blob: 11fe93f7363c9dfb0cd05f52992ac0e6dd6bba9c [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>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080041#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#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>
Jay Lan8f0ab512006-09-30 23:28:59 -070050#include <linux/tsacct_kern.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;
Andi Kleend025c9d2006-09-30 23:29:28 -070062char core_pattern[128] = "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
Christoph Lametere94b1762006-12-06 20:33:17 -0800408 mpnt = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 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 file = ERR_PTR(err);
491 if (!err) {
Trond Myklebust834f2a42005-10-18 14:20:16 -0700492 file = nameidata_to_filp(&nd, O_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 if (!IS_ERR(file)) {
494 err = deny_write_access(file);
495 if (err) {
496 fput(file);
497 file = ERR_PTR(err);
498 }
499 }
500out:
501 return file;
502 }
503 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700504 release_open_intent(&nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 path_release(&nd);
506 }
507 goto out;
508}
509
510EXPORT_SYMBOL(open_exec);
511
512int kernel_read(struct file *file, unsigned long offset,
513 char *addr, unsigned long count)
514{
515 mm_segment_t old_fs;
516 loff_t pos = offset;
517 int result;
518
519 old_fs = get_fs();
520 set_fs(get_ds());
521 /* The cast to a user pointer is valid due to the set_fs() */
522 result = vfs_read(file, (void __user *)addr, count, &pos);
523 set_fs(old_fs);
524 return result;
525}
526
527EXPORT_SYMBOL(kernel_read);
528
529static int exec_mmap(struct mm_struct *mm)
530{
531 struct task_struct *tsk;
532 struct mm_struct * old_mm, *active_mm;
533
534 /* Notify parent that we're no longer interested in the old VM */
535 tsk = current;
536 old_mm = current->mm;
537 mm_release(tsk, old_mm);
538
539 if (old_mm) {
540 /*
541 * Make sure that if there is a core dump in progress
542 * for the old mm, we get out and die instead of going
543 * through with the exec. We must hold mmap_sem around
544 * checking core_waiters and changing tsk->mm. The
545 * core-inducing thread will increment core_waiters for
546 * each thread whose ->mm == old_mm.
547 */
548 down_read(&old_mm->mmap_sem);
549 if (unlikely(old_mm->core_waiters)) {
550 up_read(&old_mm->mmap_sem);
551 return -EINTR;
552 }
553 }
554 task_lock(tsk);
555 active_mm = tsk->active_mm;
556 tsk->mm = mm;
557 tsk->active_mm = mm;
558 activate_mm(active_mm, mm);
559 task_unlock(tsk);
560 arch_pick_mmap_layout(mm);
561 if (old_mm) {
562 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200563 BUG_ON(active_mm != old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 mmput(old_mm);
565 return 0;
566 }
567 mmdrop(active_mm);
568 return 0;
569}
570
571/*
572 * This function makes sure the current process has its own signal table,
573 * so that flush_signal_handlers can later reset the handlers without
574 * disturbing other processes. (Other processes might share the signal
575 * table via the CLONE_SIGHAND option to clone().)
576 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800577static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578{
579 struct signal_struct *sig = tsk->signal;
580 struct sighand_struct *newsighand, *oldsighand = tsk->sighand;
581 spinlock_t *lock = &oldsighand->siglock;
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300582 struct task_struct *leader = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 int count;
584
585 /*
586 * If we don't share sighandlers, then we aren't sharing anything
587 * and we can just re-use it all.
588 */
589 if (atomic_read(&oldsighand->count) <= 1) {
590 BUG_ON(atomic_read(&sig->count) != 1);
591 exit_itimers(sig);
592 return 0;
593 }
594
595 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
596 if (!newsighand)
597 return -ENOMEM;
598
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700599 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 goto no_thread_group;
601
602 /*
603 * Kill all other threads in the thread group.
604 * We must hold tasklist_lock to call zap_other_threads.
605 */
606 read_lock(&tasklist_lock);
607 spin_lock_irq(lock);
608 if (sig->flags & SIGNAL_GROUP_EXIT) {
609 /*
610 * Another group action in progress, just
611 * return so that the signal is processed.
612 */
613 spin_unlock_irq(lock);
614 read_unlock(&tasklist_lock);
615 kmem_cache_free(sighand_cachep, newsighand);
616 return -EAGAIN;
617 }
Oleg Nesterov14342612006-03-28 16:10:59 -0800618
619 /*
620 * child_reaper ignores SIGKILL, change it now.
621 * Reparenting needs write_lock on tasklist_lock,
622 * so it is safe to do it under read_lock.
623 */
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -0800624 if (unlikely(tsk->group_leader == child_reaper(tsk)))
625 tsk->nsproxy->pid_ns->child_reaper = tsk;
Oleg Nesterov14342612006-03-28 16:10:59 -0800626
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700627 zap_other_threads(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 read_unlock(&tasklist_lock);
629
630 /*
631 * Account for the thread group leader hanging around:
632 */
Oleg Nesterov9e4e23b2005-10-30 15:01:37 -0800633 count = 1;
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700634 if (!thread_group_leader(tsk)) {
Oleg Nesterov9e4e23b2005-10-30 15:01:37 -0800635 count = 2;
Roland McGrath53231252005-07-12 13:58:27 -0700636 /*
637 * The SIGALRM timer survives the exec, but needs to point
638 * at us as the new group leader now. We have a race with
639 * a timer firing now getting the old leader, so we need to
640 * synchronize with any firing (by calling del_timer_sync)
641 * before we can safely let the old group leader die.
642 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700643 sig->tsk = tsk;
Oleg Nesterov932aeaf2005-10-30 15:02:17 -0800644 spin_unlock_irq(lock);
Thomas Gleixner2ff678b2006-01-09 20:52:34 -0800645 if (hrtimer_cancel(&sig->real_timer))
646 hrtimer_restart(&sig->real_timer);
Oleg Nesterov932aeaf2005-10-30 15:02:17 -0800647 spin_lock_irq(lock);
Roland McGrath53231252005-07-12 13:58:27 -0700648 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 while (atomic_read(&sig->count) > count) {
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700650 sig->group_exit_task = tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 sig->notify_count = count;
652 __set_current_state(TASK_UNINTERRUPTIBLE);
653 spin_unlock_irq(lock);
654 schedule();
655 spin_lock_irq(lock);
656 }
657 sig->group_exit_task = NULL;
658 sig->notify_count = 0;
659 spin_unlock_irq(lock);
660
661 /*
662 * At this point all other threads have exited, all we have to
663 * do is to wait for the thread group leader to become inactive,
664 * and to assume its PID:
665 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700666 if (!thread_group_leader(tsk)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 /*
668 * Wait for the thread group leader to be a zombie.
669 * It should already be zombie at this point, most
670 * of the time.
671 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700672 leader = tsk->group_leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 while (leader->exit_state != EXIT_ZOMBIE)
674 yield();
675
Roland McGrathf5e90282006-04-10 22:54:16 -0700676 /*
677 * The only record we have of the real-time age of a
678 * process, regardless of execs it's done, is start_time.
679 * All the past CPU time is accumulated in signal_struct
680 * from sister threads now dead. But in this non-leader
681 * exec, nothing survives from the original leader thread,
682 * whose birth marks the true age of this process now.
683 * When we take on its identity by switching to its PID, we
684 * also take its birthdate (always earlier than our own).
685 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700686 tsk->start_time = leader->start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -0700687
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 write_lock_irq(&tasklist_lock);
689
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700690 BUG_ON(leader->tgid != tsk->tgid);
691 BUG_ON(tsk->pid == tsk->tgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 /*
693 * An exec() starts a new thread group with the
694 * TGID of the previous thread group. Rehash the
695 * two threads with a switched PID, and release
696 * the former thread group leader:
697 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800698
699 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700700 * The old leader becomes a thread of the this thread group.
701 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800702 * is called. Odd but simple and correct.
703 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700704 detach_pid(tsk, PIDTYPE_PID);
705 tsk->pid = leader->pid;
706 attach_pid(tsk, PIDTYPE_PID, tsk->pid);
707 transfer_pid(leader, tsk, PIDTYPE_PGID);
708 transfer_pid(leader, tsk, PIDTYPE_SID);
709 list_replace_rcu(&leader->tasks, &tsk->tasks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700711 tsk->group_leader = tsk;
712 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600713
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700714 tsk->exit_signal = SIGCHLD;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800715
716 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
717 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718
719 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 }
721
722 /*
Alexander Nybergfb085cf2005-09-14 18:54:06 +0200723 * There may be one thread left which is just exiting,
724 * but it's safe to stop telling the group to kill themselves.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 */
726 sig->flags = 0;
727
728no_thread_group:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 exit_itimers(sig);
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300730 if (leader)
731 release_task(leader);
732
733 BUG_ON(atomic_read(&sig->count) != 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734
735 if (atomic_read(&oldsighand->count) == 1) {
736 /*
737 * Now that we nuked the rest of the thread group,
738 * it turns out we are not sharing sighand any more either.
739 * So we can just keep it.
740 */
741 kmem_cache_free(sighand_cachep, newsighand);
742 } else {
743 /*
744 * Move our state over to newsighand and switch it in.
745 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 atomic_set(&newsighand->count, 1);
747 memcpy(newsighand->action, oldsighand->action,
748 sizeof(newsighand->action));
749
750 write_lock_irq(&tasklist_lock);
751 spin_lock(&oldsighand->siglock);
Dave Jones513627d2006-08-27 01:23:57 -0700752 spin_lock_nested(&newsighand->siglock, SINGLE_DEPTH_NESTING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700754 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 recalc_sigpending();
756
757 spin_unlock(&newsighand->siglock);
758 spin_unlock(&oldsighand->siglock);
759 write_unlock_irq(&tasklist_lock);
760
761 if (atomic_dec_and_test(&oldsighand->count))
Oleg Nesterovaa1757f2006-03-28 16:11:12 -0800762 kmem_cache_free(sighand_cachep, oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 }
764
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700765 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 return 0;
767}
768
769/*
770 * These functions flushes out all traces of the currently running executable
771 * so that a new one can be started
772 */
773
Arjan van de Ven858119e2006-01-14 13:20:43 -0800774static void flush_old_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775{
776 long j = -1;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700777 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
779 spin_lock(&files->file_lock);
780 for (;;) {
781 unsigned long set, i;
782
783 j++;
784 i = j * __NFDBITS;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700785 fdt = files_fdtable(files);
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800786 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700788 set = fdt->close_on_exec->fds_bits[j];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 if (!set)
790 continue;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700791 fdt->close_on_exec->fds_bits[j] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 spin_unlock(&files->file_lock);
793 for ( ; set ; i++,set >>= 1) {
794 if (set & 1) {
795 sys_close(i);
796 }
797 }
798 spin_lock(&files->file_lock);
799
800 }
801 spin_unlock(&files->file_lock);
802}
803
804void get_task_comm(char *buf, struct task_struct *tsk)
805{
806 /* buf must be at least sizeof(tsk->comm) in size */
807 task_lock(tsk);
808 strncpy(buf, tsk->comm, sizeof(tsk->comm));
809 task_unlock(tsk);
810}
811
812void set_task_comm(struct task_struct *tsk, char *buf)
813{
814 task_lock(tsk);
815 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
816 task_unlock(tsk);
817}
818
819int flush_old_exec(struct linux_binprm * bprm)
820{
821 char * name;
822 int i, ch, retval;
823 struct files_struct *files;
824 char tcomm[sizeof(current->comm)];
825
826 /*
827 * Make sure we have a private signal table and that
828 * we are unassociated from the previous thread group.
829 */
830 retval = de_thread(current);
831 if (retval)
832 goto out;
833
834 /*
835 * Make sure we have private file handles. Ask the
836 * fork helper to do the work for us and the exit
837 * helper to do the cleanup of the old one.
838 */
839 files = current->files; /* refcounted so safe to hold */
840 retval = unshare_files();
841 if (retval)
842 goto out;
843 /*
844 * Release all of the old mmap stuff
845 */
846 retval = exec_mmap(bprm->mm);
847 if (retval)
848 goto mmap_failed;
849
850 bprm->mm = NULL; /* We're using it now */
851
852 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 put_files_struct(files);
854
855 current->sas_ss_sp = current->sas_ss_size = 0;
856
857 if (current->euid == current->uid && current->egid == current->gid)
858 current->mm->dumpable = 1;
Alan Coxd6e71142005-06-23 00:09:43 -0700859 else
860 current->mm->dumpable = suid_dumpable;
861
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 name = bprm->filename;
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700863
864 /* Copies the binary name from after last slash */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 for (i=0; (ch = *(name++)) != '\0';) {
866 if (ch == '/')
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700867 i = 0; /* overwrite what we wrote */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 else
869 if (i < (sizeof(tcomm) - 1))
870 tcomm[i++] = ch;
871 }
872 tcomm[i] = '\0';
873 set_task_comm(current, tcomm);
874
875 current->flags &= ~PF_RANDOMIZE;
876 flush_thread();
877
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -0800878 /* Set the new mm task size. We have to do that late because it may
879 * depend on TIF_32BIT which is only updated in flush_thread() on
880 * some architectures like powerpc
881 */
882 current->mm->task_size = TASK_SIZE;
883
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 if (bprm->e_uid != current->euid || bprm->e_gid != current->egid ||
Christoph Hellwig8c744fb2005-11-08 21:35:04 -0800885 file_permission(bprm->file, MAY_READ) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)) {
887 suid_keys(current);
Alan Coxd6e71142005-06-23 00:09:43 -0700888 current->mm->dumpable = suid_dumpable;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 }
890
891 /* An exec changes our domain. We are no longer part of the thread
892 group */
893
894 current->self_exec_id++;
895
896 flush_signal_handlers(current, 0);
897 flush_old_files(current->files);
898
899 return 0;
900
901mmap_failed:
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -0700902 reset_files_struct(current, files);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903out:
904 return retval;
905}
906
907EXPORT_SYMBOL(flush_old_exec);
908
909/*
910 * Fill the binprm structure from the inode.
911 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
912 */
913int prepare_binprm(struct linux_binprm *bprm)
914{
915 int mode;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800916 struct inode * inode = bprm->file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 int retval;
918
919 mode = inode->i_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 if (bprm->file->f_op == NULL)
921 return -EACCES;
922
923 bprm->e_uid = current->euid;
924 bprm->e_gid = current->egid;
925
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800926 if(!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 /* Set-uid? */
928 if (mode & S_ISUID) {
929 current->personality &= ~PER_CLEAR_ON_SETID;
930 bprm->e_uid = inode->i_uid;
931 }
932
933 /* Set-gid? */
934 /*
935 * If setgid is set but no group execute bit then this
936 * is a candidate for mandatory locking, not a setgid
937 * executable.
938 */
939 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
940 current->personality &= ~PER_CLEAR_ON_SETID;
941 bprm->e_gid = inode->i_gid;
942 }
943 }
944
945 /* fill in binprm security blob */
946 retval = security_bprm_set(bprm);
947 if (retval)
948 return retval;
949
950 memset(bprm->buf,0,BINPRM_BUF_SIZE);
951 return kernel_read(bprm->file,0,bprm->buf,BINPRM_BUF_SIZE);
952}
953
954EXPORT_SYMBOL(prepare_binprm);
955
Arjan van de Ven858119e2006-01-14 13:20:43 -0800956static int unsafe_exec(struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957{
958 int unsafe = 0;
959 if (p->ptrace & PT_PTRACED) {
960 if (p->ptrace & PT_PTRACE_CAP)
961 unsafe |= LSM_UNSAFE_PTRACE_CAP;
962 else
963 unsafe |= LSM_UNSAFE_PTRACE;
964 }
965 if (atomic_read(&p->fs->count) > 1 ||
966 atomic_read(&p->files->count) > 1 ||
967 atomic_read(&p->sighand->count) > 1)
968 unsafe |= LSM_UNSAFE_SHARE;
969
970 return unsafe;
971}
972
973void compute_creds(struct linux_binprm *bprm)
974{
975 int unsafe;
976
977 if (bprm->e_uid != current->uid)
978 suid_keys(current);
979 exec_keys(current);
980
981 task_lock(current);
982 unsafe = unsafe_exec(current);
983 security_bprm_apply_creds(bprm, unsafe);
984 task_unlock(current);
985 security_bprm_post_apply_creds(bprm);
986}
987
988EXPORT_SYMBOL(compute_creds);
989
990void remove_arg_zero(struct linux_binprm *bprm)
991{
992 if (bprm->argc) {
993 unsigned long offset;
994 char * kaddr;
995 struct page *page;
996
997 offset = bprm->p % PAGE_SIZE;
998 goto inside;
999
1000 while (bprm->p++, *(kaddr+offset++)) {
1001 if (offset != PAGE_SIZE)
1002 continue;
1003 offset = 0;
1004 kunmap_atomic(kaddr, KM_USER0);
1005inside:
1006 page = bprm->page[bprm->p/PAGE_SIZE];
1007 kaddr = kmap_atomic(page, KM_USER0);
1008 }
1009 kunmap_atomic(kaddr, KM_USER0);
1010 bprm->argc--;
1011 }
1012}
1013
1014EXPORT_SYMBOL(remove_arg_zero);
1015
1016/*
1017 * cycle the list of binary formats handler, until one recognizes the image
1018 */
1019int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
1020{
1021 int try,retval;
1022 struct linux_binfmt *fmt;
1023#ifdef __alpha__
1024 /* handle /sbin/loader.. */
1025 {
1026 struct exec * eh = (struct exec *) bprm->buf;
1027
1028 if (!bprm->loader && eh->fh.f_magic == 0x183 &&
1029 (eh->fh.f_flags & 0x3000) == 0x3000)
1030 {
1031 struct file * file;
1032 unsigned long loader;
1033
1034 allow_write_access(bprm->file);
1035 fput(bprm->file);
1036 bprm->file = NULL;
1037
1038 loader = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1039
1040 file = open_exec("/sbin/loader");
1041 retval = PTR_ERR(file);
1042 if (IS_ERR(file))
1043 return retval;
1044
1045 /* Remember if the application is TASO. */
1046 bprm->sh_bang = eh->ah.entry < 0x100000000UL;
1047
1048 bprm->file = file;
1049 bprm->loader = loader;
1050 retval = prepare_binprm(bprm);
1051 if (retval<0)
1052 return retval;
1053 /* should call search_binary_handler recursively here,
1054 but it does not matter */
1055 }
1056 }
1057#endif
1058 retval = security_bprm_check(bprm);
1059 if (retval)
1060 return retval;
1061
1062 /* kernel module loader fixup */
1063 /* so we don't try to load run modprobe in kernel space. */
1064 set_fs(USER_DS);
Al Viro473ae302006-04-26 14:04:08 -04001065
1066 retval = audit_bprm(bprm);
1067 if (retval)
1068 return retval;
1069
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 retval = -ENOENT;
1071 for (try=0; try<2; try++) {
1072 read_lock(&binfmt_lock);
1073 for (fmt = formats ; fmt ; fmt = fmt->next) {
1074 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1075 if (!fn)
1076 continue;
1077 if (!try_module_get(fmt->module))
1078 continue;
1079 read_unlock(&binfmt_lock);
1080 retval = fn(bprm, regs);
1081 if (retval >= 0) {
1082 put_binfmt(fmt);
1083 allow_write_access(bprm->file);
1084 if (bprm->file)
1085 fput(bprm->file);
1086 bprm->file = NULL;
1087 current->did_exec = 1;
Matt Helsley9f460802005-11-07 00:59:16 -08001088 proc_exec_connector(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 return retval;
1090 }
1091 read_lock(&binfmt_lock);
1092 put_binfmt(fmt);
1093 if (retval != -ENOEXEC || bprm->mm == NULL)
1094 break;
1095 if (!bprm->file) {
1096 read_unlock(&binfmt_lock);
1097 return retval;
1098 }
1099 }
1100 read_unlock(&binfmt_lock);
1101 if (retval != -ENOEXEC || bprm->mm == NULL) {
1102 break;
1103#ifdef CONFIG_KMOD
1104 }else{
1105#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1106 if (printable(bprm->buf[0]) &&
1107 printable(bprm->buf[1]) &&
1108 printable(bprm->buf[2]) &&
1109 printable(bprm->buf[3]))
1110 break; /* -ENOEXEC */
1111 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1112#endif
1113 }
1114 }
1115 return retval;
1116}
1117
1118EXPORT_SYMBOL(search_binary_handler);
1119
1120/*
1121 * sys_execve() executes a new program.
1122 */
1123int do_execve(char * filename,
1124 char __user *__user *argv,
1125 char __user *__user *envp,
1126 struct pt_regs * regs)
1127{
1128 struct linux_binprm *bprm;
1129 struct file *file;
1130 int retval;
1131 int i;
1132
1133 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001134 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 if (!bprm)
1136 goto out_ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137
1138 file = open_exec(filename);
1139 retval = PTR_ERR(file);
1140 if (IS_ERR(file))
1141 goto out_kfree;
1142
1143 sched_exec();
1144
1145 bprm->p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1146
1147 bprm->file = file;
1148 bprm->filename = filename;
1149 bprm->interp = filename;
1150 bprm->mm = mm_alloc();
1151 retval = -ENOMEM;
1152 if (!bprm->mm)
1153 goto out_file;
1154
1155 retval = init_new_context(current, bprm->mm);
1156 if (retval < 0)
1157 goto out_mm;
1158
1159 bprm->argc = count(argv, bprm->p / sizeof(void *));
1160 if ((retval = bprm->argc) < 0)
1161 goto out_mm;
1162
1163 bprm->envc = count(envp, bprm->p / sizeof(void *));
1164 if ((retval = bprm->envc) < 0)
1165 goto out_mm;
1166
1167 retval = security_bprm_alloc(bprm);
1168 if (retval)
1169 goto out;
1170
1171 retval = prepare_binprm(bprm);
1172 if (retval < 0)
1173 goto out;
1174
1175 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1176 if (retval < 0)
1177 goto out;
1178
1179 bprm->exec = bprm->p;
1180 retval = copy_strings(bprm->envc, envp, bprm);
1181 if (retval < 0)
1182 goto out;
1183
1184 retval = copy_strings(bprm->argc, argv, bprm);
1185 if (retval < 0)
1186 goto out;
1187
1188 retval = search_binary_handler(bprm,regs);
1189 if (retval >= 0) {
1190 free_arg_pages(bprm);
1191
1192 /* execve success */
1193 security_bprm_free(bprm);
1194 acct_update_integrals(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195 kfree(bprm);
1196 return retval;
1197 }
1198
1199out:
1200 /* Something went wrong, return the inode and free the argument pages*/
1201 for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
1202 struct page * page = bprm->page[i];
1203 if (page)
1204 __free_page(page);
1205 }
1206
1207 if (bprm->security)
1208 security_bprm_free(bprm);
1209
1210out_mm:
1211 if (bprm->mm)
1212 mmdrop(bprm->mm);
1213
1214out_file:
1215 if (bprm->file) {
1216 allow_write_access(bprm->file);
1217 fput(bprm->file);
1218 }
1219
1220out_kfree:
1221 kfree(bprm);
1222
1223out_ret:
1224 return retval;
1225}
1226
1227int set_binfmt(struct linux_binfmt *new)
1228{
1229 struct linux_binfmt *old = current->binfmt;
1230
1231 if (new) {
1232 if (!try_module_get(new->module))
1233 return -1;
1234 }
1235 current->binfmt = new;
1236 if (old)
1237 module_put(old->module);
1238 return 0;
1239}
1240
1241EXPORT_SYMBOL(set_binfmt);
1242
1243#define CORENAME_MAX_SIZE 64
1244
1245/* format_corename will inspect the pattern parameter, and output a
1246 * name into corename, which must have space for at least
1247 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1248 */
1249static void format_corename(char *corename, const char *pattern, long signr)
1250{
1251 const char *pat_ptr = pattern;
1252 char *out_ptr = corename;
1253 char *const out_end = corename + CORENAME_MAX_SIZE;
1254 int rc;
1255 int pid_in_pattern = 0;
1256
1257 /* Repeat as long as we have more pattern to process and more output
1258 space */
1259 while (*pat_ptr) {
1260 if (*pat_ptr != '%') {
1261 if (out_ptr == out_end)
1262 goto out;
1263 *out_ptr++ = *pat_ptr++;
1264 } else {
1265 switch (*++pat_ptr) {
1266 case 0:
1267 goto out;
1268 /* Double percent, output one percent */
1269 case '%':
1270 if (out_ptr == out_end)
1271 goto out;
1272 *out_ptr++ = '%';
1273 break;
1274 /* pid */
1275 case 'p':
1276 pid_in_pattern = 1;
1277 rc = snprintf(out_ptr, out_end - out_ptr,
1278 "%d", current->tgid);
1279 if (rc > out_end - out_ptr)
1280 goto out;
1281 out_ptr += rc;
1282 break;
1283 /* uid */
1284 case 'u':
1285 rc = snprintf(out_ptr, out_end - out_ptr,
1286 "%d", current->uid);
1287 if (rc > out_end - out_ptr)
1288 goto out;
1289 out_ptr += rc;
1290 break;
1291 /* gid */
1292 case 'g':
1293 rc = snprintf(out_ptr, out_end - out_ptr,
1294 "%d", current->gid);
1295 if (rc > out_end - out_ptr)
1296 goto out;
1297 out_ptr += rc;
1298 break;
1299 /* signal that caused the coredump */
1300 case 's':
1301 rc = snprintf(out_ptr, out_end - out_ptr,
1302 "%ld", signr);
1303 if (rc > out_end - out_ptr)
1304 goto out;
1305 out_ptr += rc;
1306 break;
1307 /* UNIX time of coredump */
1308 case 't': {
1309 struct timeval tv;
1310 do_gettimeofday(&tv);
1311 rc = snprintf(out_ptr, out_end - out_ptr,
1312 "%lu", tv.tv_sec);
1313 if (rc > out_end - out_ptr)
1314 goto out;
1315 out_ptr += rc;
1316 break;
1317 }
1318 /* hostname */
1319 case 'h':
1320 down_read(&uts_sem);
1321 rc = snprintf(out_ptr, out_end - out_ptr,
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001322 "%s", utsname()->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 up_read(&uts_sem);
1324 if (rc > out_end - out_ptr)
1325 goto out;
1326 out_ptr += rc;
1327 break;
1328 /* executable */
1329 case 'e':
1330 rc = snprintf(out_ptr, out_end - out_ptr,
1331 "%s", current->comm);
1332 if (rc > out_end - out_ptr)
1333 goto out;
1334 out_ptr += rc;
1335 break;
1336 default:
1337 break;
1338 }
1339 ++pat_ptr;
1340 }
1341 }
1342 /* Backward compatibility with core_uses_pid:
1343 *
1344 * If core_pattern does not include a %p (as is the default)
1345 * and core_uses_pid is set, then .%pid will be appended to
1346 * the filename */
1347 if (!pid_in_pattern
1348 && (core_uses_pid || atomic_read(&current->mm->mm_users) != 1)) {
1349 rc = snprintf(out_ptr, out_end - out_ptr,
1350 ".%d", current->tgid);
1351 if (rc > out_end - out_ptr)
1352 goto out;
1353 out_ptr += rc;
1354 }
1355 out:
1356 *out_ptr = 0;
1357}
1358
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001359static void zap_process(struct task_struct *start)
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001360{
1361 struct task_struct *t;
Oleg Nesterov281de332006-06-26 00:26:06 -07001362
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001363 start->signal->flags = SIGNAL_GROUP_EXIT;
1364 start->signal->group_stop_count = 0;
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001365
1366 t = start;
1367 do {
1368 if (t != current && t->mm) {
1369 t->mm->core_waiters++;
Oleg Nesterov281de332006-06-26 00:26:06 -07001370 sigaddset(&t->pending.signal, SIGKILL);
1371 signal_wake_up(t, 1);
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001372 }
1373 } while ((t = next_thread(t)) != start);
1374}
1375
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001376static inline int zap_threads(struct task_struct *tsk, struct mm_struct *mm,
1377 int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378{
1379 struct task_struct *g, *p;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001380 unsigned long flags;
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001381 int err = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001383 spin_lock_irq(&tsk->sighand->siglock);
1384 if (!(tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001385 tsk->signal->group_exit_code = exit_code;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001386 zap_process(tsk);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001387 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 }
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001389 spin_unlock_irq(&tsk->sighand->siglock);
1390 if (err)
1391 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001393 if (atomic_read(&mm->mm_users) == mm->core_waiters + 1)
1394 goto done;
1395
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001396 rcu_read_lock();
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001397 for_each_process(g) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001398 if (g == tsk->group_leader)
1399 continue;
1400
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001401 p = g;
1402 do {
1403 if (p->mm) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001404 if (p->mm == mm) {
1405 /*
1406 * p->sighand can't disappear, but
1407 * may be changed by de_thread()
1408 */
1409 lock_task_sighand(p, &flags);
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001410 zap_process(p);
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001411 unlock_task_sighand(p, &flags);
1412 }
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001413 break;
1414 }
1415 } while ((p = next_thread(p)) != g);
1416 }
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001417 rcu_read_unlock();
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001418done:
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001419 return mm->core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420}
1421
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001422static int coredump_wait(int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423{
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001424 struct task_struct *tsk = current;
1425 struct mm_struct *mm = tsk->mm;
1426 struct completion startup_done;
1427 struct completion *vfork_done;
Oleg Nesterov2384f552005-10-30 15:02:47 -08001428 int core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001430 init_completion(&mm->core_done);
1431 init_completion(&startup_done);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 mm->core_startup_done = &startup_done;
1433
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001434 core_waiters = zap_threads(tsk, mm, exit_code);
Oleg Nesterov2384f552005-10-30 15:02:47 -08001435 up_write(&mm->mmap_sem);
1436
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001437 if (unlikely(core_waiters < 0))
1438 goto fail;
1439
1440 /*
1441 * Make sure nobody is waiting for us to release the VM,
1442 * otherwise we can deadlock when we wait on each other
1443 */
1444 vfork_done = tsk->vfork_done;
1445 if (vfork_done) {
1446 tsk->vfork_done = NULL;
1447 complete(vfork_done);
1448 }
1449
Oleg Nesterov2384f552005-10-30 15:02:47 -08001450 if (core_waiters)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451 wait_for_completion(&startup_done);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001452fail:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 BUG_ON(mm->core_waiters);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001454 return core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455}
1456
1457int do_coredump(long signr, int exit_code, struct pt_regs * regs)
1458{
1459 char corename[CORENAME_MAX_SIZE + 1];
1460 struct mm_struct *mm = current->mm;
1461 struct linux_binfmt * binfmt;
1462 struct inode * inode;
1463 struct file * file;
1464 int retval = 0;
Alan Coxd6e71142005-06-23 00:09:43 -07001465 int fsuid = current->fsuid;
1466 int flag = 0;
Andi Kleend025c9d2006-09-30 23:29:28 -07001467 int ispipe = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468
1469 binfmt = current->binfmt;
1470 if (!binfmt || !binfmt->core_dump)
1471 goto fail;
1472 down_write(&mm->mmap_sem);
1473 if (!mm->dumpable) {
1474 up_write(&mm->mmap_sem);
1475 goto fail;
1476 }
Alan Coxd6e71142005-06-23 00:09:43 -07001477
1478 /*
1479 * We cannot trust fsuid as being the "true" uid of the
1480 * process nor do we know its entire history. We only know it
1481 * was tainted so we dump it as root in mode 2.
1482 */
1483 if (mm->dumpable == 2) { /* Setuid core dump mode */
1484 flag = O_EXCL; /* Stop rewrite attacks */
1485 current->fsuid = 0; /* Dump root private */
1486 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 mm->dumpable = 0;
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001488
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001489 retval = coredump_wait(exit_code);
1490 if (retval < 0)
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001491 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492
1493 /*
1494 * Clear any false indication of pending signals that might
1495 * be seen by the filesystem code called to write the core file.
1496 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 clear_thread_flag(TIF_SIGPENDING);
1498
1499 if (current->signal->rlim[RLIMIT_CORE].rlim_cur < binfmt->min_coredump)
1500 goto fail_unlock;
1501
1502 /*
1503 * lock_kernel() because format_corename() is controlled by sysctl, which
1504 * uses lock_kernel()
1505 */
1506 lock_kernel();
1507 format_corename(corename, core_pattern, signr);
1508 unlock_kernel();
Andi Kleend025c9d2006-09-30 23:29:28 -07001509 if (corename[0] == '|') {
1510 /* SIGPIPE can happen, but it's just never processed */
1511 if(call_usermodehelper_pipe(corename+1, NULL, NULL, &file)) {
1512 printk(KERN_INFO "Core dump to %s pipe failed\n",
1513 corename);
1514 goto fail_unlock;
1515 }
1516 ispipe = 1;
1517 } else
1518 file = filp_open(corename,
Alexey Dobriyan6d4df672006-12-06 20:40:39 -08001519 O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag,
1520 0600);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 if (IS_ERR(file))
1522 goto fail_unlock;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001523 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 if (inode->i_nlink > 1)
1525 goto close_fail; /* multiple links - don't dump */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001526 if (!ispipe && d_unhashed(file->f_path.dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 goto close_fail;
1528
Andi Kleend025c9d2006-09-30 23:29:28 -07001529 /* AK: actually i see no reason to not allow this for named pipes etc.,
1530 but keep the previous behaviour for now. */
1531 if (!ispipe && !S_ISREG(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 goto close_fail;
1533 if (!file->f_op)
1534 goto close_fail;
1535 if (!file->f_op->write)
1536 goto close_fail;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001537 if (!ispipe && do_truncate(file->f_path.dentry, 0, 0, file) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 goto close_fail;
1539
1540 retval = binfmt->core_dump(signr, regs, file);
1541
1542 if (retval)
1543 current->signal->group_exit_code |= 0x80;
1544close_fail:
1545 filp_close(file, NULL);
1546fail_unlock:
Alan Coxd6e71142005-06-23 00:09:43 -07001547 current->fsuid = fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 complete_all(&mm->core_done);
1549fail:
1550 return retval;
1551}