Contents of /alx-src/tags/kernel26-2.6.12-alx-r9/fs/exec.c
Parent Directory | Revision Log
Revision 630 -
(show annotations)
(download)
Wed Mar 4 11:03:09 2009 UTC (15 years, 6 months ago) by niro
File MIME type: text/plain
File size: 34528 byte(s)
Wed Mar 4 11:03:09 2009 UTC (15 years, 6 months ago) by niro
File MIME type: text/plain
File size: 34528 byte(s)
Tag kernel26-2.6.12-alx-r9
1 | /* |
2 | * linux/fs/exec.c |
3 | * |
4 | * Copyright (C) 1991, 1992 Linus Torvalds |
5 | */ |
6 | |
7 | /* |
8 | * #!-checking implemented by tytso. |
9 | */ |
10 | /* |
11 | * Demand-loading implemented 01.12.91 - no need to read anything but |
12 | * the header into memory. The inode of the executable is put into |
13 | * "current->executable", and page faults do the actual loading. Clean. |
14 | * |
15 | * Once more I can proudly say that linux stood up to being changed: it |
16 | * was less than 2 hours work to get demand-loading completely implemented. |
17 | * |
18 | * Demand loading changed July 1993 by Eric Youngdale. Use mmap instead, |
19 | * current->executable is only used by the procfs. This allows a dispatch |
20 | * table to check for several different types of binary formats. We keep |
21 | * trying until we recognize the file or we run out of supported binary |
22 | * formats. |
23 | */ |
24 | |
25 | #include <linux/config.h> |
26 | #include <linux/slab.h> |
27 | #include <linux/file.h> |
28 | #include <linux/mman.h> |
29 | #include <linux/a.out.h> |
30 | #include <linux/stat.h> |
31 | #include <linux/fcntl.h> |
32 | #include <linux/smp_lock.h> |
33 | #include <linux/init.h> |
34 | #include <linux/pagemap.h> |
35 | #include <linux/highmem.h> |
36 | #include <linux/spinlock.h> |
37 | #include <linux/key.h> |
38 | #include <linux/personality.h> |
39 | #include <linux/binfmts.h> |
40 | #include <linux/swap.h> |
41 | #include <linux/utsname.h> |
42 | #include <linux/module.h> |
43 | #include <linux/namei.h> |
44 | #include <linux/proc_fs.h> |
45 | #include <linux/ptrace.h> |
46 | #include <linux/mount.h> |
47 | #include <linux/security.h> |
48 | #include <linux/syscalls.h> |
49 | #include <linux/rmap.h> |
50 | #include <linux/acct.h> |
51 | |
52 | #include <asm/uaccess.h> |
53 | #include <asm/mmu_context.h> |
54 | |
55 | #ifdef CONFIG_KMOD |
56 | #include <linux/kmod.h> |
57 | #endif |
58 | |
59 | int core_uses_pid; |
60 | char core_pattern[65] = "core"; |
61 | /* The maximal length of core_pattern is also specified in sysctl.c */ |
62 | |
63 | static struct linux_binfmt *formats; |
64 | static DEFINE_RWLOCK(binfmt_lock); |
65 | |
66 | int register_binfmt(struct linux_binfmt * fmt) |
67 | { |
68 | struct linux_binfmt ** tmp = &formats; |
69 | |
70 | if (!fmt) |
71 | return -EINVAL; |
72 | if (fmt->next) |
73 | return -EBUSY; |
74 | write_lock(&binfmt_lock); |
75 | while (*tmp) { |
76 | if (fmt == *tmp) { |
77 | write_unlock(&binfmt_lock); |
78 | return -EBUSY; |
79 | } |
80 | tmp = &(*tmp)->next; |
81 | } |
82 | fmt->next = formats; |
83 | formats = fmt; |
84 | write_unlock(&binfmt_lock); |
85 | return 0; |
86 | } |
87 | |
88 | EXPORT_SYMBOL(register_binfmt); |
89 | |
90 | int unregister_binfmt(struct linux_binfmt * fmt) |
91 | { |
92 | struct linux_binfmt ** tmp = &formats; |
93 | |
94 | write_lock(&binfmt_lock); |
95 | while (*tmp) { |
96 | if (fmt == *tmp) { |
97 | *tmp = fmt->next; |
98 | write_unlock(&binfmt_lock); |
99 | return 0; |
100 | } |
101 | tmp = &(*tmp)->next; |
102 | } |
103 | write_unlock(&binfmt_lock); |
104 | return -EINVAL; |
105 | } |
106 | |
107 | EXPORT_SYMBOL(unregister_binfmt); |
108 | |
109 | static inline void put_binfmt(struct linux_binfmt * fmt) |
110 | { |
111 | module_put(fmt->module); |
112 | } |
113 | |
114 | /* |
115 | * Note that a shared library must be both readable and executable due to |
116 | * security reasons. |
117 | * |
118 | * Also note that we take the address to load from from the file itself. |
119 | */ |
120 | asmlinkage long sys_uselib(const char __user * library) |
121 | { |
122 | struct file * file; |
123 | struct nameidata nd; |
124 | int error; |
125 | |
126 | nd.intent.open.flags = FMODE_READ; |
127 | error = __user_walk(library, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd); |
128 | if (error) |
129 | goto out; |
130 | |
131 | error = -EINVAL; |
132 | if (!S_ISREG(nd.dentry->d_inode->i_mode)) |
133 | goto exit; |
134 | |
135 | error = permission(nd.dentry->d_inode, MAY_READ | MAY_EXEC, &nd); |
136 | if (error) |
137 | goto exit; |
138 | |
139 | file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); |
140 | error = PTR_ERR(file); |
141 | if (IS_ERR(file)) |
142 | goto out; |
143 | |
144 | error = -ENOEXEC; |
145 | if(file->f_op) { |
146 | struct linux_binfmt * fmt; |
147 | |
148 | read_lock(&binfmt_lock); |
149 | for (fmt = formats ; fmt ; fmt = fmt->next) { |
150 | if (!fmt->load_shlib) |
151 | continue; |
152 | if (!try_module_get(fmt->module)) |
153 | continue; |
154 | read_unlock(&binfmt_lock); |
155 | error = fmt->load_shlib(file); |
156 | read_lock(&binfmt_lock); |
157 | put_binfmt(fmt); |
158 | if (error != -ENOEXEC) |
159 | break; |
160 | } |
161 | read_unlock(&binfmt_lock); |
162 | } |
163 | fput(file); |
164 | out: |
165 | return error; |
166 | exit: |
167 | path_release(&nd); |
168 | goto out; |
169 | } |
170 | |
171 | /* |
172 | * count() counts the number of strings in array ARGV. |
173 | */ |
174 | static int count(char __user * __user * argv, int max) |
175 | { |
176 | int i = 0; |
177 | |
178 | if (argv != NULL) { |
179 | for (;;) { |
180 | char __user * p; |
181 | |
182 | if (get_user(p, argv)) |
183 | return -EFAULT; |
184 | if (!p) |
185 | break; |
186 | argv++; |
187 | if(++i > max) |
188 | return -E2BIG; |
189 | cond_resched(); |
190 | } |
191 | } |
192 | return i; |
193 | } |
194 | |
195 | /* |
196 | * 'copy_strings()' copies argument/environment strings from user |
197 | * memory to free pages in kernel mem. These are in a format ready |
198 | * to be put directly into the top of new user memory. |
199 | */ |
200 | static int copy_strings(int argc, char __user * __user * argv, |
201 | struct linux_binprm *bprm) |
202 | { |
203 | struct page *kmapped_page = NULL; |
204 | char *kaddr = NULL; |
205 | int ret; |
206 | |
207 | while (argc-- > 0) { |
208 | char __user *str; |
209 | int len; |
210 | unsigned long pos; |
211 | |
212 | if (get_user(str, argv+argc) || |
213 | !(len = strnlen_user(str, bprm->p))) { |
214 | ret = -EFAULT; |
215 | goto out; |
216 | } |
217 | |
218 | if (bprm->p < len) { |
219 | ret = -E2BIG; |
220 | goto out; |
221 | } |
222 | |
223 | bprm->p -= len; |
224 | /* XXX: add architecture specific overflow check here. */ |
225 | pos = bprm->p; |
226 | |
227 | while (len > 0) { |
228 | int i, new, err; |
229 | int offset, bytes_to_copy; |
230 | struct page *page; |
231 | |
232 | offset = pos % PAGE_SIZE; |
233 | i = pos/PAGE_SIZE; |
234 | page = bprm->page[i]; |
235 | new = 0; |
236 | if (!page) { |
237 | page = alloc_page(GFP_HIGHUSER); |
238 | bprm->page[i] = page; |
239 | if (!page) { |
240 | ret = -ENOMEM; |
241 | goto out; |
242 | } |
243 | new = 1; |
244 | } |
245 | |
246 | if (page != kmapped_page) { |
247 | if (kmapped_page) |
248 | kunmap(kmapped_page); |
249 | kmapped_page = page; |
250 | kaddr = kmap(kmapped_page); |
251 | } |
252 | if (new && offset) |
253 | memset(kaddr, 0, offset); |
254 | bytes_to_copy = PAGE_SIZE - offset; |
255 | if (bytes_to_copy > len) { |
256 | bytes_to_copy = len; |
257 | if (new) |
258 | memset(kaddr+offset+len, 0, |
259 | PAGE_SIZE-offset-len); |
260 | } |
261 | err = copy_from_user(kaddr+offset, str, bytes_to_copy); |
262 | if (err) { |
263 | ret = -EFAULT; |
264 | goto out; |
265 | } |
266 | |
267 | pos += bytes_to_copy; |
268 | str += bytes_to_copy; |
269 | len -= bytes_to_copy; |
270 | } |
271 | } |
272 | ret = 0; |
273 | out: |
274 | if (kmapped_page) |
275 | kunmap(kmapped_page); |
276 | return ret; |
277 | } |
278 | |
279 | /* |
280 | * Like copy_strings, but get argv and its values from kernel memory. |
281 | */ |
282 | int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm) |
283 | { |
284 | int r; |
285 | mm_segment_t oldfs = get_fs(); |
286 | set_fs(KERNEL_DS); |
287 | r = copy_strings(argc, (char __user * __user *)argv, bprm); |
288 | set_fs(oldfs); |
289 | return r; |
290 | } |
291 | |
292 | EXPORT_SYMBOL(copy_strings_kernel); |
293 | |
294 | #ifdef CONFIG_MMU |
295 | /* |
296 | * This routine is used to map in a page into an address space: needed by |
297 | * execve() for the initial stack and environment pages. |
298 | * |
299 | * vma->vm_mm->mmap_sem is held for writing. |
300 | */ |
301 | void install_arg_page(struct vm_area_struct *vma, |
302 | struct page *page, unsigned long address) |
303 | { |
304 | struct mm_struct *mm = vma->vm_mm; |
305 | pgd_t * pgd; |
306 | pud_t * pud; |
307 | pmd_t * pmd; |
308 | pte_t * pte; |
309 | |
310 | if (unlikely(anon_vma_prepare(vma))) |
311 | goto out_sig; |
312 | |
313 | flush_dcache_page(page); |
314 | pgd = pgd_offset(mm, address); |
315 | |
316 | spin_lock(&mm->page_table_lock); |
317 | pud = pud_alloc(mm, pgd, address); |
318 | if (!pud) |
319 | goto out; |
320 | pmd = pmd_alloc(mm, pud, address); |
321 | if (!pmd) |
322 | goto out; |
323 | pte = pte_alloc_map(mm, pmd, address); |
324 | if (!pte) |
325 | goto out; |
326 | if (!pte_none(*pte)) { |
327 | pte_unmap(pte); |
328 | goto out; |
329 | } |
330 | inc_mm_counter(mm, rss); |
331 | lru_cache_add_active(page); |
332 | set_pte_at(mm, address, pte, pte_mkdirty(pte_mkwrite(mk_pte( |
333 | page, vma->vm_page_prot)))); |
334 | page_add_anon_rmap(page, vma, address); |
335 | pte_unmap(pte); |
336 | spin_unlock(&mm->page_table_lock); |
337 | |
338 | /* no need for flush_tlb */ |
339 | return; |
340 | out: |
341 | spin_unlock(&mm->page_table_lock); |
342 | out_sig: |
343 | __free_page(page); |
344 | force_sig(SIGKILL, current); |
345 | } |
346 | |
347 | #define EXTRA_STACK_VM_PAGES 20 /* random */ |
348 | |
349 | int setup_arg_pages(struct linux_binprm *bprm, |
350 | unsigned long stack_top, |
351 | int executable_stack) |
352 | { |
353 | unsigned long stack_base; |
354 | struct vm_area_struct *mpnt; |
355 | struct mm_struct *mm = current->mm; |
356 | int i, ret; |
357 | long arg_size; |
358 | |
359 | #ifdef CONFIG_STACK_GROWSUP |
360 | /* Move the argument and environment strings to the bottom of the |
361 | * stack space. |
362 | */ |
363 | int offset, j; |
364 | char *to, *from; |
365 | |
366 | /* Start by shifting all the pages down */ |
367 | i = 0; |
368 | for (j = 0; j < MAX_ARG_PAGES; j++) { |
369 | struct page *page = bprm->page[j]; |
370 | if (!page) |
371 | continue; |
372 | bprm->page[i++] = page; |
373 | } |
374 | |
375 | /* Now move them within their pages */ |
376 | offset = bprm->p % PAGE_SIZE; |
377 | to = kmap(bprm->page[0]); |
378 | for (j = 1; j < i; j++) { |
379 | memmove(to, to + offset, PAGE_SIZE - offset); |
380 | from = kmap(bprm->page[j]); |
381 | memcpy(to + PAGE_SIZE - offset, from, offset); |
382 | kunmap(bprm->page[j - 1]); |
383 | to = from; |
384 | } |
385 | memmove(to, to + offset, PAGE_SIZE - offset); |
386 | kunmap(bprm->page[j - 1]); |
387 | |
388 | /* Limit stack size to 1GB */ |
389 | stack_base = current->signal->rlim[RLIMIT_STACK].rlim_max; |
390 | if (stack_base > (1 << 30)) |
391 | stack_base = 1 << 30; |
392 | stack_base = PAGE_ALIGN(stack_top - stack_base); |
393 | |
394 | /* Adjust bprm->p to point to the end of the strings. */ |
395 | bprm->p = stack_base + PAGE_SIZE * i - offset; |
396 | |
397 | mm->arg_start = stack_base; |
398 | arg_size = i << PAGE_SHIFT; |
399 | |
400 | /* zero pages that were copied above */ |
401 | while (i < MAX_ARG_PAGES) |
402 | bprm->page[i++] = NULL; |
403 | #else |
404 | stack_base = arch_align_stack(stack_top - MAX_ARG_PAGES*PAGE_SIZE); |
405 | stack_base = PAGE_ALIGN(stack_base); |
406 | bprm->p += stack_base; |
407 | mm->arg_start = bprm->p; |
408 | arg_size = stack_top - (PAGE_MASK & (unsigned long) mm->arg_start); |
409 | #endif |
410 | |
411 | arg_size += EXTRA_STACK_VM_PAGES * PAGE_SIZE; |
412 | |
413 | if (bprm->loader) |
414 | bprm->loader += stack_base; |
415 | bprm->exec += stack_base; |
416 | |
417 | mpnt = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL); |
418 | if (!mpnt) |
419 | return -ENOMEM; |
420 | |
421 | if (security_vm_enough_memory(arg_size >> PAGE_SHIFT)) { |
422 | kmem_cache_free(vm_area_cachep, mpnt); |
423 | return -ENOMEM; |
424 | } |
425 | |
426 | memset(mpnt, 0, sizeof(*mpnt)); |
427 | |
428 | down_write(&mm->mmap_sem); |
429 | { |
430 | mpnt->vm_mm = mm; |
431 | #ifdef CONFIG_STACK_GROWSUP |
432 | mpnt->vm_start = stack_base; |
433 | mpnt->vm_end = stack_base + arg_size; |
434 | #else |
435 | mpnt->vm_end = stack_top; |
436 | mpnt->vm_start = mpnt->vm_end - arg_size; |
437 | #endif |
438 | /* Adjust stack execute permissions; explicitly enable |
439 | * for EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X |
440 | * and leave alone (arch default) otherwise. */ |
441 | if (unlikely(executable_stack == EXSTACK_ENABLE_X)) |
442 | mpnt->vm_flags = VM_STACK_FLAGS | VM_EXEC; |
443 | else if (executable_stack == EXSTACK_DISABLE_X) |
444 | mpnt->vm_flags = VM_STACK_FLAGS & ~VM_EXEC; |
445 | else |
446 | mpnt->vm_flags = VM_STACK_FLAGS; |
447 | mpnt->vm_flags |= mm->def_flags; |
448 | mpnt->vm_page_prot = protection_map[mpnt->vm_flags & 0x7]; |
449 | if ((ret = insert_vm_struct(mm, mpnt))) { |
450 | up_write(&mm->mmap_sem); |
451 | kmem_cache_free(vm_area_cachep, mpnt); |
452 | return ret; |
453 | } |
454 | mm->stack_vm = mm->total_vm = vma_pages(mpnt); |
455 | } |
456 | |
457 | for (i = 0 ; i < MAX_ARG_PAGES ; i++) { |
458 | struct page *page = bprm->page[i]; |
459 | if (page) { |
460 | bprm->page[i] = NULL; |
461 | install_arg_page(mpnt, page, stack_base); |
462 | } |
463 | stack_base += PAGE_SIZE; |
464 | } |
465 | up_write(&mm->mmap_sem); |
466 | |
467 | return 0; |
468 | } |
469 | |
470 | EXPORT_SYMBOL(setup_arg_pages); |
471 | |
472 | #define free_arg_pages(bprm) do { } while (0) |
473 | |
474 | #else |
475 | |
476 | static inline void free_arg_pages(struct linux_binprm *bprm) |
477 | { |
478 | int i; |
479 | |
480 | for (i = 0; i < MAX_ARG_PAGES; i++) { |
481 | if (bprm->page[i]) |
482 | __free_page(bprm->page[i]); |
483 | bprm->page[i] = NULL; |
484 | } |
485 | } |
486 | |
487 | #endif /* CONFIG_MMU */ |
488 | |
489 | struct file *open_exec(const char *name) |
490 | { |
491 | struct nameidata nd; |
492 | int err; |
493 | struct file *file; |
494 | |
495 | nd.intent.open.flags = FMODE_READ; |
496 | err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd); |
497 | file = ERR_PTR(err); |
498 | |
499 | if (!err) { |
500 | struct inode *inode = nd.dentry->d_inode; |
501 | file = ERR_PTR(-EACCES); |
502 | if (!(nd.mnt->mnt_flags & MNT_NOEXEC) && |
503 | S_ISREG(inode->i_mode)) { |
504 | int err = permission(inode, MAY_EXEC, &nd); |
505 | if (!err && !(inode->i_mode & 0111)) |
506 | err = -EACCES; |
507 | file = ERR_PTR(err); |
508 | if (!err) { |
509 | file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); |
510 | if (!IS_ERR(file)) { |
511 | err = deny_write_access(file); |
512 | if (err) { |
513 | fput(file); |
514 | file = ERR_PTR(err); |
515 | } |
516 | } |
517 | out: |
518 | return file; |
519 | } |
520 | } |
521 | path_release(&nd); |
522 | } |
523 | goto out; |
524 | } |
525 | |
526 | EXPORT_SYMBOL(open_exec); |
527 | |
528 | int kernel_read(struct file *file, unsigned long offset, |
529 | char *addr, unsigned long count) |
530 | { |
531 | mm_segment_t old_fs; |
532 | loff_t pos = offset; |
533 | int result; |
534 | |
535 | old_fs = get_fs(); |
536 | set_fs(get_ds()); |
537 | /* The cast to a user pointer is valid due to the set_fs() */ |
538 | result = vfs_read(file, (void __user *)addr, count, &pos); |
539 | set_fs(old_fs); |
540 | return result; |
541 | } |
542 | |
543 | EXPORT_SYMBOL(kernel_read); |
544 | |
545 | static int exec_mmap(struct mm_struct *mm) |
546 | { |
547 | struct task_struct *tsk; |
548 | struct mm_struct * old_mm, *active_mm; |
549 | |
550 | /* Notify parent that we're no longer interested in the old VM */ |
551 | tsk = current; |
552 | old_mm = current->mm; |
553 | mm_release(tsk, old_mm); |
554 | |
555 | if (old_mm) { |
556 | /* |
557 | * Make sure that if there is a core dump in progress |
558 | * for the old mm, we get out and die instead of going |
559 | * through with the exec. We must hold mmap_sem around |
560 | * checking core_waiters and changing tsk->mm. The |
561 | * core-inducing thread will increment core_waiters for |
562 | * each thread whose ->mm == old_mm. |
563 | */ |
564 | down_read(&old_mm->mmap_sem); |
565 | if (unlikely(old_mm->core_waiters)) { |
566 | up_read(&old_mm->mmap_sem); |
567 | return -EINTR; |
568 | } |
569 | } |
570 | task_lock(tsk); |
571 | active_mm = tsk->active_mm; |
572 | tsk->mm = mm; |
573 | tsk->active_mm = mm; |
574 | activate_mm(active_mm, mm); |
575 | task_unlock(tsk); |
576 | arch_pick_mmap_layout(mm); |
577 | if (old_mm) { |
578 | up_read(&old_mm->mmap_sem); |
579 | if (active_mm != old_mm) BUG(); |
580 | mmput(old_mm); |
581 | return 0; |
582 | } |
583 | mmdrop(active_mm); |
584 | return 0; |
585 | } |
586 | |
587 | /* |
588 | * This function makes sure the current process has its own signal table, |
589 | * so that flush_signal_handlers can later reset the handlers without |
590 | * disturbing other processes. (Other processes might share the signal |
591 | * table via the CLONE_SIGHAND option to clone().) |
592 | */ |
593 | static inline int de_thread(struct task_struct *tsk) |
594 | { |
595 | struct signal_struct *sig = tsk->signal; |
596 | struct sighand_struct *newsighand, *oldsighand = tsk->sighand; |
597 | spinlock_t *lock = &oldsighand->siglock; |
598 | int count; |
599 | |
600 | /* |
601 | * If we don't share sighandlers, then we aren't sharing anything |
602 | * and we can just re-use it all. |
603 | */ |
604 | if (atomic_read(&oldsighand->count) <= 1) { |
605 | BUG_ON(atomic_read(&sig->count) != 1); |
606 | exit_itimers(sig); |
607 | return 0; |
608 | } |
609 | |
610 | newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); |
611 | if (!newsighand) |
612 | return -ENOMEM; |
613 | |
614 | if (thread_group_empty(current)) |
615 | goto no_thread_group; |
616 | |
617 | /* |
618 | * Kill all other threads in the thread group. |
619 | * We must hold tasklist_lock to call zap_other_threads. |
620 | */ |
621 | read_lock(&tasklist_lock); |
622 | spin_lock_irq(lock); |
623 | if (sig->flags & SIGNAL_GROUP_EXIT) { |
624 | /* |
625 | * Another group action in progress, just |
626 | * return so that the signal is processed. |
627 | */ |
628 | spin_unlock_irq(lock); |
629 | read_unlock(&tasklist_lock); |
630 | kmem_cache_free(sighand_cachep, newsighand); |
631 | return -EAGAIN; |
632 | } |
633 | zap_other_threads(current); |
634 | read_unlock(&tasklist_lock); |
635 | |
636 | /* |
637 | * Account for the thread group leader hanging around: |
638 | */ |
639 | count = 2; |
640 | if (thread_group_leader(current)) |
641 | count = 1; |
642 | while (atomic_read(&sig->count) > count) { |
643 | sig->group_exit_task = current; |
644 | sig->notify_count = count; |
645 | __set_current_state(TASK_UNINTERRUPTIBLE); |
646 | spin_unlock_irq(lock); |
647 | schedule(); |
648 | spin_lock_irq(lock); |
649 | } |
650 | sig->group_exit_task = NULL; |
651 | sig->notify_count = 0; |
652 | sig->real_timer.data = (unsigned long)current; |
653 | spin_unlock_irq(lock); |
654 | |
655 | /* |
656 | * At this point all other threads have exited, all we have to |
657 | * do is to wait for the thread group leader to become inactive, |
658 | * and to assume its PID: |
659 | */ |
660 | if (!thread_group_leader(current)) { |
661 | struct task_struct *leader = current->group_leader, *parent; |
662 | struct dentry *proc_dentry1, *proc_dentry2; |
663 | unsigned long exit_state, ptrace; |
664 | |
665 | /* |
666 | * Wait for the thread group leader to be a zombie. |
667 | * It should already be zombie at this point, most |
668 | * of the time. |
669 | */ |
670 | while (leader->exit_state != EXIT_ZOMBIE) |
671 | yield(); |
672 | |
673 | spin_lock(&leader->proc_lock); |
674 | spin_lock(¤t->proc_lock); |
675 | proc_dentry1 = proc_pid_unhash(current); |
676 | proc_dentry2 = proc_pid_unhash(leader); |
677 | write_lock_irq(&tasklist_lock); |
678 | |
679 | if (leader->tgid != current->tgid) |
680 | BUG(); |
681 | if (current->pid == current->tgid) |
682 | BUG(); |
683 | /* |
684 | * An exec() starts a new thread group with the |
685 | * TGID of the previous thread group. Rehash the |
686 | * two threads with a switched PID, and release |
687 | * the former thread group leader: |
688 | */ |
689 | ptrace = leader->ptrace; |
690 | parent = leader->parent; |
691 | if (unlikely(ptrace) && unlikely(parent == current)) { |
692 | /* |
693 | * Joker was ptracing his own group leader, |
694 | * and now he wants to be his own parent! |
695 | * We can't have that. |
696 | */ |
697 | ptrace = 0; |
698 | } |
699 | |
700 | ptrace_unlink(current); |
701 | ptrace_unlink(leader); |
702 | remove_parent(current); |
703 | remove_parent(leader); |
704 | |
705 | switch_exec_pids(leader, current); |
706 | |
707 | current->parent = current->real_parent = leader->real_parent; |
708 | leader->parent = leader->real_parent = child_reaper; |
709 | current->group_leader = current; |
710 | leader->group_leader = leader; |
711 | |
712 | add_parent(current, current->parent); |
713 | add_parent(leader, leader->parent); |
714 | if (ptrace) { |
715 | current->ptrace = ptrace; |
716 | __ptrace_link(current, parent); |
717 | } |
718 | |
719 | list_del(¤t->tasks); |
720 | list_add_tail(¤t->tasks, &init_task.tasks); |
721 | current->exit_signal = SIGCHLD; |
722 | exit_state = leader->exit_state; |
723 | |
724 | write_unlock_irq(&tasklist_lock); |
725 | spin_unlock(&leader->proc_lock); |
726 | spin_unlock(¤t->proc_lock); |
727 | proc_pid_flush(proc_dentry1); |
728 | proc_pid_flush(proc_dentry2); |
729 | |
730 | if (exit_state != EXIT_ZOMBIE) |
731 | BUG(); |
732 | release_task(leader); |
733 | } |
734 | |
735 | /* |
736 | * Now there are really no other threads at all, |
737 | * so it's safe to stop telling them to kill themselves. |
738 | */ |
739 | sig->flags = 0; |
740 | |
741 | no_thread_group: |
742 | BUG_ON(atomic_read(&sig->count) != 1); |
743 | exit_itimers(sig); |
744 | |
745 | if (atomic_read(&oldsighand->count) == 1) { |
746 | /* |
747 | * Now that we nuked the rest of the thread group, |
748 | * it turns out we are not sharing sighand any more either. |
749 | * So we can just keep it. |
750 | */ |
751 | kmem_cache_free(sighand_cachep, newsighand); |
752 | } else { |
753 | /* |
754 | * Move our state over to newsighand and switch it in. |
755 | */ |
756 | spin_lock_init(&newsighand->siglock); |
757 | atomic_set(&newsighand->count, 1); |
758 | memcpy(newsighand->action, oldsighand->action, |
759 | sizeof(newsighand->action)); |
760 | |
761 | write_lock_irq(&tasklist_lock); |
762 | spin_lock(&oldsighand->siglock); |
763 | spin_lock(&newsighand->siglock); |
764 | |
765 | current->sighand = newsighand; |
766 | recalc_sigpending(); |
767 | |
768 | spin_unlock(&newsighand->siglock); |
769 | spin_unlock(&oldsighand->siglock); |
770 | write_unlock_irq(&tasklist_lock); |
771 | |
772 | if (atomic_dec_and_test(&oldsighand->count)) |
773 | kmem_cache_free(sighand_cachep, oldsighand); |
774 | } |
775 | |
776 | if (!thread_group_empty(current)) |
777 | BUG(); |
778 | if (!thread_group_leader(current)) |
779 | BUG(); |
780 | return 0; |
781 | } |
782 | |
783 | /* |
784 | * These functions flushes out all traces of the currently running executable |
785 | * so that a new one can be started |
786 | */ |
787 | |
788 | static inline void flush_old_files(struct files_struct * files) |
789 | { |
790 | long j = -1; |
791 | |
792 | spin_lock(&files->file_lock); |
793 | for (;;) { |
794 | unsigned long set, i; |
795 | |
796 | j++; |
797 | i = j * __NFDBITS; |
798 | if (i >= files->max_fds || i >= files->max_fdset) |
799 | break; |
800 | set = files->close_on_exec->fds_bits[j]; |
801 | if (!set) |
802 | continue; |
803 | files->close_on_exec->fds_bits[j] = 0; |
804 | spin_unlock(&files->file_lock); |
805 | for ( ; set ; i++,set >>= 1) { |
806 | if (set & 1) { |
807 | sys_close(i); |
808 | } |
809 | } |
810 | spin_lock(&files->file_lock); |
811 | |
812 | } |
813 | spin_unlock(&files->file_lock); |
814 | } |
815 | |
816 | void get_task_comm(char *buf, struct task_struct *tsk) |
817 | { |
818 | /* buf must be at least sizeof(tsk->comm) in size */ |
819 | task_lock(tsk); |
820 | strncpy(buf, tsk->comm, sizeof(tsk->comm)); |
821 | task_unlock(tsk); |
822 | } |
823 | |
824 | void set_task_comm(struct task_struct *tsk, char *buf) |
825 | { |
826 | task_lock(tsk); |
827 | strlcpy(tsk->comm, buf, sizeof(tsk->comm)); |
828 | task_unlock(tsk); |
829 | } |
830 | |
831 | int flush_old_exec(struct linux_binprm * bprm) |
832 | { |
833 | char * name; |
834 | int i, ch, retval; |
835 | struct files_struct *files; |
836 | char tcomm[sizeof(current->comm)]; |
837 | |
838 | /* |
839 | * Make sure we have a private signal table and that |
840 | * we are unassociated from the previous thread group. |
841 | */ |
842 | retval = de_thread(current); |
843 | if (retval) |
844 | goto out; |
845 | |
846 | /* |
847 | * Make sure we have private file handles. Ask the |
848 | * fork helper to do the work for us and the exit |
849 | * helper to do the cleanup of the old one. |
850 | */ |
851 | files = current->files; /* refcounted so safe to hold */ |
852 | retval = unshare_files(); |
853 | if (retval) |
854 | goto out; |
855 | /* |
856 | * Release all of the old mmap stuff |
857 | */ |
858 | retval = exec_mmap(bprm->mm); |
859 | if (retval) |
860 | goto mmap_failed; |
861 | |
862 | bprm->mm = NULL; /* We're using it now */ |
863 | |
864 | /* This is the point of no return */ |
865 | steal_locks(files); |
866 | put_files_struct(files); |
867 | |
868 | current->sas_ss_sp = current->sas_ss_size = 0; |
869 | |
870 | if (current->euid == current->uid && current->egid == current->gid) |
871 | current->mm->dumpable = 1; |
872 | name = bprm->filename; |
873 | |
874 | /* Copies the binary name from after last slash */ |
875 | for (i=0; (ch = *(name++)) != '\0';) { |
876 | if (ch == '/') |
877 | i = 0; /* overwrite what we wrote */ |
878 | else |
879 | if (i < (sizeof(tcomm) - 1)) |
880 | tcomm[i++] = ch; |
881 | } |
882 | tcomm[i] = '\0'; |
883 | set_task_comm(current, tcomm); |
884 | |
885 | current->flags &= ~PF_RANDOMIZE; |
886 | flush_thread(); |
887 | |
888 | if (bprm->e_uid != current->euid || bprm->e_gid != current->egid || |
889 | permission(bprm->file->f_dentry->d_inode,MAY_READ, NULL) || |
890 | (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)) { |
891 | suid_keys(current); |
892 | current->mm->dumpable = 0; |
893 | } |
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 | |
905 | mmap_failed: |
906 | put_files_struct(current->files); |
907 | current->files = files; |
908 | out: |
909 | return retval; |
910 | } |
911 | |
912 | EXPORT_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 | */ |
918 | int 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 | |
965 | EXPORT_SYMBOL(prepare_binprm); |
966 | |
967 | static inline int unsafe_exec(struct task_struct *p) |
968 | { |
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 | |
984 | void 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 | |
999 | EXPORT_SYMBOL(compute_creds); |
1000 | |
1001 | void 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); |
1016 | inside: |
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 | |
1025 | EXPORT_SYMBOL(remove_arg_zero); |
1026 | |
1027 | /* |
1028 | * cycle the list of binary formats handler, until one recognizes the image |
1029 | */ |
1030 | int 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); |
1076 | retval = -ENOENT; |
1077 | for (try=0; try<2; try++) { |
1078 | read_lock(&binfmt_lock); |
1079 | for (fmt = formats ; fmt ; fmt = fmt->next) { |
1080 | int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary; |
1081 | if (!fn) |
1082 | continue; |
1083 | if (!try_module_get(fmt->module)) |
1084 | continue; |
1085 | read_unlock(&binfmt_lock); |
1086 | retval = fn(bprm, regs); |
1087 | if (retval >= 0) { |
1088 | put_binfmt(fmt); |
1089 | allow_write_access(bprm->file); |
1090 | if (bprm->file) |
1091 | fput(bprm->file); |
1092 | bprm->file = NULL; |
1093 | current->did_exec = 1; |
1094 | return retval; |
1095 | } |
1096 | read_lock(&binfmt_lock); |
1097 | put_binfmt(fmt); |
1098 | if (retval != -ENOEXEC || bprm->mm == NULL) |
1099 | break; |
1100 | if (!bprm->file) { |
1101 | read_unlock(&binfmt_lock); |
1102 | return retval; |
1103 | } |
1104 | } |
1105 | read_unlock(&binfmt_lock); |
1106 | if (retval != -ENOEXEC || bprm->mm == NULL) { |
1107 | break; |
1108 | #ifdef CONFIG_KMOD |
1109 | }else{ |
1110 | #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e)) |
1111 | if (printable(bprm->buf[0]) && |
1112 | printable(bprm->buf[1]) && |
1113 | printable(bprm->buf[2]) && |
1114 | printable(bprm->buf[3])) |
1115 | break; /* -ENOEXEC */ |
1116 | request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2])); |
1117 | #endif |
1118 | } |
1119 | } |
1120 | return retval; |
1121 | } |
1122 | |
1123 | EXPORT_SYMBOL(search_binary_handler); |
1124 | |
1125 | /* |
1126 | * sys_execve() executes a new program. |
1127 | */ |
1128 | int do_execve(char * filename, |
1129 | char __user *__user *argv, |
1130 | char __user *__user *envp, |
1131 | struct pt_regs * regs) |
1132 | { |
1133 | struct linux_binprm *bprm; |
1134 | struct file *file; |
1135 | int retval; |
1136 | int i; |
1137 | |
1138 | retval = -ENOMEM; |
1139 | bprm = kmalloc(sizeof(*bprm), GFP_KERNEL); |
1140 | if (!bprm) |
1141 | goto out_ret; |
1142 | memset(bprm, 0, sizeof(*bprm)); |
1143 | |
1144 | file = open_exec(filename); |
1145 | retval = PTR_ERR(file); |
1146 | if (IS_ERR(file)) |
1147 | goto out_kfree; |
1148 | |
1149 | sched_exec(); |
1150 | |
1151 | bprm->p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *); |
1152 | |
1153 | bprm->file = file; |
1154 | bprm->filename = filename; |
1155 | bprm->interp = filename; |
1156 | bprm->mm = mm_alloc(); |
1157 | retval = -ENOMEM; |
1158 | if (!bprm->mm) |
1159 | goto out_file; |
1160 | |
1161 | retval = init_new_context(current, bprm->mm); |
1162 | if (retval < 0) |
1163 | goto out_mm; |
1164 | |
1165 | bprm->argc = count(argv, bprm->p / sizeof(void *)); |
1166 | if ((retval = bprm->argc) < 0) |
1167 | goto out_mm; |
1168 | |
1169 | bprm->envc = count(envp, bprm->p / sizeof(void *)); |
1170 | if ((retval = bprm->envc) < 0) |
1171 | goto out_mm; |
1172 | |
1173 | retval = security_bprm_alloc(bprm); |
1174 | if (retval) |
1175 | goto out; |
1176 | |
1177 | retval = prepare_binprm(bprm); |
1178 | if (retval < 0) |
1179 | goto out; |
1180 | |
1181 | retval = copy_strings_kernel(1, &bprm->filename, bprm); |
1182 | if (retval < 0) |
1183 | goto out; |
1184 | |
1185 | bprm->exec = bprm->p; |
1186 | retval = copy_strings(bprm->envc, envp, bprm); |
1187 | if (retval < 0) |
1188 | goto out; |
1189 | |
1190 | retval = copy_strings(bprm->argc, argv, bprm); |
1191 | if (retval < 0) |
1192 | goto out; |
1193 | |
1194 | retval = search_binary_handler(bprm,regs); |
1195 | if (retval >= 0) { |
1196 | free_arg_pages(bprm); |
1197 | |
1198 | /* execve success */ |
1199 | security_bprm_free(bprm); |
1200 | acct_update_integrals(current); |
1201 | update_mem_hiwater(current); |
1202 | kfree(bprm); |
1203 | return retval; |
1204 | } |
1205 | |
1206 | out: |
1207 | /* Something went wrong, return the inode and free the argument pages*/ |
1208 | for (i = 0 ; i < MAX_ARG_PAGES ; i++) { |
1209 | struct page * page = bprm->page[i]; |
1210 | if (page) |
1211 | __free_page(page); |
1212 | } |
1213 | |
1214 | if (bprm->security) |
1215 | security_bprm_free(bprm); |
1216 | |
1217 | out_mm: |
1218 | if (bprm->mm) |
1219 | mmdrop(bprm->mm); |
1220 | |
1221 | out_file: |
1222 | if (bprm->file) { |
1223 | allow_write_access(bprm->file); |
1224 | fput(bprm->file); |
1225 | } |
1226 | |
1227 | out_kfree: |
1228 | kfree(bprm); |
1229 | |
1230 | out_ret: |
1231 | return retval; |
1232 | } |
1233 | |
1234 | int set_binfmt(struct linux_binfmt *new) |
1235 | { |
1236 | struct linux_binfmt *old = current->binfmt; |
1237 | |
1238 | if (new) { |
1239 | if (!try_module_get(new->module)) |
1240 | return -1; |
1241 | } |
1242 | current->binfmt = new; |
1243 | if (old) |
1244 | module_put(old->module); |
1245 | return 0; |
1246 | } |
1247 | |
1248 | EXPORT_SYMBOL(set_binfmt); |
1249 | |
1250 | #define CORENAME_MAX_SIZE 64 |
1251 | |
1252 | /* format_corename will inspect the pattern parameter, and output a |
1253 | * name into corename, which must have space for at least |
1254 | * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator. |
1255 | */ |
1256 | static void format_corename(char *corename, const char *pattern, long signr) |
1257 | { |
1258 | const char *pat_ptr = pattern; |
1259 | char *out_ptr = corename; |
1260 | char *const out_end = corename + CORENAME_MAX_SIZE; |
1261 | int rc; |
1262 | int pid_in_pattern = 0; |
1263 | |
1264 | /* Repeat as long as we have more pattern to process and more output |
1265 | space */ |
1266 | while (*pat_ptr) { |
1267 | if (*pat_ptr != '%') { |
1268 | if (out_ptr == out_end) |
1269 | goto out; |
1270 | *out_ptr++ = *pat_ptr++; |
1271 | } else { |
1272 | switch (*++pat_ptr) { |
1273 | case 0: |
1274 | goto out; |
1275 | /* Double percent, output one percent */ |
1276 | case '%': |
1277 | if (out_ptr == out_end) |
1278 | goto out; |
1279 | *out_ptr++ = '%'; |
1280 | break; |
1281 | /* pid */ |
1282 | case 'p': |
1283 | pid_in_pattern = 1; |
1284 | rc = snprintf(out_ptr, out_end - out_ptr, |
1285 | "%d", current->tgid); |
1286 | if (rc > out_end - out_ptr) |
1287 | goto out; |
1288 | out_ptr += rc; |
1289 | break; |
1290 | /* uid */ |
1291 | case 'u': |
1292 | rc = snprintf(out_ptr, out_end - out_ptr, |
1293 | "%d", current->uid); |
1294 | if (rc > out_end - out_ptr) |
1295 | goto out; |
1296 | out_ptr += rc; |
1297 | break; |
1298 | /* gid */ |
1299 | case 'g': |
1300 | rc = snprintf(out_ptr, out_end - out_ptr, |
1301 | "%d", current->gid); |
1302 | if (rc > out_end - out_ptr) |
1303 | goto out; |
1304 | out_ptr += rc; |
1305 | break; |
1306 | /* signal that caused the coredump */ |
1307 | case 's': |
1308 | rc = snprintf(out_ptr, out_end - out_ptr, |
1309 | "%ld", signr); |
1310 | if (rc > out_end - out_ptr) |
1311 | goto out; |
1312 | out_ptr += rc; |
1313 | break; |
1314 | /* UNIX time of coredump */ |
1315 | case 't': { |
1316 | struct timeval tv; |
1317 | do_gettimeofday(&tv); |
1318 | rc = snprintf(out_ptr, out_end - out_ptr, |
1319 | "%lu", tv.tv_sec); |
1320 | if (rc > out_end - out_ptr) |
1321 | goto out; |
1322 | out_ptr += rc; |
1323 | break; |
1324 | } |
1325 | /* hostname */ |
1326 | case 'h': |
1327 | down_read(&uts_sem); |
1328 | rc = snprintf(out_ptr, out_end - out_ptr, |
1329 | "%s", system_utsname.nodename); |
1330 | up_read(&uts_sem); |
1331 | if (rc > out_end - out_ptr) |
1332 | goto out; |
1333 | out_ptr += rc; |
1334 | break; |
1335 | /* executable */ |
1336 | case 'e': |
1337 | rc = snprintf(out_ptr, out_end - out_ptr, |
1338 | "%s", current->comm); |
1339 | if (rc > out_end - out_ptr) |
1340 | goto out; |
1341 | out_ptr += rc; |
1342 | break; |
1343 | default: |
1344 | break; |
1345 | } |
1346 | ++pat_ptr; |
1347 | } |
1348 | } |
1349 | /* Backward compatibility with core_uses_pid: |
1350 | * |
1351 | * If core_pattern does not include a %p (as is the default) |
1352 | * and core_uses_pid is set, then .%pid will be appended to |
1353 | * the filename */ |
1354 | if (!pid_in_pattern |
1355 | && (core_uses_pid || atomic_read(¤t->mm->mm_users) != 1)) { |
1356 | rc = snprintf(out_ptr, out_end - out_ptr, |
1357 | ".%d", current->tgid); |
1358 | if (rc > out_end - out_ptr) |
1359 | goto out; |
1360 | out_ptr += rc; |
1361 | } |
1362 | out: |
1363 | *out_ptr = 0; |
1364 | } |
1365 | |
1366 | static void zap_threads (struct mm_struct *mm) |
1367 | { |
1368 | struct task_struct *g, *p; |
1369 | struct task_struct *tsk = current; |
1370 | struct completion *vfork_done = tsk->vfork_done; |
1371 | int traced = 0; |
1372 | |
1373 | /* |
1374 | * Make sure nobody is waiting for us to release the VM, |
1375 | * otherwise we can deadlock when we wait on each other |
1376 | */ |
1377 | if (vfork_done) { |
1378 | tsk->vfork_done = NULL; |
1379 | complete(vfork_done); |
1380 | } |
1381 | |
1382 | read_lock(&tasklist_lock); |
1383 | do_each_thread(g,p) |
1384 | if (mm == p->mm && p != tsk) { |
1385 | force_sig_specific(SIGKILL, p); |
1386 | mm->core_waiters++; |
1387 | if (unlikely(p->ptrace) && |
1388 | unlikely(p->parent->mm == mm)) |
1389 | traced = 1; |
1390 | } |
1391 | while_each_thread(g,p); |
1392 | |
1393 | read_unlock(&tasklist_lock); |
1394 | |
1395 | if (unlikely(traced)) { |
1396 | /* |
1397 | * We are zapping a thread and the thread it ptraces. |
1398 | * If the tracee went into a ptrace stop for exit tracing, |
1399 | * we could deadlock since the tracer is waiting for this |
1400 | * coredump to finish. Detach them so they can both die. |
1401 | */ |
1402 | write_lock_irq(&tasklist_lock); |
1403 | do_each_thread(g,p) { |
1404 | if (mm == p->mm && p != tsk && |
1405 | p->ptrace && p->parent->mm == mm) { |
1406 | __ptrace_unlink(p); |
1407 | } |
1408 | } while_each_thread(g,p); |
1409 | write_unlock_irq(&tasklist_lock); |
1410 | } |
1411 | } |
1412 | |
1413 | static void coredump_wait(struct mm_struct *mm) |
1414 | { |
1415 | DECLARE_COMPLETION(startup_done); |
1416 | |
1417 | mm->core_waiters++; /* let other threads block */ |
1418 | mm->core_startup_done = &startup_done; |
1419 | |
1420 | /* give other threads a chance to run: */ |
1421 | yield(); |
1422 | |
1423 | zap_threads(mm); |
1424 | if (--mm->core_waiters) { |
1425 | up_write(&mm->mmap_sem); |
1426 | wait_for_completion(&startup_done); |
1427 | } else |
1428 | up_write(&mm->mmap_sem); |
1429 | BUG_ON(mm->core_waiters); |
1430 | } |
1431 | |
1432 | int do_coredump(long signr, int exit_code, struct pt_regs * regs) |
1433 | { |
1434 | char corename[CORENAME_MAX_SIZE + 1]; |
1435 | struct mm_struct *mm = current->mm; |
1436 | struct linux_binfmt * binfmt; |
1437 | struct inode * inode; |
1438 | struct file * file; |
1439 | int retval = 0; |
1440 | |
1441 | binfmt = current->binfmt; |
1442 | if (!binfmt || !binfmt->core_dump) |
1443 | goto fail; |
1444 | down_write(&mm->mmap_sem); |
1445 | if (!mm->dumpable) { |
1446 | up_write(&mm->mmap_sem); |
1447 | goto fail; |
1448 | } |
1449 | mm->dumpable = 0; |
1450 | init_completion(&mm->core_done); |
1451 | spin_lock_irq(¤t->sighand->siglock); |
1452 | current->signal->flags = SIGNAL_GROUP_EXIT; |
1453 | current->signal->group_exit_code = exit_code; |
1454 | spin_unlock_irq(¤t->sighand->siglock); |
1455 | coredump_wait(mm); |
1456 | |
1457 | /* |
1458 | * Clear any false indication of pending signals that might |
1459 | * be seen by the filesystem code called to write the core file. |
1460 | */ |
1461 | current->signal->group_stop_count = 0; |
1462 | clear_thread_flag(TIF_SIGPENDING); |
1463 | |
1464 | if (current->signal->rlim[RLIMIT_CORE].rlim_cur < binfmt->min_coredump) |
1465 | goto fail_unlock; |
1466 | |
1467 | /* |
1468 | * lock_kernel() because format_corename() is controlled by sysctl, which |
1469 | * uses lock_kernel() |
1470 | */ |
1471 | lock_kernel(); |
1472 | format_corename(corename, core_pattern, signr); |
1473 | unlock_kernel(); |
1474 | file = filp_open(corename, O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE, 0600); |
1475 | if (IS_ERR(file)) |
1476 | goto fail_unlock; |
1477 | inode = file->f_dentry->d_inode; |
1478 | if (inode->i_nlink > 1) |
1479 | goto close_fail; /* multiple links - don't dump */ |
1480 | if (d_unhashed(file->f_dentry)) |
1481 | goto close_fail; |
1482 | |
1483 | if (!S_ISREG(inode->i_mode)) |
1484 | goto close_fail; |
1485 | if (!file->f_op) |
1486 | goto close_fail; |
1487 | if (!file->f_op->write) |
1488 | goto close_fail; |
1489 | if (do_truncate(file->f_dentry, 0) != 0) |
1490 | goto close_fail; |
1491 | |
1492 | retval = binfmt->core_dump(signr, regs, file); |
1493 | |
1494 | if (retval) |
1495 | current->signal->group_exit_code |= 0x80; |
1496 | close_fail: |
1497 | filp_close(file, NULL); |
1498 | fail_unlock: |
1499 | complete_all(&mm->core_done); |
1500 | fail: |
1501 | return retval; |
1502 | } |