blob: 8344ba73a2a670d37d2b08a2587e149490f87eaa [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/slab.h>
26#include <linux/file.h>
27#include <linux/mman.h>
28#include <linux/a.out.h>
29#include <linux/stat.h>
30#include <linux/fcntl.h>
31#include <linux/smp_lock.h>
32#include <linux/init.h>
33#include <linux/pagemap.h>
34#include <linux/highmem.h>
35#include <linux/spinlock.h>
36#include <linux/key.h>
37#include <linux/personality.h>
38#include <linux/binfmts.h>
39#include <linux/swap.h>
40#include <linux/utsname.h>
41#include <linux/module.h>
42#include <linux/namei.h>
43#include <linux/proc_fs.h>
44#include <linux/ptrace.h>
45#include <linux/mount.h>
46#include <linux/security.h>
47#include <linux/syscalls.h>
48#include <linux/rmap.h>
49#include <linux/acct.h>
Matt Helsley9f460802005-11-07 00:59:16 -080050#include <linux/cn_proc.h>
Al Viro473ae302006-04-26 14:04:08 -040051#include <linux/audit.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070052
53#include <asm/uaccess.h>
54#include <asm/mmu_context.h>
55
56#ifdef CONFIG_KMOD
57#include <linux/kmod.h>
58#endif
59
60int core_uses_pid;
61char core_pattern[65] = "core";
Alan Coxd6e71142005-06-23 00:09:43 -070062int suid_dumpable = 0;
63
64EXPORT_SYMBOL(suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -070065/* The maximal length of core_pattern is also specified in sysctl.c */
66
67static struct linux_binfmt *formats;
68static DEFINE_RWLOCK(binfmt_lock);
69
70int register_binfmt(struct linux_binfmt * fmt)
71{
72 struct linux_binfmt ** tmp = &formats;
73
74 if (!fmt)
75 return -EINVAL;
76 if (fmt->next)
77 return -EBUSY;
78 write_lock(&binfmt_lock);
79 while (*tmp) {
80 if (fmt == *tmp) {
81 write_unlock(&binfmt_lock);
82 return -EBUSY;
83 }
84 tmp = &(*tmp)->next;
85 }
86 fmt->next = formats;
87 formats = fmt;
88 write_unlock(&binfmt_lock);
89 return 0;
90}
91
92EXPORT_SYMBOL(register_binfmt);
93
94int unregister_binfmt(struct linux_binfmt * fmt)
95{
96 struct linux_binfmt ** tmp = &formats;
97
98 write_lock(&binfmt_lock);
99 while (*tmp) {
100 if (fmt == *tmp) {
101 *tmp = fmt->next;
102 write_unlock(&binfmt_lock);
103 return 0;
104 }
105 tmp = &(*tmp)->next;
106 }
107 write_unlock(&binfmt_lock);
108 return -EINVAL;
109}
110
111EXPORT_SYMBOL(unregister_binfmt);
112
113static inline void put_binfmt(struct linux_binfmt * fmt)
114{
115 module_put(fmt->module);
116}
117
118/*
119 * Note that a shared library must be both readable and executable due to
120 * security reasons.
121 *
122 * Also note that we take the address to load from from the file itself.
123 */
124asmlinkage long sys_uselib(const char __user * library)
125{
126 struct file * file;
127 struct nameidata nd;
128 int error;
129
Oleg Drokinb5005312006-03-25 03:07:01 -0800130 error = __user_path_lookup_open(library, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131 if (error)
132 goto out;
133
134 error = -EINVAL;
135 if (!S_ISREG(nd.dentry->d_inode->i_mode))
136 goto exit;
137
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800138 error = vfs_permission(&nd, MAY_READ | MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139 if (error)
140 goto exit;
141
Trond Myklebust834f2a42005-10-18 14:20:16 -0700142 file = nameidata_to_filp(&nd, O_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 error = PTR_ERR(file);
144 if (IS_ERR(file))
145 goto out;
146
147 error = -ENOEXEC;
148 if(file->f_op) {
149 struct linux_binfmt * fmt;
150
151 read_lock(&binfmt_lock);
152 for (fmt = formats ; fmt ; fmt = fmt->next) {
153 if (!fmt->load_shlib)
154 continue;
155 if (!try_module_get(fmt->module))
156 continue;
157 read_unlock(&binfmt_lock);
158 error = fmt->load_shlib(file);
159 read_lock(&binfmt_lock);
160 put_binfmt(fmt);
161 if (error != -ENOEXEC)
162 break;
163 }
164 read_unlock(&binfmt_lock);
165 }
166 fput(file);
167out:
168 return error;
169exit:
Trond Myklebust834f2a42005-10-18 14:20:16 -0700170 release_open_intent(&nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 path_release(&nd);
172 goto out;
173}
174
175/*
176 * count() counts the number of strings in array ARGV.
177 */
178static int count(char __user * __user * argv, int max)
179{
180 int i = 0;
181
182 if (argv != NULL) {
183 for (;;) {
184 char __user * p;
185
186 if (get_user(p, argv))
187 return -EFAULT;
188 if (!p)
189 break;
190 argv++;
191 if(++i > max)
192 return -E2BIG;
193 cond_resched();
194 }
195 }
196 return i;
197}
198
199/*
200 * 'copy_strings()' copies argument/environment strings from user
201 * memory to free pages in kernel mem. These are in a format ready
202 * to be put directly into the top of new user memory.
203 */
Adrian Bunk75c96f82005-05-05 16:16:09 -0700204static int copy_strings(int argc, char __user * __user * argv,
205 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206{
207 struct page *kmapped_page = NULL;
208 char *kaddr = NULL;
209 int ret;
210
211 while (argc-- > 0) {
212 char __user *str;
213 int len;
214 unsigned long pos;
215
216 if (get_user(str, argv+argc) ||
217 !(len = strnlen_user(str, bprm->p))) {
218 ret = -EFAULT;
219 goto out;
220 }
221
222 if (bprm->p < len) {
223 ret = -E2BIG;
224 goto out;
225 }
226
227 bprm->p -= len;
228 /* XXX: add architecture specific overflow check here. */
229 pos = bprm->p;
230
231 while (len > 0) {
232 int i, new, err;
233 int offset, bytes_to_copy;
234 struct page *page;
235
236 offset = pos % PAGE_SIZE;
237 i = pos/PAGE_SIZE;
238 page = bprm->page[i];
239 new = 0;
240 if (!page) {
241 page = alloc_page(GFP_HIGHUSER);
242 bprm->page[i] = page;
243 if (!page) {
244 ret = -ENOMEM;
245 goto out;
246 }
247 new = 1;
248 }
249
250 if (page != kmapped_page) {
251 if (kmapped_page)
252 kunmap(kmapped_page);
253 kmapped_page = page;
254 kaddr = kmap(kmapped_page);
255 }
256 if (new && offset)
257 memset(kaddr, 0, offset);
258 bytes_to_copy = PAGE_SIZE - offset;
259 if (bytes_to_copy > len) {
260 bytes_to_copy = len;
261 if (new)
262 memset(kaddr+offset+len, 0,
263 PAGE_SIZE-offset-len);
264 }
265 err = copy_from_user(kaddr+offset, str, bytes_to_copy);
266 if (err) {
267 ret = -EFAULT;
268 goto out;
269 }
270
271 pos += bytes_to_copy;
272 str += bytes_to_copy;
273 len -= bytes_to_copy;
274 }
275 }
276 ret = 0;
277out:
278 if (kmapped_page)
279 kunmap(kmapped_page);
280 return ret;
281}
282
283/*
284 * Like copy_strings, but get argv and its values from kernel memory.
285 */
286int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm)
287{
288 int r;
289 mm_segment_t oldfs = get_fs();
290 set_fs(KERNEL_DS);
291 r = copy_strings(argc, (char __user * __user *)argv, bprm);
292 set_fs(oldfs);
293 return r;
294}
295
296EXPORT_SYMBOL(copy_strings_kernel);
297
298#ifdef CONFIG_MMU
299/*
300 * This routine is used to map in a page into an address space: needed by
301 * execve() for the initial stack and environment pages.
302 *
303 * vma->vm_mm->mmap_sem is held for writing.
304 */
305void install_arg_page(struct vm_area_struct *vma,
306 struct page *page, unsigned long address)
307{
308 struct mm_struct *mm = vma->vm_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 pte_t * pte;
Hugh Dickinsc74df322005-10-29 18:16:23 -0700310 spinlock_t *ptl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311
312 if (unlikely(anon_vma_prepare(vma)))
Hugh Dickinsc74df322005-10-29 18:16:23 -0700313 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314
315 flush_dcache_page(page);
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800316 pte = get_locked_pte(mm, address, &ptl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 if (!pte)
318 goto out;
319 if (!pte_none(*pte)) {
Hugh Dickinsc74df322005-10-29 18:16:23 -0700320 pte_unmap_unlock(pte, ptl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 goto out;
322 }
Hugh Dickins42946212005-10-29 18:16:05 -0700323 inc_mm_counter(mm, anon_rss);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 lru_cache_add_active(page);
325 set_pte_at(mm, address, pte, pte_mkdirty(pte_mkwrite(mk_pte(
326 page, vma->vm_page_prot))));
Nick Piggin9617d952006-01-06 00:11:12 -0800327 page_add_new_anon_rmap(page, vma, address);
Hugh Dickinsc74df322005-10-29 18:16:23 -0700328 pte_unmap_unlock(pte, ptl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329
330 /* no need for flush_tlb */
331 return;
332out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 __free_page(page);
334 force_sig(SIGKILL, current);
335}
336
337#define EXTRA_STACK_VM_PAGES 20 /* random */
338
339int setup_arg_pages(struct linux_binprm *bprm,
340 unsigned long stack_top,
341 int executable_stack)
342{
343 unsigned long stack_base;
344 struct vm_area_struct *mpnt;
345 struct mm_struct *mm = current->mm;
346 int i, ret;
347 long arg_size;
348
349#ifdef CONFIG_STACK_GROWSUP
350 /* Move the argument and environment strings to the bottom of the
351 * stack space.
352 */
353 int offset, j;
354 char *to, *from;
355
356 /* Start by shifting all the pages down */
357 i = 0;
358 for (j = 0; j < MAX_ARG_PAGES; j++) {
359 struct page *page = bprm->page[j];
360 if (!page)
361 continue;
362 bprm->page[i++] = page;
363 }
364
365 /* Now move them within their pages */
366 offset = bprm->p % PAGE_SIZE;
367 to = kmap(bprm->page[0]);
368 for (j = 1; j < i; j++) {
369 memmove(to, to + offset, PAGE_SIZE - offset);
370 from = kmap(bprm->page[j]);
371 memcpy(to + PAGE_SIZE - offset, from, offset);
372 kunmap(bprm->page[j - 1]);
373 to = from;
374 }
375 memmove(to, to + offset, PAGE_SIZE - offset);
376 kunmap(bprm->page[j - 1]);
377
378 /* Limit stack size to 1GB */
379 stack_base = current->signal->rlim[RLIMIT_STACK].rlim_max;
380 if (stack_base > (1 << 30))
381 stack_base = 1 << 30;
382 stack_base = PAGE_ALIGN(stack_top - stack_base);
383
384 /* Adjust bprm->p to point to the end of the strings. */
385 bprm->p = stack_base + PAGE_SIZE * i - offset;
386
387 mm->arg_start = stack_base;
388 arg_size = i << PAGE_SHIFT;
389
390 /* zero pages that were copied above */
391 while (i < MAX_ARG_PAGES)
392 bprm->page[i++] = NULL;
393#else
394 stack_base = arch_align_stack(stack_top - MAX_ARG_PAGES*PAGE_SIZE);
395 stack_base = PAGE_ALIGN(stack_base);
396 bprm->p += stack_base;
397 mm->arg_start = bprm->p;
398 arg_size = stack_top - (PAGE_MASK & (unsigned long) mm->arg_start);
399#endif
400
401 arg_size += EXTRA_STACK_VM_PAGES * PAGE_SIZE;
402
403 if (bprm->loader)
404 bprm->loader += stack_base;
405 bprm->exec += stack_base;
406
407 mpnt = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL);
408 if (!mpnt)
409 return -ENOMEM;
410
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 memset(mpnt, 0, sizeof(*mpnt));
412
413 down_write(&mm->mmap_sem);
414 {
415 mpnt->vm_mm = mm;
416#ifdef CONFIG_STACK_GROWSUP
417 mpnt->vm_start = stack_base;
418 mpnt->vm_end = stack_base + arg_size;
419#else
420 mpnt->vm_end = stack_top;
421 mpnt->vm_start = mpnt->vm_end - arg_size;
422#endif
423 /* Adjust stack execute permissions; explicitly enable
424 * for EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X
425 * and leave alone (arch default) otherwise. */
426 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
427 mpnt->vm_flags = VM_STACK_FLAGS | VM_EXEC;
428 else if (executable_stack == EXSTACK_DISABLE_X)
429 mpnt->vm_flags = VM_STACK_FLAGS & ~VM_EXEC;
430 else
431 mpnt->vm_flags = VM_STACK_FLAGS;
432 mpnt->vm_flags |= mm->def_flags;
433 mpnt->vm_page_prot = protection_map[mpnt->vm_flags & 0x7];
434 if ((ret = insert_vm_struct(mm, mpnt))) {
435 up_write(&mm->mmap_sem);
436 kmem_cache_free(vm_area_cachep, mpnt);
437 return ret;
438 }
439 mm->stack_vm = mm->total_vm = vma_pages(mpnt);
440 }
441
442 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
443 struct page *page = bprm->page[i];
444 if (page) {
445 bprm->page[i] = NULL;
446 install_arg_page(mpnt, page, stack_base);
447 }
448 stack_base += PAGE_SIZE;
449 }
450 up_write(&mm->mmap_sem);
451
452 return 0;
453}
454
455EXPORT_SYMBOL(setup_arg_pages);
456
457#define free_arg_pages(bprm) do { } while (0)
458
459#else
460
461static inline void free_arg_pages(struct linux_binprm *bprm)
462{
463 int i;
464
465 for (i = 0; i < MAX_ARG_PAGES; i++) {
466 if (bprm->page[i])
467 __free_page(bprm->page[i]);
468 bprm->page[i] = NULL;
469 }
470}
471
472#endif /* CONFIG_MMU */
473
474struct file *open_exec(const char *name)
475{
476 struct nameidata nd;
477 int err;
478 struct file *file;
479
Oleg Drokinb5005312006-03-25 03:07:01 -0800480 err = path_lookup_open(AT_FDCWD, name, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 file = ERR_PTR(err);
482
483 if (!err) {
484 struct inode *inode = nd.dentry->d_inode;
485 file = ERR_PTR(-EACCES);
486 if (!(nd.mnt->mnt_flags & MNT_NOEXEC) &&
487 S_ISREG(inode->i_mode)) {
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800488 int err = vfs_permission(&nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 if (!err && !(inode->i_mode & 0111))
490 err = -EACCES;
491 file = ERR_PTR(err);
492 if (!err) {
Trond Myklebust834f2a42005-10-18 14:20:16 -0700493 file = nameidata_to_filp(&nd, O_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 if (!IS_ERR(file)) {
495 err = deny_write_access(file);
496 if (err) {
497 fput(file);
498 file = ERR_PTR(err);
499 }
500 }
501out:
502 return file;
503 }
504 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700505 release_open_intent(&nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 path_release(&nd);
507 }
508 goto out;
509}
510
511EXPORT_SYMBOL(open_exec);
512
513int kernel_read(struct file *file, unsigned long offset,
514 char *addr, unsigned long count)
515{
516 mm_segment_t old_fs;
517 loff_t pos = offset;
518 int result;
519
520 old_fs = get_fs();
521 set_fs(get_ds());
522 /* The cast to a user pointer is valid due to the set_fs() */
523 result = vfs_read(file, (void __user *)addr, count, &pos);
524 set_fs(old_fs);
525 return result;
526}
527
528EXPORT_SYMBOL(kernel_read);
529
530static int exec_mmap(struct mm_struct *mm)
531{
532 struct task_struct *tsk;
533 struct mm_struct * old_mm, *active_mm;
534
535 /* Notify parent that we're no longer interested in the old VM */
536 tsk = current;
537 old_mm = current->mm;
538 mm_release(tsk, old_mm);
539
540 if (old_mm) {
541 /*
542 * Make sure that if there is a core dump in progress
543 * for the old mm, we get out and die instead of going
544 * through with the exec. We must hold mmap_sem around
545 * checking core_waiters and changing tsk->mm. The
546 * core-inducing thread will increment core_waiters for
547 * each thread whose ->mm == old_mm.
548 */
549 down_read(&old_mm->mmap_sem);
550 if (unlikely(old_mm->core_waiters)) {
551 up_read(&old_mm->mmap_sem);
552 return -EINTR;
553 }
554 }
555 task_lock(tsk);
556 active_mm = tsk->active_mm;
557 tsk->mm = mm;
558 tsk->active_mm = mm;
559 activate_mm(active_mm, mm);
560 task_unlock(tsk);
561 arch_pick_mmap_layout(mm);
562 if (old_mm) {
563 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200564 BUG_ON(active_mm != old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 mmput(old_mm);
566 return 0;
567 }
568 mmdrop(active_mm);
569 return 0;
570}
571
572/*
573 * This function makes sure the current process has its own signal table,
574 * so that flush_signal_handlers can later reset the handlers without
575 * disturbing other processes. (Other processes might share the signal
576 * table via the CLONE_SIGHAND option to clone().)
577 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800578static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579{
580 struct signal_struct *sig = tsk->signal;
581 struct sighand_struct *newsighand, *oldsighand = tsk->sighand;
582 spinlock_t *lock = &oldsighand->siglock;
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300583 struct task_struct *leader = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 int count;
585
586 /*
587 * If we don't share sighandlers, then we aren't sharing anything
588 * and we can just re-use it all.
589 */
590 if (atomic_read(&oldsighand->count) <= 1) {
591 BUG_ON(atomic_read(&sig->count) != 1);
592 exit_itimers(sig);
593 return 0;
594 }
595
596 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
597 if (!newsighand)
598 return -ENOMEM;
599
600 if (thread_group_empty(current))
601 goto no_thread_group;
602
603 /*
604 * Kill all other threads in the thread group.
605 * We must hold tasklist_lock to call zap_other_threads.
606 */
607 read_lock(&tasklist_lock);
608 spin_lock_irq(lock);
609 if (sig->flags & SIGNAL_GROUP_EXIT) {
610 /*
611 * Another group action in progress, just
612 * return so that the signal is processed.
613 */
614 spin_unlock_irq(lock);
615 read_unlock(&tasklist_lock);
616 kmem_cache_free(sighand_cachep, newsighand);
617 return -EAGAIN;
618 }
Oleg Nesterov14342612006-03-28 16:10:59 -0800619
620 /*
621 * child_reaper ignores SIGKILL, change it now.
622 * Reparenting needs write_lock on tasklist_lock,
623 * so it is safe to do it under read_lock.
624 */
625 if (unlikely(current->group_leader == child_reaper))
626 child_reaper = current;
627
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 zap_other_threads(current);
629 read_unlock(&tasklist_lock);
630
631 /*
632 * Account for the thread group leader hanging around:
633 */
Oleg Nesterov9e4e23b2005-10-30 15:01:37 -0800634 count = 1;
635 if (!thread_group_leader(current)) {
636 count = 2;
Roland McGrath53231252005-07-12 13:58:27 -0700637 /*
638 * The SIGALRM timer survives the exec, but needs to point
639 * at us as the new group leader now. We have a race with
640 * a timer firing now getting the old leader, so we need to
641 * synchronize with any firing (by calling del_timer_sync)
642 * before we can safely let the old group leader die.
643 */
Roman Zippel05cfb612006-03-26 01:38:12 -0800644 sig->tsk = current;
Oleg Nesterov932aeaf2005-10-30 15:02:17 -0800645 spin_unlock_irq(lock);
Thomas Gleixner2ff678b2006-01-09 20:52:34 -0800646 if (hrtimer_cancel(&sig->real_timer))
647 hrtimer_restart(&sig->real_timer);
Oleg Nesterov932aeaf2005-10-30 15:02:17 -0800648 spin_lock_irq(lock);
Roland McGrath53231252005-07-12 13:58:27 -0700649 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 while (atomic_read(&sig->count) > count) {
651 sig->group_exit_task = current;
652 sig->notify_count = count;
653 __set_current_state(TASK_UNINTERRUPTIBLE);
654 spin_unlock_irq(lock);
655 schedule();
656 spin_lock_irq(lock);
657 }
658 sig->group_exit_task = NULL;
659 sig->notify_count = 0;
660 spin_unlock_irq(lock);
661
662 /*
663 * At this point all other threads have exited, all we have to
664 * do is to wait for the thread group leader to become inactive,
665 * and to assume its PID:
666 */
667 if (!thread_group_leader(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 /*
669 * Wait for the thread group leader to be a zombie.
670 * It should already be zombie at this point, most
671 * of the time.
672 */
Oleg Nesterov14342612006-03-28 16:10:59 -0800673 leader = current->group_leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 while (leader->exit_state != EXIT_ZOMBIE)
675 yield();
676
Roland McGrathf5e90282006-04-10 22:54:16 -0700677 /*
678 * The only record we have of the real-time age of a
679 * process, regardless of execs it's done, is start_time.
680 * All the past CPU time is accumulated in signal_struct
681 * from sister threads now dead. But in this non-leader
682 * exec, nothing survives from the original leader thread,
683 * whose birth marks the true age of this process now.
684 * When we take on its identity by switching to its PID, we
685 * also take its birthdate (always earlier than our own).
686 */
687 current->start_time = leader->start_time;
688
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 write_lock_irq(&tasklist_lock);
690
Linus Torvaldsc2a0f592005-06-18 13:06:22 -0700691 BUG_ON(leader->tgid != current->tgid);
692 BUG_ON(current->pid == current->tgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 /*
694 * An exec() starts a new thread group with the
695 * TGID of the previous thread group. Rehash the
696 * two threads with a switched PID, and release
697 * the former thread group leader:
698 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800699
700 /* Become a process group leader with the old leader's pid.
701 * Note: The old leader also uses thispid until release_task
702 * is called. Odd but simple and correct.
703 */
704 detach_pid(current, PIDTYPE_PID);
705 current->pid = leader->pid;
706 attach_pid(current, PIDTYPE_PID, current->pid);
707 attach_pid(current, PIDTYPE_PGID, current->signal->pgrp);
708 attach_pid(current, PIDTYPE_SID, current->signal->session);
Oleg Nesterov2ceb8692006-06-26 00:26:04 -0700709 list_replace_rcu(&leader->tasks, &current->tasks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 current->group_leader = current;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600712 leader->group_leader = current;
713
714 /* Reduce leader to a thread */
715 detach_pid(leader, PIDTYPE_PGID);
716 detach_pid(leader, PIDTYPE_SID);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 current->exit_signal = SIGCHLD;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800719
720 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
721 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722
723 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 }
725
726 /*
Alexander Nybergfb085cf2005-09-14 18:54:06 +0200727 * There may be one thread left which is just exiting,
728 * but it's safe to stop telling the group to kill themselves.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 */
730 sig->flags = 0;
731
732no_thread_group:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 exit_itimers(sig);
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300734 if (leader)
735 release_task(leader);
736
737 BUG_ON(atomic_read(&sig->count) != 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738
739 if (atomic_read(&oldsighand->count) == 1) {
740 /*
741 * Now that we nuked the rest of the thread group,
742 * it turns out we are not sharing sighand any more either.
743 * So we can just keep it.
744 */
745 kmem_cache_free(sighand_cachep, newsighand);
746 } else {
747 /*
748 * Move our state over to newsighand and switch it in.
749 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 atomic_set(&newsighand->count, 1);
751 memcpy(newsighand->action, oldsighand->action,
752 sizeof(newsighand->action));
753
754 write_lock_irq(&tasklist_lock);
755 spin_lock(&oldsighand->siglock);
756 spin_lock(&newsighand->siglock);
757
Ingo Molnare56d0902006-01-08 01:01:37 -0800758 rcu_assign_pointer(current->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 recalc_sigpending();
760
761 spin_unlock(&newsighand->siglock);
762 spin_unlock(&oldsighand->siglock);
763 write_unlock_irq(&tasklist_lock);
764
765 if (atomic_dec_and_test(&oldsighand->count))
Oleg Nesterovaa1757f2006-03-28 16:11:12 -0800766 kmem_cache_free(sighand_cachep, oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 }
768
Linus Torvaldsc2a0f592005-06-18 13:06:22 -0700769 BUG_ON(!thread_group_leader(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 return 0;
771}
772
773/*
774 * These functions flushes out all traces of the currently running executable
775 * so that a new one can be started
776 */
777
Arjan van de Ven858119e2006-01-14 13:20:43 -0800778static void flush_old_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779{
780 long j = -1;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700781 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782
783 spin_lock(&files->file_lock);
784 for (;;) {
785 unsigned long set, i;
786
787 j++;
788 i = j * __NFDBITS;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700789 fdt = files_fdtable(files);
790 if (i >= fdt->max_fds || i >= fdt->max_fdset)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700792 set = fdt->close_on_exec->fds_bits[j];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 if (!set)
794 continue;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700795 fdt->close_on_exec->fds_bits[j] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 spin_unlock(&files->file_lock);
797 for ( ; set ; i++,set >>= 1) {
798 if (set & 1) {
799 sys_close(i);
800 }
801 }
802 spin_lock(&files->file_lock);
803
804 }
805 spin_unlock(&files->file_lock);
806}
807
808void get_task_comm(char *buf, struct task_struct *tsk)
809{
810 /* buf must be at least sizeof(tsk->comm) in size */
811 task_lock(tsk);
812 strncpy(buf, tsk->comm, sizeof(tsk->comm));
813 task_unlock(tsk);
814}
815
816void set_task_comm(struct task_struct *tsk, char *buf)
817{
818 task_lock(tsk);
819 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
820 task_unlock(tsk);
821}
822
823int flush_old_exec(struct linux_binprm * bprm)
824{
825 char * name;
826 int i, ch, retval;
827 struct files_struct *files;
828 char tcomm[sizeof(current->comm)];
829
830 /*
831 * Make sure we have a private signal table and that
832 * we are unassociated from the previous thread group.
833 */
834 retval = de_thread(current);
835 if (retval)
836 goto out;
837
838 /*
839 * Make sure we have private file handles. Ask the
840 * fork helper to do the work for us and the exit
841 * helper to do the cleanup of the old one.
842 */
843 files = current->files; /* refcounted so safe to hold */
844 retval = unshare_files();
845 if (retval)
846 goto out;
847 /*
848 * Release all of the old mmap stuff
849 */
850 retval = exec_mmap(bprm->mm);
851 if (retval)
852 goto mmap_failed;
853
854 bprm->mm = NULL; /* We're using it now */
855
856 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 put_files_struct(files);
858
859 current->sas_ss_sp = current->sas_ss_size = 0;
860
861 if (current->euid == current->uid && current->egid == current->gid)
862 current->mm->dumpable = 1;
Alan Coxd6e71142005-06-23 00:09:43 -0700863 else
864 current->mm->dumpable = suid_dumpable;
865
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 name = bprm->filename;
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700867
868 /* Copies the binary name from after last slash */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 for (i=0; (ch = *(name++)) != '\0';) {
870 if (ch == '/')
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700871 i = 0; /* overwrite what we wrote */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 else
873 if (i < (sizeof(tcomm) - 1))
874 tcomm[i++] = ch;
875 }
876 tcomm[i] = '\0';
877 set_task_comm(current, tcomm);
878
879 current->flags &= ~PF_RANDOMIZE;
880 flush_thread();
881
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -0800882 /* Set the new mm task size. We have to do that late because it may
883 * depend on TIF_32BIT which is only updated in flush_thread() on
884 * some architectures like powerpc
885 */
886 current->mm->task_size = TASK_SIZE;
887
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 if (bprm->e_uid != current->euid || bprm->e_gid != current->egid ||
Christoph Hellwig8c744fb2005-11-08 21:35:04 -0800889 file_permission(bprm->file, MAY_READ) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)) {
891 suid_keys(current);
Alan Coxd6e71142005-06-23 00:09:43 -0700892 current->mm->dumpable = suid_dumpable;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 }
894
895 /* An exec changes our domain. We are no longer part of the thread
896 group */
897
898 current->self_exec_id++;
899
900 flush_signal_handlers(current, 0);
901 flush_old_files(current->files);
902
903 return 0;
904
905mmap_failed:
906 put_files_struct(current->files);
907 current->files = files;
908out:
909 return retval;
910}
911
912EXPORT_SYMBOL(flush_old_exec);
913
914/*
915 * Fill the binprm structure from the inode.
916 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
917 */
918int prepare_binprm(struct linux_binprm *bprm)
919{
920 int mode;
921 struct inode * inode = bprm->file->f_dentry->d_inode;
922 int retval;
923
924 mode = inode->i_mode;
925 /*
926 * Check execute perms again - if the caller has CAP_DAC_OVERRIDE,
927 * generic_permission lets a non-executable through
928 */
929 if (!(mode & 0111)) /* with at least _one_ execute bit set */
930 return -EACCES;
931 if (bprm->file->f_op == NULL)
932 return -EACCES;
933
934 bprm->e_uid = current->euid;
935 bprm->e_gid = current->egid;
936
937 if(!(bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID)) {
938 /* Set-uid? */
939 if (mode & S_ISUID) {
940 current->personality &= ~PER_CLEAR_ON_SETID;
941 bprm->e_uid = inode->i_uid;
942 }
943
944 /* Set-gid? */
945 /*
946 * If setgid is set but no group execute bit then this
947 * is a candidate for mandatory locking, not a setgid
948 * executable.
949 */
950 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
951 current->personality &= ~PER_CLEAR_ON_SETID;
952 bprm->e_gid = inode->i_gid;
953 }
954 }
955
956 /* fill in binprm security blob */
957 retval = security_bprm_set(bprm);
958 if (retval)
959 return retval;
960
961 memset(bprm->buf,0,BINPRM_BUF_SIZE);
962 return kernel_read(bprm->file,0,bprm->buf,BINPRM_BUF_SIZE);
963}
964
965EXPORT_SYMBOL(prepare_binprm);
966
Arjan van de Ven858119e2006-01-14 13:20:43 -0800967static int unsafe_exec(struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968{
969 int unsafe = 0;
970 if (p->ptrace & PT_PTRACED) {
971 if (p->ptrace & PT_PTRACE_CAP)
972 unsafe |= LSM_UNSAFE_PTRACE_CAP;
973 else
974 unsafe |= LSM_UNSAFE_PTRACE;
975 }
976 if (atomic_read(&p->fs->count) > 1 ||
977 atomic_read(&p->files->count) > 1 ||
978 atomic_read(&p->sighand->count) > 1)
979 unsafe |= LSM_UNSAFE_SHARE;
980
981 return unsafe;
982}
983
984void compute_creds(struct linux_binprm *bprm)
985{
986 int unsafe;
987
988 if (bprm->e_uid != current->uid)
989 suid_keys(current);
990 exec_keys(current);
991
992 task_lock(current);
993 unsafe = unsafe_exec(current);
994 security_bprm_apply_creds(bprm, unsafe);
995 task_unlock(current);
996 security_bprm_post_apply_creds(bprm);
997}
998
999EXPORT_SYMBOL(compute_creds);
1000
1001void remove_arg_zero(struct linux_binprm *bprm)
1002{
1003 if (bprm->argc) {
1004 unsigned long offset;
1005 char * kaddr;
1006 struct page *page;
1007
1008 offset = bprm->p % PAGE_SIZE;
1009 goto inside;
1010
1011 while (bprm->p++, *(kaddr+offset++)) {
1012 if (offset != PAGE_SIZE)
1013 continue;
1014 offset = 0;
1015 kunmap_atomic(kaddr, KM_USER0);
1016inside:
1017 page = bprm->page[bprm->p/PAGE_SIZE];
1018 kaddr = kmap_atomic(page, KM_USER0);
1019 }
1020 kunmap_atomic(kaddr, KM_USER0);
1021 bprm->argc--;
1022 }
1023}
1024
1025EXPORT_SYMBOL(remove_arg_zero);
1026
1027/*
1028 * cycle the list of binary formats handler, until one recognizes the image
1029 */
1030int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
1031{
1032 int try,retval;
1033 struct linux_binfmt *fmt;
1034#ifdef __alpha__
1035 /* handle /sbin/loader.. */
1036 {
1037 struct exec * eh = (struct exec *) bprm->buf;
1038
1039 if (!bprm->loader && eh->fh.f_magic == 0x183 &&
1040 (eh->fh.f_flags & 0x3000) == 0x3000)
1041 {
1042 struct file * file;
1043 unsigned long loader;
1044
1045 allow_write_access(bprm->file);
1046 fput(bprm->file);
1047 bprm->file = NULL;
1048
1049 loader = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1050
1051 file = open_exec("/sbin/loader");
1052 retval = PTR_ERR(file);
1053 if (IS_ERR(file))
1054 return retval;
1055
1056 /* Remember if the application is TASO. */
1057 bprm->sh_bang = eh->ah.entry < 0x100000000UL;
1058
1059 bprm->file = file;
1060 bprm->loader = loader;
1061 retval = prepare_binprm(bprm);
1062 if (retval<0)
1063 return retval;
1064 /* should call search_binary_handler recursively here,
1065 but it does not matter */
1066 }
1067 }
1068#endif
1069 retval = security_bprm_check(bprm);
1070 if (retval)
1071 return retval;
1072
1073 /* kernel module loader fixup */
1074 /* so we don't try to load run modprobe in kernel space. */
1075 set_fs(USER_DS);
Al Viro473ae302006-04-26 14:04:08 -04001076
1077 retval = audit_bprm(bprm);
1078 if (retval)
1079 return retval;
1080
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 retval = -ENOENT;
1082 for (try=0; try<2; try++) {
1083 read_lock(&binfmt_lock);
1084 for (fmt = formats ; fmt ; fmt = fmt->next) {
1085 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1086 if (!fn)
1087 continue;
1088 if (!try_module_get(fmt->module))
1089 continue;
1090 read_unlock(&binfmt_lock);
1091 retval = fn(bprm, regs);
1092 if (retval >= 0) {
1093 put_binfmt(fmt);
1094 allow_write_access(bprm->file);
1095 if (bprm->file)
1096 fput(bprm->file);
1097 bprm->file = NULL;
1098 current->did_exec = 1;
Matt Helsley9f460802005-11-07 00:59:16 -08001099 proc_exec_connector(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 return retval;
1101 }
1102 read_lock(&binfmt_lock);
1103 put_binfmt(fmt);
1104 if (retval != -ENOEXEC || bprm->mm == NULL)
1105 break;
1106 if (!bprm->file) {
1107 read_unlock(&binfmt_lock);
1108 return retval;
1109 }
1110 }
1111 read_unlock(&binfmt_lock);
1112 if (retval != -ENOEXEC || bprm->mm == NULL) {
1113 break;
1114#ifdef CONFIG_KMOD
1115 }else{
1116#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1117 if (printable(bprm->buf[0]) &&
1118 printable(bprm->buf[1]) &&
1119 printable(bprm->buf[2]) &&
1120 printable(bprm->buf[3]))
1121 break; /* -ENOEXEC */
1122 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1123#endif
1124 }
1125 }
1126 return retval;
1127}
1128
1129EXPORT_SYMBOL(search_binary_handler);
1130
1131/*
1132 * sys_execve() executes a new program.
1133 */
1134int do_execve(char * filename,
1135 char __user *__user *argv,
1136 char __user *__user *envp,
1137 struct pt_regs * regs)
1138{
1139 struct linux_binprm *bprm;
1140 struct file *file;
1141 int retval;
1142 int i;
1143
1144 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001145 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 if (!bprm)
1147 goto out_ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148
1149 file = open_exec(filename);
1150 retval = PTR_ERR(file);
1151 if (IS_ERR(file))
1152 goto out_kfree;
1153
1154 sched_exec();
1155
1156 bprm->p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1157
1158 bprm->file = file;
1159 bprm->filename = filename;
1160 bprm->interp = filename;
1161 bprm->mm = mm_alloc();
1162 retval = -ENOMEM;
1163 if (!bprm->mm)
1164 goto out_file;
1165
1166 retval = init_new_context(current, bprm->mm);
1167 if (retval < 0)
1168 goto out_mm;
1169
1170 bprm->argc = count(argv, bprm->p / sizeof(void *));
1171 if ((retval = bprm->argc) < 0)
1172 goto out_mm;
1173
1174 bprm->envc = count(envp, bprm->p / sizeof(void *));
1175 if ((retval = bprm->envc) < 0)
1176 goto out_mm;
1177
1178 retval = security_bprm_alloc(bprm);
1179 if (retval)
1180 goto out;
1181
1182 retval = prepare_binprm(bprm);
1183 if (retval < 0)
1184 goto out;
1185
1186 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1187 if (retval < 0)
1188 goto out;
1189
1190 bprm->exec = bprm->p;
1191 retval = copy_strings(bprm->envc, envp, bprm);
1192 if (retval < 0)
1193 goto out;
1194
1195 retval = copy_strings(bprm->argc, argv, bprm);
1196 if (retval < 0)
1197 goto out;
1198
1199 retval = search_binary_handler(bprm,regs);
1200 if (retval >= 0) {
1201 free_arg_pages(bprm);
1202
1203 /* execve success */
1204 security_bprm_free(bprm);
1205 acct_update_integrals(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 kfree(bprm);
1207 return retval;
1208 }
1209
1210out:
1211 /* Something went wrong, return the inode and free the argument pages*/
1212 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
1213 struct page * page = bprm->page[i];
1214 if (page)
1215 __free_page(page);
1216 }
1217
1218 if (bprm->security)
1219 security_bprm_free(bprm);
1220
1221out_mm:
1222 if (bprm->mm)
1223 mmdrop(bprm->mm);
1224
1225out_file:
1226 if (bprm->file) {
1227 allow_write_access(bprm->file);
1228 fput(bprm->file);
1229 }
1230
1231out_kfree:
1232 kfree(bprm);
1233
1234out_ret:
1235 return retval;
1236}
1237
1238int set_binfmt(struct linux_binfmt *new)
1239{
1240 struct linux_binfmt *old = current->binfmt;
1241
1242 if (new) {
1243 if (!try_module_get(new->module))
1244 return -1;
1245 }
1246 current->binfmt = new;
1247 if (old)
1248 module_put(old->module);
1249 return 0;
1250}
1251
1252EXPORT_SYMBOL(set_binfmt);
1253
1254#define CORENAME_MAX_SIZE 64
1255
1256/* format_corename will inspect the pattern parameter, and output a
1257 * name into corename, which must have space for at least
1258 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1259 */
1260static void format_corename(char *corename, const char *pattern, long signr)
1261{
1262 const char *pat_ptr = pattern;
1263 char *out_ptr = corename;
1264 char *const out_end = corename + CORENAME_MAX_SIZE;
1265 int rc;
1266 int pid_in_pattern = 0;
1267
1268 /* Repeat as long as we have more pattern to process and more output
1269 space */
1270 while (*pat_ptr) {
1271 if (*pat_ptr != '%') {
1272 if (out_ptr == out_end)
1273 goto out;
1274 *out_ptr++ = *pat_ptr++;
1275 } else {
1276 switch (*++pat_ptr) {
1277 case 0:
1278 goto out;
1279 /* Double percent, output one percent */
1280 case '%':
1281 if (out_ptr == out_end)
1282 goto out;
1283 *out_ptr++ = '%';
1284 break;
1285 /* pid */
1286 case 'p':
1287 pid_in_pattern = 1;
1288 rc = snprintf(out_ptr, out_end - out_ptr,
1289 "%d", current->tgid);
1290 if (rc > out_end - out_ptr)
1291 goto out;
1292 out_ptr += rc;
1293 break;
1294 /* uid */
1295 case 'u':
1296 rc = snprintf(out_ptr, out_end - out_ptr,
1297 "%d", current->uid);
1298 if (rc > out_end - out_ptr)
1299 goto out;
1300 out_ptr += rc;
1301 break;
1302 /* gid */
1303 case 'g':
1304 rc = snprintf(out_ptr, out_end - out_ptr,
1305 "%d", current->gid);
1306 if (rc > out_end - out_ptr)
1307 goto out;
1308 out_ptr += rc;
1309 break;
1310 /* signal that caused the coredump */
1311 case 's':
1312 rc = snprintf(out_ptr, out_end - out_ptr,
1313 "%ld", signr);
1314 if (rc > out_end - out_ptr)
1315 goto out;
1316 out_ptr += rc;
1317 break;
1318 /* UNIX time of coredump */
1319 case 't': {
1320 struct timeval tv;
1321 do_gettimeofday(&tv);
1322 rc = snprintf(out_ptr, out_end - out_ptr,
1323 "%lu", tv.tv_sec);
1324 if (rc > out_end - out_ptr)
1325 goto out;
1326 out_ptr += rc;
1327 break;
1328 }
1329 /* hostname */
1330 case 'h':
1331 down_read(&uts_sem);
1332 rc = snprintf(out_ptr, out_end - out_ptr,
1333 "%s", system_utsname.nodename);
1334 up_read(&uts_sem);
1335 if (rc > out_end - out_ptr)
1336 goto out;
1337 out_ptr += rc;
1338 break;
1339 /* executable */
1340 case 'e':
1341 rc = snprintf(out_ptr, out_end - out_ptr,
1342 "%s", current->comm);
1343 if (rc > out_end - out_ptr)
1344 goto out;
1345 out_ptr += rc;
1346 break;
1347 default:
1348 break;
1349 }
1350 ++pat_ptr;
1351 }
1352 }
1353 /* Backward compatibility with core_uses_pid:
1354 *
1355 * If core_pattern does not include a %p (as is the default)
1356 * and core_uses_pid is set, then .%pid will be appended to
1357 * the filename */
1358 if (!pid_in_pattern
1359 && (core_uses_pid || atomic_read(&current->mm->mm_users) != 1)) {
1360 rc = snprintf(out_ptr, out_end - out_ptr,
1361 ".%d", current->tgid);
1362 if (rc > out_end - out_ptr)
1363 goto out;
1364 out_ptr += rc;
1365 }
1366 out:
1367 *out_ptr = 0;
1368}
1369
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001370static void zap_process(struct task_struct *start)
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001371{
1372 struct task_struct *t;
Oleg Nesterov281de332006-06-26 00:26:06 -07001373
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001374 start->signal->flags = SIGNAL_GROUP_EXIT;
1375 start->signal->group_stop_count = 0;
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001376
1377 t = start;
1378 do {
1379 if (t != current && t->mm) {
1380 t->mm->core_waiters++;
Oleg Nesterov281de332006-06-26 00:26:06 -07001381 sigaddset(&t->pending.signal, SIGKILL);
1382 signal_wake_up(t, 1);
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001383 }
1384 } while ((t = next_thread(t)) != start);
1385}
1386
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001387static inline int zap_threads(struct task_struct *tsk, struct mm_struct *mm,
1388 int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389{
1390 struct task_struct *g, *p;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001391 unsigned long flags;
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001392 int err = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001394 spin_lock_irq(&tsk->sighand->siglock);
1395 if (!(tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001396 tsk->signal->group_exit_code = exit_code;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001397 zap_process(tsk);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001398 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 }
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001400 spin_unlock_irq(&tsk->sighand->siglock);
1401 if (err)
1402 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001404 if (atomic_read(&mm->mm_users) == mm->core_waiters + 1)
1405 goto done;
1406
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001407 rcu_read_lock();
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001408 for_each_process(g) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001409 if (g == tsk->group_leader)
1410 continue;
1411
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001412 p = g;
1413 do {
1414 if (p->mm) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001415 if (p->mm == mm) {
1416 /*
1417 * p->sighand can't disappear, but
1418 * may be changed by de_thread()
1419 */
1420 lock_task_sighand(p, &flags);
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001421 zap_process(p);
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001422 unlock_task_sighand(p, &flags);
1423 }
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001424 break;
1425 }
1426 } while ((p = next_thread(p)) != g);
1427 }
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001428 rcu_read_unlock();
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001429done:
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001430 return mm->core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431}
1432
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001433static int coredump_wait(int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434{
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001435 struct task_struct *tsk = current;
1436 struct mm_struct *mm = tsk->mm;
1437 struct completion startup_done;
1438 struct completion *vfork_done;
Oleg Nesterov2384f552005-10-30 15:02:47 -08001439 int core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001441 init_completion(&mm->core_done);
1442 init_completion(&startup_done);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 mm->core_startup_done = &startup_done;
1444
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001445 core_waiters = zap_threads(tsk, mm, exit_code);
Oleg Nesterov2384f552005-10-30 15:02:47 -08001446 up_write(&mm->mmap_sem);
1447
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001448 if (unlikely(core_waiters < 0))
1449 goto fail;
1450
1451 /*
1452 * Make sure nobody is waiting for us to release the VM,
1453 * otherwise we can deadlock when we wait on each other
1454 */
1455 vfork_done = tsk->vfork_done;
1456 if (vfork_done) {
1457 tsk->vfork_done = NULL;
1458 complete(vfork_done);
1459 }
1460
Oleg Nesterov2384f552005-10-30 15:02:47 -08001461 if (core_waiters)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 wait_for_completion(&startup_done);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001463fail:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 BUG_ON(mm->core_waiters);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001465 return core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466}
1467
1468int do_coredump(long signr, int exit_code, struct pt_regs * regs)
1469{
1470 char corename[CORENAME_MAX_SIZE + 1];
1471 struct mm_struct *mm = current->mm;
1472 struct linux_binfmt * binfmt;
1473 struct inode * inode;
1474 struct file * file;
1475 int retval = 0;
Alan Coxd6e71142005-06-23 00:09:43 -07001476 int fsuid = current->fsuid;
1477 int flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478
1479 binfmt = current->binfmt;
1480 if (!binfmt || !binfmt->core_dump)
1481 goto fail;
1482 down_write(&mm->mmap_sem);
1483 if (!mm->dumpable) {
1484 up_write(&mm->mmap_sem);
1485 goto fail;
1486 }
Alan Coxd6e71142005-06-23 00:09:43 -07001487
1488 /*
1489 * We cannot trust fsuid as being the "true" uid of the
1490 * process nor do we know its entire history. We only know it
1491 * was tainted so we dump it as root in mode 2.
1492 */
1493 if (mm->dumpable == 2) { /* Setuid core dump mode */
1494 flag = O_EXCL; /* Stop rewrite attacks */
1495 current->fsuid = 0; /* Dump root private */
1496 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 mm->dumpable = 0;
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001498
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001499 retval = coredump_wait(exit_code);
1500 if (retval < 0)
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001501 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502
1503 /*
1504 * Clear any false indication of pending signals that might
1505 * be seen by the filesystem code called to write the core file.
1506 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 clear_thread_flag(TIF_SIGPENDING);
1508
1509 if (current->signal->rlim[RLIMIT_CORE].rlim_cur < binfmt->min_coredump)
1510 goto fail_unlock;
1511
1512 /*
1513 * lock_kernel() because format_corename() is controlled by sysctl, which
1514 * uses lock_kernel()
1515 */
1516 lock_kernel();
1517 format_corename(corename, core_pattern, signr);
1518 unlock_kernel();
Alan Coxd6e71142005-06-23 00:09:43 -07001519 file = filp_open(corename, O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag, 0600);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 if (IS_ERR(file))
1521 goto fail_unlock;
1522 inode = file->f_dentry->d_inode;
1523 if (inode->i_nlink > 1)
1524 goto close_fail; /* multiple links - don't dump */
1525 if (d_unhashed(file->f_dentry))
1526 goto close_fail;
1527
1528 if (!S_ISREG(inode->i_mode))
1529 goto close_fail;
1530 if (!file->f_op)
1531 goto close_fail;
1532 if (!file->f_op->write)
1533 goto close_fail;
NeilBrown4a301312006-01-08 01:02:39 -08001534 if (do_truncate(file->f_dentry, 0, 0, file) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 goto close_fail;
1536
1537 retval = binfmt->core_dump(signr, regs, file);
1538
1539 if (retval)
1540 current->signal->group_exit_code |= 0x80;
1541close_fail:
1542 filp_close(file, NULL);
1543fail_unlock:
Alan Coxd6e71142005-06-23 00:09:43 -07001544 current->fsuid = fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 complete_all(&mm->core_done);
1546fail:
1547 return retval;
1548}