Kernel

Kernel Git Source Tree

Root/fs/exec.c

Source at commit 9845c1745d3d531a5b9544f5322c62bfb4d4e9bc created 1 year 2 months ago. By Xiangfu, rtc: jz4740 fix hwclock give time out
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/slab.h>
26	#include <linux/file.h>
27	#include <linux/fdtable.h>
28	#include <linux/mm.h>
29	#include <linux/stat.h>
30	#include <linux/fcntl.h>
31	#include <linux/swap.h>
32	#include <linux/string.h>
33	#include <linux/init.h>
34	#include <linux/pagemap.h>
35	#include <linux/perf_event.h>
36	#include <linux/highmem.h>
37	#include <linux/spinlock.h>
38	#include <linux/key.h>
39	#include <linux/personality.h>
40	#include <linux/binfmts.h>
41	#include <linux/utsname.h>
42	#include <linux/pid_namespace.h>
43	#include <linux/module.h>
44	#include <linux/namei.h>
45	#include <linux/mount.h>
46	#include <linux/security.h>
47	#include <linux/syscalls.h>
48	#include <linux/tsacct_kern.h>
49	#include <linux/cn_proc.h>
50	#include <linux/audit.h>
51	#include <linux/tracehook.h>
52	#include <linux/kmod.h>
53	#include <linux/fsnotify.h>
54	#include <linux/fs_struct.h>
55	#include <linux/pipe_fs_i.h>
56	#include <linux/oom.h>
57	#include <linux/compat.h>
58
59	#include <asm/uaccess.h>
60	#include <asm/mmu_context.h>
61	#include <asm/tlb.h>
62	#include "internal.h"
63
64	int core_uses_pid;
65	char core_pattern[CORENAME_MAX_SIZE] = "core";
66	unsigned int core_pipe_limit;
67	int suid_dumpable = 0;
68
69	struct core_name {
70	char *corename;
71	int used, size;
72	};
73	static atomic_t call_count = ATOMIC_INIT(1);
74
75	/* The maximal length of core_pattern is also specified in sysctl.c */
76
77	static LIST_HEAD(formats);
78	static DEFINE_RWLOCK(binfmt_lock);
79
80	int __register_binfmt(struct linux_binfmt * fmt, int insert)
81	{
82	if (!fmt)
83	return -EINVAL;
84	write_lock(&binfmt_lock);
85	insert ? list_add(&fmt->lh, &formats) :
86	list_add_tail(&fmt->lh, &formats);
87	write_unlock(&binfmt_lock);
88	return 0;
89	}
90
91	EXPORT_SYMBOL(__register_binfmt);
92
93	void unregister_binfmt(struct linux_binfmt * fmt)
94	{
95	write_lock(&binfmt_lock);
96	list_del(&fmt->lh);
97	write_unlock(&binfmt_lock);
98	}
99
100	EXPORT_SYMBOL(unregister_binfmt);
101
102	static inline void put_binfmt(struct linux_binfmt * fmt)
103	{
104	module_put(fmt->module);
105	}
106
107	/*
108	* Note that a shared library must be both readable and executable due to
109	* security reasons.
110	*
111	* Also note that we take the address to load from from the file itself.
112	*/
113	SYSCALL_DEFINE1(uselib, const char __user *, library)
114	{
115	struct file *file;
116	char *tmp = getname(library);
117	int error = PTR_ERR(tmp);
118	static const struct open_flags uselib_flags = {
119	.open_flag = O_LARGEFILE \| O_RDONLY \| __FMODE_EXEC,
120	.acc_mode = MAY_READ \| MAY_EXEC \| MAY_OPEN,
121	.intent = LOOKUP_OPEN
122	};
123
124	if (IS_ERR(tmp))
125	goto out;
126
127	file = do_filp_open(AT_FDCWD, tmp, &uselib_flags, LOOKUP_FOLLOW);
128	putname(tmp);
129	error = PTR_ERR(file);
130	if (IS_ERR(file))
131	goto out;
132
133	error = -EINVAL;
134	if (!S_ISREG(file->f_path.dentry->d_inode->i_mode))
135	goto exit;
136
137	error = -EACCES;
138	if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
139	goto exit;
140
141	fsnotify_open(file);
142
143	error = -ENOEXEC;
144	if(file->f_op) {
145	struct linux_binfmt * fmt;
146
147	read_lock(&binfmt_lock);
148	list_for_each_entry(fmt, &formats, lh) {
149	if (!fmt->load_shlib)
150	continue;
151	if (!try_module_get(fmt->module))
152	continue;
153	read_unlock(&binfmt_lock);
154	error = fmt->load_shlib(file);
155	read_lock(&binfmt_lock);
156	put_binfmt(fmt);
157	if (error != -ENOEXEC)
158	break;
159	}
160	read_unlock(&binfmt_lock);
161	}
162	exit:
163	fput(file);
164	out:
165	return error;
166	}
167
168	#ifdef CONFIG_MMU
169	/*
170	* The nascent bprm->mm is not visible until exec_mmap() but it can
171	* use a lot of memory, account these pages in current->mm temporary
172	* for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
173	* change the counter back via acct_arg_size(0).
174	*/
175	static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
176	{
177	struct mm_struct *mm = current->mm;
178	long diff = (long)(pages - bprm->vma_pages);
179
180	if (!mm \|\| !diff)
181	return;
182
183	bprm->vma_pages = pages;
184	add_mm_counter(mm, MM_ANONPAGES, diff);
185	}
186
187	static struct page get_arg_page(struct linux_binprm bprm, unsigned long pos,
188	int write)
189	{
190	struct page *page;
191	int ret;
192
193	#ifdef CONFIG_STACK_GROWSUP
194	if (write) {
195	ret = expand_downwards(bprm->vma, pos);
196	if (ret < 0)
197	return NULL;
198	}
199	#endif
200	ret = get_user_pages(current, bprm->mm, pos,
201	1, write, 1, &page, NULL);
202	if (ret <= 0)
203	return NULL;
204
205	if (write) {
206	unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
207	struct rlimit *rlim;
208
209	acct_arg_size(bprm, size / PAGE_SIZE);
210
211	/*
212	* We've historically supported up to 32 pages (ARG_MAX)
213	* of argument strings even with small stacks
214	*/
215	if (size <= ARG_MAX)
216	return page;
217
218	/*
219	* Limit to 1/4-th the stack size for the argv+env strings.
220	* This ensures that:
221	* - the remaining binfmt code will not run out of stack space,
222	* - the program will have a reasonable amount of stack left
223	* to work from.
224	*/
225	rlim = current->signal->rlim;
226	if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) {
227	put_page(page);
228	return NULL;
229	}
230	}
231
232	return page;
233	}
234
235	static void put_arg_page(struct page *page)
236	{
237	put_page(page);
238	}
239
240	static void free_arg_page(struct linux_binprm *bprm, int i)
241	{
242	}
243
244	static void free_arg_pages(struct linux_binprm *bprm)
245	{
246	}
247
248	static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
249	struct page *page)
250	{
251	flush_cache_page(bprm->vma, pos, page_to_pfn(page));
252	}
253
254	static int __bprm_mm_init(struct linux_binprm *bprm)
255	{
256	int err;
257	struct vm_area_struct *vma = NULL;
258	struct mm_struct *mm = bprm->mm;
259
260	bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
261	if (!vma)
262	return -ENOMEM;
263
264	down_write(&mm->mmap_sem);
265	vma->vm_mm = mm;
266
267	/*
268	* Place the stack at the largest stack address the architecture
269	* supports. Later, we'll move this to an appropriate place. We don't
270	* use STACK_TOP because that can depend on attributes which aren't
271	* configured yet.
272	*/
273	BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
274	vma->vm_end = STACK_TOP_MAX;
275	vma->vm_start = vma->vm_end - PAGE_SIZE;
276	vma->vm_flags = VM_STACK_FLAGS \| VM_STACK_INCOMPLETE_SETUP;
277	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
278	INIT_LIST_HEAD(&vma->anon_vma_chain);
279
280	err = security_file_mmap(NULL, 0, 0, 0, vma->vm_start, 1);
281	if (err)
282	goto err;
283
284	err = insert_vm_struct(mm, vma);
285	if (err)
286	goto err;
287
288	mm->stack_vm = mm->total_vm = 1;
289	up_write(&mm->mmap_sem);
290	bprm->p = vma->vm_end - sizeof(void *);
291	return 0;
292	err:
293	up_write(&mm->mmap_sem);
294	bprm->vma = NULL;
295	kmem_cache_free(vm_area_cachep, vma);
296	return err;
297	}
298
299	static bool valid_arg_len(struct linux_binprm *bprm, long len)
300	{
301	return len <= MAX_ARG_STRLEN;
302	}
303
304	#else
305
306	static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
307	{
308	}
309
310	static struct page get_arg_page(struct linux_binprm bprm, unsigned long pos,
311	int write)
312	{
313	struct page *page;
314
315	page = bprm->page[pos / PAGE_SIZE];
316	if (!page && write) {
317	page = alloc_page(GFP_HIGHUSER\|__GFP_ZERO);
318	if (!page)
319	return NULL;
320	bprm->page[pos / PAGE_SIZE] = page;
321	}
322
323	return page;
324	}
325
326	static void put_arg_page(struct page *page)
327	{
328	}
329
330	static void free_arg_page(struct linux_binprm *bprm, int i)
331	{
332	if (bprm->page[i]) {
333	__free_page(bprm->page[i]);
334	bprm->page[i] = NULL;
335	}
336	}
337
338	static void free_arg_pages(struct linux_binprm *bprm)
339	{
340	int i;
341
342	for (i = 0; i < MAX_ARG_PAGES; i++)
343	free_arg_page(bprm, i);
344	}
345
346	static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
347	struct page *page)
348	{
349	}
350
351	static int __bprm_mm_init(struct linux_binprm *bprm)
352	{
353	bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
354	return 0;
355	}
356
357	static bool valid_arg_len(struct linux_binprm *bprm, long len)
358	{
359	return len <= bprm->p;
360	}
361
362	#endif /* CONFIG_MMU */
363
364	/*
365	* Create a new mm_struct and populate it with a temporary stack
366	* vm_area_struct. We don't have enough context at this point to set the stack
367	* flags, permissions, and offset, so we use temporary values. We'll update
368	* them later in setup_arg_pages().
369	*/
370	int bprm_mm_init(struct linux_binprm *bprm)
371	{
372	int err;
373	struct mm_struct *mm = NULL;
374
375	bprm->mm = mm = mm_alloc();
376	err = -ENOMEM;
377	if (!mm)
378	goto err;
379
380	err = init_new_context(current, mm);
381	if (err)
382	goto err;
383
384	err = __bprm_mm_init(bprm);
385	if (err)
386	goto err;
387
388	return 0;
389
390	err:
391	if (mm) {
392	bprm->mm = NULL;
393	mmdrop(mm);
394	}
395
396	return err;
397	}
398
399	struct user_arg_ptr {
400	#ifdef CONFIG_COMPAT
401	bool is_compat;
402	#endif
403	union {
404	const char __user const __user native;
405	#ifdef CONFIG_COMPAT
406	compat_uptr_t __user *compat;
407	#endif
408	} ptr;
409	};
410
411	static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
412	{
413	const char __user *native;
414
415	#ifdef CONFIG_COMPAT
416	if (unlikely(argv.is_compat)) {
417	compat_uptr_t compat;
418
419	if (get_user(compat, argv.ptr.compat + nr))
420	return ERR_PTR(-EFAULT);
421
422	return compat_ptr(compat);
423	}
424	#endif
425
426	if (get_user(native, argv.ptr.native + nr))
427	return ERR_PTR(-EFAULT);
428
429	return native;
430	}
431
432	/*
433	* count() counts the number of strings in array ARGV.
434	*/
435	static int count(struct user_arg_ptr argv, int max)
436	{
437	int i = 0;
438
439	if (argv.ptr.native != NULL) {
440	for (;;) {
441	const char __user *p = get_user_arg_ptr(argv, i);
442
443	if (!p)
444	break;
445
446	if (IS_ERR(p))
447	return -EFAULT;
448
449	if (i++ >= max)
450	return -E2BIG;
451
452	if (fatal_signal_pending(current))
453	return -ERESTARTNOHAND;
454	cond_resched();
455	}
456	}
457	return i;
458	}
459
460	/*
461	* 'copy_strings()' copies argument/environment strings from the old
462	* processes's memory to the new process's stack. The call to get_user_pages()
463	* ensures the destination page is created and not swapped out.
464	*/
465	static int copy_strings(int argc, struct user_arg_ptr argv,
466	struct linux_binprm *bprm)
467	{
468	struct page *kmapped_page = NULL;
469	char *kaddr = NULL;
470	unsigned long kpos = 0;
471	int ret;
472
473	while (argc-- > 0) {
474	const char __user *str;
475	int len;
476	unsigned long pos;
477
478	ret = -EFAULT;
479	str = get_user_arg_ptr(argv, argc);
480	if (IS_ERR(str))
481	goto out;
482
483	len = strnlen_user(str, MAX_ARG_STRLEN);
484	if (!len)
485	goto out;
486
487	ret = -E2BIG;
488	if (!valid_arg_len(bprm, len))
489	goto out;
490
491	/* We're going to work our way backwords. */
492	pos = bprm->p;
493	str += len;
494	bprm->p -= len;
495
496	while (len > 0) {
497	int offset, bytes_to_copy;
498
499	if (fatal_signal_pending(current)) {
500	ret = -ERESTARTNOHAND;
501	goto out;
502	}
503	cond_resched();
504
505	offset = pos % PAGE_SIZE;
506	if (offset == 0)
507	offset = PAGE_SIZE;
508
509	bytes_to_copy = offset;
510	if (bytes_to_copy > len)
511	bytes_to_copy = len;
512
513	offset -= bytes_to_copy;
514	pos -= bytes_to_copy;
515	str -= bytes_to_copy;
516	len -= bytes_to_copy;
517
518	if (!kmapped_page \|\| kpos != (pos & PAGE_MASK)) {
519	struct page *page;
520
521	page = get_arg_page(bprm, pos, 1);
522	if (!page) {
523	ret = -E2BIG;
524	goto out;
525	}
526
527	if (kmapped_page) {
528	flush_kernel_dcache_page(kmapped_page);
529	kunmap(kmapped_page);
530	put_arg_page(kmapped_page);
531	}
532	kmapped_page = page;
533	kaddr = kmap(kmapped_page);
534	kpos = pos & PAGE_MASK;
535	flush_arg_page(bprm, kpos, kmapped_page);
536	}
537	if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
538	ret = -EFAULT;
539	goto out;
540	}
541	}
542	}
543	ret = 0;
544	out:
545	if (kmapped_page) {
546	flush_kernel_dcache_page(kmapped_page);
547	kunmap(kmapped_page);
548	put_arg_page(kmapped_page);
549	}
550	return ret;
551	}
552
553	/*
554	* Like copy_strings, but get argv and its values from kernel memory.
555	*/
556	int copy_strings_kernel(int argc, const char const __argv,
557	struct linux_binprm *bprm)
558	{
559	int r;
560	mm_segment_t oldfs = get_fs();
561	struct user_arg_ptr argv = {
562	.ptr.native = (const char __user const __user )__argv,
563	};
564
565	set_fs(KERNEL_DS);
566	r = copy_strings(argc, argv, bprm);
567	set_fs(oldfs);
568
569	return r;
570	}
571	EXPORT_SYMBOL(copy_strings_kernel);
572
573	#ifdef CONFIG_MMU
574
575	/*
576	* During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
577	* the binfmt code determines where the new stack should reside, we shift it to
578	* its final location. The process proceeds as follows:
579	*
580	* 1) Use shift to calculate the new vma endpoints.
581	* 2) Extend vma to cover both the old and new ranges. This ensures the
582	* arguments passed to subsequent functions are consistent.
583	* 3) Move vma's page tables to the new range.
584	* 4) Free up any cleared pgd range.
585	* 5) Shrink the vma to cover only the new range.
586	*/
587	static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
588	{
589	struct mm_struct *mm = vma->vm_mm;
590	unsigned long old_start = vma->vm_start;
591	unsigned long old_end = vma->vm_end;
592	unsigned long length = old_end - old_start;
593	unsigned long new_start = old_start - shift;
594	unsigned long new_end = old_end - shift;
595	struct mmu_gather tlb;
596
597	BUG_ON(new_start > new_end);
598
599	/*
600	* ensure there are no vmas between where we want to go
601	* and where we are
602	*/
603	if (vma != find_vma(mm, new_start))
604	return -EFAULT;
605
606	/*
607	* cover the whole range: [new_start, old_end)
608	*/
609	if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
610	return -ENOMEM;
611
612	/*
613	* move the page tables downwards, on failure we rely on
614	* process cleanup to remove whatever mess we made.
615	*/
616	if (length != move_page_tables(vma, old_start,
617	vma, new_start, length))
618	return -ENOMEM;
619
620	lru_add_drain();
621	tlb_gather_mmu(&tlb, mm, 0);
622	if (new_end > old_start) {
623	/*
624	* when the old and new regions overlap clear from new_end.
625	*/
626	free_pgd_range(&tlb, new_end, old_end, new_end,
627	vma->vm_next ? vma->vm_next->vm_start : 0);
628	} else {
629	/*
630	* otherwise, clean from old_start; this is done to not touch
631	* the address space in [new_end, old_start) some architectures
632	* have constraints on va-space that make this illegal (IA64) -
633	* for the others its just a little faster.
634	*/
635	free_pgd_range(&tlb, old_start, old_end, new_end,
636	vma->vm_next ? vma->vm_next->vm_start : 0);
637	}
638	tlb_finish_mmu(&tlb, new_end, old_end);
639
640	/*
641	* Shrink the vma to just the new range. Always succeeds.
642	*/
643	vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
644
645	return 0;
646	}
647
648	/*
649	* Finalizes the stack vm_area_struct. The flags and permissions are updated,
650	* the stack is optionally relocated, and some extra space is added.
651	*/
652	int setup_arg_pages(struct linux_binprm *bprm,
653	unsigned long stack_top,
654	int executable_stack)
655	{
656	unsigned long ret;
657	unsigned long stack_shift;
658	struct mm_struct *mm = current->mm;
659	struct vm_area_struct *vma = bprm->vma;
660	struct vm_area_struct *prev = NULL;
661	unsigned long vm_flags;
662	unsigned long stack_base;
663	unsigned long stack_size;
664	unsigned long stack_expand;
665	unsigned long rlim_stack;
666
667	#ifdef CONFIG_STACK_GROWSUP
668	/* Limit stack size to 1GB */
669	stack_base = rlimit_max(RLIMIT_STACK);
670	if (stack_base > (1 << 30))
671	stack_base = 1 << 30;
672
673	/* Make sure we didn't let the argument array grow too large. */
674	if (vma->vm_end - vma->vm_start > stack_base)
675	return -ENOMEM;
676
677	stack_base = PAGE_ALIGN(stack_top - stack_base);
678
679	stack_shift = vma->vm_start - stack_base;
680	mm->arg_start = bprm->p - stack_shift;
681	bprm->p = vma->vm_end - stack_shift;
682	#else
683	stack_top = arch_align_stack(stack_top);
684	stack_top = PAGE_ALIGN(stack_top);
685
686	if (unlikely(stack_top < mmap_min_addr) \|\|
687	unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
688	return -ENOMEM;
689
690	stack_shift = vma->vm_end - stack_top;
691
692	bprm->p -= stack_shift;
693	mm->arg_start = bprm->p;
694	#endif
695
696	if (bprm->loader)
697	bprm->loader -= stack_shift;
698	bprm->exec -= stack_shift;
699
700	down_write(&mm->mmap_sem);
701	vm_flags = VM_STACK_FLAGS;
702
703	/*
704	* Adjust stack execute permissions; explicitly enable for
705	* EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
706	* (arch default) otherwise.
707	*/
708	if (unlikely(executable_stack == EXSTACK_ENABLE_X))
709	vm_flags \|= VM_EXEC;
710	else if (executable_stack == EXSTACK_DISABLE_X)
711	vm_flags &= ~VM_EXEC;
712	vm_flags \|= mm->def_flags;
713	vm_flags \|= VM_STACK_INCOMPLETE_SETUP;
714
715	ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
716	vm_flags);
717	if (ret)
718	goto out_unlock;
719	BUG_ON(prev != vma);
720
721	/* Move stack pages down in memory. */
722	if (stack_shift) {
723	ret = shift_arg_pages(vma, stack_shift);
724	if (ret)
725	goto out_unlock;
726	}
727
728	/* mprotect_fixup is overkill to remove the temporary stack flags */
729	vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
730
731	stack_expand = 131072UL; /* randomly 324k (or 264k) pages */
732	stack_size = vma->vm_end - vma->vm_start;
733	/*
734	* Align this down to a page boundary as expand_stack
735	* will align it up.
736	*/
737	rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
738	#ifdef CONFIG_STACK_GROWSUP
739	if (stack_size + stack_expand > rlim_stack)
740	stack_base = vma->vm_start + rlim_stack;
741	else
742	stack_base = vma->vm_end + stack_expand;
743	#else
744	if (stack_size + stack_expand > rlim_stack)
745	stack_base = vma->vm_end - rlim_stack;
746	else
747	stack_base = vma->vm_start - stack_expand;
748	#endif
749	current->mm->start_stack = bprm->p;
750	ret = expand_stack(vma, stack_base);
751	if (ret)
752	ret = -EFAULT;
753
754	out_unlock:
755	up_write(&mm->mmap_sem);
756	return ret;
757	}
758	EXPORT_SYMBOL(setup_arg_pages);
759
760	#endif /* CONFIG_MMU */
761
762	struct file open_exec(const char name)
763	{
764	struct file *file;
765	int err;
766	static const struct open_flags open_exec_flags = {
767	.open_flag = O_LARGEFILE \| O_RDONLY \| __FMODE_EXEC,
768	.acc_mode = MAY_EXEC \| MAY_OPEN,
769	.intent = LOOKUP_OPEN
770	};
771
772	file = do_filp_open(AT_FDCWD, name, &open_exec_flags, LOOKUP_FOLLOW);
773	if (IS_ERR(file))
774	goto out;
775
776	err = -EACCES;
777	if (!S_ISREG(file->f_path.dentry->d_inode->i_mode))
778	goto exit;
779
780	if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
781	goto exit;
782
783	fsnotify_open(file);
784
785	err = deny_write_access(file);
786	if (err)
787	goto exit;
788
789	out:
790	return file;
791
792	exit:
793	fput(file);
794	return ERR_PTR(err);
795	}
796	EXPORT_SYMBOL(open_exec);
797
798	int kernel_read(struct file *file, loff_t offset,
799	char *addr, unsigned long count)
800	{
801	mm_segment_t old_fs;
802	loff_t pos = offset;
803	int result;
804
805	old_fs = get_fs();
806	set_fs(get_ds());
807	/* The cast to a user pointer is valid due to the set_fs() */
808	result = vfs_read(file, (void __user *)addr, count, &pos);
809	set_fs(old_fs);
810	return result;
811	}
812
813	EXPORT_SYMBOL(kernel_read);
814
815	static int exec_mmap(struct mm_struct *mm)
816	{
817	struct task_struct *tsk;
818	struct mm_struct * old_mm, *active_mm;
819
820	/* Notify parent that we're no longer interested in the old VM */
821	tsk = current;
822	old_mm = current->mm;
823	sync_mm_rss(tsk, old_mm);
824	mm_release(tsk, old_mm);
825
826	if (old_mm) {
827	/*
828	* Make sure that if there is a core dump in progress
829	* for the old mm, we get out and die instead of going
830	* through with the exec. We must hold mmap_sem around
831	* checking core_state and changing tsk->mm.
832	*/
833	down_read(&old_mm->mmap_sem);
834	if (unlikely(old_mm->core_state)) {
835	up_read(&old_mm->mmap_sem);
836	return -EINTR;
837	}
838	}
839	task_lock(tsk);
840	active_mm = tsk->active_mm;
841	tsk->mm = mm;
842	tsk->active_mm = mm;
843	activate_mm(active_mm, mm);
844	task_unlock(tsk);
845	arch_pick_mmap_layout(mm);
846	if (old_mm) {
847	up_read(&old_mm->mmap_sem);
848	BUG_ON(active_mm != old_mm);
849	mm_update_next_owner(old_mm);
850	mmput(old_mm);
851	return 0;
852	}
853	mmdrop(active_mm);
854	return 0;
855	}
856
857	/*
858	* This function makes sure the current process has its own signal table,
859	* so that flush_signal_handlers can later reset the handlers without
860	* disturbing other processes. (Other processes might share the signal
861	* table via the CLONE_SIGHAND option to clone().)
862	*/
863	static int de_thread(struct task_struct *tsk)
864	{
865	struct signal_struct *sig = tsk->signal;
866	struct sighand_struct *oldsighand = tsk->sighand;
867	spinlock_t *lock = &oldsighand->siglock;
868
869	if (thread_group_empty(tsk))
870	goto no_thread_group;
871
872	/*
873	* Kill all other threads in the thread group.
874	*/
875	spin_lock_irq(lock);
876	if (signal_group_exit(sig)) {
877	/*
878	* Another group action in progress, just
879	* return so that the signal is processed.
880	*/
881	spin_unlock_irq(lock);
882	return -EAGAIN;
883	}
884
885	sig->group_exit_task = tsk;
886	sig->notify_count = zap_other_threads(tsk);
887	if (!thread_group_leader(tsk))
888	sig->notify_count--;
889
890	while (sig->notify_count) {
891	__set_current_state(TASK_UNINTERRUPTIBLE);
892	spin_unlock_irq(lock);
893	schedule();
894	spin_lock_irq(lock);
895	}
896	spin_unlock_irq(lock);
897
898	/*
899	* At this point all other threads have exited, all we have to
900	* do is to wait for the thread group leader to become inactive,
901	* and to assume its PID:
902	*/
903	if (!thread_group_leader(tsk)) {
904	struct task_struct *leader = tsk->group_leader;
905
906	sig->notify_count = -1; /* for exit_notify() */
907	for (;;) {
908	write_lock_irq(&tasklist_lock);
909	if (likely(leader->exit_state))
910	break;
911	__set_current_state(TASK_UNINTERRUPTIBLE);
912	write_unlock_irq(&tasklist_lock);
913	schedule();
914	}
915
916	/*
917	* The only record we have of the real-time age of a
918	* process, regardless of execs it's done, is start_time.
919	* All the past CPU time is accumulated in signal_struct
920	* from sister threads now dead. But in this non-leader
921	* exec, nothing survives from the original leader thread,
922	* whose birth marks the true age of this process now.
923	* When we take on its identity by switching to its PID, we
924	* also take its birthdate (always earlier than our own).
925	*/
926	tsk->start_time = leader->start_time;
927
928	BUG_ON(!same_thread_group(leader, tsk));
929	BUG_ON(has_group_leader_pid(tsk));
930	/*
931	* An exec() starts a new thread group with the
932	* TGID of the previous thread group. Rehash the
933	* two threads with a switched PID, and release
934	* the former thread group leader:
935	*/
936
937	/* Become a process group leader with the old leader's pid.
938	* The old leader becomes a thread of the this thread group.
939	* Note: The old leader also uses this pid until release_task
940	* is called. Odd but simple and correct.
941	*/
942	detach_pid(tsk, PIDTYPE_PID);
943	tsk->pid = leader->pid;
944	attach_pid(tsk, PIDTYPE_PID, task_pid(leader));
945	transfer_pid(leader, tsk, PIDTYPE_PGID);
946	transfer_pid(leader, tsk, PIDTYPE_SID);
947
948	list_replace_rcu(&leader->tasks, &tsk->tasks);
949	list_replace_init(&leader->sibling, &tsk->sibling);
950
951	tsk->group_leader = tsk;
952	leader->group_leader = tsk;
953
954	tsk->exit_signal = SIGCHLD;
955	leader->exit_signal = -1;
956
957	BUG_ON(leader->exit_state != EXIT_ZOMBIE);
958	leader->exit_state = EXIT_DEAD;
959
960	/*
961	* We are going to release_task()->ptrace_unlink() silently,
962	* the tracer can sleep in do_wait(). EXIT_DEAD guarantees
963	* the tracer wont't block again waiting for this thread.
964	*/
965	if (unlikely(leader->ptrace))
966	__wake_up_parent(leader, leader->parent);
967	write_unlock_irq(&tasklist_lock);
968
969	release_task(leader);
970	}
971
972	sig->group_exit_task = NULL;
973	sig->notify_count = 0;
974
975	no_thread_group:
976	if (current->mm)
977	setmax_mm_hiwater_rss(&sig->maxrss, current->mm);
978
979	exit_itimers(sig);
980	flush_itimer_signals();
981
982	if (atomic_read(&oldsighand->count) != 1) {
983	struct sighand_struct *newsighand;
984	/*
985	* This ->sighand is shared with the CLONE_SIGHAND
986	* but not CLONE_THREAD task, switch to the new one.
987	*/
988	newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
989	if (!newsighand)
990	return -ENOMEM;
991
992	atomic_set(&newsighand->count, 1);
993	memcpy(newsighand->action, oldsighand->action,
994	sizeof(newsighand->action));
995
996	write_lock_irq(&tasklist_lock);
997	spin_lock(&oldsighand->siglock);
998	rcu_assign_pointer(tsk->sighand, newsighand);
999	spin_unlock(&oldsighand->siglock);
1000	write_unlock_irq(&tasklist_lock);
1001
1002	__cleanup_sighand(oldsighand);
1003	}
1004
1005	BUG_ON(!thread_group_leader(tsk));
1006	return 0;
1007	}
1008
1009	/*
1010	* These functions flushes out all traces of the currently running executable
1011	* so that a new one can be started
1012	*/
1013	static void flush_old_files(struct files_struct * files)
1014	{
1015	long j = -1;
1016	struct fdtable *fdt;
1017
1018	spin_lock(&files->file_lock);
1019	for (;;) {
1020	unsigned long set, i;
1021
1022	j++;
1023	i = j * __NFDBITS;
1024	fdt = files_fdtable(files);
1025	if (i >= fdt->max_fds)
1026	break;
1027	set = fdt->close_on_exec->fds_bits[j];
1028	if (!set)
1029	continue;
1030	fdt->close_on_exec->fds_bits[j] = 0;
1031	spin_unlock(&files->file_lock);
1032	for ( ; set ; i++,set >>= 1) {
1033	if (set & 1) {
1034	sys_close(i);
1035	}
1036	}
1037	spin_lock(&files->file_lock);
1038
1039	}
1040	spin_unlock(&files->file_lock);
1041	}
1042
1043	char get_task_comm(char buf, struct task_struct *tsk)
1044	{
1045	/* buf must be at least sizeof(tsk->comm) in size */
1046	task_lock(tsk);
1047	strncpy(buf, tsk->comm, sizeof(tsk->comm));
1048	task_unlock(tsk);
1049	return buf;
1050	}
1051	EXPORT_SYMBOL_GPL(get_task_comm);
1052
1053	void set_task_comm(struct task_struct tsk, char buf)
1054	{
1055	task_lock(tsk);
1056
1057	/*
1058	* Threads may access current->comm without holding
1059	* the task lock, so write the string carefully.
1060	* Readers without a lock may see incomplete new
1061	* names but are safe from non-terminating string reads.
1062	*/
1063	memset(tsk->comm, 0, TASK_COMM_LEN);
1064	wmb();
1065	strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1066	task_unlock(tsk);
1067	perf_event_comm(tsk);
1068	}
1069
1070	int flush_old_exec(struct linux_binprm * bprm)
1071	{
1072	int retval;
1073
1074	/*
1075	* Make sure we have a private signal table and that
1076	* we are unassociated from the previous thread group.
1077	*/
1078	retval = de_thread(current);
1079	if (retval)
1080	goto out;
1081
1082	set_mm_exe_file(bprm->mm, bprm->file);
1083
1084	/*
1085	* Release all of the old mmap stuff
1086	*/
1087	acct_arg_size(bprm, 0);
1088	retval = exec_mmap(bprm->mm);
1089	if (retval)
1090	goto out;
1091
1092	bprm->mm = NULL; /* We're using it now */
1093
1094	set_fs(USER_DS);
1095	current->flags &= ~(PF_RANDOMIZE \| PF_KTHREAD);
1096	flush_thread();
1097	current->personality &= ~bprm->per_clear;
1098
1099	return 0;
1100
1101	out:
1102	return retval;
1103	}
1104	EXPORT_SYMBOL(flush_old_exec);
1105
1106	void would_dump(struct linux_binprm bprm, struct file file)
1107	{
1108	if (inode_permission(file->f_path.dentry->d_inode, MAY_READ) < 0)
1109	bprm->interp_flags \|= BINPRM_FLAGS_ENFORCE_NONDUMP;
1110	}
1111	EXPORT_SYMBOL(would_dump);
1112
1113	void setup_new_exec(struct linux_binprm * bprm)
1114	{
1115	int i, ch;
1116	const char *name;
1117	char tcomm[sizeof(current->comm)];
1118
1119	arch_pick_mmap_layout(current->mm);
1120
1121	/* This is the point of no return */
1122	current->sas_ss_sp = current->sas_ss_size = 0;
1123
1124	if (current_euid() == current_uid() && current_egid() == current_gid())
1125	set_dumpable(current->mm, 1);
1126	else
1127	set_dumpable(current->mm, suid_dumpable);
1128
1129	name = bprm->filename;
1130
1131	/* Copies the binary name from after last slash */
1132	for (i=0; (ch = *(name++)) != '\0';) {
1133	if (ch == '/')
1134	i = 0; /* overwrite what we wrote */
1135	else
1136	if (i < (sizeof(tcomm) - 1))
1137	tcomm[i++] = ch;
1138	}
1139	tcomm[i] = '\0';
1140	set_task_comm(current, tcomm);
1141
1142	/* Set the new mm task size. We have to do that late because it may
1143	* depend on TIF_32BIT which is only updated in flush_thread() on
1144	* some architectures like powerpc
1145	*/
1146	current->mm->task_size = TASK_SIZE;
1147
1148	/* install the new credentials */
1149	if (bprm->cred->uid != current_euid() \|\|
1150	bprm->cred->gid != current_egid()) {
1151	current->pdeath_signal = 0;
1152	} else {
1153	would_dump(bprm, bprm->file);
1154	if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1155	set_dumpable(current->mm, suid_dumpable);
1156	}
1157
1158	/*
1159	* Flush performance counters when crossing a
1160	* security domain:
1161	*/
1162	if (!get_dumpable(current->mm))
1163	perf_event_exit_task(current);
1164
1165	/* An exec changes our domain. We are no longer part of the thread
1166	group */
1167
1168	current->self_exec_id++;
1169
1170	flush_signal_handlers(current, 0);
1171	flush_old_files(current->files);
1172	}
1173	EXPORT_SYMBOL(setup_new_exec);
1174
1175	/*
1176	* Prepare credentials and lock ->cred_guard_mutex.
1177	* install_exec_creds() commits the new creds and drops the lock.
1178	* Or, if exec fails before, free_bprm() should release ->cred and
1179	* and unlock.
1180	*/
1181	int prepare_bprm_creds(struct linux_binprm *bprm)
1182	{
1183	if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
1184	return -ERESTARTNOINTR;
1185
1186	bprm->cred = prepare_exec_creds();
1187	if (likely(bprm->cred))
1188	return 0;
1189
1190	mutex_unlock(&current->signal->cred_guard_mutex);
1191	return -ENOMEM;
1192	}
1193
1194	void free_bprm(struct linux_binprm *bprm)
1195	{
1196	free_arg_pages(bprm);
1197	if (bprm->cred) {
1198	mutex_unlock(&current->signal->cred_guard_mutex);
1199	abort_creds(bprm->cred);
1200	}
1201	kfree(bprm);
1202	}
1203
1204	/*
1205	* install the new credentials for this executable
1206	*/
1207	void install_exec_creds(struct linux_binprm *bprm)
1208	{
1209	security_bprm_committing_creds(bprm);
1210
1211	commit_creds(bprm->cred);
1212	bprm->cred = NULL;
1213	/*
1214	* cred_guard_mutex must be held at least to this point to prevent
1215	* ptrace_attach() from altering our determination of the task's
1216	* credentials; any time after this it may be unlocked.
1217	*/
1218	security_bprm_committed_creds(bprm);
1219	mutex_unlock(&current->signal->cred_guard_mutex);
1220	}
1221	EXPORT_SYMBOL(install_exec_creds);
1222
1223	/*
1224	* determine how safe it is to execute the proposed program
1225	* - the caller must hold ->cred_guard_mutex to protect against
1226	* PTRACE_ATTACH
1227	*/
1228	int check_unsafe_exec(struct linux_binprm *bprm)
1229	{
1230	struct task_struct p = current, t;
1231	unsigned n_fs;
1232	int res = 0;
1233
1234	if (p->ptrace) {
1235	if (p->ptrace & PT_PTRACE_CAP)
1236	bprm->unsafe \|= LSM_UNSAFE_PTRACE_CAP;
1237	else
1238	bprm->unsafe \|= LSM_UNSAFE_PTRACE;
1239	}
1240
1241	n_fs = 1;
1242	spin_lock(&p->fs->lock);
1243	rcu_read_lock();
1244	for (t = next_thread(p); t != p; t = next_thread(t)) {
1245	if (t->fs == p->fs)
1246	n_fs++;
1247	}
1248	rcu_read_unlock();
1249
1250	if (p->fs->users > n_fs) {
1251	bprm->unsafe \|= LSM_UNSAFE_SHARE;
1252	} else {
1253	res = -EAGAIN;
1254	if (!p->fs->in_exec) {
1255	p->fs->in_exec = 1;
1256	res = 1;
1257	}
1258	}
1259	spin_unlock(&p->fs->lock);
1260
1261	return res;
1262	}
1263
1264	/*
1265	* Fill the binprm structure from the inode.
1266	* Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1267	*
1268	* This may be called multiple times for binary chains (scripts for example).
1269	*/
1270	int prepare_binprm(struct linux_binprm *bprm)
1271	{
1272	umode_t mode;
1273	struct inode * inode = bprm->file->f_path.dentry->d_inode;
1274	int retval;
1275
1276	mode = inode->i_mode;
1277	if (bprm->file->f_op == NULL)
1278	return -EACCES;
1279
1280	/* clear any previous set[ug]id data from a previous binary */
1281	bprm->cred->euid = current_euid();
1282	bprm->cred->egid = current_egid();
1283
1284	if (!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)) {
1285	/* Set-uid? */
1286	if (mode & S_ISUID) {
1287	bprm->per_clear \|= PER_CLEAR_ON_SETID;
1288	bprm->cred->euid = inode->i_uid;
1289	}
1290
1291	/* Set-gid? */
1292	/*
1293	* If setgid is set but no group execute bit then this
1294	* is a candidate for mandatory locking, not a setgid
1295	* executable.
1296	*/
1297	if ((mode & (S_ISGID \| S_IXGRP)) == (S_ISGID \| S_IXGRP)) {
1298	bprm->per_clear \|= PER_CLEAR_ON_SETID;
1299	bprm->cred->egid = inode->i_gid;
1300	}
1301	}
1302
1303	/* fill in binprm security blob */
1304	retval = security_bprm_set_creds(bprm);
1305	if (retval)
1306	return retval;
1307	bprm->cred_prepared = 1;
1308
1309	memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1310	return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
1311	}
1312
1313	EXPORT_SYMBOL(prepare_binprm);
1314
1315	/*
1316	* Arguments are '\0' separated strings found at the location bprm->p
1317	* points to; chop off the first by relocating brpm->p to right after
1318	* the first '\0' encountered.
1319	*/
1320	int remove_arg_zero(struct linux_binprm *bprm)
1321	{
1322	int ret = 0;
1323	unsigned long offset;
1324	char *kaddr;
1325	struct page *page;
1326
1327	if (!bprm->argc)
1328	return 0;
1329
1330	do {
1331	offset = bprm->p & ~PAGE_MASK;
1332	page = get_arg_page(bprm, bprm->p, 0);
1333	if (!page) {
1334	ret = -EFAULT;
1335	goto out;
1336	}
1337	kaddr = kmap_atomic(page, KM_USER0);
1338
1339	for (; offset < PAGE_SIZE && kaddr[offset];
1340	offset++, bprm->p++)
1341	;
1342
1343	kunmap_atomic(kaddr, KM_USER0);
1344	put_arg_page(page);
1345
1346	if (offset == PAGE_SIZE)
1347	free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1348	} while (offset == PAGE_SIZE);
1349
1350	bprm->p++;
1351	bprm->argc--;
1352	ret = 0;
1353
1354	out:
1355	return ret;
1356	}
1357	EXPORT_SYMBOL(remove_arg_zero);
1358
1359	/*
1360	* cycle the list of binary formats handler, until one recognizes the image
1361	*/
1362	int search_binary_handler(struct linux_binprm bprm,struct pt_regs regs)
1363	{
1364	unsigned int depth = bprm->recursion_depth;
1365	int try,retval;
1366	struct linux_binfmt *fmt;
1367	pid_t old_pid;
1368
1369	retval = security_bprm_check(bprm);
1370	if (retval)
1371	return retval;
1372
1373	retval = audit_bprm(bprm);
1374	if (retval)
1375	return retval;
1376
1377	/* Need to fetch pid before load_binary changes it */
1378	rcu_read_lock();
1379	old_pid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
1380	rcu_read_unlock();
1381
1382	retval = -ENOENT;
1383	for (try=0; try<2; try++) {
1384	read_lock(&binfmt_lock);
1385	list_for_each_entry(fmt, &formats, lh) {
1386	int (fn)(struct linux_binprm , struct pt_regs *) = fmt->load_binary;
1387	if (!fn)
1388	continue;
1389	if (!try_module_get(fmt->module))
1390	continue;
1391	read_unlock(&binfmt_lock);
1392	retval = fn(bprm, regs);
1393	/*
1394	* Restore the depth counter to its starting value
1395	* in this call, so we don't have to rely on every
1396	* load_binary function to restore it on return.
1397	*/
1398	bprm->recursion_depth = depth;
1399	if (retval >= 0) {
1400	if (depth == 0)
1401	ptrace_event(PTRACE_EVENT_EXEC,
1402	old_pid);
1403	put_binfmt(fmt);
1404	allow_write_access(bprm->file);
1405	if (bprm->file)
1406	fput(bprm->file);
1407	bprm->file = NULL;
1408	current->did_exec = 1;
1409	proc_exec_connector(current);
1410	return retval;
1411	}
1412	read_lock(&binfmt_lock);
1413	put_binfmt(fmt);
1414	if (retval != -ENOEXEC \|\| bprm->mm == NULL)
1415	break;
1416	if (!bprm->file) {
1417	read_unlock(&binfmt_lock);
1418	return retval;
1419	}
1420	}
1421	read_unlock(&binfmt_lock);
1422	#ifdef CONFIG_MODULES
1423	if (retval != -ENOEXEC \|\| bprm->mm == NULL) {
1424	break;
1425	} else {
1426	#define printable(c) (((c)=='\t') \|\| ((c)=='\n') \|\| (0x20<=(c) && (c)<=0x7e))
1427	if (printable(bprm->buf[0]) &&
1428	printable(bprm->buf[1]) &&
1429	printable(bprm->buf[2]) &&
1430	printable(bprm->buf[3]))
1431	break; /* -ENOEXEC */
1432	if (try)
1433	break; /* -ENOEXEC */
1434	request_module("binfmt-%04x", (unsigned short )(&bprm->buf[2]));
1435	}
1436	#else
1437	break;
1438	#endif
1439	}
1440	return retval;
1441	}
1442
1443	EXPORT_SYMBOL(search_binary_handler);
1444
1445	/*
1446	* sys_execve() executes a new program.
1447	*/
1448	static int do_execve_common(const char *filename,
1449	struct user_arg_ptr argv,
1450	struct user_arg_ptr envp,
1451	struct pt_regs *regs)
1452	{
1453	struct linux_binprm *bprm;
1454	struct file *file;
1455	struct files_struct *displaced;
1456	bool clear_in_exec;
1457	int retval;
1458	const struct cred *cred = current_cred();
1459
1460	/*
1461	* We move the actual failure in case of RLIMIT_NPROC excess from
1462	* set*uid() to execve() because too many poorly written programs
1463	* don't check setuid() return code. Here we additionally recheck
1464	* whether NPROC limit is still exceeded.
1465	*/
1466	if ((current->flags & PF_NPROC_EXCEEDED) &&
1467	atomic_read(&cred->user->processes) > rlimit(RLIMIT_NPROC)) {
1468	retval = -EAGAIN;
1469	goto out_ret;
1470	}
1471
1472	/* We're below the limit (still or again), so we don't want to make
1473	* further execve() calls fail. */
1474	current->flags &= ~PF_NPROC_EXCEEDED;
1475
1476	retval = unshare_files(&displaced);
1477	if (retval)
1478	goto out_ret;
1479
1480	retval = -ENOMEM;
1481	bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
1482	if (!bprm)
1483	goto out_files;
1484
1485	retval = prepare_bprm_creds(bprm);
1486	if (retval)
1487	goto out_free;
1488
1489	retval = check_unsafe_exec(bprm);
1490	if (retval < 0)
1491	goto out_free;
1492	clear_in_exec = retval;
1493	current->in_execve = 1;
1494
1495	file = open_exec(filename);
1496	retval = PTR_ERR(file);
1497	if (IS_ERR(file))
1498	goto out_unmark;
1499
1500	sched_exec();
1501
1502	bprm->file = file;
1503	bprm->filename = filename;
1504	bprm->interp = filename;
1505
1506	retval = bprm_mm_init(bprm);
1507	if (retval)
1508	goto out_file;
1509
1510	bprm->argc = count(argv, MAX_ARG_STRINGS);
1511	if ((retval = bprm->argc) < 0)
1512	goto out;
1513
1514	bprm->envc = count(envp, MAX_ARG_STRINGS);
1515	if ((retval = bprm->envc) < 0)
1516	goto out;
1517
1518	retval = prepare_binprm(bprm);
1519	if (retval < 0)
1520	goto out;
1521
1522	retval = copy_strings_kernel(1, &bprm->filename, bprm);
1523	if (retval < 0)
1524	goto out;
1525
1526	bprm->exec = bprm->p;
1527	retval = copy_strings(bprm->envc, envp, bprm);
1528	if (retval < 0)
1529	goto out;
1530
1531	retval = copy_strings(bprm->argc, argv, bprm);
1532	if (retval < 0)
1533	goto out;
1534
1535	retval = search_binary_handler(bprm,regs);
1536	if (retval < 0)
1537	goto out;
1538
1539	/* execve succeeded */
1540	current->fs->in_exec = 0;
1541	current->in_execve = 0;
1542	acct_update_integrals(current);
1543	free_bprm(bprm);
1544	if (displaced)
1545	put_files_struct(displaced);
1546	return retval;
1547
1548	out:
1549	if (bprm->mm) {
1550	acct_arg_size(bprm, 0);
1551	mmput(bprm->mm);
1552	}
1553
1554	out_file:
1555	if (bprm->file) {
1556	allow_write_access(bprm->file);
1557	fput(bprm->file);
1558	}
1559
1560	out_unmark:
1561	if (clear_in_exec)
1562	current->fs->in_exec = 0;
1563	current->in_execve = 0;
1564
1565	out_free:
1566	free_bprm(bprm);
1567
1568	out_files:
1569	if (displaced)
1570	reset_files_struct(displaced);
1571	out_ret:
1572	return retval;
1573	}
1574
1575	int do_execve(const char *filename,
1576	const char __user const __user __argv,
1577	const char __user const __user __envp,
1578	struct pt_regs *regs)
1579	{
1580	struct user_arg_ptr argv = { .ptr.native = __argv };
1581	struct user_arg_ptr envp = { .ptr.native = __envp };
1582	return do_execve_common(filename, argv, envp, regs);
1583	}
1584
1585	#ifdef CONFIG_COMPAT
1586	int compat_do_execve(char *filename,
1587	compat_uptr_t __user *__argv,
1588	compat_uptr_t __user *__envp,
1589	struct pt_regs *regs)
1590	{
1591	struct user_arg_ptr argv = {
1592	.is_compat = true,
1593	.ptr.compat = __argv,
1594	};
1595	struct user_arg_ptr envp = {
1596	.is_compat = true,
1597	.ptr.compat = __envp,
1598	};
1599	return do_execve_common(filename, argv, envp, regs);
1600	}
1601	#endif
1602
1603	void set_binfmt(struct linux_binfmt *new)
1604	{
1605	struct mm_struct *mm = current->mm;
1606
1607	if (mm->binfmt)
1608	module_put(mm->binfmt->module);
1609
1610	mm->binfmt = new;
1611	if (new)
1612	__module_get(new->module);
1613	}
1614
1615	EXPORT_SYMBOL(set_binfmt);
1616
1617	static int expand_corename(struct core_name *cn)
1618	{
1619	char *old_corename = cn->corename;
1620
1621	cn->size = CORENAME_MAX_SIZE * atomic_inc_return(&call_count);
1622	cn->corename = krealloc(old_corename, cn->size, GFP_KERNEL);
1623
1624	if (!cn->corename) {
1625	kfree(old_corename);
1626	return -ENOMEM;
1627	}
1628
1629	return 0;
1630	}
1631
1632	static int cn_printf(struct core_name cn, const char fmt, ...)
1633	{
1634	char *cur;
1635	int need;
1636	int ret;
1637	va_list arg;
1638
1639	va_start(arg, fmt);
1640	need = vsnprintf(NULL, 0, fmt, arg);
1641	va_end(arg);
1642
1643	if (likely(need < cn->size - cn->used - 1))
1644	goto out_printf;
1645
1646	ret = expand_corename(cn);
1647	if (ret)
1648	goto expand_fail;
1649
1650	out_printf:
1651	cur = cn->corename + cn->used;
1652	va_start(arg, fmt);
1653	vsnprintf(cur, need + 1, fmt, arg);
1654	va_end(arg);
1655	cn->used += need;
1656	return 0;
1657
1658	expand_fail:
1659	return ret;
1660	}
1661
1662	static void cn_escape(char *str)
1663	{
1664	for (; *str; str++)
1665	if (*str == '/')
1666	*str = '!';
1667	}
1668
1669	static int cn_print_exe_file(struct core_name *cn)
1670	{
1671	struct file *exe_file;
1672	char pathbuf, path;
1673	int ret;
1674
1675	exe_file = get_mm_exe_file(current->mm);
1676	if (!exe_file) {
1677	char *commstart = cn->corename + cn->used;
1678	ret = cn_printf(cn, "%s (path unknown)", current->comm);
1679	cn_escape(commstart);
1680	return ret;
1681	}
1682
1683	pathbuf = kmalloc(PATH_MAX, GFP_TEMPORARY);
1684	if (!pathbuf) {
1685	ret = -ENOMEM;
1686	goto put_exe_file;
1687	}
1688
1689	path = d_path(&exe_file->f_path, pathbuf, PATH_MAX);
1690	if (IS_ERR(path)) {
1691	ret = PTR_ERR(path);
1692	goto free_buf;
1693	}
1694
1695	cn_escape(path);
1696
1697	ret = cn_printf(cn, "%s", path);
1698
1699	free_buf:
1700	kfree(pathbuf);
1701	put_exe_file:
1702	fput(exe_file);
1703	return ret;
1704	}
1705
1706	/* format_corename will inspect the pattern parameter, and output a
1707	* name into corename, which must have space for at least
1708	* CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1709	*/
1710	static int format_corename(struct core_name *cn, long signr)
1711	{
1712	const struct cred *cred = current_cred();
1713	const char *pat_ptr = core_pattern;
1714	int ispipe = (*pat_ptr == '\|');
1715	int pid_in_pattern = 0;
1716	int err = 0;
1717
1718	cn->size = CORENAME_MAX_SIZE * atomic_read(&call_count);
1719	cn->corename = kmalloc(cn->size, GFP_KERNEL);
1720	cn->used = 0;
1721
1722	if (!cn->corename)
1723	return -ENOMEM;
1724
1725	/* Repeat as long as we have more pattern to process and more output
1726	space */
1727	while (*pat_ptr) {
1728	if (*pat_ptr != '%') {
1729	if (*pat_ptr == 0)
1730	goto out;
1731	err = cn_printf(cn, "%c", *pat_ptr++);
1732	} else {
1733	switch (*++pat_ptr) {
1734	/* single % at the end, drop that */
1735	case 0:
1736	goto out;
1737	/* Double percent, output one percent */
1738	case '%':
1739	err = cn_printf(cn, "%c", '%');
1740	break;
1741	/* pid */
1742	case 'p':
1743	pid_in_pattern = 1;
1744	err = cn_printf(cn, "%d",
1745	task_tgid_vnr(current));
1746	break;
1747	/* uid */
1748	case 'u':
1749	err = cn_printf(cn, "%d", cred->uid);
1750	break;
1751	/* gid */
1752	case 'g':
1753	err = cn_printf(cn, "%d", cred->gid);
1754	break;
1755	/* signal that caused the coredump */
1756	case 's':
1757	err = cn_printf(cn, "%ld", signr);
1758	break;
1759	/* UNIX time of coredump */
1760	case 't': {
1761	struct timeval tv;
1762	do_gettimeofday(&tv);
1763	err = cn_printf(cn, "%lu", tv.tv_sec);
1764	break;
1765	}
1766	/* hostname */
1767	case 'h': {
1768	char *namestart = cn->corename + cn->used;
1769	down_read(&uts_sem);
1770	err = cn_printf(cn, "%s",
1771	utsname()->nodename);
1772	up_read(&uts_sem);
1773	cn_escape(namestart);
1774	break;
1775	}
1776	/* executable */
1777	case 'e': {
1778	char *commstart = cn->corename + cn->used;
1779	err = cn_printf(cn, "%s", current->comm);
1780	cn_escape(commstart);
1781	break;
1782	}
1783	case 'E':
1784	err = cn_print_exe_file(cn);
1785	break;
1786	/* core limit size */
1787	case 'c':
1788	err = cn_printf(cn, "%lu",
1789	rlimit(RLIMIT_CORE));
1790	break;
1791	default:
1792	break;
1793	}
1794	++pat_ptr;
1795	}
1796
1797	if (err)
1798	return err;
1799	}
1800
1801	/* Backward compatibility with core_uses_pid:
1802	*
1803	* If core_pattern does not include a %p (as is the default)
1804	* and core_uses_pid is set, then .%pid will be appended to
1805	* the filename. Do not do this for piped commands. */
1806	if (!ispipe && !pid_in_pattern && core_uses_pid) {
1807	err = cn_printf(cn, ".%d", task_tgid_vnr(current));
1808	if (err)
1809	return err;
1810	}
1811	out:
1812	return ispipe;
1813	}
1814
1815	static int zap_process(struct task_struct *start, int exit_code)
1816	{
1817	struct task_struct *t;
1818	int nr = 0;
1819
1820	start->signal->flags = SIGNAL_GROUP_EXIT;
1821	start->signal->group_exit_code = exit_code;
1822	start->signal->group_stop_count = 0;
1823
1824	t = start;
1825	do {
1826	task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1827	if (t != current && t->mm) {
1828	sigaddset(&t->pending.signal, SIGKILL);
1829	signal_wake_up(t, 1);
1830	nr++;
1831	}
1832	} while_each_thread(start, t);
1833
1834	return nr;
1835	}
1836
1837	static inline int zap_threads(struct task_struct tsk, struct mm_struct mm,
1838	struct core_state *core_state, int exit_code)
1839	{
1840	struct task_struct g, p;
1841	unsigned long flags;
1842	int nr = -EAGAIN;
1843
1844	spin_lock_irq(&tsk->sighand->siglock);
1845	if (!signal_group_exit(tsk->signal)) {
1846	mm->core_state = core_state;
1847	nr = zap_process(tsk, exit_code);
1848	}
1849	spin_unlock_irq(&tsk->sighand->siglock);
1850	if (unlikely(nr < 0))
1851	return nr;
1852
1853	if (atomic_read(&mm->mm_users) == nr + 1)
1854	goto done;
1855	/*
1856	* We should find and kill all tasks which use this mm, and we should
1857	* count them correctly into ->nr_threads. We don't take tasklist
1858	* lock, but this is safe wrt:
1859	*
1860	* fork:
1861	* None of sub-threads can fork after zap_process(leader). All
1862	* processes which were created before this point should be
1863	* visible to zap_threads() because copy_process() adds the new
1864	* process to the tail of init_task.tasks list, and lock/unlock
1865	* of ->siglock provides a memory barrier.
1866	*
1867	* do_exit:
1868	* The caller holds mm->mmap_sem. This means that the task which
1869	* uses this mm can't pass exit_mm(), so it can't exit or clear
1870	* its ->mm.
1871	*
1872	* de_thread:
1873	* It does list_replace_rcu(&leader->tasks, &current->tasks),
1874	* we must see either old or new leader, this does not matter.
1875	* However, it can change p->sighand, so lock_task_sighand(p)
1876	* must be used. Since p->mm != NULL and we hold ->mmap_sem
1877	* it can't fail.
1878	*
1879	* Note also that "g" can be the old leader with ->mm == NULL
1880	* and already unhashed and thus removed from ->thread_group.
1881	* This is OK, __unhash_process()->list_del_rcu() does not
1882	* clear the ->next pointer, we will find the new leader via
1883	* next_thread().
1884	*/
1885	rcu_read_lock();
1886	for_each_process(g) {
1887	if (g == tsk->group_leader)
1888	continue;
1889	if (g->flags & PF_KTHREAD)
1890	continue;
1891	p = g;
1892	do {
1893	if (p->mm) {
1894	if (unlikely(p->mm == mm)) {
1895	lock_task_sighand(p, &flags);
1896	nr += zap_process(p, exit_code);
1897	unlock_task_sighand(p, &flags);
1898	}
1899	break;
1900	}
1901	} while_each_thread(g, p);
1902	}
1903	rcu_read_unlock();
1904	done:
1905	atomic_set(&core_state->nr_threads, nr);
1906	return nr;
1907	}
1908
1909	static int coredump_wait(int exit_code, struct core_state *core_state)
1910	{
1911	struct task_struct *tsk = current;
1912	struct mm_struct *mm = tsk->mm;
1913	struct completion *vfork_done;
1914	int core_waiters = -EBUSY;
1915
1916	init_completion(&core_state->startup);
1917	core_state->dumper.task = tsk;
1918	core_state->dumper.next = NULL;
1919
1920	down_write(&mm->mmap_sem);
1921	if (!mm->core_state)
1922	core_waiters = zap_threads(tsk, mm, core_state, exit_code);
1923	up_write(&mm->mmap_sem);
1924
1925	if (unlikely(core_waiters < 0))
1926	goto fail;
1927
1928	/*
1929	* Make sure nobody is waiting for us to release the VM,
1930	* otherwise we can deadlock when we wait on each other
1931	*/
1932	vfork_done = tsk->vfork_done;
1933	if (vfork_done) {
1934	tsk->vfork_done = NULL;
1935	complete(vfork_done);
1936	}
1937
1938	if (core_waiters)
1939	wait_for_completion(&core_state->startup);
1940	fail:
1941	return core_waiters;
1942	}
1943
1944	static void coredump_finish(struct mm_struct *mm)
1945	{
1946	struct core_thread curr, next;
1947	struct task_struct *task;
1948
1949	next = mm->core_state->dumper.next;
1950	while ((curr = next) != NULL) {
1951	next = curr->next;
1952	task = curr->task;
1953	/*
1954	* see exit_mm(), curr->task must not see
1955	* ->task == NULL before we read ->next.
1956	*/
1957	smp_mb();
1958	curr->task = NULL;
1959	wake_up_process(task);
1960	}
1961
1962	mm->core_state = NULL;
1963	}
1964
1965	/*
1966	* set_dumpable converts traditional three-value dumpable to two flags and
1967	* stores them into mm->flags. It modifies lower two bits of mm->flags, but
1968	* these bits are not changed atomically. So get_dumpable can observe the
1969	* intermediate state. To avoid doing unexpected behavior, get get_dumpable
1970	* return either old dumpable or new one by paying attention to the order of
1971	* modifying the bits.
1972	*
1973	* dumpable \| mm->flags (binary)
1974	* old new \| initial interim final
1975	* ---------+-----------------------
1976	* 0 1 \| 00 01 01
1977	* 0 2 \| 00 10(*) 11
1978	* 1 0 \| 01 00 00
1979	* 1 2 \| 01 11 11
1980	* 2 0 \| 11 10(*) 00
1981	* 2 1 \| 11 11 01
1982	*
1983	* (*) get_dumpable regards interim value of 10 as 11.
1984	*/
1985	void set_dumpable(struct mm_struct *mm, int value)
1986	{
1987	switch (value) {
1988	case 0:
1989	clear_bit(MMF_DUMPABLE, &mm->flags);
1990	smp_wmb();
1991	clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1992	break;
1993	case 1:
1994	set_bit(MMF_DUMPABLE, &mm->flags);
1995	smp_wmb();
1996	clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1997	break;
1998	case 2:
1999	set_bit(MMF_DUMP_SECURELY, &mm->flags);
2000	smp_wmb();
2001	set_bit(MMF_DUMPABLE, &mm->flags);
2002	break;
2003	}
2004	}
2005
2006	static int __get_dumpable(unsigned long mm_flags)
2007	{
2008	int ret;
2009
2010	ret = mm_flags & MMF_DUMPABLE_MASK;
2011	return (ret >= 2) ? 2 : ret;
2012	}
2013
2014	int get_dumpable(struct mm_struct *mm)
2015	{
2016	return __get_dumpable(mm->flags);
2017	}
2018
2019	static void wait_for_dump_helpers(struct file *file)
2020	{
2021	struct pipe_inode_info *pipe;
2022
2023	pipe = file->f_path.dentry->d_inode->i_pipe;
2024
2025	pipe_lock(pipe);
2026	pipe->readers++;
2027	pipe->writers--;
2028
2029	while ((pipe->readers > 1) && (!signal_pending(current))) {
2030	wake_up_interruptible_sync(&pipe->wait);
2031	kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
2032	pipe_wait(pipe);
2033	}
2034
2035	pipe->readers--;
2036	pipe->writers++;
2037	pipe_unlock(pipe);
2038
2039	}
2040
2041
2042	/*
2043	* umh_pipe_setup
2044	* helper function to customize the process used
2045	* to collect the core in userspace. Specifically
2046	* it sets up a pipe and installs it as fd 0 (stdin)
2047	* for the process. Returns 0 on success, or
2048	* PTR_ERR on failure.
2049	* Note that it also sets the core limit to 1. This
2050	* is a special value that we use to trap recursive
2051	* core dumps
2052	*/
2053	static int umh_pipe_setup(struct subprocess_info info, struct cred new)
2054	{
2055	struct file rp, wp;
2056	struct fdtable *fdt;
2057	struct coredump_params cp = (struct coredump_params )info->data;
2058	struct files_struct *cf = current->files;
2059
2060	wp = create_write_pipe(0);
2061	if (IS_ERR(wp))
2062	return PTR_ERR(wp);
2063
2064	rp = create_read_pipe(wp, 0);
2065	if (IS_ERR(rp)) {
2066	free_write_pipe(wp);
2067	return PTR_ERR(rp);
2068	}
2069
2070	cp->file = wp;
2071
2072	sys_close(0);
2073	fd_install(0, rp);
2074	spin_lock(&cf->file_lock);
2075	fdt = files_fdtable(cf);
2076	FD_SET(0, fdt->open_fds);
2077	FD_CLR(0, fdt->close_on_exec);
2078	spin_unlock(&cf->file_lock);
2079
2080	/* and disallow core files too */
2081	current->signal->rlim[RLIMIT_CORE] = (struct rlimit){1, 1};
2082
2083	return 0;
2084	}
2085
2086	void do_coredump(long signr, int exit_code, struct pt_regs *regs)
2087	{
2088	struct core_state core_state;
2089	struct core_name cn;
2090	struct mm_struct *mm = current->mm;
2091	struct linux_binfmt * binfmt;
2092	const struct cred *old_cred;
2093	struct cred *cred;
2094	int retval = 0;
2095	int flag = 0;
2096	int ispipe;
2097	static atomic_t core_dump_count = ATOMIC_INIT(0);
2098	struct coredump_params cprm = {
2099	.signr = signr,
2100	.regs = regs,
2101	.limit = rlimit(RLIMIT_CORE),
2102	/*
2103	* We must use the same mm->flags while dumping core to avoid
2104	* inconsistency of bit flags, since this flag is not protected
2105	* by any locks.
2106	*/
2107	.mm_flags = mm->flags,
2108	};
2109
2110	audit_core_dumps(signr);
2111
2112	binfmt = mm->binfmt;
2113	if (!binfmt \|\| !binfmt->core_dump)
2114	goto fail;
2115	if (!__get_dumpable(cprm.mm_flags))
2116	goto fail;
2117
2118	cred = prepare_creds();
2119	if (!cred)
2120	goto fail;
2121	/*
2122	* We cannot trust fsuid as being the "true" uid of the
2123	* process nor do we know its entire history. We only know it
2124	* was tainted so we dump it as root in mode 2.
2125	*/
2126	if (__get_dumpable(cprm.mm_flags) == 2) {
2127	/* Setuid core dump mode */
2128	flag = O_EXCL; /* Stop rewrite attacks */
2129	cred->fsuid = 0; /* Dump root private */
2130	}
2131
2132	retval = coredump_wait(exit_code, &core_state);
2133	if (retval < 0)
2134	goto fail_creds;
2135
2136	old_cred = override_creds(cred);
2137
2138	/*
2139	* Clear any false indication of pending signals that might
2140	* be seen by the filesystem code called to write the core file.
2141	*/
2142	clear_thread_flag(TIF_SIGPENDING);
2143
2144	ispipe = format_corename(&cn, signr);
2145
2146	if (ispipe) {
2147	int dump_count;
2148	char **helper_argv;
2149
2150	if (ispipe < 0) {
2151	printk(KERN_WARNING "format_corename failed\n");
2152	printk(KERN_WARNING "Aborting core\n");
2153	goto fail_corename;
2154	}
2155
2156	if (cprm.limit == 1) {
2157	/*
2158	* Normally core limits are irrelevant to pipes, since
2159	* we're not writing to the file system, but we use
2160	* cprm.limit of 1 here as a speacial value. Any
2161	* non-1 limit gets set to RLIM_INFINITY below, but
2162	* a limit of 0 skips the dump. This is a consistent
2163	* way to catch recursive crashes. We can still crash
2164	* if the core_pattern binary sets RLIM_CORE = !1
2165	* but it runs as root, and can do lots of stupid things
2166	* Note that we use task_tgid_vnr here to grab the pid
2167	* of the process group leader. That way we get the
2168	* right pid if a thread in a multi-threaded
2169	* core_pattern process dies.
2170	*/
2171	printk(KERN_WARNING
2172	"Process %d(%s) has RLIMIT_CORE set to 1\n",
2173	task_tgid_vnr(current), current->comm);
2174	printk(KERN_WARNING "Aborting core\n");
2175	goto fail_unlock;
2176	}
2177	cprm.limit = RLIM_INFINITY;
2178
2179	dump_count = atomic_inc_return(&core_dump_count);
2180	if (core_pipe_limit && (core_pipe_limit < dump_count)) {
2181	printk(KERN_WARNING "Pid %d(%s) over core_pipe_limit\n",
2182	task_tgid_vnr(current), current->comm);
2183	printk(KERN_WARNING "Skipping core dump\n");
2184	goto fail_dropcount;
2185	}
2186
2187	helper_argv = argv_split(GFP_KERNEL, cn.corename+1, NULL);
2188	if (!helper_argv) {
2189	printk(KERN_WARNING "%s failed to allocate memory\n",
2190	__func__);
2191	goto fail_dropcount;
2192	}
2193
2194	retval = call_usermodehelper_fns(helper_argv[0], helper_argv,
2195	NULL, UMH_WAIT_EXEC, umh_pipe_setup,
2196	NULL, &cprm);
2197	argv_free(helper_argv);
2198	if (retval) {
2199	printk(KERN_INFO "Core dump to %s pipe failed\n",
2200	cn.corename);
2201	goto close_fail;
2202	}
2203	} else {
2204	struct inode *inode;
2205
2206	if (cprm.limit < binfmt->min_coredump)
2207	goto fail_unlock;
2208
2209	cprm.file = filp_open(cn.corename,
2210	O_CREAT \| 2 \| O_NOFOLLOW \| O_LARGEFILE \| flag,
2211	0600);
2212	if (IS_ERR(cprm.file))
2213	goto fail_unlock;
2214
2215	inode = cprm.file->f_path.dentry->d_inode;
2216	if (inode->i_nlink > 1)
2217	goto close_fail;
2218	if (d_unhashed(cprm.file->f_path.dentry))
2219	goto close_fail;
2220	/*
2221	* AK: actually i see no reason to not allow this for named
2222	* pipes etc, but keep the previous behaviour for now.
2223	*/
2224	if (!S_ISREG(inode->i_mode))
2225	goto close_fail;
2226	/*
2227	* Dont allow local users get cute and trick others to coredump
2228	* into their pre-created files.
2229	*/
2230	if (inode->i_uid != current_fsuid())
2231	goto close_fail;
2232	if (!cprm.file->f_op \|\| !cprm.file->f_op->write)
2233	goto close_fail;
2234	if (do_truncate(cprm.file->f_path.dentry, 0, 0, cprm.file))
2235	goto close_fail;
2236	}
2237
2238	retval = binfmt->core_dump(&cprm);
2239	if (retval)
2240	current->signal->group_exit_code \|= 0x80;
2241
2242	if (ispipe && core_pipe_limit)
2243	wait_for_dump_helpers(cprm.file);
2244	close_fail:
2245	if (cprm.file)
2246	filp_close(cprm.file, NULL);
2247	fail_dropcount:
2248	if (ispipe)
2249	atomic_dec(&core_dump_count);
2250	fail_unlock:
2251	kfree(cn.corename);
2252	fail_corename:
2253	coredump_finish(mm);
2254	revert_creds(old_cred);
2255	fail_creds:
2256	put_cred(cred);
2257	fail:
2258	return;
2259	}
2260
2261	/*
2262	* Core dumping helper functions. These are the only things you should
2263	* do on a core-file: use only these functions to write out all the
2264	* necessary info.
2265	*/
2266	int dump_write(struct file file, const void addr, int nr)
2267	{
2268	return access_ok(VERIFY_READ, addr, nr) && file->f_op->write(file, addr, nr, &file->f_pos) == nr;
2269	}
2270	EXPORT_SYMBOL(dump_write);
2271
2272	int dump_seek(struct file *file, loff_t off)
2273	{
2274	int ret = 1;
2275
2276	if (file->f_op->llseek && file->f_op->llseek != no_llseek) {
2277	if (file->f_op->llseek(file, off, SEEK_CUR) < 0)
2278	return 0;
2279	} else {
2280	char buf = (char )get_zeroed_page(GFP_KERNEL);
2281
2282	if (!buf)
2283	return 0;
2284	while (off > 0) {
2285	unsigned long n = off;
2286
2287	if (n > PAGE_SIZE)
2288	n = PAGE_SIZE;
2289	if (!dump_write(file, buf, n)) {
2290	ret = 0;
2291	break;
2292	}
2293	off -= n;
2294	}
2295	free_page((unsigned long)buf);
2296	}
2297	return ret;
2298	}
2299	EXPORT_SYMBOL(dump_seek);
2300

Download this file

Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master

Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9

Kernel

Kernel Git Source Tree

Root/fs/exec.c

interactive