blob: d07858c0b7c4e3885c2d6b31521f3e5668b94007 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/exec.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7/*
8 * #!-checking implemented by tytso.
9 */
10/*
11 * Demand-loading implemented 01.12.91 - no need to read anything but
12 * the header into memory. The inode of the executable is put into
13 * "current->executable", and page faults do the actual loading. Clean.
14 *
15 * Once more I can proudly say that linux stood up to being changed: it
16 * was less than 2 hours work to get demand-loading completely implemented.
17 *
18 * Demand loading changed July 1993 by Eric Youngdale. Use mmap instead,
19 * current->executable is only used by the procfs. This allows a dispatch
20 * table to check for several different types of binary formats. We keep
21 * trying until we recognize the file or we run out of supported binary
22 * formats.
23 */
24
25#include <linux/config.h>
26#include <linux/slab.h>
27#include <linux/file.h>
28#include <linux/mman.h>
29#include <linux/a.out.h>
30#include <linux/stat.h>
31#include <linux/fcntl.h>
32#include <linux/smp_lock.h>
33#include <linux/init.h>
34#include <linux/pagemap.h>
35#include <linux/highmem.h>
36#include <linux/spinlock.h>
37#include <linux/key.h>
38#include <linux/personality.h>
39#include <linux/binfmts.h>
40#include <linux/swap.h>
41#include <linux/utsname.h>
42#include <linux/module.h>
43#include <linux/namei.h>
44#include <linux/proc_fs.h>
45#include <linux/ptrace.h>
46#include <linux/mount.h>
47#include <linux/security.h>
48#include <linux/syscalls.h>
49#include <linux/rmap.h>
50#include <linux/acct.h>
Matt Helsley9f460802005-11-07 00:59:16 -080051#include <linux/cn_proc.h>
Al Viro473ae302006-04-26 14:04:08 -040052#include <linux/audit.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
54#include <asm/uaccess.h>
55#include <asm/mmu_context.h>
56
57#ifdef CONFIG_KMOD
58#include <linux/kmod.h>
59#endif
60
61int core_uses_pid;
62char core_pattern[65] = "core";
Alan Coxd6e71142005-06-23 00:09:43 -070063int suid_dumpable = 0;
64
65EXPORT_SYMBOL(suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -070066/* The maximal length of core_pattern is also specified in sysctl.c */
67
68static struct linux_binfmt *formats;
69static DEFINE_RWLOCK(binfmt_lock);
70
71int register_binfmt(struct linux_binfmt * fmt)
72{
73 struct linux_binfmt ** tmp = &formats;
74
75 if (!fmt)
76 return -EINVAL;
77 if (fmt->next)
78 return -EBUSY;
79 write_lock(&binfmt_lock);
80 while (*tmp) {
81 if (fmt == *tmp) {
82 write_unlock(&binfmt_lock);
83 return -EBUSY;
84 }
85 tmp = &(*tmp)->next;
86 }
87 fmt->next = formats;
88 formats = fmt;
89 write_unlock(&binfmt_lock);
90 return 0;
91}
92
93EXPORT_SYMBOL(register_binfmt);
94
95int unregister_binfmt(struct linux_binfmt * fmt)
96{
97 struct linux_binfmt ** tmp = &formats;
98
99 write_lock(&binfmt_lock);
100 while (*tmp) {
101 if (fmt == *tmp) {
102 *tmp = fmt->next;
103 write_unlock(&binfmt_lock);
104 return 0;
105 }
106 tmp = &(*tmp)->next;
107 }
108 write_unlock(&binfmt_lock);
109 return -EINVAL;
110}
111
112EXPORT_SYMBOL(unregister_binfmt);
113
114static inline void put_binfmt(struct linux_binfmt * fmt)
115{
116 module_put(fmt->module);
117}
118
119/*
120 * Note that a shared library must be both readable and executable due to
121 * security reasons.
122 *
123 * Also note that we take the address to load from from the file itself.
124 */
125asmlinkage long sys_uselib(const char __user * library)
126{
127 struct file * file;
128 struct nameidata nd;
129 int error;
130
Oleg Drokinb5005312006-03-25 03:07:01 -0800131 error = __user_path_lookup_open(library, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 if (error)
133 goto out;
134
135 error = -EINVAL;
136 if (!S_ISREG(nd.dentry->d_inode->i_mode))
137 goto exit;
138
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800139 error = vfs_permission(&nd, MAY_READ | MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 if (error)
141 goto exit;
142
Trond Myklebust834f2a42005-10-18 14:20:16 -0700143 file = nameidata_to_filp(&nd, O_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 error = PTR_ERR(file);
145 if (IS_ERR(file))
146 goto out;
147
148 error = -ENOEXEC;
149 if(file->f_op) {
150 struct linux_binfmt * fmt;
151
152 read_lock(&binfmt_lock);
153 for (fmt = formats ; fmt ; fmt = fmt->next) {
154 if (!fmt->load_shlib)
155 continue;
156 if (!try_module_get(fmt->module))
157 continue;
158 read_unlock(&binfmt_lock);
159 error = fmt->load_shlib(file);
160 read_lock(&binfmt_lock);
161 put_binfmt(fmt);
162 if (error != -ENOEXEC)
163 break;
164 }
165 read_unlock(&binfmt_lock);
166 }
167 fput(file);
168out:
169 return error;
170exit:
Trond Myklebust834f2a42005-10-18 14:20:16 -0700171 release_open_intent(&nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 path_release(&nd);
173 goto out;
174}
175
176/*
177 * count() counts the number of strings in array ARGV.
178 */
179static int count(char __user * __user * argv, int max)
180{
181 int i = 0;
182
183 if (argv != NULL) {
184 for (;;) {
185 char __user * p;
186
187 if (get_user(p, argv))
188 return -EFAULT;
189 if (!p)
190 break;
191 argv++;
192 if(++i > max)
193 return -E2BIG;
194 cond_resched();
195 }
196 }
197 return i;
198}
199
200/*
201 * 'copy_strings()' copies argument/environment strings from user
202 * memory to free pages in kernel mem. These are in a format ready
203 * to be put directly into the top of new user memory.
204 */
Adrian Bunk75c96f82005-05-05 16:16:09 -0700205static int copy_strings(int argc, char __user * __user * argv,
206 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207{
208 struct page *kmapped_page = NULL;
209 char *kaddr = NULL;
210 int ret;
211
212 while (argc-- > 0) {
213 char __user *str;
214 int len;
215 unsigned long pos;
216
217 if (get_user(str, argv+argc) ||
218 !(len = strnlen_user(str, bprm->p))) {
219 ret = -EFAULT;
220 goto out;
221 }
222
223 if (bprm->p < len) {
224 ret = -E2BIG;
225 goto out;
226 }
227
228 bprm->p -= len;
229 /* XXX: add architecture specific overflow check here. */
230 pos = bprm->p;
231
232 while (len > 0) {
233 int i, new, err;
234 int offset, bytes_to_copy;
235 struct page *page;
236
237 offset = pos % PAGE_SIZE;
238 i = pos/PAGE_SIZE;
239 page = bprm->page[i];
240 new = 0;
241 if (!page) {
242 page = alloc_page(GFP_HIGHUSER);
243 bprm->page[i] = page;
244 if (!page) {
245 ret = -ENOMEM;
246 goto out;
247 }
248 new = 1;
249 }
250
251 if (page != kmapped_page) {
252 if (kmapped_page)
253 kunmap(kmapped_page);
254 kmapped_page = page;
255 kaddr = kmap(kmapped_page);
256 }
257 if (new && offset)
258 memset(kaddr, 0, offset);
259 bytes_to_copy = PAGE_SIZE - offset;
260 if (bytes_to_copy > len) {
261 bytes_to_copy = len;
262 if (new)
263 memset(kaddr+offset+len, 0,
264 PAGE_SIZE-offset-len);
265 }
266 err = copy_from_user(kaddr+offset, str, bytes_to_copy);
267 if (err) {
268 ret = -EFAULT;
269 goto out;
270 }
271
272 pos += bytes_to_copy;
273 str += bytes_to_copy;
274 len -= bytes_to_copy;
275 }
276 }
277 ret = 0;
278out:
279 if (kmapped_page)
280 kunmap(kmapped_page);
281 return ret;
282}
283
284/*
285 * Like copy_strings, but get argv and its values from kernel memory.
286 */
287int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm)
288{
289 int r;
290 mm_segment_t oldfs = get_fs();
291 set_fs(KERNEL_DS);
292 r = copy_strings(argc, (char __user * __user *)argv, bprm);
293 set_fs(oldfs);
294 return r;
295}
296
297EXPORT_SYMBOL(copy_strings_kernel);
298
299#ifdef CONFIG_MMU
300/*
301 * This routine is used to map in a page into an address space: needed by
302 * execve() for the initial stack and environment pages.
303 *
304 * vma->vm_mm->mmap_sem is held for writing.
305 */
306void install_arg_page(struct vm_area_struct *vma,
307 struct page *page, unsigned long address)
308{
309 struct mm_struct *mm = vma->vm_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 pte_t * pte;
Hugh Dickinsc74df322005-10-29 18:16:23 -0700311 spinlock_t *ptl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
313 if (unlikely(anon_vma_prepare(vma)))
Hugh Dickinsc74df322005-10-29 18:16:23 -0700314 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315
316 flush_dcache_page(page);
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800317 pte = get_locked_pte(mm, address, &ptl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 if (!pte)
319 goto out;
320 if (!pte_none(*pte)) {
Hugh Dickinsc74df322005-10-29 18:16:23 -0700321 pte_unmap_unlock(pte, ptl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 goto out;
323 }
Hugh Dickins42946212005-10-29 18:16:05 -0700324 inc_mm_counter(mm, anon_rss);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 lru_cache_add_active(page);
326 set_pte_at(mm, address, pte, pte_mkdirty(pte_mkwrite(mk_pte(
327 page, vma->vm_page_prot))));
Nick Piggin9617d952006-01-06 00:11:12 -0800328 page_add_new_anon_rmap(page, vma, address);
Hugh Dickinsc74df322005-10-29 18:16:23 -0700329 pte_unmap_unlock(pte, ptl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330
331 /* no need for flush_tlb */
332 return;
333out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 __free_page(page);
335 force_sig(SIGKILL, current);
336}
337
338#define EXTRA_STACK_VM_PAGES 20 /* random */
339
340int setup_arg_pages(struct linux_binprm *bprm,
341 unsigned long stack_top,
342 int executable_stack)
343{
344 unsigned long stack_base;
345 struct vm_area_struct *mpnt;
346 struct mm_struct *mm = current->mm;
347 int i, ret;
348 long arg_size;
349
350#ifdef CONFIG_STACK_GROWSUP
351 /* Move the argument and environment strings to the bottom of the
352 * stack space.
353 */
354 int offset, j;
355 char *to, *from;
356
357 /* Start by shifting all the pages down */
358 i = 0;
359 for (j = 0; j < MAX_ARG_PAGES; j++) {
360 struct page *page = bprm->page[j];
361 if (!page)
362 continue;
363 bprm->page[i++] = page;
364 }
365
366 /* Now move them within their pages */
367 offset = bprm->p % PAGE_SIZE;
368 to = kmap(bprm->page[0]);
369 for (j = 1; j < i; j++) {
370 memmove(to, to + offset, PAGE_SIZE - offset);
371 from = kmap(bprm->page[j]);
372 memcpy(to + PAGE_SIZE - offset, from, offset);
373 kunmap(bprm->page[j - 1]);
374 to = from;
375 }
376 memmove(to, to + offset, PAGE_SIZE - offset);
377 kunmap(bprm->page[j - 1]);
378
379 /* Limit stack size to 1GB */
380 stack_base = current->signal->rlim[RLIMIT_STACK].rlim_max;
381 if (stack_base > (1 << 30))
382 stack_base = 1 << 30;
383 stack_base = PAGE_ALIGN(stack_top - stack_base);
384
385 /* Adjust bprm->p to point to the end of the strings. */
386 bprm->p = stack_base + PAGE_SIZE * i - offset;
387
388 mm->arg_start = stack_base;
389 arg_size = i << PAGE_SHIFT;
390
391 /* zero pages that were copied above */
392 while (i < MAX_ARG_PAGES)
393 bprm->page[i++] = NULL;
394#else
395 stack_base = arch_align_stack(stack_top - MAX_ARG_PAGES*PAGE_SIZE);
396 stack_base = PAGE_ALIGN(stack_base);
397 bprm->p += stack_base;
398 mm->arg_start = bprm->p;
399 arg_size = stack_top - (PAGE_MASK & (unsigned long) mm->arg_start);
400#endif
401
402 arg_size += EXTRA_STACK_VM_PAGES * PAGE_SIZE;
403
404 if (bprm->loader)
405 bprm->loader += stack_base;
406 bprm->exec += stack_base;
407
408 mpnt = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL);
409 if (!mpnt)
410 return -ENOMEM;
411
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 memset(mpnt, 0, sizeof(*mpnt));
413
414 down_write(&mm->mmap_sem);
415 {
416 mpnt->vm_mm = mm;
417#ifdef CONFIG_STACK_GROWSUP
418 mpnt->vm_start = stack_base;
419 mpnt->vm_end = stack_base + arg_size;
420#else
421 mpnt->vm_end = stack_top;
422 mpnt->vm_start = mpnt->vm_end - arg_size;
423#endif
424 /* Adjust stack execute permissions; explicitly enable
425 * for EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X
426 * and leave alone (arch default) otherwise. */
427 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
428 mpnt->vm_flags = VM_STACK_FLAGS | VM_EXEC;
429 else if (executable_stack == EXSTACK_DISABLE_X)
430 mpnt->vm_flags = VM_STACK_FLAGS & ~VM_EXEC;
431 else
432 mpnt->vm_flags = VM_STACK_FLAGS;
433 mpnt->vm_flags |= mm->def_flags;
434 mpnt->vm_page_prot = protection_map[mpnt->vm_flags & 0x7];
435 if ((ret = insert_vm_struct(mm, mpnt))) {
436 up_write(&mm->mmap_sem);
437 kmem_cache_free(vm_area_cachep, mpnt);
438 return ret;
439 }
440 mm->stack_vm = mm->total_vm = vma_pages(mpnt);
441 }
442
443 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
444 struct page *page = bprm->page[i];
445 if (page) {
446 bprm->page[i] = NULL;
447 install_arg_page(mpnt, page, stack_base);
448 }
449 stack_base += PAGE_SIZE;
450 }
451 up_write(&mm->mmap_sem);
452
453 return 0;
454}
455
456EXPORT_SYMBOL(setup_arg_pages);
457
458#define free_arg_pages(bprm) do { } while (0)
459
460#else
461
462static inline void free_arg_pages(struct linux_binprm *bprm)
463{
464 int i;
465
466 for (i = 0; i < MAX_ARG_PAGES; i++) {
467 if (bprm->page[i])
468 __free_page(bprm->page[i]);
469 bprm->page[i] = NULL;
470 }
471}
472
473#endif /* CONFIG_MMU */
474
475struct file *open_exec(const char *name)
476{
477 struct nameidata nd;
478 int err;
479 struct file *file;
480
Oleg Drokinb5005312006-03-25 03:07:01 -0800481 err = path_lookup_open(AT_FDCWD, name, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 file = ERR_PTR(err);
483
484 if (!err) {
485 struct inode *inode = nd.dentry->d_inode;
486 file = ERR_PTR(-EACCES);
487 if (!(nd.mnt->mnt_flags & MNT_NOEXEC) &&
488 S_ISREG(inode->i_mode)) {
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800489 int err = vfs_permission(&nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 if (!err && !(inode->i_mode & 0111))
491 err = -EACCES;
492 file = ERR_PTR(err);
493 if (!err) {
Trond Myklebust834f2a42005-10-18 14:20:16 -0700494 file = nameidata_to_filp(&nd, O_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 if (!IS_ERR(file)) {
496 err = deny_write_access(file);
497 if (err) {
498 fput(file);
499 file = ERR_PTR(err);
500 }
501 }
502out:
503 return file;
504 }
505 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700506 release_open_intent(&nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 path_release(&nd);
508 }
509 goto out;
510}
511
512EXPORT_SYMBOL(open_exec);
513
514int kernel_read(struct file *file, unsigned long offset,
515 char *addr, unsigned long count)
516{
517 mm_segment_t old_fs;
518 loff_t pos = offset;
519 int result;
520
521 old_fs = get_fs();
522 set_fs(get_ds());
523 /* The cast to a user pointer is valid due to the set_fs() */
524 result = vfs_read(file, (void __user *)addr, count, &pos);
525 set_fs(old_fs);
526 return result;
527}
528
529EXPORT_SYMBOL(kernel_read);
530
531static int exec_mmap(struct mm_struct *mm)
532{
533 struct task_struct *tsk;
534 struct mm_struct * old_mm, *active_mm;
535
536 /* Notify parent that we're no longer interested in the old VM */
537 tsk = current;
538 old_mm = current->mm;
539 mm_release(tsk, old_mm);
540
541 if (old_mm) {
542 /*
543 * Make sure that if there is a core dump in progress
544 * for the old mm, we get out and die instead of going
545 * through with the exec. We must hold mmap_sem around
546 * checking core_waiters and changing tsk->mm. The
547 * core-inducing thread will increment core_waiters for
548 * each thread whose ->mm == old_mm.
549 */
550 down_read(&old_mm->mmap_sem);
551 if (unlikely(old_mm->core_waiters)) {
552 up_read(&old_mm->mmap_sem);
553 return -EINTR;
554 }
555 }
556 task_lock(tsk);
557 active_mm = tsk->active_mm;
558 tsk->mm = mm;
559 tsk->active_mm = mm;
560 activate_mm(active_mm, mm);
561 task_unlock(tsk);
562 arch_pick_mmap_layout(mm);
563 if (old_mm) {
564 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200565 BUG_ON(active_mm != old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 mmput(old_mm);
567 return 0;
568 }
569 mmdrop(active_mm);
570 return 0;
571}
572
573/*
574 * This function makes sure the current process has its own signal table,
575 * so that flush_signal_handlers can later reset the handlers without
576 * disturbing other processes. (Other processes might share the signal
577 * table via the CLONE_SIGHAND option to clone().)
578 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800579static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580{
581 struct signal_struct *sig = tsk->signal;
582 struct sighand_struct *newsighand, *oldsighand = tsk->sighand;
583 spinlock_t *lock = &oldsighand->siglock;
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300584 struct task_struct *leader = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 int count;
586
587 /*
588 * If we don't share sighandlers, then we aren't sharing anything
589 * and we can just re-use it all.
590 */
591 if (atomic_read(&oldsighand->count) <= 1) {
592 BUG_ON(atomic_read(&sig->count) != 1);
593 exit_itimers(sig);
594 return 0;
595 }
596
597 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
598 if (!newsighand)
599 return -ENOMEM;
600
601 if (thread_group_empty(current))
602 goto no_thread_group;
603
604 /*
605 * Kill all other threads in the thread group.
606 * We must hold tasklist_lock to call zap_other_threads.
607 */
608 read_lock(&tasklist_lock);
609 spin_lock_irq(lock);
610 if (sig->flags & SIGNAL_GROUP_EXIT) {
611 /*
612 * Another group action in progress, just
613 * return so that the signal is processed.
614 */
615 spin_unlock_irq(lock);
616 read_unlock(&tasklist_lock);
617 kmem_cache_free(sighand_cachep, newsighand);
618 return -EAGAIN;
619 }
Oleg Nesterov14342612006-03-28 16:10:59 -0800620
621 /*
622 * child_reaper ignores SIGKILL, change it now.
623 * Reparenting needs write_lock on tasklist_lock,
624 * so it is safe to do it under read_lock.
625 */
626 if (unlikely(current->group_leader == child_reaper))
627 child_reaper = current;
628
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 zap_other_threads(current);
630 read_unlock(&tasklist_lock);
631
632 /*
633 * Account for the thread group leader hanging around:
634 */
Oleg Nesterov9e4e23b2005-10-30 15:01:37 -0800635 count = 1;
636 if (!thread_group_leader(current)) {
637 count = 2;
Roland McGrath53231252005-07-12 13:58:27 -0700638 /*
639 * The SIGALRM timer survives the exec, but needs to point
640 * at us as the new group leader now. We have a race with
641 * a timer firing now getting the old leader, so we need to
642 * synchronize with any firing (by calling del_timer_sync)
643 * before we can safely let the old group leader die.
644 */
Roman Zippel05cfb612006-03-26 01:38:12 -0800645 sig->tsk = current;
Oleg Nesterov932aeaf2005-10-30 15:02:17 -0800646 spin_unlock_irq(lock);
Thomas Gleixner2ff678b2006-01-09 20:52:34 -0800647 if (hrtimer_cancel(&sig->real_timer))
648 hrtimer_restart(&sig->real_timer);
Oleg Nesterov932aeaf2005-10-30 15:02:17 -0800649 spin_lock_irq(lock);
Roland McGrath53231252005-07-12 13:58:27 -0700650 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 while (atomic_read(&sig->count) > count) {
652 sig->group_exit_task = current;
653 sig->notify_count = count;
654 __set_current_state(TASK_UNINTERRUPTIBLE);
655 spin_unlock_irq(lock);
656 schedule();
657 spin_lock_irq(lock);
658 }
659 sig->group_exit_task = NULL;
660 sig->notify_count = 0;
661 spin_unlock_irq(lock);
662
663 /*
664 * At this point all other threads have exited, all we have to
665 * do is to wait for the thread group leader to become inactive,
666 * and to assume its PID:
667 */
668 if (!thread_group_leader(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 struct dentry *proc_dentry1, *proc_dentry2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670
671 /*
672 * Wait for the thread group leader to be a zombie.
673 * It should already be zombie at this point, most
674 * of the time.
675 */
Oleg Nesterov14342612006-03-28 16:10:59 -0800676 leader = current->group_leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 while (leader->exit_state != EXIT_ZOMBIE)
678 yield();
679
Roland McGrathf5e90282006-04-10 22:54:16 -0700680 /*
681 * The only record we have of the real-time age of a
682 * process, regardless of execs it's done, is start_time.
683 * All the past CPU time is accumulated in signal_struct
684 * from sister threads now dead. But in this non-leader
685 * exec, nothing survives from the original leader thread,
686 * whose birth marks the true age of this process now.
687 * When we take on its identity by switching to its PID, we
688 * also take its birthdate (always earlier than our own).
689 */
690 current->start_time = leader->start_time;
691
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 spin_lock(&leader->proc_lock);
693 spin_lock(&current->proc_lock);
694 proc_dentry1 = proc_pid_unhash(current);
695 proc_dentry2 = proc_pid_unhash(leader);
696 write_lock_irq(&tasklist_lock);
697
Linus Torvaldsc2a0f592005-06-18 13:06:22 -0700698 BUG_ON(leader->tgid != current->tgid);
699 BUG_ON(current->pid == current->tgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 /*
701 * An exec() starts a new thread group with the
702 * TGID of the previous thread group. Rehash the
703 * two threads with a switched PID, and release
704 * the former thread group leader:
705 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800706
707 /* Become a process group leader with the old leader's pid.
708 * Note: The old leader also uses thispid until release_task
709 * is called. Odd but simple and correct.
710 */
711 detach_pid(current, PIDTYPE_PID);
712 current->pid = leader->pid;
713 attach_pid(current, PIDTYPE_PID, current->pid);
714 attach_pid(current, PIDTYPE_PGID, current->signal->pgrp);
715 attach_pid(current, PIDTYPE_SID, current->signal->session);
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -0700716 list_add_tail_rcu(&current->tasks, &init_task.tasks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 current->group_leader = current;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600719 leader->group_leader = current;
720
721 /* Reduce leader to a thread */
722 detach_pid(leader, PIDTYPE_PGID);
723 detach_pid(leader, PIDTYPE_SID);
724 list_del_init(&leader->tasks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 current->exit_signal = SIGCHLD;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800727
728 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
729 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730
731 write_unlock_irq(&tasklist_lock);
732 spin_unlock(&leader->proc_lock);
733 spin_unlock(&current->proc_lock);
734 proc_pid_flush(proc_dentry1);
735 proc_pid_flush(proc_dentry2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 }
737
738 /*
Alexander Nybergfb085cf2005-09-14 18:54:06 +0200739 * There may be one thread left which is just exiting,
740 * but it's safe to stop telling the group to kill themselves.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 */
742 sig->flags = 0;
743
744no_thread_group:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 exit_itimers(sig);
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300746 if (leader)
747 release_task(leader);
748
749 BUG_ON(atomic_read(&sig->count) != 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
751 if (atomic_read(&oldsighand->count) == 1) {
752 /*
753 * Now that we nuked the rest of the thread group,
754 * it turns out we are not sharing sighand any more either.
755 * So we can just keep it.
756 */
757 kmem_cache_free(sighand_cachep, newsighand);
758 } else {
759 /*
760 * Move our state over to newsighand and switch it in.
761 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 atomic_set(&newsighand->count, 1);
763 memcpy(newsighand->action, oldsighand->action,
764 sizeof(newsighand->action));
765
766 write_lock_irq(&tasklist_lock);
767 spin_lock(&oldsighand->siglock);
768 spin_lock(&newsighand->siglock);
769
Ingo Molnare56d0902006-01-08 01:01:37 -0800770 rcu_assign_pointer(current->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 recalc_sigpending();
772
773 spin_unlock(&newsighand->siglock);
774 spin_unlock(&oldsighand->siglock);
775 write_unlock_irq(&tasklist_lock);
776
777 if (atomic_dec_and_test(&oldsighand->count))
Oleg Nesterovaa1757f2006-03-28 16:11:12 -0800778 kmem_cache_free(sighand_cachep, oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 }
780
Linus Torvaldsc2a0f592005-06-18 13:06:22 -0700781 BUG_ON(!thread_group_leader(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 return 0;
783}
784
785/*
786 * These functions flushes out all traces of the currently running executable
787 * so that a new one can be started
788 */
789
Arjan van de Ven858119e2006-01-14 13:20:43 -0800790static void flush_old_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791{
792 long j = -1;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700793 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
795 spin_lock(&files->file_lock);
796 for (;;) {
797 unsigned long set, i;
798
799 j++;
800 i = j * __NFDBITS;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700801 fdt = files_fdtable(files);
802 if (i >= fdt->max_fds || i >= fdt->max_fdset)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700804 set = fdt->close_on_exec->fds_bits[j];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 if (!set)
806 continue;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700807 fdt->close_on_exec->fds_bits[j] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 spin_unlock(&files->file_lock);
809 for ( ; set ; i++,set >>= 1) {
810 if (set & 1) {
811 sys_close(i);
812 }
813 }
814 spin_lock(&files->file_lock);
815
816 }
817 spin_unlock(&files->file_lock);
818}
819
820void get_task_comm(char *buf, struct task_struct *tsk)
821{
822 /* buf must be at least sizeof(tsk->comm) in size */
823 task_lock(tsk);
824 strncpy(buf, tsk->comm, sizeof(tsk->comm));
825 task_unlock(tsk);
826}
827
828void set_task_comm(struct task_struct *tsk, char *buf)
829{
830 task_lock(tsk);
831 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
832 task_unlock(tsk);
833}
834
835int flush_old_exec(struct linux_binprm * bprm)
836{
837 char * name;
838 int i, ch, retval;
839 struct files_struct *files;
840 char tcomm[sizeof(current->comm)];
841
842 /*
843 * Make sure we have a private signal table and that
844 * we are unassociated from the previous thread group.
845 */
846 retval = de_thread(current);
847 if (retval)
848 goto out;
849
850 /*
851 * Make sure we have private file handles. Ask the
852 * fork helper to do the work for us and the exit
853 * helper to do the cleanup of the old one.
854 */
855 files = current->files; /* refcounted so safe to hold */
856 retval = unshare_files();
857 if (retval)
858 goto out;
859 /*
860 * Release all of the old mmap stuff
861 */
862 retval = exec_mmap(bprm->mm);
863 if (retval)
864 goto mmap_failed;
865
866 bprm->mm = NULL; /* We're using it now */
867
868 /* This is the point of no return */
869 steal_locks(files);
870 put_files_struct(files);
871
872 current->sas_ss_sp = current->sas_ss_size = 0;
873
874 if (current->euid == current->uid && current->egid == current->gid)
875 current->mm->dumpable = 1;
Alan Coxd6e71142005-06-23 00:09:43 -0700876 else
877 current->mm->dumpable = suid_dumpable;
878
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 name = bprm->filename;
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700880
881 /* Copies the binary name from after last slash */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 for (i=0; (ch = *(name++)) != '\0';) {
883 if (ch == '/')
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700884 i = 0; /* overwrite what we wrote */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 else
886 if (i < (sizeof(tcomm) - 1))
887 tcomm[i++] = ch;
888 }
889 tcomm[i] = '\0';
890 set_task_comm(current, tcomm);
891
892 current->flags &= ~PF_RANDOMIZE;
893 flush_thread();
894
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -0800895 /* Set the new mm task size. We have to do that late because it may
896 * depend on TIF_32BIT which is only updated in flush_thread() on
897 * some architectures like powerpc
898 */
899 current->mm->task_size = TASK_SIZE;
900
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 if (bprm->e_uid != current->euid || bprm->e_gid != current->egid ||
Christoph Hellwig8c744fb2005-11-08 21:35:04 -0800902 file_permission(bprm->file, MAY_READ) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)) {
904 suid_keys(current);
Alan Coxd6e71142005-06-23 00:09:43 -0700905 current->mm->dumpable = suid_dumpable;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 }
907
908 /* An exec changes our domain. We are no longer part of the thread
909 group */
910
911 current->self_exec_id++;
912
913 flush_signal_handlers(current, 0);
914 flush_old_files(current->files);
915
916 return 0;
917
918mmap_failed:
919 put_files_struct(current->files);
920 current->files = files;
921out:
922 return retval;
923}
924
925EXPORT_SYMBOL(flush_old_exec);
926
927/*
928 * Fill the binprm structure from the inode.
929 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
930 */
931int prepare_binprm(struct linux_binprm *bprm)
932{
933 int mode;
934 struct inode * inode = bprm->file->f_dentry->d_inode;
935 int retval;
936
937 mode = inode->i_mode;
938 /*
939 * Check execute perms again - if the caller has CAP_DAC_OVERRIDE,
940 * generic_permission lets a non-executable through
941 */
942 if (!(mode & 0111)) /* with at least _one_ execute bit set */
943 return -EACCES;
944 if (bprm->file->f_op == NULL)
945 return -EACCES;
946
947 bprm->e_uid = current->euid;
948 bprm->e_gid = current->egid;
949
950 if(!(bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID)) {
951 /* Set-uid? */
952 if (mode & S_ISUID) {
953 current->personality &= ~PER_CLEAR_ON_SETID;
954 bprm->e_uid = inode->i_uid;
955 }
956
957 /* Set-gid? */
958 /*
959 * If setgid is set but no group execute bit then this
960 * is a candidate for mandatory locking, not a setgid
961 * executable.
962 */
963 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
964 current->personality &= ~PER_CLEAR_ON_SETID;
965 bprm->e_gid = inode->i_gid;
966 }
967 }
968
969 /* fill in binprm security blob */
970 retval = security_bprm_set(bprm);
971 if (retval)
972 return retval;
973
974 memset(bprm->buf,0,BINPRM_BUF_SIZE);
975 return kernel_read(bprm->file,0,bprm->buf,BINPRM_BUF_SIZE);
976}
977
978EXPORT_SYMBOL(prepare_binprm);
979
Arjan van de Ven858119e2006-01-14 13:20:43 -0800980static int unsafe_exec(struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981{
982 int unsafe = 0;
983 if (p->ptrace & PT_PTRACED) {
984 if (p->ptrace & PT_PTRACE_CAP)
985 unsafe |= LSM_UNSAFE_PTRACE_CAP;
986 else
987 unsafe |= LSM_UNSAFE_PTRACE;
988 }
989 if (atomic_read(&p->fs->count) > 1 ||
990 atomic_read(&p->files->count) > 1 ||
991 atomic_read(&p->sighand->count) > 1)
992 unsafe |= LSM_UNSAFE_SHARE;
993
994 return unsafe;
995}
996
997void compute_creds(struct linux_binprm *bprm)
998{
999 int unsafe;
1000
1001 if (bprm->e_uid != current->uid)
1002 suid_keys(current);
1003 exec_keys(current);
1004
1005 task_lock(current);
1006 unsafe = unsafe_exec(current);
1007 security_bprm_apply_creds(bprm, unsafe);
1008 task_unlock(current);
1009 security_bprm_post_apply_creds(bprm);
1010}
1011
1012EXPORT_SYMBOL(compute_creds);
1013
1014void remove_arg_zero(struct linux_binprm *bprm)
1015{
1016 if (bprm->argc) {
1017 unsigned long offset;
1018 char * kaddr;
1019 struct page *page;
1020
1021 offset = bprm->p % PAGE_SIZE;
1022 goto inside;
1023
1024 while (bprm->p++, *(kaddr+offset++)) {
1025 if (offset != PAGE_SIZE)
1026 continue;
1027 offset = 0;
1028 kunmap_atomic(kaddr, KM_USER0);
1029inside:
1030 page = bprm->page[bprm->p/PAGE_SIZE];
1031 kaddr = kmap_atomic(page, KM_USER0);
1032 }
1033 kunmap_atomic(kaddr, KM_USER0);
1034 bprm->argc--;
1035 }
1036}
1037
1038EXPORT_SYMBOL(remove_arg_zero);
1039
1040/*
1041 * cycle the list of binary formats handler, until one recognizes the image
1042 */
1043int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
1044{
1045 int try,retval;
1046 struct linux_binfmt *fmt;
1047#ifdef __alpha__
1048 /* handle /sbin/loader.. */
1049 {
1050 struct exec * eh = (struct exec *) bprm->buf;
1051
1052 if (!bprm->loader && eh->fh.f_magic == 0x183 &&
1053 (eh->fh.f_flags & 0x3000) == 0x3000)
1054 {
1055 struct file * file;
1056 unsigned long loader;
1057
1058 allow_write_access(bprm->file);
1059 fput(bprm->file);
1060 bprm->file = NULL;
1061
1062 loader = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1063
1064 file = open_exec("/sbin/loader");
1065 retval = PTR_ERR(file);
1066 if (IS_ERR(file))
1067 return retval;
1068
1069 /* Remember if the application is TASO. */
1070 bprm->sh_bang = eh->ah.entry < 0x100000000UL;
1071
1072 bprm->file = file;
1073 bprm->loader = loader;
1074 retval = prepare_binprm(bprm);
1075 if (retval<0)
1076 return retval;
1077 /* should call search_binary_handler recursively here,
1078 but it does not matter */
1079 }
1080 }
1081#endif
1082 retval = security_bprm_check(bprm);
1083 if (retval)
1084 return retval;
1085
1086 /* kernel module loader fixup */
1087 /* so we don't try to load run modprobe in kernel space. */
1088 set_fs(USER_DS);
Al Viro473ae302006-04-26 14:04:08 -04001089
1090 retval = audit_bprm(bprm);
1091 if (retval)
1092 return retval;
1093
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 retval = -ENOENT;
1095 for (try=0; try<2; try++) {
1096 read_lock(&binfmt_lock);
1097 for (fmt = formats ; fmt ; fmt = fmt->next) {
1098 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1099 if (!fn)
1100 continue;
1101 if (!try_module_get(fmt->module))
1102 continue;
1103 read_unlock(&binfmt_lock);
1104 retval = fn(bprm, regs);
1105 if (retval >= 0) {
1106 put_binfmt(fmt);
1107 allow_write_access(bprm->file);
1108 if (bprm->file)
1109 fput(bprm->file);
1110 bprm->file = NULL;
1111 current->did_exec = 1;
Matt Helsley9f460802005-11-07 00:59:16 -08001112 proc_exec_connector(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 return retval;
1114 }
1115 read_lock(&binfmt_lock);
1116 put_binfmt(fmt);
1117 if (retval != -ENOEXEC || bprm->mm == NULL)
1118 break;
1119 if (!bprm->file) {
1120 read_unlock(&binfmt_lock);
1121 return retval;
1122 }
1123 }
1124 read_unlock(&binfmt_lock);
1125 if (retval != -ENOEXEC || bprm->mm == NULL) {
1126 break;
1127#ifdef CONFIG_KMOD
1128 }else{
1129#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1130 if (printable(bprm->buf[0]) &&
1131 printable(bprm->buf[1]) &&
1132 printable(bprm->buf[2]) &&
1133 printable(bprm->buf[3]))
1134 break; /* -ENOEXEC */
1135 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1136#endif
1137 }
1138 }
1139 return retval;
1140}
1141
1142EXPORT_SYMBOL(search_binary_handler);
1143
1144/*
1145 * sys_execve() executes a new program.
1146 */
1147int do_execve(char * filename,
1148 char __user *__user *argv,
1149 char __user *__user *envp,
1150 struct pt_regs * regs)
1151{
1152 struct linux_binprm *bprm;
1153 struct file *file;
1154 int retval;
1155 int i;
1156
1157 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001158 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 if (!bprm)
1160 goto out_ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161
1162 file = open_exec(filename);
1163 retval = PTR_ERR(file);
1164 if (IS_ERR(file))
1165 goto out_kfree;
1166
1167 sched_exec();
1168
1169 bprm->p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1170
1171 bprm->file = file;
1172 bprm->filename = filename;
1173 bprm->interp = filename;
1174 bprm->mm = mm_alloc();
1175 retval = -ENOMEM;
1176 if (!bprm->mm)
1177 goto out_file;
1178
1179 retval = init_new_context(current, bprm->mm);
1180 if (retval < 0)
1181 goto out_mm;
1182
1183 bprm->argc = count(argv, bprm->p / sizeof(void *));
1184 if ((retval = bprm->argc) < 0)
1185 goto out_mm;
1186
1187 bprm->envc = count(envp, bprm->p / sizeof(void *));
1188 if ((retval = bprm->envc) < 0)
1189 goto out_mm;
1190
1191 retval = security_bprm_alloc(bprm);
1192 if (retval)
1193 goto out;
1194
1195 retval = prepare_binprm(bprm);
1196 if (retval < 0)
1197 goto out;
1198
1199 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1200 if (retval < 0)
1201 goto out;
1202
1203 bprm->exec = bprm->p;
1204 retval = copy_strings(bprm->envc, envp, bprm);
1205 if (retval < 0)
1206 goto out;
1207
1208 retval = copy_strings(bprm->argc, argv, bprm);
1209 if (retval < 0)
1210 goto out;
1211
1212 retval = search_binary_handler(bprm,regs);
1213 if (retval >= 0) {
1214 free_arg_pages(bprm);
1215
1216 /* execve success */
1217 security_bprm_free(bprm);
1218 acct_update_integrals(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 kfree(bprm);
1220 return retval;
1221 }
1222
1223out:
1224 /* Something went wrong, return the inode and free the argument pages*/
1225 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
1226 struct page * page = bprm->page[i];
1227 if (page)
1228 __free_page(page);
1229 }
1230
1231 if (bprm->security)
1232 security_bprm_free(bprm);
1233
1234out_mm:
1235 if (bprm->mm)
1236 mmdrop(bprm->mm);
1237
1238out_file:
1239 if (bprm->file) {
1240 allow_write_access(bprm->file);
1241 fput(bprm->file);
1242 }
1243
1244out_kfree:
1245 kfree(bprm);
1246
1247out_ret:
1248 return retval;
1249}
1250
1251int set_binfmt(struct linux_binfmt *new)
1252{
1253 struct linux_binfmt *old = current->binfmt;
1254
1255 if (new) {
1256 if (!try_module_get(new->module))
1257 return -1;
1258 }
1259 current->binfmt = new;
1260 if (old)
1261 module_put(old->module);
1262 return 0;
1263}
1264
1265EXPORT_SYMBOL(set_binfmt);
1266
1267#define CORENAME_MAX_SIZE 64
1268
1269/* format_corename will inspect the pattern parameter, and output a
1270 * name into corename, which must have space for at least
1271 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1272 */
1273static void format_corename(char *corename, const char *pattern, long signr)
1274{
1275 const char *pat_ptr = pattern;
1276 char *out_ptr = corename;
1277 char *const out_end = corename + CORENAME_MAX_SIZE;
1278 int rc;
1279 int pid_in_pattern = 0;
1280
1281 /* Repeat as long as we have more pattern to process and more output
1282 space */
1283 while (*pat_ptr) {
1284 if (*pat_ptr != '%') {
1285 if (out_ptr == out_end)
1286 goto out;
1287 *out_ptr++ = *pat_ptr++;
1288 } else {
1289 switch (*++pat_ptr) {
1290 case 0:
1291 goto out;
1292 /* Double percent, output one percent */
1293 case '%':
1294 if (out_ptr == out_end)
1295 goto out;
1296 *out_ptr++ = '%';
1297 break;
1298 /* pid */
1299 case 'p':
1300 pid_in_pattern = 1;
1301 rc = snprintf(out_ptr, out_end - out_ptr,
1302 "%d", current->tgid);
1303 if (rc > out_end - out_ptr)
1304 goto out;
1305 out_ptr += rc;
1306 break;
1307 /* uid */
1308 case 'u':
1309 rc = snprintf(out_ptr, out_end - out_ptr,
1310 "%d", current->uid);
1311 if (rc > out_end - out_ptr)
1312 goto out;
1313 out_ptr += rc;
1314 break;
1315 /* gid */
1316 case 'g':
1317 rc = snprintf(out_ptr, out_end - out_ptr,
1318 "%d", current->gid);
1319 if (rc > out_end - out_ptr)
1320 goto out;
1321 out_ptr += rc;
1322 break;
1323 /* signal that caused the coredump */
1324 case 's':
1325 rc = snprintf(out_ptr, out_end - out_ptr,
1326 "%ld", signr);
1327 if (rc > out_end - out_ptr)
1328 goto out;
1329 out_ptr += rc;
1330 break;
1331 /* UNIX time of coredump */
1332 case 't': {
1333 struct timeval tv;
1334 do_gettimeofday(&tv);
1335 rc = snprintf(out_ptr, out_end - out_ptr,
1336 "%lu", tv.tv_sec);
1337 if (rc > out_end - out_ptr)
1338 goto out;
1339 out_ptr += rc;
1340 break;
1341 }
1342 /* hostname */
1343 case 'h':
1344 down_read(&uts_sem);
1345 rc = snprintf(out_ptr, out_end - out_ptr,
1346 "%s", system_utsname.nodename);
1347 up_read(&uts_sem);
1348 if (rc > out_end - out_ptr)
1349 goto out;
1350 out_ptr += rc;
1351 break;
1352 /* executable */
1353 case 'e':
1354 rc = snprintf(out_ptr, out_end - out_ptr,
1355 "%s", current->comm);
1356 if (rc > out_end - out_ptr)
1357 goto out;
1358 out_ptr += rc;
1359 break;
1360 default:
1361 break;
1362 }
1363 ++pat_ptr;
1364 }
1365 }
1366 /* Backward compatibility with core_uses_pid:
1367 *
1368 * If core_pattern does not include a %p (as is the default)
1369 * and core_uses_pid is set, then .%pid will be appended to
1370 * the filename */
1371 if (!pid_in_pattern
1372 && (core_uses_pid || atomic_read(&current->mm->mm_users) != 1)) {
1373 rc = snprintf(out_ptr, out_end - out_ptr,
1374 ".%d", current->tgid);
1375 if (rc > out_end - out_ptr)
1376 goto out;
1377 out_ptr += rc;
1378 }
1379 out:
1380 *out_ptr = 0;
1381}
1382
1383static void zap_threads (struct mm_struct *mm)
1384{
1385 struct task_struct *g, *p;
1386 struct task_struct *tsk = current;
1387 struct completion *vfork_done = tsk->vfork_done;
1388 int traced = 0;
1389
1390 /*
1391 * Make sure nobody is waiting for us to release the VM,
1392 * otherwise we can deadlock when we wait on each other
1393 */
1394 if (vfork_done) {
1395 tsk->vfork_done = NULL;
1396 complete(vfork_done);
1397 }
1398
1399 read_lock(&tasklist_lock);
1400 do_each_thread(g,p)
1401 if (mm == p->mm && p != tsk) {
1402 force_sig_specific(SIGKILL, p);
1403 mm->core_waiters++;
1404 if (unlikely(p->ptrace) &&
1405 unlikely(p->parent->mm == mm))
1406 traced = 1;
1407 }
1408 while_each_thread(g,p);
1409
1410 read_unlock(&tasklist_lock);
1411
1412 if (unlikely(traced)) {
1413 /*
1414 * We are zapping a thread and the thread it ptraces.
1415 * If the tracee went into a ptrace stop for exit tracing,
1416 * we could deadlock since the tracer is waiting for this
1417 * coredump to finish. Detach them so they can both die.
1418 */
1419 write_lock_irq(&tasklist_lock);
1420 do_each_thread(g,p) {
1421 if (mm == p->mm && p != tsk &&
1422 p->ptrace && p->parent->mm == mm) {
Oleg Nesterov5ecfbae2006-02-15 22:50:10 +03001423 __ptrace_detach(p, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424 }
1425 } while_each_thread(g,p);
1426 write_unlock_irq(&tasklist_lock);
1427 }
1428}
1429
1430static void coredump_wait(struct mm_struct *mm)
1431{
1432 DECLARE_COMPLETION(startup_done);
Oleg Nesterov2384f552005-10-30 15:02:47 -08001433 int core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 mm->core_startup_done = &startup_done;
1436
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 zap_threads(mm);
Oleg Nesterov2384f552005-10-30 15:02:47 -08001438 core_waiters = mm->core_waiters;
1439 up_write(&mm->mmap_sem);
1440
1441 if (core_waiters)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 wait_for_completion(&startup_done);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 BUG_ON(mm->core_waiters);
1444}
1445
1446int do_coredump(long signr, int exit_code, struct pt_regs * regs)
1447{
1448 char corename[CORENAME_MAX_SIZE + 1];
1449 struct mm_struct *mm = current->mm;
1450 struct linux_binfmt * binfmt;
1451 struct inode * inode;
1452 struct file * file;
1453 int retval = 0;
Alan Coxd6e71142005-06-23 00:09:43 -07001454 int fsuid = current->fsuid;
1455 int flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456
1457 binfmt = current->binfmt;
1458 if (!binfmt || !binfmt->core_dump)
1459 goto fail;
1460 down_write(&mm->mmap_sem);
1461 if (!mm->dumpable) {
1462 up_write(&mm->mmap_sem);
1463 goto fail;
1464 }
Alan Coxd6e71142005-06-23 00:09:43 -07001465
1466 /*
1467 * We cannot trust fsuid as being the "true" uid of the
1468 * process nor do we know its entire history. We only know it
1469 * was tainted so we dump it as root in mode 2.
1470 */
1471 if (mm->dumpable == 2) { /* Setuid core dump mode */
1472 flag = O_EXCL; /* Stop rewrite attacks */
1473 current->fsuid = 0; /* Dump root private */
1474 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 mm->dumpable = 0;
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001476
1477 retval = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 spin_lock_irq(&current->sighand->siglock);
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001479 if (!(current->signal->flags & SIGNAL_GROUP_EXIT)) {
1480 current->signal->flags = SIGNAL_GROUP_EXIT;
1481 current->signal->group_exit_code = exit_code;
Oleg Nesterovbb6f6db2006-01-08 01:03:13 -08001482 current->signal->group_stop_count = 0;
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001483 retval = 0;
1484 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 spin_unlock_irq(&current->sighand->siglock);
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001486 if (retval) {
1487 up_write(&mm->mmap_sem);
1488 goto fail;
1489 }
1490
1491 init_completion(&mm->core_done);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 coredump_wait(mm);
1493
1494 /*
1495 * Clear any false indication of pending signals that might
1496 * be seen by the filesystem code called to write the core file.
1497 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 clear_thread_flag(TIF_SIGPENDING);
1499
1500 if (current->signal->rlim[RLIMIT_CORE].rlim_cur < binfmt->min_coredump)
1501 goto fail_unlock;
1502
1503 /*
1504 * lock_kernel() because format_corename() is controlled by sysctl, which
1505 * uses lock_kernel()
1506 */
1507 lock_kernel();
1508 format_corename(corename, core_pattern, signr);
1509 unlock_kernel();
Alan Coxd6e71142005-06-23 00:09:43 -07001510 file = filp_open(corename, O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag, 0600);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 if (IS_ERR(file))
1512 goto fail_unlock;
1513 inode = file->f_dentry->d_inode;
1514 if (inode->i_nlink > 1)
1515 goto close_fail; /* multiple links - don't dump */
1516 if (d_unhashed(file->f_dentry))
1517 goto close_fail;
1518
1519 if (!S_ISREG(inode->i_mode))
1520 goto close_fail;
1521 if (!file->f_op)
1522 goto close_fail;
1523 if (!file->f_op->write)
1524 goto close_fail;
NeilBrown4a301312006-01-08 01:02:39 -08001525 if (do_truncate(file->f_dentry, 0, 0, file) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 goto close_fail;
1527
1528 retval = binfmt->core_dump(signr, regs, file);
1529
1530 if (retval)
1531 current->signal->group_exit_code |= 0x80;
1532close_fail:
1533 filp_close(file, NULL);
1534fail_unlock:
Alan Coxd6e71142005-06-23 00:09:43 -07001535 current->fsuid = fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 complete_all(&mm->core_done);
1537fail:
1538 return retval;
1539}