blob: e857c7260b1f88b71b56a42a7f25821073b0aa41 [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 Sarai613cc2b2016-12-21 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>
Ingo Molnar6e84f312017-02-08 18:51:29 +010035#include <linux/sched/mm.h>
Ingo Molnarf7ccbae2017-02-08 18:51:30 +010036#include <linux/sched/coredump.h>
Ingo Molnar3f07c012017-02-08 18:51:30 +010037#include <linux/sched/signal.h>
Ingo Molnar6a3827d2017-02-08 18:51:31 +010038#include <linux/sched/numa_balancing.h>
Hugh Dickinsca5b1722008-07-28 15:46:18 -070039#include <linux/pagemap.h>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020040#include <linux/perf_event.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#include <linux/highmem.h>
42#include <linux/spinlock.h>
43#include <linux/key.h>
44#include <linux/personality.h>
45#include <linux/binfmts.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070046#include <linux/utsname.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080047#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#include <linux/module.h>
49#include <linux/namei.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050#include <linux/mount.h>
51#include <linux/security.h>
52#include <linux/syscalls.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070053#include <linux/tsacct_kern.h>
Matt Helsley9f460802005-11-07 00:59:16 -080054#include <linux/cn_proc.h>
Al Viro473ae302006-04-26 14:04:08 -040055#include <linux/audit.h>
Roland McGrath6341c392008-07-25 19:45:44 -070056#include <linux/tracehook.h>
Johannes Berg5f4123b2008-07-09 10:28:40 +020057#include <linux/kmod.h>
Eric Paris6110e3a2008-12-17 13:53:20 -050058#include <linux/fsnotify.h>
Al Viro5ad4e532009-03-29 19:50:06 -040059#include <linux/fs_struct.h>
Neil Horman61be2282009-09-23 15:56:58 -070060#include <linux/pipe_fs_i.h>
Ying Han3d5992d2010-10-26 14:21:23 -070061#include <linux/oom.h>
Oleg Nesterov0e028462011-03-06 18:02:54 +010062#include <linux/compat.h>
Mimi Zoharb44a7df2015-12-28 16:02:29 -050063#include <linux/vmalloc.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070064
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080065#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070066#include <asm/mmu_context.h>
Ollie Wildb6a2fea2007-07-19 01:48:16 -070067#include <asm/tlb.h>
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080068
69#include <trace/events/task.h>
David Howellsa6f76f22008-11-14 10:39:24 +110070#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070071
David Smith4ff16c22012-02-07 10:11:05 -060072#include <trace/events/sched.h>
73
Alan Coxd6e71142005-06-23 00:09:43 -070074int suid_dumpable = 0;
75
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070076static LIST_HEAD(formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070077static DEFINE_RWLOCK(binfmt_lock);
78
Al Viro8fc3dc52012-03-17 03:05:16 -040079void __register_binfmt(struct linux_binfmt * fmt, int insert)
Linus Torvalds1da177e2005-04-16 15:20:36 -070080{
Al Viro8fc3dc52012-03-17 03:05:16 -040081 BUG_ON(!fmt);
Oleg Nesterov92eaa562013-09-11 14:24:42 -070082 if (WARN_ON(!fmt->load_binary))
83 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -070084 write_lock(&binfmt_lock);
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070085 insert ? list_add(&fmt->lh, &formats) :
86 list_add_tail(&fmt->lh, &formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070088}
89
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070090EXPORT_SYMBOL(__register_binfmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
Alexey Dobriyanf6b450d2007-10-16 23:26:04 -070092void unregister_binfmt(struct linux_binfmt * fmt)
Linus Torvalds1da177e2005-04-16 15:20:36 -070093{
Linus Torvalds1da177e2005-04-16 15:20:36 -070094 write_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070095 list_del(&fmt->lh);
Linus Torvalds1da177e2005-04-16 15:20:36 -070096 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070097}
98
99EXPORT_SYMBOL(unregister_binfmt);
100
101static inline void put_binfmt(struct linux_binfmt * fmt)
102{
103 module_put(fmt->module);
104}
105
Eric W. Biederman90f85722015-06-29 14:42:03 -0500106bool path_noexec(const struct path *path)
107{
108 return (path->mnt->mnt_flags & MNT_NOEXEC) ||
109 (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
110}
111
Josh Triplett69369a72014-04-03 14:48:27 -0700112#ifdef CONFIG_USELIB
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113/*
114 * Note that a shared library must be both readable and executable due to
115 * security reasons.
116 *
117 * Also note that we take the address to load from from the file itself.
118 */
Heiko Carstens1e7bfb22009-01-14 14:14:29 +0100119SYSCALL_DEFINE1(uselib, const char __user *, library)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120{
Al Viro72c2d532013-09-22 16:27:52 -0400121 struct linux_binfmt *fmt;
Al Viro964bd182008-07-26 03:33:14 -0400122 struct file *file;
Jeff Layton91a27b22012-10-10 15:25:28 -0400123 struct filename *tmp = getname(library);
Al Viro964bd182008-07-26 03:33:14 -0400124 int error = PTR_ERR(tmp);
Al Viro47c805d2011-02-23 17:44:09 -0500125 static const struct open_flags uselib_flags = {
126 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
Al Viro62fb4a12015-12-26 22:33:24 -0500127 .acc_mode = MAY_READ | MAY_EXEC,
Al Virof9652e12013-06-11 08:23:01 +0400128 .intent = LOOKUP_OPEN,
129 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500130 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131
Al Viro6e8341a2009-04-06 11:16:22 -0400132 if (IS_ERR(tmp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 goto out;
134
Al Virof9652e12013-06-11 08:23:01 +0400135 file = do_filp_open(AT_FDCWD, tmp, &uselib_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400136 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 error = PTR_ERR(file);
138 if (IS_ERR(file))
139 goto out;
140
Al Viro6e8341a2009-04-06 11:16:22 -0400141 error = -EINVAL;
Al Viro496ad9a2013-01-23 17:07:38 -0500142 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400143 goto exit;
144
145 error = -EACCES;
Eric W. Biederman90f85722015-06-29 14:42:03 -0500146 if (path_noexec(&file->f_path))
Al Viro6e8341a2009-04-06 11:16:22 -0400147 goto exit;
148
Eric Paris2a12a9d2009-12-17 21:24:21 -0500149 fsnotify_open(file);
Eric Paris6110e3a2008-12-17 13:53:20 -0500150
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 error = -ENOEXEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152
Al Viro72c2d532013-09-22 16:27:52 -0400153 read_lock(&binfmt_lock);
154 list_for_each_entry(fmt, &formats, lh) {
155 if (!fmt->load_shlib)
156 continue;
157 if (!try_module_get(fmt->module))
158 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 read_unlock(&binfmt_lock);
Al Viro72c2d532013-09-22 16:27:52 -0400160 error = fmt->load_shlib(file);
161 read_lock(&binfmt_lock);
162 put_binfmt(fmt);
163 if (error != -ENOEXEC)
164 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 }
Al Viro72c2d532013-09-22 16:27:52 -0400166 read_unlock(&binfmt_lock);
Al Viro6e8341a2009-04-06 11:16:22 -0400167exit:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 fput(file);
169out:
170 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171}
Josh Triplett69369a72014-04-03 14:48:27 -0700172#endif /* #ifdef CONFIG_USELIB */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700174#ifdef CONFIG_MMU
Oleg Nesterovae6b5852011-03-06 18:03:11 +0100175/*
176 * The nascent bprm->mm is not visible until exec_mmap() but it can
177 * use a lot of memory, account these pages in current->mm temporary
178 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
179 * change the counter back via acct_arg_size(0).
180 */
Oleg Nesterov0e028462011-03-06 18:02:54 +0100181static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100182{
183 struct mm_struct *mm = current->mm;
184 long diff = (long)(pages - bprm->vma_pages);
185
186 if (!mm || !diff)
187 return;
188
189 bprm->vma_pages = pages;
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100190 add_mm_counter(mm, MM_ANONPAGES, diff);
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100191}
192
Oleg Nesterov0e028462011-03-06 18:02:54 +0100193static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700194 int write)
195{
196 struct page *page;
197 int ret;
Lorenzo Stoakes9beae1e2016-10-13 01:20:17 +0100198 unsigned int gup_flags = FOLL_FORCE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700199
200#ifdef CONFIG_STACK_GROWSUP
201 if (write) {
Michal Hockod05f3162011-05-24 17:11:44 -0700202 ret = expand_downwards(bprm->vma, pos);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700203 if (ret < 0)
204 return NULL;
205 }
206#endif
Lorenzo Stoakes9beae1e2016-10-13 01:20:17 +0100207
208 if (write)
209 gup_flags |= FOLL_WRITE;
210
Dave Hansen1e987792016-02-12 13:01:54 -0800211 /*
212 * We are doing an exec(). 'current' is the process
213 * doing the exec and bprm->mm is the new process's mm.
214 */
Lorenzo Stoakes9beae1e2016-10-13 01:20:17 +0100215 ret = get_user_pages_remote(current, bprm->mm, pos, 1, gup_flags,
Lorenzo Stoakes5b56d492016-12-14 15:06:52 -0800216 &page, NULL, NULL);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700217 if (ret <= 0)
218 return NULL;
219
220 if (write) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700221 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800222 struct rlimit *rlim;
223
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100224 acct_arg_size(bprm, size / PAGE_SIZE);
225
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800226 /*
227 * We've historically supported up to 32 pages (ARG_MAX)
228 * of argument strings even with small stacks
229 */
230 if (size <= ARG_MAX)
231 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700232
233 /*
234 * Limit to 1/4-th the stack size for the argv+env strings.
235 * This ensures that:
236 * - the remaining binfmt code will not run out of stack space,
237 * - the program will have a reasonable amount of stack left
238 * to work from.
239 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800240 rlim = current->signal->rlim;
Jiri Slabyd554ed892010-03-05 13:42:42 -0800241 if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700242 put_page(page);
243 return NULL;
244 }
245 }
246
247 return page;
248}
249
250static void put_arg_page(struct page *page)
251{
252 put_page(page);
253}
254
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700255static void free_arg_pages(struct linux_binprm *bprm)
256{
257}
258
259static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
260 struct page *page)
261{
262 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
263}
264
265static int __bprm_mm_init(struct linux_binprm *bprm)
266{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800267 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700268 struct vm_area_struct *vma = NULL;
269 struct mm_struct *mm = bprm->mm;
270
271 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
272 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800273 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700274
Michal Hockof268dfe2016-05-23 16:26:02 -0700275 if (down_write_killable(&mm->mmap_sem)) {
276 err = -EINTR;
277 goto err_free;
278 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700279 vma->vm_mm = mm;
280
281 /*
282 * Place the stack at the largest stack address the architecture
283 * supports. Later, we'll move this to an appropriate place. We don't
284 * use STACK_TOP because that can depend on attributes which aren't
285 * configured yet.
286 */
Michal Hockoaacb3d12011-07-26 16:08:40 -0700287 BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700288 vma->vm_end = STACK_TOP_MAX;
289 vma->vm_start = vma->vm_end - PAGE_SIZE;
Cyrill Gorcunovd9104d12013-09-11 14:22:24 -0700290 vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP;
Coly Li3ed75eb2007-10-18 23:39:15 -0700291 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Rik van Riel5beb4932010-03-05 13:42:07 -0800292 INIT_LIST_HEAD(&vma->anon_vma_chain);
Tavis Ormandy462e635e2010-12-09 15:29:42 +0100293
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700294 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800295 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700296 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700297
298 mm->stack_vm = mm->total_vm = 1;
Dave Hansenfe3d1972014-11-14 07:18:29 -0800299 arch_bprm_mm_init(mm, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700300 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700301 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700302 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700303err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800304 up_write(&mm->mmap_sem);
Michal Hockof268dfe2016-05-23 16:26:02 -0700305err_free:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800306 bprm->vma = NULL;
307 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700308 return err;
309}
310
311static bool valid_arg_len(struct linux_binprm *bprm, long len)
312{
313 return len <= MAX_ARG_STRLEN;
314}
315
316#else
317
Oleg Nesterov0e028462011-03-06 18:02:54 +0100318static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100319{
320}
321
Oleg Nesterov0e028462011-03-06 18:02:54 +0100322static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700323 int write)
324{
325 struct page *page;
326
327 page = bprm->page[pos / PAGE_SIZE];
328 if (!page && write) {
329 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
330 if (!page)
331 return NULL;
332 bprm->page[pos / PAGE_SIZE] = page;
333 }
334
335 return page;
336}
337
338static void put_arg_page(struct page *page)
339{
340}
341
342static void free_arg_page(struct linux_binprm *bprm, int i)
343{
344 if (bprm->page[i]) {
345 __free_page(bprm->page[i]);
346 bprm->page[i] = NULL;
347 }
348}
349
350static void free_arg_pages(struct linux_binprm *bprm)
351{
352 int i;
353
354 for (i = 0; i < MAX_ARG_PAGES; i++)
355 free_arg_page(bprm, i);
356}
357
358static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
359 struct page *page)
360{
361}
362
363static int __bprm_mm_init(struct linux_binprm *bprm)
364{
365 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
366 return 0;
367}
368
369static bool valid_arg_len(struct linux_binprm *bprm, long len)
370{
371 return len <= bprm->p;
372}
373
374#endif /* CONFIG_MMU */
375
376/*
377 * Create a new mm_struct and populate it with a temporary stack
378 * vm_area_struct. We don't have enough context at this point to set the stack
379 * flags, permissions, and offset, so we use temporary values. We'll update
380 * them later in setup_arg_pages().
381 */
Yuanhan Liu9cc64ce2013-02-20 13:16:01 +1100382static int bprm_mm_init(struct linux_binprm *bprm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700383{
384 int err;
385 struct mm_struct *mm = NULL;
386
387 bprm->mm = mm = mm_alloc();
388 err = -ENOMEM;
389 if (!mm)
390 goto err;
391
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700392 err = __bprm_mm_init(bprm);
393 if (err)
394 goto err;
395
396 return 0;
397
398err:
399 if (mm) {
400 bprm->mm = NULL;
401 mmdrop(mm);
402 }
403
404 return err;
405}
406
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100407struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100408#ifdef CONFIG_COMPAT
409 bool is_compat;
410#endif
411 union {
412 const char __user *const __user *native;
413#ifdef CONFIG_COMPAT
Al Viro38b983b2012-09-30 13:38:55 -0400414 const compat_uptr_t __user *compat;
Oleg Nesterov0e028462011-03-06 18:02:54 +0100415#endif
416 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100417};
418
419static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100420{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100421 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100422
Oleg Nesterov0e028462011-03-06 18:02:54 +0100423#ifdef CONFIG_COMPAT
424 if (unlikely(argv.is_compat)) {
425 compat_uptr_t compat;
426
427 if (get_user(compat, argv.ptr.compat + nr))
428 return ERR_PTR(-EFAULT);
429
430 return compat_ptr(compat);
431 }
432#endif
433
434 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100435 return ERR_PTR(-EFAULT);
436
Oleg Nesterov0e028462011-03-06 18:02:54 +0100437 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100438}
439
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440/*
441 * count() counts the number of strings in array ARGV.
442 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100443static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444{
445 int i = 0;
446
Oleg Nesterov0e028462011-03-06 18:02:54 +0100447 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100449 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 if (!p)
452 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100453
454 if (IS_ERR(p))
455 return -EFAULT;
456
Xi Wang6d92d4f2013-01-11 14:31:48 -0800457 if (i >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 return -E2BIG;
Xi Wang6d92d4f2013-01-11 14:31:48 -0800459 ++i;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700460
461 if (fatal_signal_pending(current))
462 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 cond_resched();
464 }
465 }
466 return i;
467}
468
469/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700470 * 'copy_strings()' copies argument/environment strings from the old
471 * processes's memory to the new process's stack. The call to get_user_pages()
472 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100474static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700475 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476{
477 struct page *kmapped_page = NULL;
478 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700479 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 int ret;
481
482 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100483 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 int len;
485 unsigned long pos;
486
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100487 ret = -EFAULT;
488 str = get_user_arg_ptr(argv, argc);
489 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100492 len = strnlen_user(str, MAX_ARG_STRLEN);
493 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100495
496 ret = -E2BIG;
497 if (!valid_arg_len(bprm, len))
498 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700500 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700502 str += len;
503 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
505 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Roland McGrath9aea5a62010-09-07 19:37:06 -0700508 if (fatal_signal_pending(current)) {
509 ret = -ERESTARTNOHAND;
510 goto out;
511 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700512 cond_resched();
513
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700515 if (offset == 0)
516 offset = PAGE_SIZE;
517
518 bytes_to_copy = offset;
519 if (bytes_to_copy > len)
520 bytes_to_copy = len;
521
522 offset -= bytes_to_copy;
523 pos -= bytes_to_copy;
524 str -= bytes_to_copy;
525 len -= bytes_to_copy;
526
527 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
528 struct page *page;
529
530 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700532 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 goto out;
534 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700536 if (kmapped_page) {
537 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700539 put_arg_page(kmapped_page);
540 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 kmapped_page = page;
542 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700543 kpos = pos & PAGE_MASK;
544 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700546 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 ret = -EFAULT;
548 goto out;
549 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 }
551 }
552 ret = 0;
553out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700554 if (kmapped_page) {
555 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700557 put_arg_page(kmapped_page);
558 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 return ret;
560}
561
562/*
563 * Like copy_strings, but get argv and its values from kernel memory.
564 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100565int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100566 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567{
568 int r;
569 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100570 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100571 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100572 };
573
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100575 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100577
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 return r;
579}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580EXPORT_SYMBOL(copy_strings_kernel);
581
582#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700583
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700585 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
586 * the binfmt code determines where the new stack should reside, we shift it to
587 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700589 * 1) Use shift to calculate the new vma endpoints.
590 * 2) Extend vma to cover both the old and new ranges. This ensures the
591 * arguments passed to subsequent functions are consistent.
592 * 3) Move vma's page tables to the new range.
593 * 4) Free up any cleared pgd range.
594 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700596static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597{
598 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700599 unsigned long old_start = vma->vm_start;
600 unsigned long old_end = vma->vm_end;
601 unsigned long length = old_end - old_start;
602 unsigned long new_start = old_start - shift;
603 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700604 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700606 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700608 /*
609 * ensure there are no vmas between where we want to go
610 * and where we are
611 */
612 if (vma != find_vma(mm, new_start))
613 return -EFAULT;
614
615 /*
616 * cover the whole range: [new_start, old_end)
617 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800618 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
619 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700620
621 /*
622 * move the page tables downwards, on failure we rely on
623 * process cleanup to remove whatever mess we made.
624 */
625 if (length != move_page_tables(vma, old_start,
Michel Lespinasse38a76012012-10-08 16:31:50 -0700626 vma, new_start, length, false))
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700627 return -ENOMEM;
628
629 lru_add_drain();
Linus Torvalds2b047252013-08-15 11:42:25 -0700630 tlb_gather_mmu(&tlb, mm, old_start, old_end);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700631 if (new_end > old_start) {
632 /*
633 * when the old and new regions overlap clear from new_end.
634 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700635 free_pgd_range(&tlb, new_end, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700636 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700637 } else {
638 /*
639 * otherwise, clean from old_start; this is done to not touch
640 * the address space in [new_end, old_start) some architectures
641 * have constraints on va-space that make this illegal (IA64) -
642 * for the others its just a little faster.
643 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700644 free_pgd_range(&tlb, old_start, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700645 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 }
Linus Torvalds2b047252013-08-15 11:42:25 -0700647 tlb_finish_mmu(&tlb, old_start, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700649 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800650 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700651 */
652 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
653
654 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655}
656
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700657/*
658 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
659 * the stack is optionally relocated, and some extra space is added.
660 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661int setup_arg_pages(struct linux_binprm *bprm,
662 unsigned long stack_top,
663 int executable_stack)
664{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700665 unsigned long ret;
666 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700668 struct vm_area_struct *vma = bprm->vma;
669 struct vm_area_struct *prev = NULL;
670 unsigned long vm_flags;
671 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800672 unsigned long stack_size;
673 unsigned long stack_expand;
674 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675
676#ifdef CONFIG_STACK_GROWSUP
James Hogand71f2902014-05-13 23:58:24 +0100677 /* Limit stack size */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800678 stack_base = rlimit_max(RLIMIT_STACK);
James Hogand71f2902014-05-13 23:58:24 +0100679 if (stack_base > STACK_SIZE_MAX)
680 stack_base = STACK_SIZE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681
Helge Dellerd045c772015-05-11 22:01:27 +0200682 /* Add space for stack randomization. */
683 stack_base += (STACK_RND_MASK << PAGE_SHIFT);
684
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700685 /* Make sure we didn't let the argument array grow too large. */
686 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 return -ENOMEM;
688
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700689 stack_base = PAGE_ALIGN(stack_top - stack_base);
690
691 stack_shift = vma->vm_start - stack_base;
692 mm->arg_start = bprm->p - stack_shift;
693 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700695 stack_top = arch_align_stack(stack_top);
696 stack_top = PAGE_ALIGN(stack_top);
Roland McGrath1b528182010-09-07 19:35:49 -0700697
698 if (unlikely(stack_top < mmap_min_addr) ||
699 unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
700 return -ENOMEM;
701
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700702 stack_shift = vma->vm_end - stack_top;
703
704 bprm->p -= stack_shift;
705 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700707
708 if (bprm->loader)
709 bprm->loader -= stack_shift;
710 bprm->exec -= stack_shift;
711
Michal Hockof268dfe2016-05-23 16:26:02 -0700712 if (down_write_killable(&mm->mmap_sem))
713 return -EINTR;
714
Hugh Dickins96a8e132008-07-10 21:19:20 +0100715 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700716
717 /*
718 * Adjust stack execute permissions; explicitly enable for
719 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
720 * (arch default) otherwise.
721 */
722 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
723 vm_flags |= VM_EXEC;
724 else if (executable_stack == EXSTACK_DISABLE_X)
725 vm_flags &= ~VM_EXEC;
726 vm_flags |= mm->def_flags;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700727 vm_flags |= VM_STACK_INCOMPLETE_SETUP;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700728
729 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
730 vm_flags);
731 if (ret)
732 goto out_unlock;
733 BUG_ON(prev != vma);
734
735 /* Move stack pages down in memory. */
736 if (stack_shift) {
737 ret = shift_arg_pages(vma, stack_shift);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800738 if (ret)
739 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 }
741
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700742 /* mprotect_fixup is overkill to remove the temporary stack flags */
743 vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
744
Michael Neuling5ef097d2010-03-05 13:42:57 -0800745 stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */
Michael Neuling803bf5e2010-02-10 13:56:42 -0800746 stack_size = vma->vm_end - vma->vm_start;
747 /*
748 * Align this down to a page boundary as expand_stack
749 * will align it up.
750 */
751 rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700752#ifdef CONFIG_STACK_GROWSUP
Michael Neuling803bf5e2010-02-10 13:56:42 -0800753 if (stack_size + stack_expand > rlim_stack)
754 stack_base = vma->vm_start + rlim_stack;
755 else
756 stack_base = vma->vm_end + stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700757#else
Michael Neuling803bf5e2010-02-10 13:56:42 -0800758 if (stack_size + stack_expand > rlim_stack)
759 stack_base = vma->vm_end - rlim_stack;
760 else
761 stack_base = vma->vm_start - stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700762#endif
Eric B Munson3af9e852010-05-18 15:30:49 +0100763 current->mm->start_stack = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700764 ret = expand_stack(vma, stack_base);
765 if (ret)
766 ret = -EFAULT;
767
768out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 up_write(&mm->mmap_sem);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800770 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772EXPORT_SYMBOL(setup_arg_pages);
773
Nicolas Pitre7e7ec6a2016-07-24 11:30:18 -0400774#else
775
776/*
777 * Transfer the program arguments and environment from the holding pages
778 * onto the stack. The provided stack pointer is adjusted accordingly.
779 */
780int transfer_args_to_stack(struct linux_binprm *bprm,
781 unsigned long *sp_location)
782{
783 unsigned long index, stop, sp;
784 int ret = 0;
785
786 stop = bprm->p >> PAGE_SHIFT;
787 sp = *sp_location;
788
789 for (index = MAX_ARG_PAGES - 1; index >= stop; index--) {
790 unsigned int offset = index == stop ? bprm->p & ~PAGE_MASK : 0;
791 char *src = kmap(bprm->page[index]) + offset;
792 sp -= PAGE_SIZE - offset;
793 if (copy_to_user((void *) sp, src, PAGE_SIZE - offset) != 0)
794 ret = -EFAULT;
795 kunmap(bprm->page[index]);
796 if (ret)
797 goto out;
798 }
799
800 *sp_location = sp;
801
802out:
803 return ret;
804}
805EXPORT_SYMBOL(transfer_args_to_stack);
806
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807#endif /* CONFIG_MMU */
808
David Drysdale51f39a12014-12-12 16:57:29 -0800809static struct file *do_open_execat(int fd, struct filename *name, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200812 int err;
David Drysdale51f39a12014-12-12 16:57:29 -0800813 struct open_flags open_exec_flags = {
Al Viro47c805d2011-02-23 17:44:09 -0500814 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
Al Viro62fb4a12015-12-26 22:33:24 -0500815 .acc_mode = MAY_EXEC,
Al Virof9652e12013-06-11 08:23:01 +0400816 .intent = LOOKUP_OPEN,
817 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500818 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819
David Drysdale51f39a12014-12-12 16:57:29 -0800820 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
821 return ERR_PTR(-EINVAL);
822 if (flags & AT_SYMLINK_NOFOLLOW)
823 open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW;
824 if (flags & AT_EMPTY_PATH)
825 open_exec_flags.lookup_flags |= LOOKUP_EMPTY;
826
827 file = do_filp_open(fd, name, &open_exec_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400828 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200829 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200831 err = -EACCES;
Al Viro496ad9a2013-01-23 17:07:38 -0500832 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400833 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200834
Eric W. Biederman90f85722015-06-29 14:42:03 -0500835 if (path_noexec(&file->f_path))
Al Viro6e8341a2009-04-06 11:16:22 -0400836 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200837
838 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400839 if (err)
840 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
David Drysdale51f39a12014-12-12 16:57:29 -0800842 if (name->name[0] != '\0')
843 fsnotify_open(file);
844
Al Viro6e8341a2009-04-06 11:16:22 -0400845out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200846 return file;
847
Al Viro6e8341a2009-04-06 11:16:22 -0400848exit:
849 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200850 return ERR_PTR(err);
851}
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800852
853struct file *open_exec(const char *name)
854{
Paul Moore51689102015-01-22 00:00:03 -0500855 struct filename *filename = getname_kernel(name);
856 struct file *f = ERR_CAST(filename);
857
858 if (!IS_ERR(filename)) {
859 f = do_open_execat(AT_FDCWD, filename, 0);
860 putname(filename);
861 }
862 return f;
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800863}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864EXPORT_SYMBOL(open_exec);
865
Mimi Zohar6777d772009-08-21 14:32:48 -0400866int kernel_read(struct file *file, loff_t offset,
867 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868{
869 mm_segment_t old_fs;
870 loff_t pos = offset;
871 int result;
872
873 old_fs = get_fs();
874 set_fs(get_ds());
875 /* The cast to a user pointer is valid due to the set_fs() */
876 result = vfs_read(file, (void __user *)addr, count, &pos);
877 set_fs(old_fs);
878 return result;
879}
880
881EXPORT_SYMBOL(kernel_read);
882
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500883int kernel_read_file(struct file *file, void **buf, loff_t *size,
Mimi Zoharbc8ca5b2016-01-24 10:07:32 -0500884 loff_t max_size, enum kernel_read_file_id id)
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500885{
886 loff_t i_size, pos;
887 ssize_t bytes = 0;
888 int ret;
889
890 if (!S_ISREG(file_inode(file)->i_mode) || max_size < 0)
891 return -EINVAL;
892
Mimi Zohar39eeb4f2016-01-30 22:23:26 -0500893 ret = security_kernel_read_file(file, id);
894 if (ret)
895 return ret;
896
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200897 ret = deny_write_access(file);
898 if (ret)
899 return ret;
900
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500901 i_size = i_size_read(file_inode(file));
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200902 if (max_size > 0 && i_size > max_size) {
903 ret = -EFBIG;
904 goto out;
905 }
906 if (i_size <= 0) {
907 ret = -EINVAL;
908 goto out;
909 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500910
Stephen Boyda098ecd2016-08-02 14:04:28 -0700911 if (id != READING_FIRMWARE_PREALLOC_BUFFER)
912 *buf = vmalloc(i_size);
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200913 if (!*buf) {
914 ret = -ENOMEM;
915 goto out;
916 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500917
918 pos = 0;
919 while (pos < i_size) {
920 bytes = kernel_read(file, pos, (char *)(*buf) + pos,
921 i_size - pos);
922 if (bytes < 0) {
923 ret = bytes;
924 goto out;
925 }
926
927 if (bytes == 0)
928 break;
929 pos += bytes;
930 }
931
932 if (pos != i_size) {
933 ret = -EIO;
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200934 goto out_free;
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500935 }
936
Mimi Zoharbc8ca5b2016-01-24 10:07:32 -0500937 ret = security_kernel_post_read_file(file, *buf, i_size, id);
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500938 if (!ret)
939 *size = pos;
940
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200941out_free:
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500942 if (ret < 0) {
Stephen Boyda098ecd2016-08-02 14:04:28 -0700943 if (id != READING_FIRMWARE_PREALLOC_BUFFER) {
944 vfree(*buf);
945 *buf = NULL;
946 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500947 }
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200948
949out:
950 allow_write_access(file);
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500951 return ret;
952}
953EXPORT_SYMBOL_GPL(kernel_read_file);
954
Mimi Zohar09596b92015-11-19 12:39:22 -0500955int kernel_read_file_from_path(char *path, void **buf, loff_t *size,
956 loff_t max_size, enum kernel_read_file_id id)
957{
958 struct file *file;
959 int ret;
960
961 if (!path || !*path)
962 return -EINVAL;
963
964 file = filp_open(path, O_RDONLY, 0);
965 if (IS_ERR(file))
966 return PTR_ERR(file);
967
968 ret = kernel_read_file(file, buf, size, max_size, id);
969 fput(file);
970 return ret;
971}
972EXPORT_SYMBOL_GPL(kernel_read_file_from_path);
973
Mimi Zoharb844f0e2016-02-01 08:36:21 -0500974int kernel_read_file_from_fd(int fd, void **buf, loff_t *size, loff_t max_size,
975 enum kernel_read_file_id id)
976{
977 struct fd f = fdget(fd);
978 int ret = -EBADF;
979
980 if (!f.file)
981 goto out;
982
983 ret = kernel_read_file(f.file, buf, size, max_size, id);
984out:
985 fdput(f);
986 return ret;
987}
988EXPORT_SYMBOL_GPL(kernel_read_file_from_fd);
989
Al Viro3dc20cb2013-04-13 20:31:37 -0400990ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len)
991{
Al Viroec695572014-02-04 19:08:21 -0500992 ssize_t res = vfs_read(file, (void __user *)addr, len, &pos);
Al Viro3dc20cb2013-04-13 20:31:37 -0400993 if (res > 0)
994 flush_icache_range(addr, addr + len);
995 return res;
996}
997EXPORT_SYMBOL(read_code);
998
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999static int exec_mmap(struct mm_struct *mm)
1000{
1001 struct task_struct *tsk;
Davidlohr Bueso615d6e82014-04-07 15:37:25 -07001002 struct mm_struct *old_mm, *active_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
1004 /* Notify parent that we're no longer interested in the old VM */
1005 tsk = current;
1006 old_mm = current->mm;
1007 mm_release(tsk, old_mm);
1008
1009 if (old_mm) {
Konstantin Khlebnikov4fe7efd2012-06-20 12:53:01 -07001010 sync_mm_rss(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 /*
1012 * Make sure that if there is a core dump in progress
1013 * for the old mm, we get out and die instead of going
1014 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001015 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 */
1017 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001018 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 up_read(&old_mm->mmap_sem);
1020 return -EINTR;
1021 }
1022 }
1023 task_lock(tsk);
1024 active_mm = tsk->active_mm;
1025 tsk->mm = mm;
1026 tsk->active_mm = mm;
1027 activate_mm(active_mm, mm);
Davidlohr Bueso615d6e82014-04-07 15:37:25 -07001028 tsk->mm->vmacache_seqnum = 0;
1029 vmacache_flush(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 if (old_mm) {
1032 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +02001033 BUG_ON(active_mm != old_mm);
Oleg Nesterov701085b2012-03-19 17:04:01 +01001034 setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +01001035 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 mmput(old_mm);
1037 return 0;
1038 }
1039 mmdrop(active_mm);
1040 return 0;
1041}
1042
1043/*
1044 * This function makes sure the current process has its own signal table,
1045 * so that flush_signal_handlers can later reset the handlers without
1046 * disturbing other processes. (Other processes might share the signal
1047 * table via the CLONE_SIGHAND option to clone().)
1048 */
Arjan van de Ven858119e2006-01-14 13:20:43 -08001049static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050{
1051 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001052 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 spinlock_t *lock = &oldsighand->siglock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001055 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 goto no_thread_group;
1057
1058 /*
1059 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001062 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 /*
1064 * Another group action in progress, just
1065 * return so that the signal is processed.
1066 */
1067 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 return -EAGAIN;
1069 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070
Oleg Nesterovd3441932010-05-26 14:43:11 -07001071 sig->group_exit_task = tsk;
1072 sig->notify_count = zap_other_threads(tsk);
1073 if (!thread_group_leader(tsk))
1074 sig->notify_count--;
1075
1076 while (sig->notify_count) {
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001077 __set_current_state(TASK_KILLABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 spin_unlock_irq(lock);
1079 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001080 if (unlikely(__fatal_signal_pending(tsk)))
1081 goto killed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 spin_lock_irq(lock);
1083 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 spin_unlock_irq(lock);
1085
1086 /*
1087 * At this point all other threads have exited, all we have to
1088 * do is to wait for the thread group leader to become inactive,
1089 * and to assume its PID:
1090 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001091 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -08001092 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001093
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001094 for (;;) {
Ingo Molnar780de9d2017-02-02 11:50:56 +01001095 cgroup_threadgroup_change_begin(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001096 write_lock_irq(&tasklist_lock);
Kirill Tkhaidfcce792015-04-16 12:48:01 -07001097 /*
1098 * Do this under tasklist_lock to ensure that
1099 * exit_notify() can't miss ->group_exit_task
1100 */
1101 sig->notify_count = -1;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001102 if (likely(leader->exit_state))
1103 break;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001104 __set_current_state(TASK_KILLABLE);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001105 write_unlock_irq(&tasklist_lock);
Ingo Molnar780de9d2017-02-02 11:50:56 +01001106 cgroup_threadgroup_change_end(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001107 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001108 if (unlikely(__fatal_signal_pending(tsk)))
1109 goto killed;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001110 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111
Roland McGrathf5e90282006-04-10 22:54:16 -07001112 /*
1113 * The only record we have of the real-time age of a
1114 * process, regardless of execs it's done, is start_time.
1115 * All the past CPU time is accumulated in signal_struct
1116 * from sister threads now dead. But in this non-leader
1117 * exec, nothing survives from the original leader thread,
1118 * whose birth marks the true age of this process now.
1119 * When we take on its identity by switching to its PID, we
1120 * also take its birthdate (always earlier than our own).
1121 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001122 tsk->start_time = leader->start_time;
Oleg Nesterov266b7a02013-07-03 15:08:35 -07001123 tsk->real_start_time = leader->real_start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -07001124
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07001125 BUG_ON(!same_thread_group(leader, tsk));
1126 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 /*
1128 * An exec() starts a new thread group with the
1129 * TGID of the previous thread group. Rehash the
1130 * two threads with a switched PID, and release
1131 * the former thread group leader:
1132 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -08001133
1134 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -07001135 * The old leader becomes a thread of the this thread group.
1136 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -08001137 * is called. Odd but simple and correct.
1138 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001139 tsk->pid = leader->pid;
Oleg Nesterov3f418542013-07-03 15:08:25 -07001140 change_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001141 transfer_pid(leader, tsk, PIDTYPE_PGID);
1142 transfer_pid(leader, tsk, PIDTYPE_SID);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001143
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001144 list_replace_rcu(&leader->tasks, &tsk->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001145 list_replace_init(&leader->sibling, &tsk->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001147 tsk->group_leader = tsk;
1148 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -06001149
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001150 tsk->exit_signal = SIGCHLD;
Oleg Nesterov087806b2011-06-22 23:10:26 +02001151 leader->exit_signal = -1;
Oleg Nesterov962b5642005-11-23 13:37:43 -08001152
1153 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
1154 leader->exit_state = EXIT_DEAD;
Oleg Nesteroveac1b5e2011-07-21 20:00:43 +02001155
1156 /*
1157 * We are going to release_task()->ptrace_unlink() silently,
1158 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1159 * the tracer wont't block again waiting for this thread.
1160 */
1161 if (unlikely(leader->ptrace))
1162 __wake_up_parent(leader, leader->parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163 write_unlock_irq(&tasklist_lock);
Ingo Molnar780de9d2017-02-02 11:50:56 +01001164 cgroup_threadgroup_change_end(tsk);
Oleg Nesterov81879262008-12-01 14:18:16 -08001165
1166 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001167 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001169 sig->group_exit_task = NULL;
1170 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171
1172no_thread_group:
Oleg Nesterove6368252012-03-19 17:03:22 +01001173 /* we have changed execution domain */
1174 tsk->exit_signal = SIGCHLD;
1175
Nicolas Pitrebaa73d92016-11-11 00:10:10 -05001176#ifdef CONFIG_POSIX_TIMERS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +04001178 flush_itimer_signals();
Nicolas Pitrebaa73d92016-11-11 00:10:10 -05001179#endif
Oleg Nesterov329f7db2005-11-07 21:12:43 +03001180
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001181 if (atomic_read(&oldsighand->count) != 1) {
1182 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001184 * This ->sighand is shared with the CLONE_SIGHAND
1185 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001187 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1188 if (!newsighand)
1189 return -ENOMEM;
1190
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 atomic_set(&newsighand->count, 1);
1192 memcpy(newsighand->action, oldsighand->action,
1193 sizeof(newsighand->action));
1194
1195 write_lock_irq(&tasklist_lock);
1196 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001197 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 spin_unlock(&oldsighand->siglock);
1199 write_unlock_irq(&tasklist_lock);
1200
Davide Libenzifba2afa2007-05-10 22:23:13 -07001201 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 }
1203
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001204 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 return 0;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001206
1207killed:
1208 /* protects against exit_notify() and __exit_signal() */
1209 read_lock(&tasklist_lock);
1210 sig->group_exit_task = NULL;
1211 sig->notify_count = 0;
1212 read_unlock(&tasklist_lock);
1213 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214}
Oleg Nesterov0840a902007-10-16 23:27:22 -07001215
Andrew Morton59714d62008-02-04 22:27:21 -08001216char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217{
1218 /* buf must be at least sizeof(tsk->comm) in size */
1219 task_lock(tsk);
1220 strncpy(buf, tsk->comm, sizeof(tsk->comm));
1221 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -08001222 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223}
J Freyensee7d74f492011-05-06 16:56:47 -07001224EXPORT_SYMBOL_GPL(get_task_comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225
Al Viro6a6d27d2012-08-21 09:56:33 -04001226/*
1227 * These functions flushes out all traces of the currently running executable
1228 * so that a new one can be started
1229 */
1230
Adrian Hunter82b89772014-05-28 11:45:04 +03001231void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232{
1233 task_lock(tsk);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001234 trace_task_rename(tsk, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1236 task_unlock(tsk);
Adrian Hunter82b89772014-05-28 11:45:04 +03001237 perf_event_comm(tsk, exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238}
1239
1240int flush_old_exec(struct linux_binprm * bprm)
1241{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001242 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243
1244 /*
1245 * Make sure we have a private signal table and that
1246 * we are unassociated from the previous thread group.
1247 */
1248 retval = de_thread(current);
1249 if (retval)
1250 goto out;
1251
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001252 /*
1253 * Must be called _before_ exec_mmap() as bprm->mm is
1254 * not visibile until then. This also enables the update
1255 * to be lockless.
1256 */
Matt Helsley925d1c42008-04-29 01:01:36 -07001257 set_mm_exe_file(bprm->mm, bprm->file);
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001258
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 * Release all of the old mmap stuff
1261 */
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001262 acct_arg_size(bprm, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 retval = exec_mmap(bprm->mm);
1264 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -04001265 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266
1267 bprm->mm = NULL; /* We're using it now */
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001268
Mathias Krausedac853a2011-06-09 20:05:18 +02001269 set_fs(USER_DS);
Zhang Yib88fae62014-01-23 15:55:57 -08001270 current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD |
1271 PF_NOFREEZE | PF_NO_SETAFFINITY);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001272 flush_thread();
1273 current->personality &= ~bprm->per_clear;
1274
Aleksa Sarai613cc2b2016-12-21 16:26:24 +11001275 /*
1276 * We have to apply CLOEXEC before we change whether the process is
1277 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1278 * trying to access the should-be-closed file descriptors of a process
1279 * undergoing exec(2).
1280 */
1281 do_close_on_exec(current->files);
Linus Torvalds221af7f2010-01-28 22:14:42 -08001282 return 0;
1283
1284out:
1285 return retval;
1286}
1287EXPORT_SYMBOL(flush_old_exec);
1288
Al Viro1b5d7832011-06-19 12:49:47 -04001289void would_dump(struct linux_binprm *bprm, struct file *file)
1290{
Eric W. Biedermanf84df2a2016-11-16 22:06:51 -06001291 struct inode *inode = file_inode(file);
1292 if (inode_permission(inode, MAY_READ) < 0) {
1293 struct user_namespace *old, *user_ns;
Al Viro1b5d7832011-06-19 12:49:47 -04001294 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
Eric W. Biedermanf84df2a2016-11-16 22:06:51 -06001295
1296 /* Ensure mm->user_ns contains the executable */
1297 user_ns = old = bprm->mm->user_ns;
1298 while ((user_ns != &init_user_ns) &&
1299 !privileged_wrt_inode_uidgid(user_ns, inode))
1300 user_ns = user_ns->parent;
1301
1302 if (old != user_ns) {
1303 bprm->mm->user_ns = get_user_ns(user_ns);
1304 put_user_ns(old);
1305 }
1306 }
Al Viro1b5d7832011-06-19 12:49:47 -04001307}
1308EXPORT_SYMBOL(would_dump);
1309
Linus Torvalds221af7f2010-01-28 22:14:42 -08001310void setup_new_exec(struct linux_binprm * bprm)
1311{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001312 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313
1314 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 current->sas_ss_sp = current->sas_ss_size = 0;
1316
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001317 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
Kees Cooke579d2c2013-02-27 17:03:15 -08001318 set_dumpable(current->mm, SUID_DUMP_USER);
Alan Coxd6e71142005-06-23 00:09:43 -07001319 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001320 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001321
Peter Zijlstrae041e322014-05-21 17:32:19 +02001322 perf_event_exec();
Adrian Hunter82b89772014-05-28 11:45:04 +03001323 __set_task_comm(current, kbasename(bprm->filename), true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001325 /* Set the new mm task size. We have to do that late because it may
1326 * depend on TIF_32BIT which is only updated in flush_thread() on
1327 * some architectures like powerpc
1328 */
1329 current->mm->task_size = TASK_SIZE;
1330
David Howellsa6f76f22008-11-14 10:39:24 +11001331 /* install the new credentials */
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001332 if (!uid_eq(bprm->cred->uid, current_euid()) ||
1333 !gid_eq(bprm->cred->gid, current_egid())) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001334 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001335 } else {
Al Viro1b5d7832011-06-19 12:49:47 -04001336 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1337 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 }
1339
1340 /* An exec changes our domain. We are no longer part of the thread
1341 group */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 current->self_exec_id++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 flush_signal_handlers(current, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001345EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346
David Howellsa6f76f22008-11-14 10:39:24 +11001347/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001348 * Prepare credentials and lock ->cred_guard_mutex.
1349 * install_exec_creds() commits the new creds and drops the lock.
1350 * Or, if exec fails before, free_bprm() should release ->cred and
1351 * and unlock.
1352 */
1353int prepare_bprm_creds(struct linux_binprm *bprm)
1354{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001355 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001356 return -ERESTARTNOINTR;
1357
1358 bprm->cred = prepare_exec_creds();
1359 if (likely(bprm->cred))
1360 return 0;
1361
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001362 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001363 return -ENOMEM;
1364}
1365
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001366static void free_bprm(struct linux_binprm *bprm)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001367{
1368 free_arg_pages(bprm);
1369 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001370 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001371 abort_creds(bprm->cred);
1372 }
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001373 if (bprm->file) {
1374 allow_write_access(bprm->file);
1375 fput(bprm->file);
1376 }
Kees Cookb66c5982012-12-20 15:05:16 -08001377 /* If a binfmt changed the interp, free it. */
1378 if (bprm->interp != bprm->filename)
1379 kfree(bprm->interp);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001380 kfree(bprm);
1381}
1382
Kees Cookb66c5982012-12-20 15:05:16 -08001383int bprm_change_interp(char *interp, struct linux_binprm *bprm)
1384{
1385 /* If a binfmt changed the interp, free it first. */
1386 if (bprm->interp != bprm->filename)
1387 kfree(bprm->interp);
1388 bprm->interp = kstrdup(interp, GFP_KERNEL);
1389 if (!bprm->interp)
1390 return -ENOMEM;
1391 return 0;
1392}
1393EXPORT_SYMBOL(bprm_change_interp);
1394
Oleg Nesterova2a84742009-09-05 11:17:13 -07001395/*
David Howellsa6f76f22008-11-14 10:39:24 +11001396 * install the new credentials for this executable
1397 */
1398void install_exec_creds(struct linux_binprm *bprm)
1399{
1400 security_bprm_committing_creds(bprm);
1401
1402 commit_creds(bprm->cred);
1403 bprm->cred = NULL;
Stephane Eranian2976b102013-06-20 11:36:28 +02001404
1405 /*
1406 * Disable monitoring for regular users
1407 * when executing setuid binaries. Must
1408 * wait until new credentials are committed
1409 * by commit_creds() above
1410 */
1411 if (get_dumpable(current->mm) != SUID_DUMP_USER)
1412 perf_event_exit_task(current);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001413 /*
1414 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001415 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001416 * credentials; any time after this it may be unlocked.
1417 */
David Howellsa6f76f22008-11-14 10:39:24 +11001418 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001419 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001420}
1421EXPORT_SYMBOL(install_exec_creds);
1422
1423/*
1424 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001425 * - the caller must hold ->cred_guard_mutex to protect against
Kees Cookc2e1f2e2014-06-05 00:23:17 -07001426 * PTRACE_ATTACH or seccomp thread-sync
David Howellsa6f76f22008-11-14 10:39:24 +11001427 */
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001428static void check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001429{
David Howells0bf2f3a2009-02-06 11:45:46 +00001430 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001431 unsigned n_fs;
David Howellsa6f76f22008-11-14 10:39:24 +11001432
Eric W. Biederman9227dd22017-01-23 17:26:31 +13001433 if (p->ptrace)
1434 bprm->unsafe |= LSM_UNSAFE_PTRACE;
David Howellsa6f76f22008-11-14 10:39:24 +11001435
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001436 /*
1437 * This isn't strictly necessary, but it makes it harder for LSMs to
1438 * mess up.
1439 */
Kees Cook1d4457f2014-05-21 15:23:46 -07001440 if (task_no_new_privs(current))
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001441 bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS;
1442
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001443 t = p;
David Howells0bf2f3a2009-02-06 11:45:46 +00001444 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001445 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001446 rcu_read_lock();
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001447 while_each_thread(p, t) {
David Howells0bf2f3a2009-02-06 11:45:46 +00001448 if (t->fs == p->fs)
1449 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001450 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001451 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001452
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001453 if (p->fs->users > n_fs)
David Howellsa6f76f22008-11-14 10:39:24 +11001454 bprm->unsafe |= LSM_UNSAFE_SHARE;
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001455 else
1456 p->fs->in_exec = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001457 spin_unlock(&p->fs->lock);
David Howellsa6f76f22008-11-14 10:39:24 +11001458}
1459
Jann Horn8b01fc82015-04-19 02:48:39 +02001460static void bprm_fill_uid(struct linux_binprm *bprm)
1461{
1462 struct inode *inode;
1463 unsigned int mode;
1464 kuid_t uid;
1465 kgid_t gid;
1466
Kees Cookcb6fd682016-05-17 12:14:39 -07001467 /*
1468 * Since this can be called multiple times (via prepare_binprm),
1469 * we must clear any previous work done when setting set[ug]id
1470 * bits from any earlier bprm->file uses (for example when run
1471 * first for a setuid script then again for its interpreter).
1472 */
Jann Horn8b01fc82015-04-19 02:48:39 +02001473 bprm->cred->euid = current_euid();
1474 bprm->cred->egid = current_egid();
1475
Andy Lutomirski380cf5b2016-06-23 16:41:05 -05001476 if (!mnt_may_suid(bprm->file->f_path.mnt))
Jann Horn8b01fc82015-04-19 02:48:39 +02001477 return;
1478
1479 if (task_no_new_privs(current))
1480 return;
1481
Vivek Goyalfea6d2a62017-02-13 15:45:26 -05001482 inode = bprm->file->f_path.dentry->d_inode;
Jann Horn8b01fc82015-04-19 02:48:39 +02001483 mode = READ_ONCE(inode->i_mode);
1484 if (!(mode & (S_ISUID|S_ISGID)))
1485 return;
1486
1487 /* Be careful if suid/sgid is set */
Al Viro59551022016-01-22 15:40:57 -05001488 inode_lock(inode);
Jann Horn8b01fc82015-04-19 02:48:39 +02001489
1490 /* reload atomically mode/uid/gid now that lock held */
1491 mode = inode->i_mode;
1492 uid = inode->i_uid;
1493 gid = inode->i_gid;
Al Viro59551022016-01-22 15:40:57 -05001494 inode_unlock(inode);
Jann Horn8b01fc82015-04-19 02:48:39 +02001495
1496 /* We ignore suid/sgid if there are no mappings for them in the ns */
1497 if (!kuid_has_mapping(bprm->cred->user_ns, uid) ||
1498 !kgid_has_mapping(bprm->cred->user_ns, gid))
1499 return;
1500
1501 if (mode & S_ISUID) {
1502 bprm->per_clear |= PER_CLEAR_ON_SETID;
1503 bprm->cred->euid = uid;
1504 }
1505
1506 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
1507 bprm->per_clear |= PER_CLEAR_ON_SETID;
1508 bprm->cred->egid = gid;
1509 }
1510}
1511
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001512/*
1513 * Fill the binprm structure from the inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001515 *
1516 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517 */
1518int prepare_binprm(struct linux_binprm *bprm)
1519{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 int retval;
1521
Jann Horn8b01fc82015-04-19 02:48:39 +02001522 bprm_fill_uid(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523
1524 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001525 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 if (retval)
1527 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001528 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529
David Howellsa6f76f22008-11-14 10:39:24 +11001530 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1531 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532}
1533
1534EXPORT_SYMBOL(prepare_binprm);
1535
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001536/*
1537 * Arguments are '\0' separated strings found at the location bprm->p
1538 * points to; chop off the first by relocating brpm->p to right after
1539 * the first '\0' encountered.
1540 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001541int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001543 int ret = 0;
1544 unsigned long offset;
1545 char *kaddr;
1546 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001548 if (!bprm->argc)
1549 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001551 do {
1552 offset = bprm->p & ~PAGE_MASK;
1553 page = get_arg_page(bprm, bprm->p, 0);
1554 if (!page) {
1555 ret = -EFAULT;
1556 goto out;
1557 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001558 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001559
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001560 for (; offset < PAGE_SIZE && kaddr[offset];
1561 offset++, bprm->p++)
1562 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001563
Cong Wange8e3c3d2011-11-25 23:14:27 +08001564 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001565 put_arg_page(page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001566 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001567
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001568 bprm->p++;
1569 bprm->argc--;
1570 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001571
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001572out:
1573 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575EXPORT_SYMBOL(remove_arg_zero);
1576
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001577#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578/*
1579 * cycle the list of binary formats handler, until one recognizes the image
1580 */
Al Viro3c456bf2012-10-20 21:53:31 -04001581int search_binary_handler(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582{
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001583 bool need_retry = IS_ENABLED(CONFIG_MODULES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584 struct linux_binfmt *fmt;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001585 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586
Kees Cookd7402692012-12-17 16:03:20 -08001587 /* This allows 4 levels of binfmt rewrites before failing hard. */
Oleg Nesterov131b2f92013-09-11 14:24:39 -07001588 if (bprm->recursion_depth > 5)
Kees Cookd7402692012-12-17 16:03:20 -08001589 return -ELOOP;
1590
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 retval = security_bprm_check(bprm);
1592 if (retval)
1593 return retval;
1594
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 retval = -ENOENT;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001596 retry:
1597 read_lock(&binfmt_lock);
1598 list_for_each_entry(fmt, &formats, lh) {
1599 if (!try_module_get(fmt->module))
1600 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 read_unlock(&binfmt_lock);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001602 bprm->recursion_depth++;
1603 retval = fmt->load_binary(bprm);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001604 read_lock(&binfmt_lock);
1605 put_binfmt(fmt);
Al Viro19d860a2014-05-04 20:11:36 -04001606 bprm->recursion_depth--;
1607 if (retval < 0 && !bprm->mm) {
1608 /* we got to flush_old_exec() and failed after it */
1609 read_unlock(&binfmt_lock);
1610 force_sigsegv(SIGSEGV, current);
1611 return retval;
1612 }
1613 if (retval != -ENOEXEC || !bprm->file) {
1614 read_unlock(&binfmt_lock);
1615 return retval;
1616 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617 }
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001618 read_unlock(&binfmt_lock);
1619
Al Viro19d860a2014-05-04 20:11:36 -04001620 if (need_retry) {
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001621 if (printable(bprm->buf[0]) && printable(bprm->buf[1]) &&
1622 printable(bprm->buf[2]) && printable(bprm->buf[3]))
1623 return retval;
Oleg Nesterov4e0621a2013-09-11 14:24:45 -07001624 if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0)
1625 return retval;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001626 need_retry = false;
1627 goto retry;
1628 }
1629
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 return retval;
1631}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632EXPORT_SYMBOL(search_binary_handler);
1633
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001634static int exec_binprm(struct linux_binprm *bprm)
1635{
1636 pid_t old_pid, old_vpid;
1637 int ret;
1638
1639 /* Need to fetch pid before load_binary changes it */
1640 old_pid = current->pid;
1641 rcu_read_lock();
1642 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
1643 rcu_read_unlock();
1644
1645 ret = search_binary_handler(bprm);
1646 if (ret >= 0) {
Linus Torvalds3eaded862013-11-21 19:18:14 -08001647 audit_bprm(bprm);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001648 trace_sched_process_exec(current, old_pid, bprm);
1649 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
Oleg Nesterov9beb2662013-09-11 14:24:40 -07001650 proc_exec_connector(current);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001651 }
1652
1653 return ret;
1654}
1655
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656/*
1657 * sys_execve() executes a new program.
1658 */
David Drysdale51f39a12014-12-12 16:57:29 -08001659static int do_execveat_common(int fd, struct filename *filename,
1660 struct user_arg_ptr argv,
1661 struct user_arg_ptr envp,
1662 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663{
David Drysdale51f39a12014-12-12 16:57:29 -08001664 char *pathbuf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 struct linux_binprm *bprm;
1666 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001667 struct files_struct *displaced;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001669
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001670 if (IS_ERR(filename))
1671 return PTR_ERR(filename);
1672
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001673 /*
1674 * We move the actual failure in case of RLIMIT_NPROC excess from
1675 * set*uid() to execve() because too many poorly written programs
1676 * don't check setuid() return code. Here we additionally recheck
1677 * whether NPROC limit is still exceeded.
1678 */
1679 if ((current->flags & PF_NPROC_EXCEEDED) &&
Oleg Nesterovbd9d43f2013-07-03 15:08:34 -07001680 atomic_read(&current_user()->processes) > rlimit(RLIMIT_NPROC)) {
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001681 retval = -EAGAIN;
1682 goto out_ret;
1683 }
1684
1685 /* We're below the limit (still or again), so we don't want to make
1686 * further execve() calls fail. */
1687 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688
Al Viro3b125382008-04-22 05:31:30 -04001689 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001690 if (retval)
1691 goto out_ret;
1692
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001694 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001696 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697
Oleg Nesterova2a84742009-09-05 11:17:13 -07001698 retval = prepare_bprm_creds(bprm);
1699 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001700 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001701
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001702 check_unsafe_exec(bprm);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001703 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001704
David Drysdale51f39a12014-12-12 16:57:29 -08001705 file = do_open_execat(fd, filename, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 retval = PTR_ERR(file);
1707 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001708 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709
1710 sched_exec();
1711
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 bprm->file = file;
David Drysdale51f39a12014-12-12 16:57:29 -08001713 if (fd == AT_FDCWD || filename->name[0] == '/') {
1714 bprm->filename = filename->name;
1715 } else {
1716 if (filename->name[0] == '\0')
1717 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d", fd);
1718 else
1719 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d/%s",
1720 fd, filename->name);
1721 if (!pathbuf) {
1722 retval = -ENOMEM;
1723 goto out_unmark;
1724 }
1725 /*
1726 * Record that a name derived from an O_CLOEXEC fd will be
1727 * inaccessible after exec. Relies on having exclusive access to
1728 * current->files (due to unshare_files above).
1729 */
1730 if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt)))
1731 bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE;
1732 bprm->filename = pathbuf;
1733 }
1734 bprm->interp = bprm->filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001735
1736 retval = bprm_mm_init(bprm);
1737 if (retval)
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001738 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001740 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001742 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001744 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 goto out;
1747
1748 retval = prepare_binprm(bprm);
1749 if (retval < 0)
1750 goto out;
1751
1752 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1753 if (retval < 0)
1754 goto out;
1755
1756 bprm->exec = bprm->p;
1757 retval = copy_strings(bprm->envc, envp, bprm);
1758 if (retval < 0)
1759 goto out;
1760
1761 retval = copy_strings(bprm->argc, argv, bprm);
1762 if (retval < 0)
1763 goto out;
1764
Eric W. Biedermanf84df2a2016-11-16 22:06:51 -06001765 would_dump(bprm, bprm->file);
1766
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001767 retval = exec_binprm(bprm);
David Howellsa6f76f22008-11-14 10:39:24 +11001768 if (retval < 0)
1769 goto out;
1770
1771 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001772 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001773 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001774 acct_update_integrals(current);
Rik van Riel82727012013-10-07 11:29:28 +01001775 task_numa_free(current);
David Howellsa6f76f22008-11-14 10:39:24 +11001776 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001777 kfree(pathbuf);
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001778 putname(filename);
David Howellsa6f76f22008-11-14 10:39:24 +11001779 if (displaced)
1780 put_files_struct(displaced);
1781 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782
1783out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001784 if (bprm->mm) {
1785 acct_arg_size(bprm, 0);
1786 mmput(bprm->mm);
1787 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788
Al Viro498052b2009-03-30 07:20:30 -04001789out_unmark:
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001790 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001791 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001792
1793out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001794 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001795 kfree(pathbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796
Al Virofd8328b2008-04-22 05:11:59 -04001797out_files:
Al Viro3b125382008-04-22 05:31:30 -04001798 if (displaced)
1799 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800out_ret:
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001801 putname(filename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 return retval;
1803}
1804
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001805int do_execve(struct filename *filename,
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001806 const char __user *const __user *__argv,
Al Viroda3d4c52012-10-20 21:49:33 -04001807 const char __user *const __user *__envp)
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001808{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001809 struct user_arg_ptr argv = { .ptr.native = __argv };
1810 struct user_arg_ptr envp = { .ptr.native = __envp };
David Drysdale51f39a12014-12-12 16:57:29 -08001811 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1812}
1813
1814int do_execveat(int fd, struct filename *filename,
1815 const char __user *const __user *__argv,
1816 const char __user *const __user *__envp,
1817 int flags)
1818{
1819 struct user_arg_ptr argv = { .ptr.native = __argv };
1820 struct user_arg_ptr envp = { .ptr.native = __envp };
1821
1822 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001823}
1824
Oleg Nesterov0e028462011-03-06 18:02:54 +01001825#ifdef CONFIG_COMPAT
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001826static int compat_do_execve(struct filename *filename,
Al Viro38b983b2012-09-30 13:38:55 -04001827 const compat_uptr_t __user *__argv,
Al Virod03d26e2012-10-20 21:46:25 -04001828 const compat_uptr_t __user *__envp)
Oleg Nesterov0e028462011-03-06 18:02:54 +01001829{
1830 struct user_arg_ptr argv = {
1831 .is_compat = true,
1832 .ptr.compat = __argv,
1833 };
1834 struct user_arg_ptr envp = {
1835 .is_compat = true,
1836 .ptr.compat = __envp,
1837 };
David Drysdale51f39a12014-12-12 16:57:29 -08001838 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1839}
1840
1841static int compat_do_execveat(int fd, struct filename *filename,
1842 const compat_uptr_t __user *__argv,
1843 const compat_uptr_t __user *__envp,
1844 int flags)
1845{
1846 struct user_arg_ptr argv = {
1847 .is_compat = true,
1848 .ptr.compat = __argv,
1849 };
1850 struct user_arg_ptr envp = {
1851 .is_compat = true,
1852 .ptr.compat = __envp,
1853 };
1854 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterov0e028462011-03-06 18:02:54 +01001855}
1856#endif
1857
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001858void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001860 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001862 if (mm->binfmt)
1863 module_put(mm->binfmt->module);
1864
1865 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001866 if (new)
1867 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869EXPORT_SYMBOL(set_binfmt);
1870
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001871/*
Oleg Nesterov7288e112014-01-23 15:55:32 -08001872 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001873 */
1874void set_dumpable(struct mm_struct *mm, int value)
1875{
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001876 unsigned long old, new;
1877
Oleg Nesterov7288e112014-01-23 15:55:32 -08001878 if (WARN_ON((unsigned)value > SUID_DUMP_ROOT))
1879 return;
1880
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001881 do {
1882 old = ACCESS_ONCE(mm->flags);
Oleg Nesterov7288e112014-01-23 15:55:32 -08001883 new = (old & ~MMF_DUMPABLE_MASK) | value;
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001884 } while (cmpxchg(&mm->flags, old, new) != old);
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001885}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001886
Al Viro38b983b2012-09-30 13:38:55 -04001887SYSCALL_DEFINE3(execve,
1888 const char __user *, filename,
1889 const char __user *const __user *, argv,
1890 const char __user *const __user *, envp)
1891{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001892 return do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001893}
David Drysdale51f39a12014-12-12 16:57:29 -08001894
1895SYSCALL_DEFINE5(execveat,
1896 int, fd, const char __user *, filename,
1897 const char __user *const __user *, argv,
1898 const char __user *const __user *, envp,
1899 int, flags)
1900{
1901 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1902
1903 return do_execveat(fd,
1904 getname_flags(filename, lookup_flags, NULL),
1905 argv, envp, flags);
1906}
1907
Al Viro38b983b2012-09-30 13:38:55 -04001908#ifdef CONFIG_COMPAT
Heiko Carstens625b1d72014-03-04 10:53:50 +01001909COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename,
1910 const compat_uptr_t __user *, argv,
1911 const compat_uptr_t __user *, envp)
Al Viro38b983b2012-09-30 13:38:55 -04001912{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001913 return compat_do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001914}
David Drysdale51f39a12014-12-12 16:57:29 -08001915
1916COMPAT_SYSCALL_DEFINE5(execveat, int, fd,
1917 const char __user *, filename,
1918 const compat_uptr_t __user *, argv,
1919 const compat_uptr_t __user *, envp,
1920 int, flags)
1921{
1922 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1923
1924 return compat_do_execveat(fd,
1925 getname_flags(filename, lookup_flags, NULL),
1926 argv, envp, flags);
1927}
Al Viro38b983b2012-09-30 13:38:55 -04001928#endif