1/*
2 * linux/fs/ext3/xattr.c
3 *
4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
5 *
6 * Fix by Harrison Xing <harrison@mountainviewdata.com>.
7 * Ext3 code with a lot of help from Eric Jarman <ejarman@acm.org>.
8 * Extended attributes for symlinks and special files added per
9 *  suggestion of Luka Renko <luka.renko@hermes.si>.
10 * xattr consolidation Copyright (c) 2004 James Morris <jmorris@redhat.com>,
11 *  Red Hat Inc.
12 * ea-in-inode support by Alex Tomas <alex@clusterfs.com> aka bzzz
13 *  and Andreas Gruenbacher <agruen@suse.de>.
14 */
15
16/*
17 * Extended attributes are stored directly in inodes (on file systems with
18 * inodes bigger than 128 bytes) and on additional disk blocks. The i_file_acl
19 * field contains the block number if an inode uses an additional block. All
20 * attributes must fit in the inode and one additional block. Blocks that
21 * contain the identical set of attributes may be shared among several inodes.
22 * Identical blocks are detected by keeping a cache of blocks that have
23 * recently been accessed.
24 *
25 * The attributes in inodes and on blocks have a different header; the entries
26 * are stored in the same format:
27 *
28 *   +------------------+
29 *   | header           |
30 *   | entry 1          | |
31 *   | entry 2          | | growing downwards
32 *   | entry 3          | v
33 *   | four null bytes  |
34 *   | . . .            |
35 *   | value 1          | ^
36 *   | value 3          | | growing upwards
37 *   | value 2          | |
38 *   +------------------+
39 *
40 * The header is followed by multiple entry descriptors. In disk blocks, the
41 * entry descriptors are kept sorted. In inodes, they are unsorted. The
42 * attribute values are aligned to the end of the block in no specific order.
43 *
44 * Locking strategy
45 * ----------------
46 * EXT3_I(inode)->i_file_acl is protected by EXT3_I(inode)->xattr_sem.
47 * EA blocks are only changed if they are exclusive to an inode, so
48 * holding xattr_sem also means that nothing but the EA block's reference
49 * count can change. Multiple writers to the same block are synchronized
50 * by the buffer lock.
51 */
52
53#include <linux/init.h>
54#include <linux/fs.h>
55#include <linux/slab.h>
56#include <linux/ext3_jbd.h>
57#include <linux/ext3_fs.h>
58#include <linux/mbcache.h>
59#include <linux/quotaops.h>
60#include <linux/rwsem.h>
61#include "xattr.h"
62#include "acl.h"
63
64#define BHDR(bh) ((struct ext3_xattr_header *)((bh)->b_data))
65#define ENTRY(ptr) ((struct ext3_xattr_entry *)(ptr))
66#define BFIRST(bh) ENTRY(BHDR(bh)+1)
67#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0)
68
69#define IHDR(inode, raw_inode) \
70	((struct ext3_xattr_ibody_header *) \
71		((void *)raw_inode + \
72		 EXT3_GOOD_OLD_INODE_SIZE + \
73		 EXT3_I(inode)->i_extra_isize))
74#define IFIRST(hdr) ((struct ext3_xattr_entry *)((hdr)+1))
75
76#ifdef EXT3_XATTR_DEBUG
77# define ea_idebug(inode, f...) do { \
78		printk(KERN_DEBUG "inode %s:%lu: ", \
79			inode->i_sb->s_id, inode->i_ino); \
80		printk(f); \
81		printk("\n"); \
82	} while (0)
83# define ea_bdebug(bh, f...) do { \
84		char b[BDEVNAME_SIZE]; \
85		printk(KERN_DEBUG "block %s:%lu: ", \
86			bdevname(bh->b_bdev, b), \
87			(unsigned long) bh->b_blocknr); \
88		printk(f); \
89		printk("\n"); \
90	} while (0)
91#else
92# define ea_idebug(f...)
93# define ea_bdebug(f...)
94#endif
95
96static void ext3_xattr_cache_insert(struct buffer_head *);
97static struct buffer_head *ext3_xattr_cache_find(struct inode *,
98						 struct ext3_xattr_header *,
99						 struct mb_cache_entry **);
100static void ext3_xattr_rehash(struct ext3_xattr_header *,
101			      struct ext3_xattr_entry *);
102static int ext3_xattr_list(struct dentry *dentry, char *buffer,
103			   size_t buffer_size);
104
105static struct mb_cache *ext3_xattr_cache;
106
107static const struct xattr_handler *ext3_xattr_handler_map[] = {
108	[EXT3_XATTR_INDEX_USER]		     = &ext3_xattr_user_handler,
109#ifdef CONFIG_EXT3_FS_POSIX_ACL
110	[EXT3_XATTR_INDEX_POSIX_ACL_ACCESS]  = &ext3_xattr_acl_access_handler,
111	[EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext3_xattr_acl_default_handler,
112#endif
113	[EXT3_XATTR_INDEX_TRUSTED]	     = &ext3_xattr_trusted_handler,
114#ifdef CONFIG_EXT3_FS_SECURITY
115	[EXT3_XATTR_INDEX_SECURITY]	     = &ext3_xattr_security_handler,
116#endif
117};
118
119const struct xattr_handler *ext3_xattr_handlers[] = {
120	&ext3_xattr_user_handler,
121	&ext3_xattr_trusted_handler,
122#ifdef CONFIG_EXT3_FS_POSIX_ACL
123	&ext3_xattr_acl_access_handler,
124	&ext3_xattr_acl_default_handler,
125#endif
126#ifdef CONFIG_EXT3_FS_SECURITY
127	&ext3_xattr_security_handler,
128#endif
129	NULL
130};
131
132static inline const struct xattr_handler *
133ext3_xattr_handler(int name_index)
134{
135	const struct xattr_handler *handler = NULL;
136
137	if (name_index > 0 && name_index < ARRAY_SIZE(ext3_xattr_handler_map))
138		handler = ext3_xattr_handler_map[name_index];
139	return handler;
140}
141
142/*
143 * Inode operation listxattr()
144 *
145 * dentry->d_inode->i_mutex: don't care
146 */
147ssize_t
148ext3_listxattr(struct dentry *dentry, char *buffer, size_t size)
149{
150	return ext3_xattr_list(dentry, buffer, size);
151}
152
153static int
154ext3_xattr_check_names(struct ext3_xattr_entry *entry, void *end)
155{
156	while (!IS_LAST_ENTRY(entry)) {
157		struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(entry);
158		if ((void *)next >= end)
159			return -EIO;
160		entry = next;
161	}
162	return 0;
163}
164
165static inline int
166ext3_xattr_check_block(struct buffer_head *bh)
167{
168	int error;
169
170	if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) ||
171	    BHDR(bh)->h_blocks != cpu_to_le32(1))
172		return -EIO;
173	error = ext3_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
174	return error;
175}
176
177static inline int
178ext3_xattr_check_entry(struct ext3_xattr_entry *entry, size_t size)
179{
180	size_t value_size = le32_to_cpu(entry->e_value_size);
181
182	if (entry->e_value_block != 0 || value_size > size ||
183	    le16_to_cpu(entry->e_value_offs) + value_size > size)
184		return -EIO;
185	return 0;
186}
187
188static int
189ext3_xattr_find_entry(struct ext3_xattr_entry **pentry, int name_index,
190		      const char *name, size_t size, int sorted)
191{
192	struct ext3_xattr_entry *entry;
193	size_t name_len;
194	int cmp = 1;
195
196	if (name == NULL)
197		return -EINVAL;
198	name_len = strlen(name);
199	entry = *pentry;
200	for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) {
201		cmp = name_index - entry->e_name_index;
202		if (!cmp)
203			cmp = name_len - entry->e_name_len;
204		if (!cmp)
205			cmp = memcmp(name, entry->e_name, name_len);
206		if (cmp <= 0 && (sorted || cmp == 0))
207			break;
208	}
209	*pentry = entry;
210	if (!cmp && ext3_xattr_check_entry(entry, size))
211			return -EIO;
212	return cmp ? -ENODATA : 0;
213}
214
215static int
216ext3_xattr_block_get(struct inode *inode, int name_index, const char *name,
217		     void *buffer, size_t buffer_size)
218{
219	struct buffer_head *bh = NULL;
220	struct ext3_xattr_entry *entry;
221	size_t size;
222	int error;
223
224	ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
225		  name_index, name, buffer, (long)buffer_size);
226
227	error = -ENODATA;
228	if (!EXT3_I(inode)->i_file_acl)
229		goto cleanup;
230	ea_idebug(inode, "reading block %u", EXT3_I(inode)->i_file_acl);
231	bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
232	if (!bh)
233		goto cleanup;
234	ea_bdebug(bh, "b_count=%d, refcount=%d",
235		atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
236	if (ext3_xattr_check_block(bh)) {
237bad_block:	ext3_error(inode->i_sb, __func__,
238			   "inode %lu: bad block "E3FSBLK, inode->i_ino,
239			   EXT3_I(inode)->i_file_acl);
240		error = -EIO;
241		goto cleanup;
242	}
243	ext3_xattr_cache_insert(bh);
244	entry = BFIRST(bh);
245	error = ext3_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
246	if (error == -EIO)
247		goto bad_block;
248	if (error)
249		goto cleanup;
250	size = le32_to_cpu(entry->e_value_size);
251	if (buffer) {
252		error = -ERANGE;
253		if (size > buffer_size)
254			goto cleanup;
255		memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
256		       size);
257	}
258	error = size;
259
260cleanup:
261	brelse(bh);
262	return error;
263}
264
265static int
266ext3_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
267		     void *buffer, size_t buffer_size)
268{
269	struct ext3_xattr_ibody_header *header;
270	struct ext3_xattr_entry *entry;
271	struct ext3_inode *raw_inode;
272	struct ext3_iloc iloc;
273	size_t size;
274	void *end;
275	int error;
276
277	if (!ext3_test_inode_state(inode, EXT3_STATE_XATTR))
278		return -ENODATA;
279	error = ext3_get_inode_loc(inode, &iloc);
280	if (error)
281		return error;
282	raw_inode = ext3_raw_inode(&iloc);
283	header = IHDR(inode, raw_inode);
284	entry = IFIRST(header);
285	end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
286	error = ext3_xattr_check_names(entry, end);
287	if (error)
288		goto cleanup;
289	error = ext3_xattr_find_entry(&entry, name_index, name,
290				      end - (void *)entry, 0);
291	if (error)
292		goto cleanup;
293	size = le32_to_cpu(entry->e_value_size);
294	if (buffer) {
295		error = -ERANGE;
296		if (size > buffer_size)
297			goto cleanup;
298		memcpy(buffer, (void *)IFIRST(header) +
299		       le16_to_cpu(entry->e_value_offs), size);
300	}
301	error = size;
302
303cleanup:
304	brelse(iloc.bh);
305	return error;
306}
307
308/*
309 * ext3_xattr_get()
310 *
311 * Copy an extended attribute into the buffer
312 * provided, or compute the buffer size required.
313 * Buffer is NULL to compute the size of the buffer required.
314 *
315 * Returns a negative error number on failure, or the number of bytes
316 * used / required on success.
317 */
318int
319ext3_xattr_get(struct inode *inode, int name_index, const char *name,
320	       void *buffer, size_t buffer_size)
321{
322	int error;
323
324	down_read(&EXT3_I(inode)->xattr_sem);
325	error = ext3_xattr_ibody_get(inode, name_index, name, buffer,
326				     buffer_size);
327	if (error == -ENODATA)
328		error = ext3_xattr_block_get(inode, name_index, name, buffer,
329					     buffer_size);
330	up_read(&EXT3_I(inode)->xattr_sem);
331	return error;
332}
333
334static int
335ext3_xattr_list_entries(struct dentry *dentry, struct ext3_xattr_entry *entry,
336			char *buffer, size_t buffer_size)
337{
338	size_t rest = buffer_size;
339
340	for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) {
341		const struct xattr_handler *handler =
342			ext3_xattr_handler(entry->e_name_index);
343
344		if (handler) {
345			size_t size = handler->list(dentry, buffer, rest,
346						    entry->e_name,
347						    entry->e_name_len,
348						    handler->flags);
349			if (buffer) {
350				if (size > rest)
351					return -ERANGE;
352				buffer += size;
353			}
354			rest -= size;
355		}
356	}
357	return buffer_size - rest;
358}
359
360static int
361ext3_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
362{
363	struct inode *inode = dentry->d_inode;
364	struct buffer_head *bh = NULL;
365	int error;
366
367	ea_idebug(inode, "buffer=%p, buffer_size=%ld",
368		  buffer, (long)buffer_size);
369
370	error = 0;
371	if (!EXT3_I(inode)->i_file_acl)
372		goto cleanup;
373	ea_idebug(inode, "reading block %u", EXT3_I(inode)->i_file_acl);
374	bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
375	error = -EIO;
376	if (!bh)
377		goto cleanup;
378	ea_bdebug(bh, "b_count=%d, refcount=%d",
379		atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
380	if (ext3_xattr_check_block(bh)) {
381		ext3_error(inode->i_sb, __func__,
382			   "inode %lu: bad block "E3FSBLK, inode->i_ino,
383			   EXT3_I(inode)->i_file_acl);
384		error = -EIO;
385		goto cleanup;
386	}
387	ext3_xattr_cache_insert(bh);
388	error = ext3_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
389
390cleanup:
391	brelse(bh);
392
393	return error;
394}
395
396static int
397ext3_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
398{
399	struct inode *inode = dentry->d_inode;
400	struct ext3_xattr_ibody_header *header;
401	struct ext3_inode *raw_inode;
402	struct ext3_iloc iloc;
403	void *end;
404	int error;
405
406	if (!ext3_test_inode_state(inode, EXT3_STATE_XATTR))
407		return 0;
408	error = ext3_get_inode_loc(inode, &iloc);
409	if (error)
410		return error;
411	raw_inode = ext3_raw_inode(&iloc);
412	header = IHDR(inode, raw_inode);
413	end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
414	error = ext3_xattr_check_names(IFIRST(header), end);
415	if (error)
416		goto cleanup;
417	error = ext3_xattr_list_entries(dentry, IFIRST(header),
418					buffer, buffer_size);
419
420cleanup:
421	brelse(iloc.bh);
422	return error;
423}
424
425/*
426 * ext3_xattr_list()
427 *
428 * Copy a list of attribute names into the buffer
429 * provided, or compute the buffer size required.
430 * Buffer is NULL to compute the size of the buffer required.
431 *
432 * Returns a negative error number on failure, or the number of bytes
433 * used / required on success.
434 */
435static int
436ext3_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
437{
438	int i_error, b_error;
439
440	down_read(&EXT3_I(dentry->d_inode)->xattr_sem);
441	i_error = ext3_xattr_ibody_list(dentry, buffer, buffer_size);
442	if (i_error < 0) {
443		b_error = 0;
444	} else {
445		if (buffer) {
446			buffer += i_error;
447			buffer_size -= i_error;
448		}
449		b_error = ext3_xattr_block_list(dentry, buffer, buffer_size);
450		if (b_error < 0)
451			i_error = 0;
452	}
453	up_read(&EXT3_I(dentry->d_inode)->xattr_sem);
454	return i_error + b_error;
455}
456
457/*
458 * If the EXT3_FEATURE_COMPAT_EXT_ATTR feature of this file system is
459 * not set, set it.
460 */
461static void ext3_xattr_update_super_block(handle_t *handle,
462					  struct super_block *sb)
463{
464	if (EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR))
465		return;
466
467	if (ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh) == 0) {
468		EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR);
469		ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh);
470	}
471}
472
473/*
474 * Release the xattr block BH: If the reference count is > 1, decrement
475 * it; otherwise free the block.
476 */
477static void
478ext3_xattr_release_block(handle_t *handle, struct inode *inode,
479			 struct buffer_head *bh)
480{
481	struct mb_cache_entry *ce = NULL;
482	int error = 0;
483
484	ce = mb_cache_entry_get(ext3_xattr_cache, bh->b_bdev, bh->b_blocknr);
485	error = ext3_journal_get_write_access(handle, bh);
486	if (error)
487		 goto out;
488
489	lock_buffer(bh);
490
491	if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
492		ea_bdebug(bh, "refcount now=0; freeing");
493		if (ce)
494			mb_cache_entry_free(ce);
495		ext3_free_blocks(handle, inode, bh->b_blocknr, 1);
496		get_bh(bh);
497		ext3_forget(handle, 1, inode, bh, bh->b_blocknr);
498	} else {
499		le32_add_cpu(&BHDR(bh)->h_refcount, -1);
500		error = ext3_journal_dirty_metadata(handle, bh);
501		if (IS_SYNC(inode))
502			handle->h_sync = 1;
503		dquot_free_block(inode, 1);
504		ea_bdebug(bh, "refcount now=%d; releasing",
505			  le32_to_cpu(BHDR(bh)->h_refcount));
506		if (ce)
507			mb_cache_entry_release(ce);
508	}
509	unlock_buffer(bh);
510out:
511	ext3_std_error(inode->i_sb, error);
512	return;
513}
514
515struct ext3_xattr_info {
516	int name_index;
517	const char *name;
518	const void *value;
519	size_t value_len;
520};
521
522struct ext3_xattr_search {
523	struct ext3_xattr_entry *first;
524	void *base;
525	void *end;
526	struct ext3_xattr_entry *here;
527	int not_found;
528};
529
530static int
531ext3_xattr_set_entry(struct ext3_xattr_info *i, struct ext3_xattr_search *s)
532{
533	struct ext3_xattr_entry *last;
534	size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
535
536	/* Compute min_offs and last. */
537	last = s->first;
538	for (; !IS_LAST_ENTRY(last); last = EXT3_XATTR_NEXT(last)) {
539		if (!last->e_value_block && last->e_value_size) {
540			size_t offs = le16_to_cpu(last->e_value_offs);
541			if (offs < min_offs)
542				min_offs = offs;
543		}
544	}
545	free = min_offs - ((void *)last - s->base) - sizeof(__u32);
546	if (!s->not_found) {
547		if (!s->here->e_value_block && s->here->e_value_size) {
548			size_t size = le32_to_cpu(s->here->e_value_size);
549			free += EXT3_XATTR_SIZE(size);
550		}
551		free += EXT3_XATTR_LEN(name_len);
552	}
553	if (i->value) {
554		if (free < EXT3_XATTR_SIZE(i->value_len) ||
555		    free < EXT3_XATTR_LEN(name_len) +
556			   EXT3_XATTR_SIZE(i->value_len))
557			return -ENOSPC;
558	}
559
560	if (i->value && s->not_found) {
561		/* Insert the new name. */
562		size_t size = EXT3_XATTR_LEN(name_len);
563		size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
564		memmove((void *)s->here + size, s->here, rest);
565		memset(s->here, 0, size);
566		s->here->e_name_index = i->name_index;
567		s->here->e_name_len = name_len;
568		memcpy(s->here->e_name, i->name, name_len);
569	} else {
570		if (!s->here->e_value_block && s->here->e_value_size) {
571			void *first_val = s->base + min_offs;
572			size_t offs = le16_to_cpu(s->here->e_value_offs);
573			void *val = s->base + offs;
574			size_t size = EXT3_XATTR_SIZE(
575				le32_to_cpu(s->here->e_value_size));
576
577			if (i->value && size == EXT3_XATTR_SIZE(i->value_len)) {
578				/* The old and the new value have the same
579				   size. Just replace. */
580				s->here->e_value_size =
581					cpu_to_le32(i->value_len);
582				memset(val + size - EXT3_XATTR_PAD, 0,
583				       EXT3_XATTR_PAD); /* Clear pad bytes. */
584				memcpy(val, i->value, i->value_len);
585				return 0;
586			}
587
588			/* Remove the old value. */
589			memmove(first_val + size, first_val, val - first_val);
590			memset(first_val, 0, size);
591			s->here->e_value_size = 0;
592			s->here->e_value_offs = 0;
593			min_offs += size;
594
595			/* Adjust all value offsets. */
596			last = s->first;
597			while (!IS_LAST_ENTRY(last)) {
598				size_t o = le16_to_cpu(last->e_value_offs);
599				if (!last->e_value_block &&
600				    last->e_value_size && o < offs)
601					last->e_value_offs =
602						cpu_to_le16(o + size);
603				last = EXT3_XATTR_NEXT(last);
604			}
605		}
606		if (!i->value) {
607			/* Remove the old name. */
608			size_t size = EXT3_XATTR_LEN(name_len);
609			last = ENTRY((void *)last - size);
610			memmove(s->here, (void *)s->here + size,
611				(void *)last - (void *)s->here + sizeof(__u32));
612			memset(last, 0, size);
613		}
614	}
615
616	if (i->value) {
617		/* Insert the new value. */
618		s->here->e_value_size = cpu_to_le32(i->value_len);
619		if (i->value_len) {
620			size_t size = EXT3_XATTR_SIZE(i->value_len);
621			void *val = s->base + min_offs - size;
622			s->here->e_value_offs = cpu_to_le16(min_offs - size);
623			memset(val + size - EXT3_XATTR_PAD, 0,
624			       EXT3_XATTR_PAD); /* Clear the pad bytes. */
625			memcpy(val, i->value, i->value_len);
626		}
627	}
628	return 0;
629}
630
631struct ext3_xattr_block_find {
632	struct ext3_xattr_search s;
633	struct buffer_head *bh;
634};
635
636static int
637ext3_xattr_block_find(struct inode *inode, struct ext3_xattr_info *i,
638		      struct ext3_xattr_block_find *bs)
639{
640	struct super_block *sb = inode->i_sb;
641	int error;
642
643	ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
644		  i->name_index, i->name, i->value, (long)i->value_len);
645
646	if (EXT3_I(inode)->i_file_acl) {
647		/* The inode already has an extended attribute block. */
648		bs->bh = sb_bread(sb, EXT3_I(inode)->i_file_acl);
649		error = -EIO;
650		if (!bs->bh)
651			goto cleanup;
652		ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
653			atomic_read(&(bs->bh->b_count)),
654			le32_to_cpu(BHDR(bs->bh)->h_refcount));
655		if (ext3_xattr_check_block(bs->bh)) {
656			ext3_error(sb, __func__,
657				"inode %lu: bad block "E3FSBLK, inode->i_ino,
658				EXT3_I(inode)->i_file_acl);
659			error = -EIO;
660			goto cleanup;
661		}
662		/* Find the named attribute. */
663		bs->s.base = BHDR(bs->bh);
664		bs->s.first = BFIRST(bs->bh);
665		bs->s.end = bs->bh->b_data + bs->bh->b_size;
666		bs->s.here = bs->s.first;
667		error = ext3_xattr_find_entry(&bs->s.here, i->name_index,
668					      i->name, bs->bh->b_size, 1);
669		if (error && error != -ENODATA)
670			goto cleanup;
671		bs->s.not_found = error;
672	}
673	error = 0;
674
675cleanup:
676	return error;
677}
678
679static int
680ext3_xattr_block_set(handle_t *handle, struct inode *inode,
681		     struct ext3_xattr_info *i,
682		     struct ext3_xattr_block_find *bs)
683{
684	struct super_block *sb = inode->i_sb;
685	struct buffer_head *new_bh = NULL;
686	struct ext3_xattr_search *s = &bs->s;
687	struct mb_cache_entry *ce = NULL;
688	int error = 0;
689
690#define header(x) ((struct ext3_xattr_header *)(x))
691
692	if (i->value && i->value_len > sb->s_blocksize)
693		return -ENOSPC;
694	if (s->base) {
695		ce = mb_cache_entry_get(ext3_xattr_cache, bs->bh->b_bdev,
696					bs->bh->b_blocknr);
697		error = ext3_journal_get_write_access(handle, bs->bh);
698		if (error)
699			goto cleanup;
700		lock_buffer(bs->bh);
701
702		if (header(s->base)->h_refcount == cpu_to_le32(1)) {
703			if (ce) {
704				mb_cache_entry_free(ce);
705				ce = NULL;
706			}
707			ea_bdebug(bs->bh, "modifying in-place");
708			error = ext3_xattr_set_entry(i, s);
709			if (!error) {
710				if (!IS_LAST_ENTRY(s->first))
711					ext3_xattr_rehash(header(s->base),
712							  s->here);
713				ext3_xattr_cache_insert(bs->bh);
714			}
715			unlock_buffer(bs->bh);
716			if (error == -EIO)
717				goto bad_block;
718			if (!error)
719				error = ext3_journal_dirty_metadata(handle,
720								    bs->bh);
721			if (error)
722				goto cleanup;
723			goto inserted;
724		} else {
725			int offset = (char *)s->here - bs->bh->b_data;
726
727			unlock_buffer(bs->bh);
728			journal_release_buffer(handle, bs->bh);
729
730			if (ce) {
731				mb_cache_entry_release(ce);
732				ce = NULL;
733			}
734			ea_bdebug(bs->bh, "cloning");
735			s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
736			error = -ENOMEM;
737			if (s->base == NULL)
738				goto cleanup;
739			memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
740			s->first = ENTRY(header(s->base)+1);
741			header(s->base)->h_refcount = cpu_to_le32(1);
742			s->here = ENTRY(s->base + offset);
743			s->end = s->base + bs->bh->b_size;
744		}
745	} else {
746		/* Allocate a buffer where we construct the new block. */
747		s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
748		/* assert(header == s->base) */
749		error = -ENOMEM;
750		if (s->base == NULL)
751			goto cleanup;
752		header(s->base)->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC);
753		header(s->base)->h_blocks = cpu_to_le32(1);
754		header(s->base)->h_refcount = cpu_to_le32(1);
755		s->first = ENTRY(header(s->base)+1);
756		s->here = ENTRY(header(s->base)+1);
757		s->end = s->base + sb->s_blocksize;
758	}
759
760	error = ext3_xattr_set_entry(i, s);
761	if (error == -EIO)
762		goto bad_block;
763	if (error)
764		goto cleanup;
765	if (!IS_LAST_ENTRY(s->first))
766		ext3_xattr_rehash(header(s->base), s->here);
767
768inserted:
769	if (!IS_LAST_ENTRY(s->first)) {
770		new_bh = ext3_xattr_cache_find(inode, header(s->base), &ce);
771		if (new_bh) {
772			/* We found an identical block in the cache. */
773			if (new_bh == bs->bh)
774				ea_bdebug(new_bh, "keeping");
775			else {
776				/* The old block is released after updating
777				   the inode. */
778				error = dquot_alloc_block(inode, 1);
779				if (error)
780					goto cleanup;
781				error = ext3_journal_get_write_access(handle,
782								      new_bh);
783				if (error)
784					goto cleanup_dquot;
785				lock_buffer(new_bh);
786				le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
787				ea_bdebug(new_bh, "reusing; refcount now=%d",
788					le32_to_cpu(BHDR(new_bh)->h_refcount));
789				unlock_buffer(new_bh);
790				error = ext3_journal_dirty_metadata(handle,
791								    new_bh);
792				if (error)
793					goto cleanup_dquot;
794			}
795			mb_cache_entry_release(ce);
796			ce = NULL;
797		} else if (bs->bh && s->base == bs->bh->b_data) {
798			/* We were modifying this block in-place. */
799			ea_bdebug(bs->bh, "keeping this block");
800			new_bh = bs->bh;
801			get_bh(new_bh);
802		} else {
803			/* We need to allocate a new block */
804			ext3_fsblk_t goal = ext3_group_first_block_no(sb,
805						EXT3_I(inode)->i_block_group);
806			ext3_fsblk_t block = ext3_new_block(handle, inode,
807							goal, &error);
808			if (error)
809				goto cleanup;
810			ea_idebug(inode, "creating block %d", block);
811
812			new_bh = sb_getblk(sb, block);
813			if (!new_bh) {
814getblk_failed:
815				ext3_free_blocks(handle, inode, block, 1);
816				error = -EIO;
817				goto cleanup;
818			}
819			lock_buffer(new_bh);
820			error = ext3_journal_get_create_access(handle, new_bh);
821			if (error) {
822				unlock_buffer(new_bh);
823				goto getblk_failed;
824			}
825			memcpy(new_bh->b_data, s->base, new_bh->b_size);
826			set_buffer_uptodate(new_bh);
827			unlock_buffer(new_bh);
828			ext3_xattr_cache_insert(new_bh);
829			error = ext3_journal_dirty_metadata(handle, new_bh);
830			if (error)
831				goto cleanup;
832		}
833	}
834
835	/* Update the inode. */
836	EXT3_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
837
838	/* Drop the previous xattr block. */
839	if (bs->bh && bs->bh != new_bh)
840		ext3_xattr_release_block(handle, inode, bs->bh);
841	error = 0;
842
843cleanup:
844	if (ce)
845		mb_cache_entry_release(ce);
846	brelse(new_bh);
847	if (!(bs->bh && s->base == bs->bh->b_data))
848		kfree(s->base);
849
850	return error;
851
852cleanup_dquot:
853	dquot_free_block(inode, 1);
854	goto cleanup;
855
856bad_block:
857	ext3_error(inode->i_sb, __func__,
858		   "inode %lu: bad block "E3FSBLK, inode->i_ino,
859		   EXT3_I(inode)->i_file_acl);
860	goto cleanup;
861
862#undef header
863}
864
865struct ext3_xattr_ibody_find {
866	struct ext3_xattr_search s;
867	struct ext3_iloc iloc;
868};
869
870static int
871ext3_xattr_ibody_find(struct inode *inode, struct ext3_xattr_info *i,
872		      struct ext3_xattr_ibody_find *is)
873{
874	struct ext3_xattr_ibody_header *header;
875	struct ext3_inode *raw_inode;
876	int error;
877
878	if (EXT3_I(inode)->i_extra_isize == 0)
879		return 0;
880	raw_inode = ext3_raw_inode(&is->iloc);
881	header = IHDR(inode, raw_inode);
882	is->s.base = is->s.first = IFIRST(header);
883	is->s.here = is->s.first;
884	is->s.end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
885	if (ext3_test_inode_state(inode, EXT3_STATE_XATTR)) {
886		error = ext3_xattr_check_names(IFIRST(header), is->s.end);
887		if (error)
888			return error;
889		/* Find the named attribute. */
890		error = ext3_xattr_find_entry(&is->s.here, i->name_index,
891					      i->name, is->s.end -
892					      (void *)is->s.base, 0);
893		if (error && error != -ENODATA)
894			return error;
895		is->s.not_found = error;
896	}
897	return 0;
898}
899
900static int
901ext3_xattr_ibody_set(handle_t *handle, struct inode *inode,
902		     struct ext3_xattr_info *i,
903		     struct ext3_xattr_ibody_find *is)
904{
905	struct ext3_xattr_ibody_header *header;
906	struct ext3_xattr_search *s = &is->s;
907	int error;
908
909	if (EXT3_I(inode)->i_extra_isize == 0)
910		return -ENOSPC;
911	error = ext3_xattr_set_entry(i, s);
912	if (error)
913		return error;
914	header = IHDR(inode, ext3_raw_inode(&is->iloc));
915	if (!IS_LAST_ENTRY(s->first)) {
916		header->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC);
917		ext3_set_inode_state(inode, EXT3_STATE_XATTR);
918	} else {
919		header->h_magic = cpu_to_le32(0);
920		ext3_clear_inode_state(inode, EXT3_STATE_XATTR);
921	}
922	return 0;
923}
924
925/*
926 * ext3_xattr_set_handle()
927 *
928 * Create, replace or remove an extended attribute for this inode. Buffer
929 * is NULL to remove an existing extended attribute, and non-NULL to
930 * either replace an existing extended attribute, or create a new extended
931 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
932 * specify that an extended attribute must exist and must not exist
933 * previous to the call, respectively.
934 *
935 * Returns 0, or a negative error number on failure.
936 */
937int
938ext3_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
939		      const char *name, const void *value, size_t value_len,
940		      int flags)
941{
942	struct ext3_xattr_info i = {
943		.name_index = name_index,
944		.name = name,
945		.value = value,
946		.value_len = value_len,
947
948	};
949	struct ext3_xattr_ibody_find is = {
950		.s = { .not_found = -ENODATA, },
951	};
952	struct ext3_xattr_block_find bs = {
953		.s = { .not_found = -ENODATA, },
954	};
955	int error;
956
957	if (!name)
958		return -EINVAL;
959	if (strlen(name) > 255)
960		return -ERANGE;
961	down_write(&EXT3_I(inode)->xattr_sem);
962	error = ext3_get_inode_loc(inode, &is.iloc);
963	if (error)
964		goto cleanup;
965
966	error = ext3_journal_get_write_access(handle, is.iloc.bh);
967	if (error)
968		goto cleanup;
969
970	if (ext3_test_inode_state(inode, EXT3_STATE_NEW)) {
971		struct ext3_inode *raw_inode = ext3_raw_inode(&is.iloc);
972		memset(raw_inode, 0, EXT3_SB(inode->i_sb)->s_inode_size);
973		ext3_clear_inode_state(inode, EXT3_STATE_NEW);
974	}
975
976	error = ext3_xattr_ibody_find(inode, &i, &is);
977	if (error)
978		goto cleanup;
979	if (is.s.not_found)
980		error = ext3_xattr_block_find(inode, &i, &bs);
981	if (error)
982		goto cleanup;
983	if (is.s.not_found && bs.s.not_found) {
984		error = -ENODATA;
985		if (flags & XATTR_REPLACE)
986			goto cleanup;
987		error = 0;
988		if (!value)
989			goto cleanup;
990	} else {
991		error = -EEXIST;
992		if (flags & XATTR_CREATE)
993			goto cleanup;
994	}
995	if (!value) {
996		if (!is.s.not_found)
997			error = ext3_xattr_ibody_set(handle, inode, &i, &is);
998		else if (!bs.s.not_found)
999			error = ext3_xattr_block_set(handle, inode, &i, &bs);
1000	} else {
1001		error = ext3_xattr_ibody_set(handle, inode, &i, &is);
1002		if (!error && !bs.s.not_found) {
1003			i.value = NULL;
1004			error = ext3_xattr_block_set(handle, inode, &i, &bs);
1005		} else if (error == -ENOSPC) {
1006			if (EXT3_I(inode)->i_file_acl && !bs.s.base) {
1007				error = ext3_xattr_block_find(inode, &i, &bs);
1008				if (error)
1009					goto cleanup;
1010			}
1011			error = ext3_xattr_block_set(handle, inode, &i, &bs);
1012			if (error)
1013				goto cleanup;
1014			if (!is.s.not_found) {
1015				i.value = NULL;
1016				error = ext3_xattr_ibody_set(handle, inode, &i,
1017							     &is);
1018			}
1019		}
1020	}
1021	if (!error) {
1022		ext3_xattr_update_super_block(handle, inode->i_sb);
1023		inode->i_ctime = CURRENT_TIME_SEC;
1024		error = ext3_mark_iloc_dirty(handle, inode, &is.iloc);
1025		/*
1026		 * The bh is consumed by ext3_mark_iloc_dirty, even with
1027		 * error != 0.
1028		 */
1029		is.iloc.bh = NULL;
1030		if (IS_SYNC(inode))
1031			handle->h_sync = 1;
1032	}
1033
1034cleanup:
1035	brelse(is.iloc.bh);
1036	brelse(bs.bh);
1037	up_write(&EXT3_I(inode)->xattr_sem);
1038	return error;
1039}
1040
1041/*
1042 * ext3_xattr_set()
1043 *
1044 * Like ext3_xattr_set_handle, but start from an inode. This extended
1045 * attribute modification is a filesystem transaction by itself.
1046 *
1047 * Returns 0, or a negative error number on failure.
1048 */
1049int
1050ext3_xattr_set(struct inode *inode, int name_index, const char *name,
1051	       const void *value, size_t value_len, int flags)
1052{
1053	handle_t *handle;
1054	int error, retries = 0;
1055
1056retry:
1057	handle = ext3_journal_start(inode, EXT3_DATA_TRANS_BLOCKS(inode->i_sb));
1058	if (IS_ERR(handle)) {
1059		error = PTR_ERR(handle);
1060	} else {
1061		int error2;
1062
1063		error = ext3_xattr_set_handle(handle, inode, name_index, name,
1064					      value, value_len, flags);
1065		error2 = ext3_journal_stop(handle);
1066		if (error == -ENOSPC &&
1067		    ext3_should_retry_alloc(inode->i_sb, &retries))
1068			goto retry;
1069		if (error == 0)
1070			error = error2;
1071	}
1072
1073	return error;
1074}
1075
1076/*
1077 * ext3_xattr_delete_inode()
1078 *
1079 * Free extended attribute resources associated with this inode. This
1080 * is called immediately before an inode is freed. We have exclusive
1081 * access to the inode.
1082 */
1083void
1084ext3_xattr_delete_inode(handle_t *handle, struct inode *inode)
1085{
1086	struct buffer_head *bh = NULL;
1087
1088	if (!EXT3_I(inode)->i_file_acl)
1089		goto cleanup;
1090	bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
1091	if (!bh) {
1092		ext3_error(inode->i_sb, __func__,
1093			"inode %lu: block "E3FSBLK" read error", inode->i_ino,
1094			EXT3_I(inode)->i_file_acl);
1095		goto cleanup;
1096	}
1097	if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) ||
1098	    BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1099		ext3_error(inode->i_sb, __func__,
1100			"inode %lu: bad block "E3FSBLK, inode->i_ino,
1101			EXT3_I(inode)->i_file_acl);
1102		goto cleanup;
1103	}
1104	ext3_xattr_release_block(handle, inode, bh);
1105	EXT3_I(inode)->i_file_acl = 0;
1106
1107cleanup:
1108	brelse(bh);
1109}
1110
1111/*
1112 * ext3_xattr_put_super()
1113 *
1114 * This is called when a file system is unmounted.
1115 */
1116void
1117ext3_xattr_put_super(struct super_block *sb)
1118{
1119	mb_cache_shrink(sb->s_bdev);
1120}
1121
1122/*
1123 * ext3_xattr_cache_insert()
1124 *
1125 * Create a new entry in the extended attribute cache, and insert
1126 * it unless such an entry is already in the cache.
1127 *
1128 * Returns 0, or a negative error number on failure.
1129 */
1130static void
1131ext3_xattr_cache_insert(struct buffer_head *bh)
1132{
1133	__u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1134	struct mb_cache_entry *ce;
1135	int error;
1136
1137	ce = mb_cache_entry_alloc(ext3_xattr_cache, GFP_NOFS);
1138	if (!ce) {
1139		ea_bdebug(bh, "out of memory");
1140		return;
1141	}
1142	error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1143	if (error) {
1144		mb_cache_entry_free(ce);
1145		if (error == -EBUSY) {
1146			ea_bdebug(bh, "already in cache");
1147			error = 0;
1148		}
1149	} else {
1150		ea_bdebug(bh, "inserting [%x]", (int)hash);
1151		mb_cache_entry_release(ce);
1152	}
1153}
1154
1155/*
1156 * ext3_xattr_cmp()
1157 *
1158 * Compare two extended attribute blocks for equality.
1159 *
1160 * Returns 0 if the blocks are equal, 1 if they differ, and
1161 * a negative error number on errors.
1162 */
1163static int
1164ext3_xattr_cmp(struct ext3_xattr_header *header1,
1165	       struct ext3_xattr_header *header2)
1166{
1167	struct ext3_xattr_entry *entry1, *entry2;
1168
1169	entry1 = ENTRY(header1+1);
1170	entry2 = ENTRY(header2+1);
1171	while (!IS_LAST_ENTRY(entry1)) {
1172		if (IS_LAST_ENTRY(entry2))
1173			return 1;
1174		if (entry1->e_hash != entry2->e_hash ||
1175		    entry1->e_name_index != entry2->e_name_index ||
1176		    entry1->e_name_len != entry2->e_name_len ||
1177		    entry1->e_value_size != entry2->e_value_size ||
1178		    memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1179			return 1;
1180		if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1181			return -EIO;
1182		if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1183			   (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1184			   le32_to_cpu(entry1->e_value_size)))
1185			return 1;
1186
1187		entry1 = EXT3_XATTR_NEXT(entry1);
1188		entry2 = EXT3_XATTR_NEXT(entry2);
1189	}
1190	if (!IS_LAST_ENTRY(entry2))
1191		return 1;
1192	return 0;
1193}
1194
1195/*
1196 * ext3_xattr_cache_find()
1197 *
1198 * Find an identical extended attribute block.
1199 *
1200 * Returns a pointer to the block found, or NULL if such a block was
1201 * not found or an error occurred.
1202 */
1203static struct buffer_head *
1204ext3_xattr_cache_find(struct inode *inode, struct ext3_xattr_header *header,
1205		      struct mb_cache_entry **pce)
1206{
1207	__u32 hash = le32_to_cpu(header->h_hash);
1208	struct mb_cache_entry *ce;
1209
1210	if (!header->h_hash)
1211		return NULL;  /* never share */
1212	ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1213again:
1214	ce = mb_cache_entry_find_first(ext3_xattr_cache, inode->i_sb->s_bdev,
1215				       hash);
1216	while (ce) {
1217		struct buffer_head *bh;
1218
1219		if (IS_ERR(ce)) {
1220			if (PTR_ERR(ce) == -EAGAIN)
1221				goto again;
1222			break;
1223		}
1224		bh = sb_bread(inode->i_sb, ce->e_block);
1225		if (!bh) {
1226			ext3_error(inode->i_sb, __func__,
1227				"inode %lu: block %lu read error",
1228				inode->i_ino, (unsigned long) ce->e_block);
1229		} else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1230				EXT3_XATTR_REFCOUNT_MAX) {
1231			ea_idebug(inode, "block %lu refcount %d>=%d",
1232				  (unsigned long) ce->e_block,
1233				  le32_to_cpu(BHDR(bh)->h_refcount),
1234					  EXT3_XATTR_REFCOUNT_MAX);
1235		} else if (ext3_xattr_cmp(header, BHDR(bh)) == 0) {
1236			*pce = ce;
1237			return bh;
1238		}
1239		brelse(bh);
1240		ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1241	}
1242	return NULL;
1243}
1244
1245#define NAME_HASH_SHIFT 5
1246#define VALUE_HASH_SHIFT 16
1247
1248/*
1249 * ext3_xattr_hash_entry()
1250 *
1251 * Compute the hash of an extended attribute.
1252 */
1253static inline void ext3_xattr_hash_entry(struct ext3_xattr_header *header,
1254					 struct ext3_xattr_entry *entry)
1255{
1256	__u32 hash = 0;
1257	char *name = entry->e_name;
1258	int n;
1259
1260	for (n=0; n < entry->e_name_len; n++) {
1261		hash = (hash << NAME_HASH_SHIFT) ^
1262		       (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1263		       *name++;
1264	}
1265
1266	if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1267		__le32 *value = (__le32 *)((char *)header +
1268			le16_to_cpu(entry->e_value_offs));
1269		for (n = (le32_to_cpu(entry->e_value_size) +
1270		     EXT3_XATTR_ROUND) >> EXT3_XATTR_PAD_BITS; n; n--) {
1271			hash = (hash << VALUE_HASH_SHIFT) ^
1272			       (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1273			       le32_to_cpu(*value++);
1274		}
1275	}
1276	entry->e_hash = cpu_to_le32(hash);
1277}
1278
1279#undef NAME_HASH_SHIFT
1280#undef VALUE_HASH_SHIFT
1281
1282#define BLOCK_HASH_SHIFT 16
1283
1284/*
1285 * ext3_xattr_rehash()
1286 *
1287 * Re-compute the extended attribute hash value after an entry has changed.
1288 */
1289static void ext3_xattr_rehash(struct ext3_xattr_header *header,
1290			      struct ext3_xattr_entry *entry)
1291{
1292	struct ext3_xattr_entry *here;
1293	__u32 hash = 0;
1294
1295	ext3_xattr_hash_entry(header, entry);
1296	here = ENTRY(header+1);
1297	while (!IS_LAST_ENTRY(here)) {
1298		if (!here->e_hash) {
1299			/* Block is not shared if an entry's hash value == 0 */
1300			hash = 0;
1301			break;
1302		}
1303		hash = (hash << BLOCK_HASH_SHIFT) ^
1304		       (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1305		       le32_to_cpu(here->e_hash);
1306		here = EXT3_XATTR_NEXT(here);
1307	}
1308	header->h_hash = cpu_to_le32(hash);
1309}
1310
1311#undef BLOCK_HASH_SHIFT
1312
1313int __init
1314init_ext3_xattr(void)
1315{
1316	ext3_xattr_cache = mb_cache_create("ext3_xattr", 6);
1317	if (!ext3_xattr_cache)
1318		return -ENOMEM;
1319	return 0;
1320}
1321
1322void
1323exit_ext3_xattr(void)
1324{
1325	if (ext3_xattr_cache)
1326		mb_cache_destroy(ext3_xattr_cache);
1327	ext3_xattr_cache = NULL;
1328}
1329