval = exec_mmap(bprm->mm);
1258 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -04001259 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260
1261 bprm->mm = NULL; /* We're using it now */
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001262
Mathias Krausedac853a2011-06-09 20:05:18 +02001263 set_fs(USER_DS);
Zhang Yib88fae62014-01-23 15:55:57 -08001264 current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD |
1265 PF_NOFREEZE | PF_NO_SETAFFINITY);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001266 flush_thread();
1267 current->personality &= ~bprm->per_clear;
1268
Aleksa Saraic1df5a62016-12-21 16:26:24 +11001269 /*
1270 * We have to apply CLOEXEC before we change whether the process is
1271 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1272 * trying to access the should-be-closed file descriptors of a process
1273 * undergoing exec(2).
1274 */
1275 do_close_on_exec(current->files);
Linus Torvalds221af7f2010-01-28 22:14:42 -08001276 return 0;
1277
1278out:
1279 return retval;
1280}
1281EXPORT_SYMBOL(flush_old_exec);
1282
Al Viro1b5d7832011-06-19 12:49:47 -04001283void would_dump(struct linux_binprm *bprm, struct file *file)
1284{
Eric W. Biederman21245b82016-11-16 22:06:51 -06001285 struct inode *inode = file_inode(file);
Dmitry Shmidtcd082872017-01-31 13:51:21 -08001286 if (inode_permission2(file->f_path.mnt, inode, MAY_READ) < 0) {
Eric W. Biederman21245b82016-11-16 22:06:51 -06001287 struct user_namespace *old, *user_ns;
Al Viro1b5d7832011-06-19 12:49:47 -04001288 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
Eric W. Biederman21245b82016-11-16 22:06:51 -06001289
1290 /* Ensure mm->user_ns contains the executable */
1291 user_ns = old = bprm->mm->user_ns;
1292 while ((user_ns != &init_user_ns) &&
1293 !privileged_wrt_inode_uidgid(user_ns, inode))
1294 user_ns = user_ns->parent;
1295
1296 if (old != user_ns) {
1297 bprm->mm->user_ns = get_user_ns(user_ns);
1298 put_user_ns(old);
1299 }
1300 }
Al Viro1b5d7832011-06-19 12:49:47 -04001301}
1302EXPORT_SYMBOL(would_dump);
1303
Linus Torvalds221af7f2010-01-28 22:14:42 -08001304void setup_new_exec(struct linux_binprm * bprm)
1305{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001306 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307
1308 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309 current->sas_ss_sp = current->sas_ss_size = 0;
1310
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001311 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
Kees Cooke579d2c2013-02-27 17:03:15 -08001312 set_dumpable(current->mm, SUID_DUMP_USER);
Alan Coxd6e71142005-06-23 00:09:43 -07001313 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001314 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001315
Peter Zijlstrae041e322014-05-21 17:32:19 +02001316 perf_event_exec();
Adrian Hunter82b89772014-05-28 11:45:04 +03001317 __set_task_comm(current, kbasename(bprm->filename), true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001319 /* Set the new mm task size. We have to do that late because it may
1320 * depend on TIF_32BIT which is only updated in flush_thread() on
1321 * some architectures like powerpc
1322 */
1323 current->mm->task_size = TASK_SIZE;
1324
David Howellsa6f76f22008-11-14 10:39:24 +11001325 /* install the new credentials */
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001326 if (!uid_eq(bprm->cred->uid, current_euid()) ||
1327 !gid_eq(bprm->cred->gid, current_egid())) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001328 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001329 } else {
Al Viro1b5d7832011-06-19 12:49:47 -04001330 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1331 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 }
1333
1334 /* An exec changes our domain. We are no longer part of the thread
1335 group */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 current->self_exec_id++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 flush_signal_handlers(current, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001339EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340
David Howellsa6f76f22008-11-14 10:39:24 +11001341/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001342 * Prepare credentials and lock ->cred_guard_mutex.
1343 * install_exec_creds() commits the new creds and drops the lock.
1344 * Or, if exec fails before, free_bprm() should release ->cred and
1345 * and unlock.
1346 */
1347int prepare_bprm_creds(struct linux_binprm *bprm)
1348{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001349 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001350 return -ERESTARTNOINTR;
1351
1352 bprm->cred = prepare_exec_creds();
1353 if (likely(bprm->cred))
1354 return 0;
1355
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001356 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001357 return -ENOMEM;
1358}
1359
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001360static void free_bprm(struct linux_binprm *bprm)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001361{
1362 free_arg_pages(bprm);
1363 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001364 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001365 abort_creds(bprm->cred);
1366 }
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001367 if (bprm->file) {
1368 allow_write_access(bprm->file);
1369 fput(bprm->file);
1370 }
Kees Cookb66c5982012-12-20 15:05:16 -08001371 /* If a binfmt changed the interp, free it. */
1372 if (bprm->interp != bprm->filename)
1373 kfree(bprm->interp);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001374 kfree(bprm);
1375}
1376
Kees Cookb66c5982012-12-20 15:05:16 -08001377int bprm_change_interp(char *interp, struct linux_binprm *bprm)
1378{
1379 /* If a binfmt changed the interp, free it first. */
1380 if (bprm->interp != bprm->filename)
1381 kfree(bprm->interp);
1382 bprm->interp = kstrdup(interp, GFP_KERNEL);
1383 if (!bprm->interp)
1384 return -ENOMEM;
1385 return 0;
1386}
1387EXPORT_SYMBOL(bprm_change_interp);
1388
Oleg Nesterova2a84742009-09-05 11:17:13 -07001389/*
David Howellsa6f76f22008-11-14 10:39:24 +11001390 * install the new credentials for this executable
1391 */
1392void install_exec_creds(struct linux_binprm *bprm)
1393{
1394 security_bprm_committing_creds(bprm);
1395
1396 commit_creds(bprm->cred);
1397 bprm->cred = NULL;
Stephane Eranian2976b102013-06-20 11:36:28 +02001398
1399 /*
1400 * Disable monitoring for regular users
1401 * when executing setuid binaries. Must
1402 * wait until new credentials are committed
1403 * by commit_creds() above
1404 */
1405 if (get_dumpable(current->mm) != SUID_DUMP_USER)
1406 perf_event_exit_task(current);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001407 /*
1408 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001409 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001410 * credentials; any time after this it may be unlocked.
1411 */
David Howellsa6f76f22008-11-14 10:39:24 +11001412 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001413 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001414}
1415EXPORT_SYMBOL(install_exec_creds);
1416
1417/*
1418 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001419 * - the caller must hold ->cred_guard_mutex to protect against
Kees Cookc2e1f2e2014-06-05 00:23:17 -07001420 * PTRACE_ATTACH or seccomp thread-sync
David Howellsa6f76f22008-11-14 10:39:24 +11001421 */
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001422static void check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001423{
David Howells0bf2f3a2009-02-06 11:45:46 +00001424 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001425 unsigned n_fs;
David Howellsa6f76f22008-11-14 10:39:24 +11001426
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001427 if (p->ptrace) {
Eric W. Biedermane747b4a2016-11-14 18:48:07 -06001428 if (ptracer_capable(p, current_user_ns()))
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001429 bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP;
1430 else
1431 bprm->unsafe |= LSM_UNSAFE_PTRACE;
1432 }
David Howellsa6f76f22008-11-14 10:39:24 +11001433
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001434 /*
1435 * This isn't strictly necessary, but it makes it harder for LSMs to
1436 * mess up.
1437 */
Kees Cook1d4457f2014-05-21 15:23:46 -07001438 if (task_no_new_privs(current))
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001439 bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS;
1440
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001441 t = p;
David Howells0bf2f3a2009-02-06 11:45:46 +00001442 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001443 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001444 rcu_read_lock();
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001445 while_each_thread(p, t) {
David Howells0bf2f3a2009-02-06 11:45:46 +00001446 if (t->fs == p->fs)
1447 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001448 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001449 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001450
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001451 if (p->fs->users > n_fs)
David Howellsa6f76f22008-11-14 10:39:24 +11001452 bprm->unsafe |= LSM_UNSAFE_SHARE;
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001453 else
1454 p->fs->in_exec = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001455 spin_unlock(&p->fs->lock);
David Howellsa6f76f22008-11-14 10:39:24 +11001456}
1457
Jann Horn8b01fc82015-04-19 02:48:39 +02001458static void bprm_fill_uid(struct linux_binprm *bprm)
1459{
1460 struct inode *inode;
1461 unsigned int mode;
1462 kuid_t uid;
1463 kgid_t gid;
1464
Kees Cookcb6fd682016-05-17 12:14:39 -07001465 /*
1466 * Since this can be called multiple times (via prepare_binprm),
1467 * we must clear any previous work done when setting set[ug]id
1468 * bits from any earlier bprm->file uses (for example when run
1469 * first for a setuid script then again for its interpreter).
1470 */
Jann Horn8b01fc82015-04-19 02:48:39 +02001471 bprm->cred->euid = current_euid();
1472 bprm->cred->egid = current_egid();
1473
Andy Lutomirski380cf5b2016-06-23 16:41:05 -05001474 if (!mnt_may_suid(bprm->file->f_path.mnt))
Jann Horn8b01fc82015-04-19 02:48:39 +02001475 return;
1476
1477 if (task_no_new_privs(current))
1478 return;
1479
1480 inode = file_inode(bprm->file);
1481 mode = READ_ONCE(inode->i_mode);
1482 if (!(mode & (S_ISUID|S_ISGID)))
1483 return;
1484
1485 /* Be careful if suid/sgid is set */
Al Viro59551022016-01-22 15:40:57 -05001486 inode_lock(inode);
Jann Horn8b01fc82015-04-19 02:48:39 +02001487
1488 /* reload atomically mode/uid/gid now that lock held */
1489 mode = inode->i_mode;
1490 uid = inode->i_uid;
1491 gid = inode->i_gid;
Al Viro59551022016-01-22 15:40:57 -05001492 inode_unlock(inode);
Jann Horn8b01fc82015-04-19 02:48:39 +02001493
1494 /* We ignore suid/sgid if there are no mappings for them in the ns */
1495 if (!kuid_has_mapping(bprm->cred->user_ns, uid) ||
1496 !kgid_has_mapping(bprm->cred->user_ns, gid))
1497 return;
1498
1499 if (mode & S_ISUID) {
1500 bprm->per_clear |= PER_CLEAR_ON_SETID;
1501 bprm->cred->euid = uid;
1502 }
1503
1504 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
1505 bprm->per_clear |= PER_CLEAR_ON_SETID;
1506 bprm->cred->egid = gid;
1507 }
1508}
1509
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001510/*
1511 * Fill the binprm structure from the inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001513 *
1514 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 */
1516int prepare_binprm(struct linux_binprm *bprm)
1517{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 int retval;
1519
Jann Horn8b01fc82015-04-19 02:48:39 +02001520 bprm_fill_uid(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521
1522 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001523 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 if (retval)
1525 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001526 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527
David Howellsa6f76f22008-11-14 10:39:24 +11001528 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1529 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530}
1531
1532EXPORT_SYMBOL(prepare_binprm);
1533
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001534/*
1535 * Arguments are '\0' separated strings found at the location bprm->p
1536 * points to; chop off the first by relocating brpm->p to right after
1537 * the first '\0' encountered.
1538 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001539int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001541 int ret = 0;
1542 unsigned long offset;
1543 char *kaddr;
1544 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001546 if (!bprm->argc)
1547 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001549 do {
1550 offset = bprm->p & ~PAGE_MASK;
1551 page = get_arg_page(bprm, bprm->p, 0);
1552 if (!page) {
1553 ret = -EFAULT;
1554 goto out;
1555 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001556 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001557
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001558 for (; offset < PAGE_SIZE && kaddr[offset];
1559 offset++, bprm->p++)
1560 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001561
Cong Wange8e3c3d2011-11-25 23:14:27 +08001562 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001563 put_arg_page(page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001564 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001565
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001566 bprm->p++;
1567 bprm->argc--;
1568 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001569
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001570out:
1571 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573EXPORT_SYMBOL(remove_arg_zero);
1574
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001575#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576/*
1577 * cycle the list of binary formats handler, until one recognizes the image
1578 */
Al Viro3c456bf2012-10-20 21:53:31 -04001579int search_binary_handler(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580{
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001581 bool need_retry = IS_ENABLED(CONFIG_MODULES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 struct linux_binfmt *fmt;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001583 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584
Kees Cookd7402692012-12-17 16:03:20 -08001585 /* This allows 4 levels of binfmt rewrites before failing hard. */
Oleg Nesterov131b2f92013-09-11 14:24:39 -07001586 if (bprm->recursion_depth > 5)
Kees Cookd7402692012-12-17 16:03:20 -08001587 return -ELOOP;
1588
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 retval = security_bprm_check(bprm);
1590 if (retval)
1591 return retval;
1592
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 retval = -ENOENT;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001594 retry:
1595 read_lock(&binfmt_lock);
1596 list_for_each_entry(fmt, &formats, lh) {
1597 if (!try_module_get(fmt->module))
1598 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 read_unlock(&binfmt_lock);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001600 bprm->recursion_depth++;
1601 retval = fmt->load_binary(bprm);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001602 read_lock(&binfmt_lock);
1603 put_binfmt(fmt);
Al Viro19d860a2014-05-04 20:11:36 -04001604 bprm->recursion_depth--;
1605 if (retval < 0 && !bprm->mm) {
1606 /* we got to flush_old_exec() and failed after it */
1607 read_unlock(&binfmt_lock);
1608 force_sigsegv(SIGSEGV, current);
1609 return retval;
1610 }
1611 if (retval != -ENOEXEC || !bprm->file) {
1612 read_unlock(&binfmt_lock);
1613 return retval;
1614 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 }
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001616 read_unlock(&binfmt_lock);
1617
Al Viro19d860a2014-05-04 20:11:36 -04001618 if (need_retry) {
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001619 if (printable(bprm->buf[0]) && printable(bprm->buf[1]) &&
1620 printable(bprm->buf[2]) && printable(bprm->buf[3]))
1621 return retval;
Oleg Nesterov4e0621a2013-09-11 14:24:45 -07001622 if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0)
1623 return retval;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001624 need_retry = false;
1625 goto retry;
1626 }
1627
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 return retval;
1629}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630EXPORT_SYMBOL(search_binary_handler);
1631
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001632static int exec_binprm(struct linux_binprm *bprm)
1633{
1634 pid_t old_pid, old_vpid;
1635 int ret;
1636
1637 /* Need to fetch pid before load_binary changes it */
1638 old_pid = current->pid;
1639 rcu_read_lock();
1640 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
1641 rcu_read_unlock();
1642
1643 ret = search_binary_handler(bprm);
1644 if (ret >= 0) {
Linus Torvalds3eaded862013-11-21 19:18:14 -08001645 audit_bprm(bprm);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001646 trace_sched_process_exec(current, old_pid, bprm);
1647 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
Oleg Nesterov9beb2662013-09-11 14:24:40 -07001648 proc_exec_connector(current);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001649 }
1650
1651 return ret;
1652}
1653
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654/*
1655 * sys_execve() executes a new program.
1656 */
David Drysdale51f39a12014-12-12 16:57:29 -08001657static int do_execveat_common(int fd, struct filename *filename,
1658 struct user_arg_ptr argv,
1659 struct user_arg_ptr envp,
1660 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661{
David Drysdale51f39a12014-12-12 16:57:29 -08001662 char *pathbuf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 struct linux_binprm *bprm;
1664 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001665 struct files_struct *displaced;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001667
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001668 if (IS_ERR(filename))
1669 return PTR_ERR(filename);
1670
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001671 /*
1672 * We move the actual failure in case of RLIMIT_NPROC excess from
1673 * set*uid() to execve() because too many poorly written programs
1674 * don't check setuid() return code. Here we additionally recheck
1675 * whether NPROC limit is still exceeded.
1676 */
1677 if ((current->flags & PF_NPROC_EXCEEDED) &&
Oleg Nesterovbd9d43f2013-07-03 15:08:34 -07001678 atomic_read(&current_user()->processes) > rlimit(RLIMIT_NPROC)) {
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001679 retval = -EAGAIN;
1680 goto out_ret;
1681 }
1682
1683 /* We're below the limit (still or again), so we don't want to make
1684 * further execve() calls fail. */
1685 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686
Al Viro3b125382008-04-22 05:31:30 -04001687 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001688 if (retval)
1689 goto out_ret;
1690
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001692 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001694 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695
Oleg Nesterova2a84742009-09-05 11:17:13 -07001696 retval = prepare_bprm_creds(bprm);
1697 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001698 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001699
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001700 check_unsafe_exec(bprm);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001701 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001702
David Drysdale51f39a12014-12-12 16:57:29 -08001703 file = do_open_execat(fd, filename, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 retval = PTR_ERR(file);
1705 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001706 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707
1708 sched_exec();
1709
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 bprm->file = file;
David Drysdale51f39a12014-12-12 16:57:29 -08001711 if (fd == AT_FDCWD || filename->name[0] == '/') {
1712 bprm->filename = filename->name;
1713 } else {
1714 if (filename->name[0] == '\0')
1715 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d", fd);
1716 else
1717 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d/%s",
1718 fd, filename->name);
1719 if (!pathbuf) {
1720 retval = -ENOMEM;
1721 goto out_unmark;
1722 }
1723 /*
1724 * Record that a name derived from an O_CLOEXEC fd will be
1725 * inaccessible after exec. Relies on having exclusive access to
1726 * current->files (due to unshare_files above).
1727 */
1728 if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt)))
1729 bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE;
1730 bprm->filename = pathbuf;
1731 }
1732 bprm->interp = bprm->filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001733
1734 retval = bprm_mm_init(bprm);
1735 if (retval)
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001736 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001738 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001740 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001742 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 goto out;
1745
1746 retval = prepare_binprm(bprm);
1747 if (retval < 0)
1748 goto out;
1749
1750 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1751 if (retval < 0)
1752 goto out;
1753
1754 bprm->exec = bprm->p;
1755 retval = copy_strings(bprm->envc, envp, bprm);
1756 if (retval < 0)
1757 goto out;
1758
1759 retval = copy_strings(bprm->argc, argv, bprm);
1760 if (retval < 0)
1761 goto out;
1762
Eric W. Biederman21245b82016-11-16 22:06:51 -06001763 would_dump(bprm, bprm->file);
1764
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001765 retval = exec_binprm(bprm);
David Howellsa6f76f22008-11-14 10:39:24 +11001766 if (retval < 0)
1767 goto out;
1768
1769 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001770 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001771 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001772 acct_update_integrals(current);
Rik van Riel82727012013-10-07 11:29:28 +01001773 task_numa_free(current);
David Howellsa6f76f22008-11-14 10:39:24 +11001774 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001775 kfree(pathbuf);
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001776 putname(filename);
David Howellsa6f76f22008-11-14 10:39:24 +11001777 if (displaced)
1778 put_files_struct(displaced);
1779 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780
1781out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001782 if (bprm->mm) {
1783 acct_arg_size(bprm, 0);
1784 mmput(bprm->mm);
1785 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786
Al Viro498052b2009-03-30 07:20:30 -04001787out_unmark:
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001788 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001789 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001790
1791out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001792 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001793 kfree(pathbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794
Al Virofd8328b2008-04-22 05:11:59 -04001795out_files:
Al Viro3b125382008-04-22 05:31:30 -04001796 if (displaced)
1797 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798out_ret:
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001799 putname(filename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 return retval;
1801}
1802
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001803int do_execve(struct filename *filename,
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001804 const char __user *const __user *__argv,
Al Viroda3d4c52012-10-20 21:49:33 -04001805 const char __user *const __user *__envp)
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001806{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001807 struct user_arg_ptr argv = { .ptr.native = __argv };
1808 struct user_arg_ptr envp = { .ptr.native = __envp };
David Drysdale51f39a12014-12-12 16:57:29 -08001809 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1810}
1811
1812int do_execveat(int fd, struct filename *filename,
1813 const char __user *const __user *__argv,
1814 const char __user *const __user *__envp,
1815 int flags)
1816{
1817 struct user_arg_ptr argv = { .ptr.native = __argv };
1818 struct user_arg_ptr envp = { .ptr.native = __envp };
1819
1820 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001821}
1822
Oleg Nesterov0e028462011-03-06 18:02:54 +01001823#ifdef CONFIG_COMPAT
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001824static int compat_do_execve(struct filename *filename,
Al Viro38b983b2012-09-30 13:38:55 -04001825 const compat_uptr_t __user *__argv,
Al Virod03d26e2012-10-20 21:46:25 -04001826 const compat_uptr_t __user *__envp)
Oleg Nesterov0e028462011-03-06 18:02:54 +01001827{
1828 struct user_arg_ptr argv = {
1829 .is_compat = true,
1830 .ptr.compat = __argv,
1831 };
1832 struct user_arg_ptr envp = {
1833 .is_compat = true,
1834 .ptr.compat = __envp,
1835 };
David Drysdale51f39a12014-12-12 16:57:29 -08001836 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1837}
1838
1839static int compat_do_execveat(int fd, struct filename *filename,
1840 const compat_uptr_t __user *__argv,
1841 const compat_uptr_t __user *__envp,
1842 int flags)
1843{
1844 struct user_arg_ptr argv = {
1845 .is_compat = true,
1846 .ptr.compat = __argv,
1847 };
1848 struct user_arg_ptr envp = {
1849 .is_compat = true,
1850 .ptr.compat = __envp,
1851 };
1852 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterov0e028462011-03-06 18:02:54 +01001853}
1854#endif
1855
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001856void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001858 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001860 if (mm->binfmt)
1861 module_put(mm->binfmt->module);
1862
1863 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001864 if (new)
1865 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867EXPORT_SYMBOL(set_binfmt);
1868
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001869/*
Oleg Nesterov7288e112014-01-23 15:55:32 -08001870 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001871 */
1872void set_dumpable(struct mm_struct *mm, int value)
1873{
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001874 unsigned long old, new;
1875
Oleg Nesterov7288e112014-01-23 15:55:32 -08001876 if (WARN_ON((unsigned)value > SUID_DUMP_ROOT))
1877 return;
1878
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001879 do {
1880 old = ACCESS_ONCE(mm->flags);
Oleg Nesterov7288e112014-01-23 15:55:32 -08001881 new = (old & ~MMF_DUMPABLE_MASK) | value;
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001882 } while (cmpxchg(&mm->flags, old, new) != old);
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001883}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001884
Al Viro38b983b2012-09-30 13:38:55 -04001885SYSCALL_DEFINE3(execve,
1886 const char __user *, filename,
1887 const char __user *const __user *, argv,
1888 const char __user *const __user *, envp)
1889{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001890 return do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001891}
David Drysdale51f39a12014-12-12 16:57:29 -08001892
1893SYSCALL_DEFINE5(execveat,
1894 int, fd, const char __user *, filename,
1895 const char __user *const __user *, argv,
1896 const char __user *const __user *, envp,
1897 int, flags)
1898{
1899 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1900
1901 return do_execveat(fd,
1902 getname_flags(filename, lookup_flags, NULL),
1903 argv, envp, flags);
1904}
1905
Al Viro38b983b2012-09-30 13:38:55 -04001906#ifdef CONFIG_COMPAT
Heiko Carstens625b1d72014-03-04 10:53:50 +01001907COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename,
1908 const compat_uptr_t __user *, argv,
1909 const compat_uptr_t __user *, envp)
Al Viro38b983b2012-09-30 13:38:55 -04001910{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001911 return compat_do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001912}
David Drysdale51f39a12014-12-12 16:57:29 -08001913
1914COMPAT_SYSCALL_DEFINE5(execveat, int, fd,
1915 const char __user *, filename,
1916 const compat_uptr_t __user *, argv,
1917 const compat_uptr_t __user *, envp,
1918 int, flags)
1919{
1920 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1921
1922 return compat_do_execveat(fd,
1923 getname_flags(filename, lookup_flags, NULL),
1924 argv, envp, flags);
1925}
Al Viro38b983b2012-09-30 13:38:55 -04001926#endif