Linux Audio

Check our new training course

Embedded Linux training

Mar 31-Apr 8, 2025
Register
Loading...
v3.1
 
   1/*
   2 * linux/fs/ext4/xattr.c
   3 *
   4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
   5 *
   6 * Fix by Harrison Xing <harrison@mountainviewdata.com>.
   7 * Ext4 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 * EXT4_I(inode)->i_file_acl is protected by EXT4_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/mbcache.h>
  57#include <linux/quotaops.h>
  58#include <linux/rwsem.h>
  59#include "ext4_jbd2.h"
  60#include "ext4.h"
  61#include "xattr.h"
  62#include "acl.h"
  63
  64#define BHDR(bh) ((struct ext4_xattr_header *)((bh)->b_data))
  65#define ENTRY(ptr) ((struct ext4_xattr_entry *)(ptr))
  66#define BFIRST(bh) ENTRY(BHDR(bh)+1)
  67#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0)
  68
  69#ifdef EXT4_XATTR_DEBUG
  70# define ea_idebug(inode, f...) do { \
  71		printk(KERN_DEBUG "inode %s:%lu: ", \
  72			inode->i_sb->s_id, inode->i_ino); \
  73		printk(f); \
  74		printk("\n"); \
  75	} while (0)
  76# define ea_bdebug(bh, f...) do { \
  77		char b[BDEVNAME_SIZE]; \
  78		printk(KERN_DEBUG "block %s:%lu: ", \
  79			bdevname(bh->b_bdev, b), \
  80			(unsigned long) bh->b_blocknr); \
  81		printk(f); \
  82		printk("\n"); \
  83	} while (0)
  84#else
  85# define ea_idebug(f...)
  86# define ea_bdebug(f...)
  87#endif
  88
  89static void ext4_xattr_cache_insert(struct buffer_head *);
  90static struct buffer_head *ext4_xattr_cache_find(struct inode *,
  91						 struct ext4_xattr_header *,
  92						 struct mb_cache_entry **);
  93static void ext4_xattr_rehash(struct ext4_xattr_header *,
  94			      struct ext4_xattr_entry *);
  95static int ext4_xattr_list(struct dentry *dentry, char *buffer,
  96			   size_t buffer_size);
  97
  98static struct mb_cache *ext4_xattr_cache;
  99
 100static const struct xattr_handler *ext4_xattr_handler_map[] = {
 101	[EXT4_XATTR_INDEX_USER]		     = &ext4_xattr_user_handler,
 102#ifdef CONFIG_EXT4_FS_POSIX_ACL
 103	[EXT4_XATTR_INDEX_POSIX_ACL_ACCESS]  = &ext4_xattr_acl_access_handler,
 104	[EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext4_xattr_acl_default_handler,
 105#endif
 106	[EXT4_XATTR_INDEX_TRUSTED]	     = &ext4_xattr_trusted_handler,
 107#ifdef CONFIG_EXT4_FS_SECURITY
 108	[EXT4_XATTR_INDEX_SECURITY]	     = &ext4_xattr_security_handler,
 109#endif
 110};
 111
 112const struct xattr_handler *ext4_xattr_handlers[] = {
 113	&ext4_xattr_user_handler,
 114	&ext4_xattr_trusted_handler,
 115#ifdef CONFIG_EXT4_FS_POSIX_ACL
 116	&ext4_xattr_acl_access_handler,
 117	&ext4_xattr_acl_default_handler,
 118#endif
 119#ifdef CONFIG_EXT4_FS_SECURITY
 120	&ext4_xattr_security_handler,
 121#endif
 122	NULL
 123};
 124
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 125static inline const struct xattr_handler *
 126ext4_xattr_handler(int name_index)
 127{
 128	const struct xattr_handler *handler = NULL;
 129
 130	if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
 131		handler = ext4_xattr_handler_map[name_index];
 132	return handler;
 133}
 134
 135/*
 136 * Inode operation listxattr()
 137 *
 138 * dentry->d_inode->i_mutex: don't care
 139 */
 140ssize_t
 141ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
 142{
 143	return ext4_xattr_list(dentry, buffer, size);
 144}
 145
 146static int
 147ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
 
 148{
 149	while (!IS_LAST_ENTRY(entry)) {
 150		struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
 
 
 
 151		if ((void *)next >= end)
 152			return -EIO;
 153		entry = next;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 154	}
 
 155	return 0;
 156}
 157
 158static inline int
 159ext4_xattr_check_block(struct buffer_head *bh)
 
 160{
 161	int error;
 
 
 
 162
 163	if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
 164	    BHDR(bh)->h_blocks != cpu_to_le32(1))
 165		return -EIO;
 166	error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
 
 
 
 
 
 
 
 
 
 
 
 167	return error;
 168}
 169
 170static inline int
 171ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
 
 
 
 
 
 172{
 173	size_t value_size = le32_to_cpu(entry->e_value_size);
 174
 175	if (entry->e_value_block != 0 || value_size > size ||
 176	    le16_to_cpu(entry->e_value_offs) + value_size > size)
 177		return -EIO;
 178	return 0;
 
 
 
 
 
 179}
 180
 
 
 
 181static int
 182ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
 183		      const char *name, size_t size, int sorted)
 184{
 185	struct ext4_xattr_entry *entry;
 186	size_t name_len;
 187	int cmp = 1;
 188
 189	if (name == NULL)
 190		return -EINVAL;
 191	name_len = strlen(name);
 192	entry = *pentry;
 193	for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
 
 
 
 
 194		cmp = name_index - entry->e_name_index;
 195		if (!cmp)
 196			cmp = name_len - entry->e_name_len;
 197		if (!cmp)
 198			cmp = memcmp(name, entry->e_name, name_len);
 199		if (cmp <= 0 && (sorted || cmp == 0))
 200			break;
 201	}
 202	*pentry = entry;
 203	if (!cmp && ext4_xattr_check_entry(entry, size))
 204			return -EIO;
 205	return cmp ? -ENODATA : 0;
 206}
 207
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 208static int
 209ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
 210		     void *buffer, size_t buffer_size)
 211{
 212	struct buffer_head *bh = NULL;
 213	struct ext4_xattr_entry *entry;
 214	size_t size;
 
 215	int error;
 
 216
 217	ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
 218		  name_index, name, buffer, (long)buffer_size);
 219
 220	error = -ENODATA;
 221	if (!EXT4_I(inode)->i_file_acl)
 222		goto cleanup;
 223	ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
 
 224	bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
 225	if (!bh)
 226		goto cleanup;
 227	ea_bdebug(bh, "b_count=%d, refcount=%d",
 228		atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
 229	if (ext4_xattr_check_block(bh)) {
 230bad_block:
 231		EXT4_ERROR_INODE(inode, "bad block %llu",
 232				 EXT4_I(inode)->i_file_acl);
 233		error = -EIO;
 234		goto cleanup;
 235	}
 236	ext4_xattr_cache_insert(bh);
 237	entry = BFIRST(bh);
 238	error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
 239	if (error == -EIO)
 240		goto bad_block;
 241	if (error)
 242		goto cleanup;
 243	size = le32_to_cpu(entry->e_value_size);
 
 
 
 244	if (buffer) {
 245		error = -ERANGE;
 246		if (size > buffer_size)
 247			goto cleanup;
 248		memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
 249		       size);
 
 
 
 
 
 
 
 
 
 
 
 250	}
 251	error = size;
 252
 253cleanup:
 254	brelse(bh);
 255	return error;
 256}
 257
 258static int
 259ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
 260		     void *buffer, size_t buffer_size)
 261{
 262	struct ext4_xattr_ibody_header *header;
 263	struct ext4_xattr_entry *entry;
 264	struct ext4_inode *raw_inode;
 265	struct ext4_iloc iloc;
 266	size_t size;
 267	void *end;
 268	int error;
 269
 270	if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
 271		return -ENODATA;
 272	error = ext4_get_inode_loc(inode, &iloc);
 273	if (error)
 274		return error;
 275	raw_inode = ext4_raw_inode(&iloc);
 276	header = IHDR(inode, raw_inode);
 277	entry = IFIRST(header);
 278	end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
 279	error = ext4_xattr_check_names(entry, end);
 280	if (error)
 281		goto cleanup;
 282	error = ext4_xattr_find_entry(&entry, name_index, name,
 283				      end - (void *)entry, 0);
 284	if (error)
 285		goto cleanup;
 286	size = le32_to_cpu(entry->e_value_size);
 
 
 
 287	if (buffer) {
 288		error = -ERANGE;
 289		if (size > buffer_size)
 290			goto cleanup;
 291		memcpy(buffer, (void *)IFIRST(header) +
 292		       le16_to_cpu(entry->e_value_offs), size);
 
 
 
 
 
 
 
 
 
 
 
 293	}
 294	error = size;
 295
 296cleanup:
 297	brelse(iloc.bh);
 298	return error;
 299}
 300
 301/*
 302 * ext4_xattr_get()
 303 *
 304 * Copy an extended attribute into the buffer
 305 * provided, or compute the buffer size required.
 306 * Buffer is NULL to compute the size of the buffer required.
 307 *
 308 * Returns a negative error number on failure, or the number of bytes
 309 * used / required on success.
 310 */
 311int
 312ext4_xattr_get(struct inode *inode, int name_index, const char *name,
 313	       void *buffer, size_t buffer_size)
 314{
 315	int error;
 316
 
 
 
 
 
 
 317	down_read(&EXT4_I(inode)->xattr_sem);
 318	error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
 319				     buffer_size);
 320	if (error == -ENODATA)
 321		error = ext4_xattr_block_get(inode, name_index, name, buffer,
 322					     buffer_size);
 323	up_read(&EXT4_I(inode)->xattr_sem);
 324	return error;
 325}
 326
 327static int
 328ext4_xattr_list_entries(struct dentry *dentry, struct ext4_xattr_entry *entry,
 329			char *buffer, size_t buffer_size)
 330{
 331	size_t rest = buffer_size;
 332
 333	for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
 334		const struct xattr_handler *handler =
 335			ext4_xattr_handler(entry->e_name_index);
 336
 337		if (handler) {
 338			size_t size = handler->list(dentry, buffer, rest,
 339						    entry->e_name,
 340						    entry->e_name_len,
 341						    handler->flags);
 342			if (buffer) {
 343				if (size > rest)
 344					return -ERANGE;
 345				buffer += size;
 
 
 
 
 346			}
 347			rest -= size;
 348		}
 349	}
 350	return buffer_size - rest;
 351}
 352
 353static int
 354ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
 355{
 356	struct inode *inode = dentry->d_inode;
 357	struct buffer_head *bh = NULL;
 358	int error;
 359
 360	ea_idebug(inode, "buffer=%p, buffer_size=%ld",
 361		  buffer, (long)buffer_size);
 362
 363	error = 0;
 364	if (!EXT4_I(inode)->i_file_acl)
 365		goto cleanup;
 366	ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
 
 367	bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
 368	error = -EIO;
 369	if (!bh)
 370		goto cleanup;
 371	ea_bdebug(bh, "b_count=%d, refcount=%d",
 372		atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
 373	if (ext4_xattr_check_block(bh)) {
 374		EXT4_ERROR_INODE(inode, "bad block %llu",
 375				 EXT4_I(inode)->i_file_acl);
 376		error = -EIO;
 377		goto cleanup;
 378	}
 379	ext4_xattr_cache_insert(bh);
 380	error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
 381
 382cleanup:
 383	brelse(bh);
 384
 385	return error;
 386}
 387
 388static int
 389ext4_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
 390{
 391	struct inode *inode = dentry->d_inode;
 392	struct ext4_xattr_ibody_header *header;
 393	struct ext4_inode *raw_inode;
 394	struct ext4_iloc iloc;
 395	void *end;
 396	int error;
 397
 398	if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
 399		return 0;
 400	error = ext4_get_inode_loc(inode, &iloc);
 401	if (error)
 402		return error;
 403	raw_inode = ext4_raw_inode(&iloc);
 404	header = IHDR(inode, raw_inode);
 405	end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
 406	error = ext4_xattr_check_names(IFIRST(header), end);
 407	if (error)
 408		goto cleanup;
 409	error = ext4_xattr_list_entries(dentry, IFIRST(header),
 410					buffer, buffer_size);
 411
 412cleanup:
 413	brelse(iloc.bh);
 414	return error;
 415}
 416
 417/*
 418 * ext4_xattr_list()
 
 
 419 *
 420 * Copy a list of attribute names into the buffer
 421 * provided, or compute the buffer size required.
 422 * Buffer is NULL to compute the size of the buffer required.
 423 *
 424 * Returns a negative error number on failure, or the number of bytes
 425 * used / required on success.
 426 */
 427static int
 428ext4_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
 429{
 430	int ret, ret2;
 431
 432	down_read(&EXT4_I(dentry->d_inode)->xattr_sem);
 433	ret = ret2 = ext4_xattr_ibody_list(dentry, buffer, buffer_size);
 434	if (ret < 0)
 435		goto errout;
 436	if (buffer) {
 437		buffer += ret;
 438		buffer_size -= ret;
 439	}
 440	ret = ext4_xattr_block_list(dentry, buffer, buffer_size);
 441	if (ret < 0)
 442		goto errout;
 443	ret += ret2;
 444errout:
 445	up_read(&EXT4_I(dentry->d_inode)->xattr_sem);
 446	return ret;
 447}
 448
 449/*
 450 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
 451 * not set, set it.
 452 */
 453static void ext4_xattr_update_super_block(handle_t *handle,
 454					  struct super_block *sb)
 455{
 456	if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
 457		return;
 458
 
 459	if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
 460		EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
 461		ext4_handle_dirty_super(handle, sb);
 462	}
 463}
 464
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 465/*
 466 * Release the xattr block BH: If the reference count is > 1, decrement
 467 * it; otherwise free the block.
 468 */
 469static void
 470ext4_xattr_release_block(handle_t *handle, struct inode *inode,
 471			 struct buffer_head *bh)
 
 
 472{
 473	struct mb_cache_entry *ce = NULL;
 
 474	int error = 0;
 475
 476	ce = mb_cache_entry_get(ext4_xattr_cache, bh->b_bdev, bh->b_blocknr);
 477	error = ext4_journal_get_write_access(handle, bh);
 478	if (error)
 479		goto out;
 480
 481	lock_buffer(bh);
 482	if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
 
 
 483		ea_bdebug(bh, "refcount now=0; freeing");
 484		if (ce)
 485			mb_cache_entry_free(ce);
 
 
 
 
 
 486		get_bh(bh);
 
 
 
 
 
 
 
 
 
 487		ext4_free_blocks(handle, inode, bh, 0, 1,
 488				 EXT4_FREE_BLOCKS_METADATA |
 489				 EXT4_FREE_BLOCKS_FORGET);
 490	} else {
 491		le32_add_cpu(&BHDR(bh)->h_refcount, -1);
 492		error = ext4_handle_dirty_metadata(handle, inode, bh);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 493		if (IS_SYNC(inode))
 494			ext4_handle_sync(handle);
 495		dquot_free_block(inode, 1);
 496		ea_bdebug(bh, "refcount now=%d; releasing",
 497			  le32_to_cpu(BHDR(bh)->h_refcount));
 498		if (ce)
 499			mb_cache_entry_release(ce);
 500	}
 501	unlock_buffer(bh);
 502out:
 503	ext4_std_error(inode->i_sb, error);
 504	return;
 505}
 506
 507/*
 508 * Find the available free space for EAs. This also returns the total number of
 509 * bytes used by EA entries.
 510 */
 511static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
 512				    size_t *min_offs, void *base, int *total)
 513{
 514	for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
 515		*total += EXT4_XATTR_LEN(last->e_name_len);
 516		if (!last->e_value_block && last->e_value_size) {
 517			size_t offs = le16_to_cpu(last->e_value_offs);
 518			if (offs < *min_offs)
 519				*min_offs = offs;
 520		}
 
 
 521	}
 522	return (*min_offs - ((void *)last - base) - sizeof(__u32));
 523}
 524
 525struct ext4_xattr_info {
 526	int name_index;
 527	const char *name;
 528	const void *value;
 529	size_t value_len;
 530};
 
 
 
 
 
 
 
 531
 532struct ext4_xattr_search {
 533	struct ext4_xattr_entry *first;
 534	void *base;
 535	void *end;
 536	struct ext4_xattr_entry *here;
 537	int not_found;
 538};
 
 
 
 
 
 
 
 
 
 
 
 539
 540static int
 541ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 542{
 543	struct ext4_xattr_entry *last;
 544	size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 545
 546	/* Compute min_offs and last. */
 547	last = s->first;
 548	for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
 549		if (!last->e_value_block && last->e_value_size) {
 550			size_t offs = le16_to_cpu(last->e_value_offs);
 551			if (offs < min_offs)
 552				min_offs = offs;
 553		}
 554	}
 555	free = min_offs - ((void *)last - s->base) - sizeof(__u32);
 556	if (!s->not_found) {
 557		if (!s->here->e_value_block && s->here->e_value_size) {
 558			size_t size = le32_to_cpu(s->here->e_value_size);
 559			free += EXT4_XATTR_SIZE(size);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 560		}
 561		free += EXT4_XATTR_LEN(name_len);
 562	}
 563	if (i->value) {
 564		if (free < EXT4_XATTR_SIZE(i->value_len) ||
 565		    free < EXT4_XATTR_LEN(name_len) +
 566			   EXT4_XATTR_SIZE(i->value_len))
 567			return -ENOSPC;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 568	}
 569
 570	if (i->value && s->not_found) {
 571		/* Insert the new name. */
 572		size_t size = EXT4_XATTR_LEN(name_len);
 573		size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
 574		memmove((void *)s->here + size, s->here, rest);
 575		memset(s->here, 0, size);
 576		s->here->e_name_index = i->name_index;
 577		s->here->e_name_len = name_len;
 578		memcpy(s->here->e_name, i->name, name_len);
 
 
 
 
 
 
 
 
 
 579	} else {
 580		if (!s->here->e_value_block && s->here->e_value_size) {
 581			void *first_val = s->base + min_offs;
 582			size_t offs = le16_to_cpu(s->here->e_value_offs);
 583			void *val = s->base + offs;
 584			size_t size = EXT4_XATTR_SIZE(
 585				le32_to_cpu(s->here->e_value_size));
 586
 587			if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
 588				/* The old and the new value have the same
 589				   size. Just replace. */
 590				s->here->e_value_size =
 591					cpu_to_le32(i->value_len);
 592				memset(val + size - EXT4_XATTR_PAD, 0,
 593				       EXT4_XATTR_PAD); /* Clear pad bytes. */
 594				memcpy(val, i->value, i->value_len);
 595				return 0;
 596			}
 597
 598			/* Remove the old value. */
 599			memmove(first_val + size, first_val, val - first_val);
 600			memset(first_val, 0, size);
 601			s->here->e_value_size = 0;
 602			s->here->e_value_offs = 0;
 603			min_offs += size;
 604
 605			/* Adjust all value offsets. */
 606			last = s->first;
 607			while (!IS_LAST_ENTRY(last)) {
 608				size_t o = le16_to_cpu(last->e_value_offs);
 609				if (!last->e_value_block &&
 610				    last->e_value_size && o < offs)
 611					last->e_value_offs =
 612						cpu_to_le16(o + size);
 613				last = EXT4_XATTR_NEXT(last);
 614			}
 615		}
 616		if (!i->value) {
 617			/* Remove the old name. */
 618			size_t size = EXT4_XATTR_LEN(name_len);
 619			last = ENTRY((void *)last - size);
 620			memmove(s->here, (void *)s->here + size,
 621				(void *)last - (void *)s->here + sizeof(__u32));
 622			memset(last, 0, size);
 623		}
 624	}
 625
 
 626	if (i->value) {
 627		/* Insert the new value. */
 628		s->here->e_value_size = cpu_to_le32(i->value_len);
 629		if (i->value_len) {
 630			size_t size = EXT4_XATTR_SIZE(i->value_len);
 631			void *val = s->base + min_offs - size;
 632			s->here->e_value_offs = cpu_to_le16(min_offs - size);
 633			memset(val + size - EXT4_XATTR_PAD, 0,
 634			       EXT4_XATTR_PAD); /* Clear the pad bytes. */
 635			memcpy(val, i->value, i->value_len);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 636		}
 
 637	}
 638	return 0;
 
 
 
 
 
 
 
 
 639}
 640
 641struct ext4_xattr_block_find {
 642	struct ext4_xattr_search s;
 643	struct buffer_head *bh;
 644};
 645
 646static int
 647ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
 648		      struct ext4_xattr_block_find *bs)
 649{
 650	struct super_block *sb = inode->i_sb;
 651	int error;
 652
 653	ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
 654		  i->name_index, i->name, i->value, (long)i->value_len);
 655
 656	if (EXT4_I(inode)->i_file_acl) {
 657		/* The inode already has an extended attribute block. */
 658		bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
 659		error = -EIO;
 660		if (!bs->bh)
 661			goto cleanup;
 662		ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
 663			atomic_read(&(bs->bh->b_count)),
 664			le32_to_cpu(BHDR(bs->bh)->h_refcount));
 665		if (ext4_xattr_check_block(bs->bh)) {
 666			EXT4_ERROR_INODE(inode, "bad block %llu",
 667					 EXT4_I(inode)->i_file_acl);
 668			error = -EIO;
 669			goto cleanup;
 670		}
 671		/* Find the named attribute. */
 672		bs->s.base = BHDR(bs->bh);
 673		bs->s.first = BFIRST(bs->bh);
 674		bs->s.end = bs->bh->b_data + bs->bh->b_size;
 675		bs->s.here = bs->s.first;
 676		error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
 677					      i->name, bs->bh->b_size, 1);
 678		if (error && error != -ENODATA)
 679			goto cleanup;
 680		bs->s.not_found = error;
 681	}
 682	error = 0;
 683
 684cleanup:
 685	return error;
 686}
 687
 688static int
 689ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 690		     struct ext4_xattr_info *i,
 691		     struct ext4_xattr_block_find *bs)
 692{
 693	struct super_block *sb = inode->i_sb;
 694	struct buffer_head *new_bh = NULL;
 695	struct ext4_xattr_search *s = &bs->s;
 
 696	struct mb_cache_entry *ce = NULL;
 697	int error = 0;
 
 
 
 
 
 698
 699#define header(x) ((struct ext4_xattr_header *)(x))
 700
 701	if (i->value && i->value_len > sb->s_blocksize)
 702		return -ENOSPC;
 703	if (s->base) {
 704		ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
 705					bs->bh->b_blocknr);
 706		error = ext4_journal_get_write_access(handle, bs->bh);
 707		if (error)
 708			goto cleanup;
 709		lock_buffer(bs->bh);
 710
 711		if (header(s->base)->h_refcount == cpu_to_le32(1)) {
 712			if (ce) {
 713				mb_cache_entry_free(ce);
 714				ce = NULL;
 715			}
 
 
 
 
 
 
 716			ea_bdebug(bs->bh, "modifying in-place");
 717			error = ext4_xattr_set_entry(i, s);
 718			if (!error) {
 719				if (!IS_LAST_ENTRY(s->first))
 720					ext4_xattr_rehash(header(s->base),
 721							  s->here);
 722				ext4_xattr_cache_insert(bs->bh);
 723			}
 724			unlock_buffer(bs->bh);
 725			if (error == -EIO)
 726				goto bad_block;
 727			if (!error)
 728				error = ext4_handle_dirty_metadata(handle,
 729								   inode,
 730								   bs->bh);
 731			if (error)
 732				goto cleanup;
 733			goto inserted;
 734		} else {
 735			int offset = (char *)s->here - bs->bh->b_data;
 736
 737			unlock_buffer(bs->bh);
 738			ext4_handle_release_buffer(handle, bs->bh);
 739			if (ce) {
 740				mb_cache_entry_release(ce);
 741				ce = NULL;
 742			}
 743			ea_bdebug(bs->bh, "cloning");
 744			s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
 745			error = -ENOMEM;
 746			if (s->base == NULL)
 747				goto cleanup;
 748			memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
 749			s->first = ENTRY(header(s->base)+1);
 750			header(s->base)->h_refcount = cpu_to_le32(1);
 751			s->here = ENTRY(s->base + offset);
 752			s->end = s->base + bs->bh->b_size;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 753		}
 754	} else {
 755		/* Allocate a buffer where we construct the new block. */
 756		s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
 757		/* assert(header == s->base) */
 758		error = -ENOMEM;
 759		if (s->base == NULL)
 760			goto cleanup;
 761		header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
 762		header(s->base)->h_blocks = cpu_to_le32(1);
 763		header(s->base)->h_refcount = cpu_to_le32(1);
 764		s->first = ENTRY(header(s->base)+1);
 765		s->here = ENTRY(header(s->base)+1);
 766		s->end = s->base + sb->s_blocksize;
 767	}
 768
 769	error = ext4_xattr_set_entry(i, s);
 770	if (error == -EIO)
 771		goto bad_block;
 772	if (error)
 773		goto cleanup;
 774	if (!IS_LAST_ENTRY(s->first))
 775		ext4_xattr_rehash(header(s->base), s->here);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 776
 777inserted:
 778	if (!IS_LAST_ENTRY(s->first)) {
 779		new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
 
 780		if (new_bh) {
 781			/* We found an identical block in the cache. */
 782			if (new_bh == bs->bh)
 783				ea_bdebug(new_bh, "keeping");
 784			else {
 
 
 
 
 785				/* The old block is released after updating
 786				   the inode. */
 787				error = dquot_alloc_block(inode, 1);
 
 788				if (error)
 789					goto cleanup;
 
 790				error = ext4_journal_get_write_access(handle,
 791								      new_bh);
 792				if (error)
 793					goto cleanup_dquot;
 794				lock_buffer(new_bh);
 795				le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 796				ea_bdebug(new_bh, "reusing; refcount now=%d",
 797					le32_to_cpu(BHDR(new_bh)->h_refcount));
 
 798				unlock_buffer(new_bh);
 799				error = ext4_handle_dirty_metadata(handle,
 800								   inode,
 801								   new_bh);
 802				if (error)
 803					goto cleanup_dquot;
 804			}
 805			mb_cache_entry_release(ce);
 
 806			ce = NULL;
 807		} else if (bs->bh && s->base == bs->bh->b_data) {
 808			/* We were modifying this block in-place. */
 809			ea_bdebug(bs->bh, "keeping this block");
 
 810			new_bh = bs->bh;
 811			get_bh(new_bh);
 812		} else {
 813			/* We need to allocate a new block */
 814			ext4_fsblk_t goal, block;
 815
 
 
 816			goal = ext4_group_first_block_no(sb,
 817						EXT4_I(inode)->i_block_group);
 818
 819			/* non-extent files can't have physical blocks past 2^32 */
 820			if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
 821				goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
 822
 823			block = ext4_new_meta_blocks(handle, inode, goal, 0,
 824						     NULL, &error);
 825			if (error)
 826				goto cleanup;
 827
 828			if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
 829				BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
 830
 831			ea_idebug(inode, "creating block %d", block);
 
 832
 833			new_bh = sb_getblk(sb, block);
 834			if (!new_bh) {
 
 835getblk_failed:
 836				ext4_free_blocks(handle, inode, NULL, block, 1,
 837						 EXT4_FREE_BLOCKS_METADATA);
 838				error = -EIO;
 839				goto cleanup;
 840			}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 841			lock_buffer(new_bh);
 842			error = ext4_journal_get_create_access(handle, new_bh);
 843			if (error) {
 844				unlock_buffer(new_bh);
 
 845				goto getblk_failed;
 846			}
 847			memcpy(new_bh->b_data, s->base, new_bh->b_size);
 
 848			set_buffer_uptodate(new_bh);
 849			unlock_buffer(new_bh);
 850			ext4_xattr_cache_insert(new_bh);
 851			error = ext4_handle_dirty_metadata(handle,
 852							   inode, new_bh);
 853			if (error)
 854				goto cleanup;
 855		}
 856	}
 857
 
 
 
 858	/* Update the inode. */
 859	EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
 860
 861	/* Drop the previous xattr block. */
 862	if (bs->bh && bs->bh != new_bh)
 863		ext4_xattr_release_block(handle, inode, bs->bh);
 
 
 
 
 
 
 864	error = 0;
 865
 866cleanup:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 867	if (ce)
 868		mb_cache_entry_release(ce);
 869	brelse(new_bh);
 870	if (!(bs->bh && s->base == bs->bh->b_data))
 871		kfree(s->base);
 872
 873	return error;
 874
 875cleanup_dquot:
 876	dquot_free_block(inode, 1);
 877	goto cleanup;
 878
 879bad_block:
 880	EXT4_ERROR_INODE(inode, "bad block %llu",
 881			 EXT4_I(inode)->i_file_acl);
 882	goto cleanup;
 883
 884#undef header
 885}
 886
 887struct ext4_xattr_ibody_find {
 888	struct ext4_xattr_search s;
 889	struct ext4_iloc iloc;
 890};
 891
 892static int
 893ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
 894		      struct ext4_xattr_ibody_find *is)
 895{
 896	struct ext4_xattr_ibody_header *header;
 897	struct ext4_inode *raw_inode;
 898	int error;
 899
 900	if (EXT4_I(inode)->i_extra_isize == 0)
 901		return 0;
 902	raw_inode = ext4_raw_inode(&is->iloc);
 903	header = IHDR(inode, raw_inode);
 904	is->s.base = is->s.first = IFIRST(header);
 905	is->s.here = is->s.first;
 906	is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
 907	if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
 908		error = ext4_xattr_check_names(IFIRST(header), is->s.end);
 909		if (error)
 910			return error;
 911		/* Find the named attribute. */
 912		error = ext4_xattr_find_entry(&is->s.here, i->name_index,
 913					      i->name, is->s.end -
 914					      (void *)is->s.base, 0);
 915		if (error && error != -ENODATA)
 916			return error;
 917		is->s.not_found = error;
 918	}
 919	return 0;
 920}
 921
 922static int
 923ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
 924		     struct ext4_xattr_info *i,
 925		     struct ext4_xattr_ibody_find *is)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 926{
 927	struct ext4_xattr_ibody_header *header;
 928	struct ext4_xattr_search *s = &is->s;
 929	int error;
 930
 931	if (EXT4_I(inode)->i_extra_isize == 0)
 932		return -ENOSPC;
 933	error = ext4_xattr_set_entry(i, s);
 934	if (error)
 935		return error;
 936	header = IHDR(inode, ext4_raw_inode(&is->iloc));
 937	if (!IS_LAST_ENTRY(s->first)) {
 938		header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
 939		ext4_set_inode_state(inode, EXT4_STATE_XATTR);
 940	} else {
 941		header->h_magic = cpu_to_le32(0);
 942		ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
 943	}
 944	return 0;
 945}
 946
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 947/*
 948 * ext4_xattr_set_handle()
 949 *
 950 * Create, replace or remove an extended attribute for this inode.  Value
 951 * is NULL to remove an existing extended attribute, and non-NULL to
 952 * either replace an existing extended attribute, or create a new extended
 953 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
 954 * specify that an extended attribute must exist and must not exist
 955 * previous to the call, respectively.
 956 *
 957 * Returns 0, or a negative error number on failure.
 958 */
 959int
 960ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
 961		      const char *name, const void *value, size_t value_len,
 962		      int flags)
 963{
 964	struct ext4_xattr_info i = {
 965		.name_index = name_index,
 966		.name = name,
 967		.value = value,
 968		.value_len = value_len,
 969
 970	};
 971	struct ext4_xattr_ibody_find is = {
 972		.s = { .not_found = -ENODATA, },
 973	};
 974	struct ext4_xattr_block_find bs = {
 975		.s = { .not_found = -ENODATA, },
 976	};
 977	unsigned long no_expand;
 978	int error;
 979
 980	if (!name)
 981		return -EINVAL;
 982	if (strlen(name) > 255)
 983		return -ERANGE;
 984	down_write(&EXT4_I(inode)->xattr_sem);
 985	no_expand = ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND);
 986	ext4_set_inode_state(inode, EXT4_STATE_NO_EXPAND);
 987
 988	error = ext4_get_inode_loc(inode, &is.iloc);
 989	if (error)
 990		goto cleanup;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 991
 992	error = ext4_journal_get_write_access(handle, is.iloc.bh);
 993	if (error)
 994		goto cleanup;
 995
 996	if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
 997		struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
 998		memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
 999		ext4_clear_inode_state(inode, EXT4_STATE_NEW);
1000	}
1001
1002	error = ext4_xattr_ibody_find(inode, &i, &is);
1003	if (error)
1004		goto cleanup;
1005	if (is.s.not_found)
1006		error = ext4_xattr_block_find(inode, &i, &bs);
1007	if (error)
1008		goto cleanup;
1009	if (is.s.not_found && bs.s.not_found) {
1010		error = -ENODATA;
1011		if (flags & XATTR_REPLACE)
1012			goto cleanup;
1013		error = 0;
1014		if (!value)
1015			goto cleanup;
1016	} else {
1017		error = -EEXIST;
1018		if (flags & XATTR_CREATE)
1019			goto cleanup;
1020	}
 
1021	if (!value) {
1022		if (!is.s.not_found)
1023			error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1024		else if (!bs.s.not_found)
1025			error = ext4_xattr_block_set(handle, inode, &i, &bs);
1026	} else {
 
 
 
 
 
 
 
 
 
 
 
 
1027		error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1028		if (!error && !bs.s.not_found) {
1029			i.value = NULL;
1030			error = ext4_xattr_block_set(handle, inode, &i, &bs);
1031		} else if (error == -ENOSPC) {
1032			if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1033				error = ext4_xattr_block_find(inode, &i, &bs);
1034				if (error)
1035					goto cleanup;
1036			}
1037			error = ext4_xattr_block_set(handle, inode, &i, &bs);
1038			if (error)
1039				goto cleanup;
1040			if (!is.s.not_found) {
1041				i.value = NULL;
1042				error = ext4_xattr_ibody_set(handle, inode, &i,
1043							     &is);
 
 
 
 
 
 
 
 
 
 
1044			}
1045		}
1046	}
1047	if (!error) {
1048		ext4_xattr_update_super_block(handle, inode->i_sb);
1049		inode->i_ctime = ext4_current_time(inode);
1050		if (!value)
1051			ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1052		error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1053		/*
1054		 * The bh is consumed by ext4_mark_iloc_dirty, even with
1055		 * error != 0.
1056		 */
1057		is.iloc.bh = NULL;
1058		if (IS_SYNC(inode))
1059			ext4_handle_sync(handle);
1060	}
1061
1062cleanup:
1063	brelse(is.iloc.bh);
1064	brelse(bs.bh);
1065	if (no_expand == 0)
1066		ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1067	up_write(&EXT4_I(inode)->xattr_sem);
1068	return error;
1069}
1070
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1071/*
1072 * ext4_xattr_set()
1073 *
1074 * Like ext4_xattr_set_handle, but start from an inode. This extended
1075 * attribute modification is a filesystem transaction by itself.
1076 *
1077 * Returns 0, or a negative error number on failure.
1078 */
1079int
1080ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1081	       const void *value, size_t value_len, int flags)
1082{
1083	handle_t *handle;
 
1084	int error, retries = 0;
 
 
 
 
 
1085
1086retry:
1087	handle = ext4_journal_start(inode, EXT4_DATA_TRANS_BLOCKS(inode->i_sb));
 
 
 
 
 
1088	if (IS_ERR(handle)) {
1089		error = PTR_ERR(handle);
1090	} else {
1091		int error2;
1092
1093		error = ext4_xattr_set_handle(handle, inode, name_index, name,
1094					      value, value_len, flags);
1095		error2 = ext4_journal_stop(handle);
1096		if (error == -ENOSPC &&
1097		    ext4_should_retry_alloc(inode->i_sb, &retries))
1098			goto retry;
1099		if (error == 0)
1100			error = error2;
1101	}
1102
1103	return error;
1104}
1105
1106/*
1107 * Shift the EA entries in the inode to create space for the increased
1108 * i_extra_isize.
1109 */
1110static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1111				     int value_offs_shift, void *to,
1112				     void *from, size_t n, int blocksize)
1113{
1114	struct ext4_xattr_entry *last = entry;
1115	int new_offs;
1116
 
 
 
1117	/* Adjust the value offsets of the entries */
1118	for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1119		if (!last->e_value_block && last->e_value_size) {
1120			new_offs = le16_to_cpu(last->e_value_offs) +
1121							value_offs_shift;
1122			BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1123				 > blocksize);
1124			last->e_value_offs = cpu_to_le16(new_offs);
1125		}
1126	}
1127	/* Shift the entries by n bytes */
1128	memmove(to, from, n);
1129}
1130
1131/*
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1132 * Expand an inode by new_extra_isize bytes when EAs are present.
1133 * Returns 0 on success or negative error number on failure.
1134 */
1135int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1136			       struct ext4_inode *raw_inode, handle_t *handle)
1137{
1138	struct ext4_xattr_ibody_header *header;
1139	struct ext4_xattr_entry *entry, *last, *first;
1140	struct buffer_head *bh = NULL;
1141	struct ext4_xattr_ibody_find *is = NULL;
1142	struct ext4_xattr_block_find *bs = NULL;
1143	char *buffer = NULL, *b_entry_name = NULL;
1144	size_t min_offs, free;
1145	int total_ino, total_blk;
1146	void *base, *start, *end;
1147	int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1148	int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1149
1150	down_write(&EXT4_I(inode)->xattr_sem);
1151retry:
1152	if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1153		up_write(&EXT4_I(inode)->xattr_sem);
1154		return 0;
1155	}
1156
1157	header = IHDR(inode, raw_inode);
1158	entry = IFIRST(header);
1159
1160	/*
1161	 * Check if enough free space is available in the inode to shift the
1162	 * entries ahead by new_extra_isize.
1163	 */
1164
1165	base = start = entry;
1166	end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1167	min_offs = end - base;
1168	last = entry;
1169	total_ino = sizeof(struct ext4_xattr_ibody_header);
1170
1171	free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1172	if (free >= new_extra_isize) {
1173		entry = IFIRST(header);
1174		ext4_xattr_shift_entries(entry,	EXT4_I(inode)->i_extra_isize
1175				- new_extra_isize, (void *)raw_inode +
1176				EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1177				(void *)header, total_ino,
1178				inode->i_sb->s_blocksize);
1179		EXT4_I(inode)->i_extra_isize = new_extra_isize;
1180		error = 0;
1181		goto cleanup;
1182	}
 
 
 
1183
1184	/*
1185	 * Enough free space isn't available in the inode, check if
1186	 * EA block can hold new_extra_isize bytes.
1187	 */
1188	if (EXT4_I(inode)->i_file_acl) {
1189		bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1190		error = -EIO;
1191		if (!bh)
1192			goto cleanup;
1193		if (ext4_xattr_check_block(bh)) {
1194			EXT4_ERROR_INODE(inode, "bad block %llu",
1195					 EXT4_I(inode)->i_file_acl);
1196			error = -EIO;
1197			goto cleanup;
1198		}
1199		base = BHDR(bh);
1200		first = BFIRST(bh);
1201		end = bh->b_data + bh->b_size;
1202		min_offs = end - base;
1203		free = ext4_xattr_free_space(first, &min_offs, base,
1204					     &total_blk);
1205		if (free < new_extra_isize) {
 
1206			if (!tried_min_extra_isize && s_min_extra_isize) {
1207				tried_min_extra_isize++;
1208				new_extra_isize = s_min_extra_isize;
1209				brelse(bh);
1210				goto retry;
1211			}
1212			error = -1;
1213			goto cleanup;
1214		}
1215	} else {
1216		free = inode->i_sb->s_blocksize;
1217	}
1218
1219	while (new_extra_isize > 0) {
1220		size_t offs, size, entry_size;
1221		struct ext4_xattr_entry *small_entry = NULL;
1222		struct ext4_xattr_info i = {
1223			.value = NULL,
1224			.value_len = 0,
1225		};
1226		unsigned int total_size;  /* EA entry size + value size */
1227		unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1228		unsigned int min_total_size = ~0U;
1229
1230		is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1231		bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1232		if (!is || !bs) {
1233			error = -ENOMEM;
1234			goto cleanup;
1235		}
1236
1237		is->s.not_found = -ENODATA;
1238		bs->s.not_found = -ENODATA;
1239		is->iloc.bh = NULL;
1240		bs->bh = NULL;
1241
1242		last = IFIRST(header);
1243		/* Find the entry best suited to be pushed into EA block */
1244		entry = NULL;
1245		for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1246			total_size =
1247			EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1248					EXT4_XATTR_LEN(last->e_name_len);
1249			if (total_size <= free && total_size < min_total_size) {
1250				if (total_size < new_extra_isize) {
1251					small_entry = last;
1252				} else {
1253					entry = last;
1254					min_total_size = total_size;
1255				}
1256			}
1257		}
1258
1259		if (entry == NULL) {
1260			if (small_entry) {
1261				entry = small_entry;
1262			} else {
1263				if (!tried_min_extra_isize &&
1264				    s_min_extra_isize) {
1265					tried_min_extra_isize++;
1266					new_extra_isize = s_min_extra_isize;
1267					goto retry;
1268				}
1269				error = -1;
1270				goto cleanup;
1271			}
1272		}
1273		offs = le16_to_cpu(entry->e_value_offs);
1274		size = le32_to_cpu(entry->e_value_size);
1275		entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1276		i.name_index = entry->e_name_index,
1277		buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1278		b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1279		if (!buffer || !b_entry_name) {
1280			error = -ENOMEM;
1281			goto cleanup;
1282		}
1283		/* Save the entry name and the entry value */
1284		memcpy(buffer, (void *)IFIRST(header) + offs,
1285		       EXT4_XATTR_SIZE(size));
1286		memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1287		b_entry_name[entry->e_name_len] = '\0';
1288		i.name = b_entry_name;
1289
1290		error = ext4_get_inode_loc(inode, &is->iloc);
1291		if (error)
1292			goto cleanup;
1293
1294		error = ext4_xattr_ibody_find(inode, &i, is);
1295		if (error)
1296			goto cleanup;
1297
1298		/* Remove the chosen entry from the inode */
1299		error = ext4_xattr_ibody_set(handle, inode, &i, is);
1300		if (error)
1301			goto cleanup;
1302
1303		entry = IFIRST(header);
1304		if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1305			shift_bytes = new_extra_isize;
1306		else
1307			shift_bytes = entry_size + size;
1308		/* Adjust the offsets and shift the remaining entries ahead */
1309		ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1310			shift_bytes, (void *)raw_inode +
1311			EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1312			(void *)header, total_ino - entry_size,
1313			inode->i_sb->s_blocksize);
1314
1315		extra_isize += shift_bytes;
1316		new_extra_isize -= shift_bytes;
1317		EXT4_I(inode)->i_extra_isize = extra_isize;
1318
1319		i.name = b_entry_name;
1320		i.value = buffer;
1321		i.value_len = size;
1322		error = ext4_xattr_block_find(inode, &i, bs);
1323		if (error)
1324			goto cleanup;
1325
1326		/* Add entry which was removed from the inode into the block */
1327		error = ext4_xattr_block_set(handle, inode, &i, bs);
1328		if (error)
1329			goto cleanup;
1330		kfree(b_entry_name);
1331		kfree(buffer);
1332		b_entry_name = NULL;
1333		buffer = NULL;
1334		brelse(is->iloc.bh);
1335		kfree(is);
1336		kfree(bs);
1337	}
1338	brelse(bh);
1339	up_write(&EXT4_I(inode)->xattr_sem);
1340	return 0;
 
 
 
 
1341
1342cleanup:
1343	kfree(b_entry_name);
1344	kfree(buffer);
1345	if (is)
1346		brelse(is->iloc.bh);
1347	kfree(is);
1348	kfree(bs);
1349	brelse(bh);
1350	up_write(&EXT4_I(inode)->xattr_sem);
1351	return error;
1352}
1353
 
 
1354
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1355
1356/*
1357 * ext4_xattr_delete_inode()
1358 *
1359 * Free extended attribute resources associated with this inode. This
1360 * is called immediately before an inode is freed. We have exclusive
1361 * access to the inode.
 
 
1362 */
1363void
1364ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
 
1365{
1366	struct buffer_head *bh = NULL;
 
 
 
 
 
1367
1368	if (!EXT4_I(inode)->i_file_acl)
1369		goto cleanup;
1370	bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1371	if (!bh) {
1372		EXT4_ERROR_INODE(inode, "block %llu read error",
1373				 EXT4_I(inode)->i_file_acl);
1374		goto cleanup;
1375	}
1376	if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1377	    BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1378		EXT4_ERROR_INODE(inode, "bad block %llu",
1379				 EXT4_I(inode)->i_file_acl);
1380		goto cleanup;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1381	}
1382	ext4_xattr_release_block(handle, inode, bh);
1383	EXT4_I(inode)->i_file_acl = 0;
1384
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1385cleanup:
 
1386	brelse(bh);
 
1387}
1388
1389/*
1390 * ext4_xattr_put_super()
1391 *
1392 * This is called when a file system is unmounted.
1393 */
1394void
1395ext4_xattr_put_super(struct super_block *sb)
1396{
1397	mb_cache_shrink(sb->s_bdev);
 
 
 
 
 
 
 
1398}
1399
1400/*
1401 * ext4_xattr_cache_insert()
1402 *
1403 * Create a new entry in the extended attribute cache, and insert
1404 * it unless such an entry is already in the cache.
1405 *
1406 * Returns 0, or a negative error number on failure.
1407 */
1408static void
1409ext4_xattr_cache_insert(struct buffer_head *bh)
 
1410{
1411	__u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1412	struct mb_cache_entry *ce;
 
 
1413	int error;
1414
1415	ce = mb_cache_entry_alloc(ext4_xattr_cache, GFP_NOFS);
1416	if (!ce) {
1417		ea_bdebug(bh, "out of memory");
1418		return;
1419	}
1420	error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1421	if (error) {
1422		mb_cache_entry_free(ce);
1423		if (error == -EBUSY) {
1424			ea_bdebug(bh, "already in cache");
1425			error = 0;
1426		}
1427	} else {
1428		ea_bdebug(bh, "inserting [%x]", (int)hash);
1429		mb_cache_entry_release(ce);
1430	}
1431}
1432
1433/*
1434 * ext4_xattr_cmp()
1435 *
1436 * Compare two extended attribute blocks for equality.
1437 *
1438 * Returns 0 if the blocks are equal, 1 if they differ, and
1439 * a negative error number on errors.
1440 */
1441static int
1442ext4_xattr_cmp(struct ext4_xattr_header *header1,
1443	       struct ext4_xattr_header *header2)
1444{
1445	struct ext4_xattr_entry *entry1, *entry2;
1446
1447	entry1 = ENTRY(header1+1);
1448	entry2 = ENTRY(header2+1);
1449	while (!IS_LAST_ENTRY(entry1)) {
1450		if (IS_LAST_ENTRY(entry2))
1451			return 1;
1452		if (entry1->e_hash != entry2->e_hash ||
1453		    entry1->e_name_index != entry2->e_name_index ||
1454		    entry1->e_name_len != entry2->e_name_len ||
1455		    entry1->e_value_size != entry2->e_value_size ||
 
1456		    memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1457			return 1;
1458		if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1459			return -EIO;
1460		if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1461			   (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1462			   le32_to_cpu(entry1->e_value_size)))
1463			return 1;
1464
1465		entry1 = EXT4_XATTR_NEXT(entry1);
1466		entry2 = EXT4_XATTR_NEXT(entry2);
1467	}
1468	if (!IS_LAST_ENTRY(entry2))
1469		return 1;
1470	return 0;
1471}
1472
1473/*
1474 * ext4_xattr_cache_find()
1475 *
1476 * Find an identical extended attribute block.
1477 *
1478 * Returns a pointer to the block found, or NULL if such a block was
1479 * not found or an error occurred.
1480 */
1481static struct buffer_head *
1482ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1483		      struct mb_cache_entry **pce)
 
1484{
1485	__u32 hash = le32_to_cpu(header->h_hash);
1486	struct mb_cache_entry *ce;
 
1487
 
 
1488	if (!header->h_hash)
1489		return NULL;  /* never share */
1490	ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1491again:
1492	ce = mb_cache_entry_find_first(ext4_xattr_cache, inode->i_sb->s_bdev,
1493				       hash);
1494	while (ce) {
1495		struct buffer_head *bh;
1496
1497		if (IS_ERR(ce)) {
1498			if (PTR_ERR(ce) == -EAGAIN)
1499				goto again;
1500			break;
1501		}
1502		bh = sb_bread(inode->i_sb, ce->e_block);
1503		if (!bh) {
1504			EXT4_ERROR_INODE(inode, "block %lu read error",
1505					 (unsigned long) ce->e_block);
1506		} else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1507				EXT4_XATTR_REFCOUNT_MAX) {
1508			ea_idebug(inode, "block %lu refcount %d>=%d",
1509				  (unsigned long) ce->e_block,
1510				  le32_to_cpu(BHDR(bh)->h_refcount),
1511					  EXT4_XATTR_REFCOUNT_MAX);
1512		} else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1513			*pce = ce;
1514			return bh;
1515		}
1516		brelse(bh);
1517		ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1518	}
1519	return NULL;
1520}
1521
1522#define NAME_HASH_SHIFT 5
1523#define VALUE_HASH_SHIFT 16
1524
1525/*
1526 * ext4_xattr_hash_entry()
1527 *
1528 * Compute the hash of an extended attribute.
1529 */
1530static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1531					 struct ext4_xattr_entry *entry)
1532{
1533	__u32 hash = 0;
1534	char *name = entry->e_name;
1535	int n;
1536
1537	for (n = 0; n < entry->e_name_len; n++) {
1538		hash = (hash << NAME_HASH_SHIFT) ^
1539		       (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1540		       *name++;
1541	}
1542
1543	if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1544		__le32 *value = (__le32 *)((char *)header +
1545			le16_to_cpu(entry->e_value_offs));
1546		for (n = (le32_to_cpu(entry->e_value_size) +
1547		     EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1548			hash = (hash << VALUE_HASH_SHIFT) ^
1549			       (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1550			       le32_to_cpu(*value++);
1551		}
1552	}
1553	entry->e_hash = cpu_to_le32(hash);
1554}
1555
1556#undef NAME_HASH_SHIFT
1557#undef VALUE_HASH_SHIFT
1558
1559#define BLOCK_HASH_SHIFT 16
1560
1561/*
1562 * ext4_xattr_rehash()
1563 *
1564 * Re-compute the extended attribute hash value after an entry has changed.
1565 */
1566static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1567			      struct ext4_xattr_entry *entry)
1568{
1569	struct ext4_xattr_entry *here;
1570	__u32 hash = 0;
1571
1572	ext4_xattr_hash_entry(header, entry);
1573	here = ENTRY(header+1);
1574	while (!IS_LAST_ENTRY(here)) {
1575		if (!here->e_hash) {
1576			/* Block is not shared if an entry's hash value == 0 */
1577			hash = 0;
1578			break;
1579		}
1580		hash = (hash << BLOCK_HASH_SHIFT) ^
1581		       (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1582		       le32_to_cpu(here->e_hash);
1583		here = EXT4_XATTR_NEXT(here);
1584	}
1585	header->h_hash = cpu_to_le32(hash);
1586}
1587
1588#undef BLOCK_HASH_SHIFT
1589
1590int __init
1591ext4_init_xattr(void)
 
 
1592{
1593	ext4_xattr_cache = mb_cache_create("ext4_xattr", 6);
1594	if (!ext4_xattr_cache)
1595		return -ENOMEM;
1596	return 0;
1597}
1598
1599void
1600ext4_exit_xattr(void)
1601{
1602	if (ext4_xattr_cache)
1603		mb_cache_destroy(ext4_xattr_cache);
1604	ext4_xattr_cache = NULL;
1605}
v4.17
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * linux/fs/ext4/xattr.c
   4 *
   5 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
   6 *
   7 * Fix by Harrison Xing <harrison@mountainviewdata.com>.
   8 * Ext4 code with a lot of help from Eric Jarman <ejarman@acm.org>.
   9 * Extended attributes for symlinks and special files added per
  10 *  suggestion of Luka Renko <luka.renko@hermes.si>.
  11 * xattr consolidation Copyright (c) 2004 James Morris <jmorris@redhat.com>,
  12 *  Red Hat Inc.
  13 * ea-in-inode support by Alex Tomas <alex@clusterfs.com> aka bzzz
  14 *  and Andreas Gruenbacher <agruen@suse.de>.
  15 */
  16
  17/*
  18 * Extended attributes are stored directly in inodes (on file systems with
  19 * inodes bigger than 128 bytes) and on additional disk blocks. The i_file_acl
  20 * field contains the block number if an inode uses an additional block. All
  21 * attributes must fit in the inode and one additional block. Blocks that
  22 * contain the identical set of attributes may be shared among several inodes.
  23 * Identical blocks are detected by keeping a cache of blocks that have
  24 * recently been accessed.
  25 *
  26 * The attributes in inodes and on blocks have a different header; the entries
  27 * are stored in the same format:
  28 *
  29 *   +------------------+
  30 *   | header           |
  31 *   | entry 1          | |
  32 *   | entry 2          | | growing downwards
  33 *   | entry 3          | v
  34 *   | four null bytes  |
  35 *   | . . .            |
  36 *   | value 1          | ^
  37 *   | value 3          | | growing upwards
  38 *   | value 2          | |
  39 *   +------------------+
  40 *
  41 * The header is followed by multiple entry descriptors. In disk blocks, the
  42 * entry descriptors are kept sorted. In inodes, they are unsorted. The
  43 * attribute values are aligned to the end of the block in no specific order.
  44 *
  45 * Locking strategy
  46 * ----------------
  47 * EXT4_I(inode)->i_file_acl is protected by EXT4_I(inode)->xattr_sem.
  48 * EA blocks are only changed if they are exclusive to an inode, so
  49 * holding xattr_sem also means that nothing but the EA block's reference
  50 * count can change. Multiple writers to the same block are synchronized
  51 * by the buffer lock.
  52 */
  53
  54#include <linux/init.h>
  55#include <linux/fs.h>
  56#include <linux/slab.h>
  57#include <linux/mbcache.h>
  58#include <linux/quotaops.h>
  59#include <linux/iversion.h>
  60#include "ext4_jbd2.h"
  61#include "ext4.h"
  62#include "xattr.h"
  63#include "acl.h"
  64
 
 
 
 
 
  65#ifdef EXT4_XATTR_DEBUG
  66# define ea_idebug(inode, fmt, ...)					\
  67	printk(KERN_DEBUG "inode %s:%lu: " fmt "\n",			\
  68	       inode->i_sb->s_id, inode->i_ino, ##__VA_ARGS__)
  69# define ea_bdebug(bh, fmt, ...)					\
  70	printk(KERN_DEBUG "block %pg:%lu: " fmt "\n",			\
  71	       bh->b_bdev, (unsigned long)bh->b_blocknr, ##__VA_ARGS__)
 
 
 
 
 
 
 
 
  72#else
  73# define ea_idebug(inode, fmt, ...)	no_printk(fmt, ##__VA_ARGS__)
  74# define ea_bdebug(bh, fmt, ...)	no_printk(fmt, ##__VA_ARGS__)
  75#endif
  76
  77static void ext4_xattr_block_cache_insert(struct mb_cache *,
  78					  struct buffer_head *);
  79static struct buffer_head *
  80ext4_xattr_block_cache_find(struct inode *, struct ext4_xattr_header *,
  81			    struct mb_cache_entry **);
  82static __le32 ext4_xattr_hash_entry(char *name, size_t name_len, __le32 *value,
  83				    size_t value_count);
  84static void ext4_xattr_rehash(struct ext4_xattr_header *);
 
 
  85
  86static const struct xattr_handler * const ext4_xattr_handler_map[] = {
  87	[EXT4_XATTR_INDEX_USER]		     = &ext4_xattr_user_handler,
  88#ifdef CONFIG_EXT4_FS_POSIX_ACL
  89	[EXT4_XATTR_INDEX_POSIX_ACL_ACCESS]  = &posix_acl_access_xattr_handler,
  90	[EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &posix_acl_default_xattr_handler,
  91#endif
  92	[EXT4_XATTR_INDEX_TRUSTED]	     = &ext4_xattr_trusted_handler,
  93#ifdef CONFIG_EXT4_FS_SECURITY
  94	[EXT4_XATTR_INDEX_SECURITY]	     = &ext4_xattr_security_handler,
  95#endif
  96};
  97
  98const struct xattr_handler *ext4_xattr_handlers[] = {
  99	&ext4_xattr_user_handler,
 100	&ext4_xattr_trusted_handler,
 101#ifdef CONFIG_EXT4_FS_POSIX_ACL
 102	&posix_acl_access_xattr_handler,
 103	&posix_acl_default_xattr_handler,
 104#endif
 105#ifdef CONFIG_EXT4_FS_SECURITY
 106	&ext4_xattr_security_handler,
 107#endif
 108	NULL
 109};
 110
 111#define EA_BLOCK_CACHE(inode)	(((struct ext4_sb_info *) \
 112				inode->i_sb->s_fs_info)->s_ea_block_cache)
 113
 114#define EA_INODE_CACHE(inode)	(((struct ext4_sb_info *) \
 115				inode->i_sb->s_fs_info)->s_ea_inode_cache)
 116
 117static int
 118ext4_expand_inode_array(struct ext4_xattr_inode_array **ea_inode_array,
 119			struct inode *inode);
 120
 121#ifdef CONFIG_LOCKDEP
 122void ext4_xattr_inode_set_class(struct inode *ea_inode)
 123{
 124	lockdep_set_subclass(&ea_inode->i_rwsem, 1);
 125}
 126#endif
 127
 128static __le32 ext4_xattr_block_csum(struct inode *inode,
 129				    sector_t block_nr,
 130				    struct ext4_xattr_header *hdr)
 131{
 132	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
 133	__u32 csum;
 134	__le64 dsk_block_nr = cpu_to_le64(block_nr);
 135	__u32 dummy_csum = 0;
 136	int offset = offsetof(struct ext4_xattr_header, h_checksum);
 137
 138	csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&dsk_block_nr,
 139			   sizeof(dsk_block_nr));
 140	csum = ext4_chksum(sbi, csum, (__u8 *)hdr, offset);
 141	csum = ext4_chksum(sbi, csum, (__u8 *)&dummy_csum, sizeof(dummy_csum));
 142	offset += sizeof(dummy_csum);
 143	csum = ext4_chksum(sbi, csum, (__u8 *)hdr + offset,
 144			   EXT4_BLOCK_SIZE(inode->i_sb) - offset);
 145
 146	return cpu_to_le32(csum);
 147}
 148
 149static int ext4_xattr_block_csum_verify(struct inode *inode,
 150					struct buffer_head *bh)
 151{
 152	struct ext4_xattr_header *hdr = BHDR(bh);
 153	int ret = 1;
 154
 155	if (ext4_has_metadata_csum(inode->i_sb)) {
 156		lock_buffer(bh);
 157		ret = (hdr->h_checksum == ext4_xattr_block_csum(inode,
 158							bh->b_blocknr, hdr));
 159		unlock_buffer(bh);
 160	}
 161	return ret;
 162}
 163
 164static void ext4_xattr_block_csum_set(struct inode *inode,
 165				      struct buffer_head *bh)
 166{
 167	if (ext4_has_metadata_csum(inode->i_sb))
 168		BHDR(bh)->h_checksum = ext4_xattr_block_csum(inode,
 169						bh->b_blocknr, BHDR(bh));
 170}
 171
 172static inline const struct xattr_handler *
 173ext4_xattr_handler(int name_index)
 174{
 175	const struct xattr_handler *handler = NULL;
 176
 177	if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
 178		handler = ext4_xattr_handler_map[name_index];
 179	return handler;
 180}
 181
 
 
 
 
 
 
 
 
 
 
 
 182static int
 183ext4_xattr_check_entries(struct ext4_xattr_entry *entry, void *end,
 184			 void *value_start)
 185{
 186	struct ext4_xattr_entry *e = entry;
 187
 188	/* Find the end of the names list */
 189	while (!IS_LAST_ENTRY(e)) {
 190		struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(e);
 191		if ((void *)next >= end)
 192			return -EFSCORRUPTED;
 193		e = next;
 194	}
 195
 196	/* Check the values */
 197	while (!IS_LAST_ENTRY(entry)) {
 198		u32 size = le32_to_cpu(entry->e_value_size);
 199
 200		if (size > EXT4_XATTR_SIZE_MAX)
 201			return -EFSCORRUPTED;
 202
 203		if (size != 0 && entry->e_value_inum == 0) {
 204			u16 offs = le16_to_cpu(entry->e_value_offs);
 205			void *value;
 206
 207			/*
 208			 * The value cannot overlap the names, and the value
 209			 * with padding cannot extend beyond 'end'.  Check both
 210			 * the padded and unpadded sizes, since the size may
 211			 * overflow to 0 when adding padding.
 212			 */
 213			if (offs > end - value_start)
 214				return -EFSCORRUPTED;
 215			value = value_start + offs;
 216			if (value < (void *)e + sizeof(u32) ||
 217			    size > end - value ||
 218			    EXT4_XATTR_SIZE(size) > end - value)
 219				return -EFSCORRUPTED;
 220		}
 221		entry = EXT4_XATTR_NEXT(entry);
 222	}
 223
 224	return 0;
 225}
 226
 227static inline int
 228__ext4_xattr_check_block(struct inode *inode, struct buffer_head *bh,
 229			 const char *function, unsigned int line)
 230{
 231	int error = -EFSCORRUPTED;
 232
 233	if (buffer_verified(bh))
 234		return 0;
 235
 236	if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
 237	    BHDR(bh)->h_blocks != cpu_to_le32(1))
 238		goto errout;
 239	error = -EFSBADCRC;
 240	if (!ext4_xattr_block_csum_verify(inode, bh))
 241		goto errout;
 242	error = ext4_xattr_check_entries(BFIRST(bh), bh->b_data + bh->b_size,
 243					 bh->b_data);
 244errout:
 245	if (error)
 246		__ext4_error_inode(inode, function, line, 0,
 247				   "corrupted xattr block %llu",
 248				   (unsigned long long) bh->b_blocknr);
 249	else
 250		set_buffer_verified(bh);
 251	return error;
 252}
 253
 254#define ext4_xattr_check_block(inode, bh) \
 255	__ext4_xattr_check_block((inode), (bh),  __func__, __LINE__)
 256
 257
 258static int
 259__xattr_check_inode(struct inode *inode, struct ext4_xattr_ibody_header *header,
 260			 void *end, const char *function, unsigned int line)
 261{
 262	int error = -EFSCORRUPTED;
 263
 264	if (end - (void *)header < sizeof(*header) + sizeof(u32) ||
 265	    (header->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC)))
 266		goto errout;
 267	error = ext4_xattr_check_entries(IFIRST(header), end, IFIRST(header));
 268errout:
 269	if (error)
 270		__ext4_error_inode(inode, function, line, 0,
 271				   "corrupted in-inode xattr");
 272	return error;
 273}
 274
 275#define xattr_check_inode(inode, header, end) \
 276	__xattr_check_inode((inode), (header), (end), __func__, __LINE__)
 277
 278static int
 279xattr_find_entry(struct inode *inode, struct ext4_xattr_entry **pentry,
 280		 void *end, int name_index, const char *name, int sorted)
 281{
 282	struct ext4_xattr_entry *entry, *next;
 283	size_t name_len;
 284	int cmp = 1;
 285
 286	if (name == NULL)
 287		return -EINVAL;
 288	name_len = strlen(name);
 289	for (entry = *pentry; !IS_LAST_ENTRY(entry); entry = next) {
 290		next = EXT4_XATTR_NEXT(entry);
 291		if ((void *) next >= end) {
 292			EXT4_ERROR_INODE(inode, "corrupted xattr entries");
 293			return -EFSCORRUPTED;
 294		}
 295		cmp = name_index - entry->e_name_index;
 296		if (!cmp)
 297			cmp = name_len - entry->e_name_len;
 298		if (!cmp)
 299			cmp = memcmp(name, entry->e_name, name_len);
 300		if (cmp <= 0 && (sorted || cmp == 0))
 301			break;
 302	}
 303	*pentry = entry;
 
 
 304	return cmp ? -ENODATA : 0;
 305}
 306
 307static u32
 308ext4_xattr_inode_hash(struct ext4_sb_info *sbi, const void *buffer, size_t size)
 309{
 310	return ext4_chksum(sbi, sbi->s_csum_seed, buffer, size);
 311}
 312
 313static u64 ext4_xattr_inode_get_ref(struct inode *ea_inode)
 314{
 315	return ((u64)ea_inode->i_ctime.tv_sec << 32) |
 316		(u32) inode_peek_iversion_raw(ea_inode);
 317}
 318
 319static void ext4_xattr_inode_set_ref(struct inode *ea_inode, u64 ref_count)
 320{
 321	ea_inode->i_ctime.tv_sec = (u32)(ref_count >> 32);
 322	inode_set_iversion_raw(ea_inode, ref_count & 0xffffffff);
 323}
 324
 325static u32 ext4_xattr_inode_get_hash(struct inode *ea_inode)
 326{
 327	return (u32)ea_inode->i_atime.tv_sec;
 328}
 329
 330static void ext4_xattr_inode_set_hash(struct inode *ea_inode, u32 hash)
 331{
 332	ea_inode->i_atime.tv_sec = hash;
 333}
 334
 335/*
 336 * Read the EA value from an inode.
 337 */
 338static int ext4_xattr_inode_read(struct inode *ea_inode, void *buf, size_t size)
 339{
 340	int blocksize = 1 << ea_inode->i_blkbits;
 341	int bh_count = (size + blocksize - 1) >> ea_inode->i_blkbits;
 342	int tail_size = (size % blocksize) ?: blocksize;
 343	struct buffer_head *bhs_inline[8];
 344	struct buffer_head **bhs = bhs_inline;
 345	int i, ret;
 346
 347	if (bh_count > ARRAY_SIZE(bhs_inline)) {
 348		bhs = kmalloc_array(bh_count, sizeof(*bhs), GFP_NOFS);
 349		if (!bhs)
 350			return -ENOMEM;
 351	}
 352
 353	ret = ext4_bread_batch(ea_inode, 0 /* block */, bh_count,
 354			       true /* wait */, bhs);
 355	if (ret)
 356		goto free_bhs;
 357
 358	for (i = 0; i < bh_count; i++) {
 359		/* There shouldn't be any holes in ea_inode. */
 360		if (!bhs[i]) {
 361			ret = -EFSCORRUPTED;
 362			goto put_bhs;
 363		}
 364		memcpy((char *)buf + blocksize * i, bhs[i]->b_data,
 365		       i < bh_count - 1 ? blocksize : tail_size);
 366	}
 367	ret = 0;
 368put_bhs:
 369	for (i = 0; i < bh_count; i++)
 370		brelse(bhs[i]);
 371free_bhs:
 372	if (bhs != bhs_inline)
 373		kfree(bhs);
 374	return ret;
 375}
 376
 377#define EXT4_XATTR_INODE_GET_PARENT(inode) ((__u32)(inode)->i_mtime.tv_sec)
 378
 379static int ext4_xattr_inode_iget(struct inode *parent, unsigned long ea_ino,
 380				 u32 ea_inode_hash, struct inode **ea_inode)
 381{
 382	struct inode *inode;
 383	int err;
 384
 385	inode = ext4_iget(parent->i_sb, ea_ino);
 386	if (IS_ERR(inode)) {
 387		err = PTR_ERR(inode);
 388		ext4_error(parent->i_sb,
 389			   "error while reading EA inode %lu err=%d", ea_ino,
 390			   err);
 391		return err;
 392	}
 393
 394	if (is_bad_inode(inode)) {
 395		ext4_error(parent->i_sb,
 396			   "error while reading EA inode %lu is_bad_inode",
 397			   ea_ino);
 398		err = -EIO;
 399		goto error;
 400	}
 401
 402	if (!(EXT4_I(inode)->i_flags & EXT4_EA_INODE_FL)) {
 403		ext4_error(parent->i_sb,
 404			   "EA inode %lu does not have EXT4_EA_INODE_FL flag",
 405			    ea_ino);
 406		err = -EINVAL;
 407		goto error;
 408	}
 409
 410	ext4_xattr_inode_set_class(inode);
 411
 412	/*
 413	 * Check whether this is an old Lustre-style xattr inode. Lustre
 414	 * implementation does not have hash validation, rather it has a
 415	 * backpointer from ea_inode to the parent inode.
 416	 */
 417	if (ea_inode_hash != ext4_xattr_inode_get_hash(inode) &&
 418	    EXT4_XATTR_INODE_GET_PARENT(inode) == parent->i_ino &&
 419	    inode->i_generation == parent->i_generation) {
 420		ext4_set_inode_state(inode, EXT4_STATE_LUSTRE_EA_INODE);
 421		ext4_xattr_inode_set_ref(inode, 1);
 422	} else {
 423		inode_lock(inode);
 424		inode->i_flags |= S_NOQUOTA;
 425		inode_unlock(inode);
 426	}
 427
 428	*ea_inode = inode;
 429	return 0;
 430error:
 431	iput(inode);
 432	return err;
 433}
 434
 435static int
 436ext4_xattr_inode_verify_hashes(struct inode *ea_inode,
 437			       struct ext4_xattr_entry *entry, void *buffer,
 438			       size_t size)
 439{
 440	u32 hash;
 441
 442	/* Verify stored hash matches calculated hash. */
 443	hash = ext4_xattr_inode_hash(EXT4_SB(ea_inode->i_sb), buffer, size);
 444	if (hash != ext4_xattr_inode_get_hash(ea_inode))
 445		return -EFSCORRUPTED;
 446
 447	if (entry) {
 448		__le32 e_hash, tmp_data;
 449
 450		/* Verify entry hash. */
 451		tmp_data = cpu_to_le32(hash);
 452		e_hash = ext4_xattr_hash_entry(entry->e_name, entry->e_name_len,
 453					       &tmp_data, 1);
 454		if (e_hash != entry->e_hash)
 455			return -EFSCORRUPTED;
 456	}
 457	return 0;
 458}
 459
 460/*
 461 * Read xattr value from the EA inode.
 462 */
 463static int
 464ext4_xattr_inode_get(struct inode *inode, struct ext4_xattr_entry *entry,
 465		     void *buffer, size_t size)
 466{
 467	struct mb_cache *ea_inode_cache = EA_INODE_CACHE(inode);
 468	struct inode *ea_inode;
 469	int err;
 470
 471	err = ext4_xattr_inode_iget(inode, le32_to_cpu(entry->e_value_inum),
 472				    le32_to_cpu(entry->e_hash), &ea_inode);
 473	if (err) {
 474		ea_inode = NULL;
 475		goto out;
 476	}
 477
 478	if (i_size_read(ea_inode) != size) {
 479		ext4_warning_inode(ea_inode,
 480				   "ea_inode file size=%llu entry size=%zu",
 481				   i_size_read(ea_inode), size);
 482		err = -EFSCORRUPTED;
 483		goto out;
 484	}
 485
 486	err = ext4_xattr_inode_read(ea_inode, buffer, size);
 487	if (err)
 488		goto out;
 489
 490	if (!ext4_test_inode_state(ea_inode, EXT4_STATE_LUSTRE_EA_INODE)) {
 491		err = ext4_xattr_inode_verify_hashes(ea_inode, entry, buffer,
 492						     size);
 493		if (err) {
 494			ext4_warning_inode(ea_inode,
 495					   "EA inode hash validation failed");
 496			goto out;
 497		}
 498
 499		if (ea_inode_cache)
 500			mb_cache_entry_create(ea_inode_cache, GFP_NOFS,
 501					ext4_xattr_inode_get_hash(ea_inode),
 502					ea_inode->i_ino, true /* reusable */);
 503	}
 504out:
 505	iput(ea_inode);
 506	return err;
 507}
 508
 509static int
 510ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
 511		     void *buffer, size_t buffer_size)
 512{
 513	struct buffer_head *bh = NULL;
 514	struct ext4_xattr_entry *entry;
 515	size_t size;
 516	void *end;
 517	int error;
 518	struct mb_cache *ea_block_cache = EA_BLOCK_CACHE(inode);
 519
 520	ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
 521		  name_index, name, buffer, (long)buffer_size);
 522
 523	error = -ENODATA;
 524	if (!EXT4_I(inode)->i_file_acl)
 525		goto cleanup;
 526	ea_idebug(inode, "reading block %llu",
 527		  (unsigned long long)EXT4_I(inode)->i_file_acl);
 528	bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
 529	if (!bh)
 530		goto cleanup;
 531	ea_bdebug(bh, "b_count=%d, refcount=%d",
 532		atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
 533	error = ext4_xattr_check_block(inode, bh);
 534	if (error)
 
 
 
 535		goto cleanup;
 536	ext4_xattr_block_cache_insert(ea_block_cache, bh);
 
 537	entry = BFIRST(bh);
 538	end = bh->b_data + bh->b_size;
 539	error = xattr_find_entry(inode, &entry, end, name_index, name, 1);
 
 540	if (error)
 541		goto cleanup;
 542	size = le32_to_cpu(entry->e_value_size);
 543	error = -ERANGE;
 544	if (unlikely(size > EXT4_XATTR_SIZE_MAX))
 545		goto cleanup;
 546	if (buffer) {
 
 547		if (size > buffer_size)
 548			goto cleanup;
 549		if (entry->e_value_inum) {
 550			error = ext4_xattr_inode_get(inode, entry, buffer,
 551						     size);
 552			if (error)
 553				goto cleanup;
 554		} else {
 555			u16 offset = le16_to_cpu(entry->e_value_offs);
 556			void *p = bh->b_data + offset;
 557
 558			if (unlikely(p + size > end))
 559				goto cleanup;
 560			memcpy(buffer, p, size);
 561		}
 562	}
 563	error = size;
 564
 565cleanup:
 566	brelse(bh);
 567	return error;
 568}
 569
 570int
 571ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
 572		     void *buffer, size_t buffer_size)
 573{
 574	struct ext4_xattr_ibody_header *header;
 575	struct ext4_xattr_entry *entry;
 576	struct ext4_inode *raw_inode;
 577	struct ext4_iloc iloc;
 578	size_t size;
 579	void *end;
 580	int error;
 581
 582	if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
 583		return -ENODATA;
 584	error = ext4_get_inode_loc(inode, &iloc);
 585	if (error)
 586		return error;
 587	raw_inode = ext4_raw_inode(&iloc);
 588	header = IHDR(inode, raw_inode);
 
 589	end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
 590	error = xattr_check_inode(inode, header, end);
 591	if (error)
 592		goto cleanup;
 593	entry = IFIRST(header);
 594	error = xattr_find_entry(inode, &entry, end, name_index, name, 0);
 595	if (error)
 596		goto cleanup;
 597	size = le32_to_cpu(entry->e_value_size);
 598	error = -ERANGE;
 599	if (unlikely(size > EXT4_XATTR_SIZE_MAX))
 600		goto cleanup;
 601	if (buffer) {
 
 602		if (size > buffer_size)
 603			goto cleanup;
 604		if (entry->e_value_inum) {
 605			error = ext4_xattr_inode_get(inode, entry, buffer,
 606						     size);
 607			if (error)
 608				goto cleanup;
 609		} else {
 610			u16 offset = le16_to_cpu(entry->e_value_offs);
 611			void *p = (void *)IFIRST(header) + offset;
 612
 613			if (unlikely(p + size > end))
 614				goto cleanup;
 615			memcpy(buffer, p, size);
 616		}
 617	}
 618	error = size;
 619
 620cleanup:
 621	brelse(iloc.bh);
 622	return error;
 623}
 624
 625/*
 626 * ext4_xattr_get()
 627 *
 628 * Copy an extended attribute into the buffer
 629 * provided, or compute the buffer size required.
 630 * Buffer is NULL to compute the size of the buffer required.
 631 *
 632 * Returns a negative error number on failure, or the number of bytes
 633 * used / required on success.
 634 */
 635int
 636ext4_xattr_get(struct inode *inode, int name_index, const char *name,
 637	       void *buffer, size_t buffer_size)
 638{
 639	int error;
 640
 641	if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
 642		return -EIO;
 643
 644	if (strlen(name) > 255)
 645		return -ERANGE;
 646
 647	down_read(&EXT4_I(inode)->xattr_sem);
 648	error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
 649				     buffer_size);
 650	if (error == -ENODATA)
 651		error = ext4_xattr_block_get(inode, name_index, name, buffer,
 652					     buffer_size);
 653	up_read(&EXT4_I(inode)->xattr_sem);
 654	return error;
 655}
 656
 657static int
 658ext4_xattr_list_entries(struct dentry *dentry, struct ext4_xattr_entry *entry,
 659			char *buffer, size_t buffer_size)
 660{
 661	size_t rest = buffer_size;
 662
 663	for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
 664		const struct xattr_handler *handler =
 665			ext4_xattr_handler(entry->e_name_index);
 666
 667		if (handler && (!handler->list || handler->list(dentry))) {
 668			const char *prefix = handler->prefix ?: handler->name;
 669			size_t prefix_len = strlen(prefix);
 670			size_t size = prefix_len + entry->e_name_len + 1;
 671
 672			if (buffer) {
 673				if (size > rest)
 674					return -ERANGE;
 675				memcpy(buffer, prefix, prefix_len);
 676				buffer += prefix_len;
 677				memcpy(buffer, entry->e_name, entry->e_name_len);
 678				buffer += entry->e_name_len;
 679				*buffer++ = 0;
 680			}
 681			rest -= size;
 682		}
 683	}
 684	return buffer_size - rest;  /* total size */
 685}
 686
 687static int
 688ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
 689{
 690	struct inode *inode = d_inode(dentry);
 691	struct buffer_head *bh = NULL;
 692	int error;
 693
 694	ea_idebug(inode, "buffer=%p, buffer_size=%ld",
 695		  buffer, (long)buffer_size);
 696
 697	error = 0;
 698	if (!EXT4_I(inode)->i_file_acl)
 699		goto cleanup;
 700	ea_idebug(inode, "reading block %llu",
 701		  (unsigned long long)EXT4_I(inode)->i_file_acl);
 702	bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
 703	error = -EIO;
 704	if (!bh)
 705		goto cleanup;
 706	ea_bdebug(bh, "b_count=%d, refcount=%d",
 707		atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
 708	error = ext4_xattr_check_block(inode, bh);
 709	if (error)
 
 
 710		goto cleanup;
 711	ext4_xattr_block_cache_insert(EA_BLOCK_CACHE(inode), bh);
 
 712	error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
 713
 714cleanup:
 715	brelse(bh);
 716
 717	return error;
 718}
 719
 720static int
 721ext4_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
 722{
 723	struct inode *inode = d_inode(dentry);
 724	struct ext4_xattr_ibody_header *header;
 725	struct ext4_inode *raw_inode;
 726	struct ext4_iloc iloc;
 727	void *end;
 728	int error;
 729
 730	if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
 731		return 0;
 732	error = ext4_get_inode_loc(inode, &iloc);
 733	if (error)
 734		return error;
 735	raw_inode = ext4_raw_inode(&iloc);
 736	header = IHDR(inode, raw_inode);
 737	end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
 738	error = xattr_check_inode(inode, header, end);
 739	if (error)
 740		goto cleanup;
 741	error = ext4_xattr_list_entries(dentry, IFIRST(header),
 742					buffer, buffer_size);
 743
 744cleanup:
 745	brelse(iloc.bh);
 746	return error;
 747}
 748
 749/*
 750 * Inode operation listxattr()
 751 *
 752 * d_inode(dentry)->i_rwsem: don't care
 753 *
 754 * Copy a list of attribute names into the buffer
 755 * provided, or compute the buffer size required.
 756 * Buffer is NULL to compute the size of the buffer required.
 757 *
 758 * Returns a negative error number on failure, or the number of bytes
 759 * used / required on success.
 760 */
 761ssize_t
 762ext4_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
 763{
 764	int ret, ret2;
 765
 766	down_read(&EXT4_I(d_inode(dentry))->xattr_sem);
 767	ret = ret2 = ext4_xattr_ibody_list(dentry, buffer, buffer_size);
 768	if (ret < 0)
 769		goto errout;
 770	if (buffer) {
 771		buffer += ret;
 772		buffer_size -= ret;
 773	}
 774	ret = ext4_xattr_block_list(dentry, buffer, buffer_size);
 775	if (ret < 0)
 776		goto errout;
 777	ret += ret2;
 778errout:
 779	up_read(&EXT4_I(d_inode(dentry))->xattr_sem);
 780	return ret;
 781}
 782
 783/*
 784 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
 785 * not set, set it.
 786 */
 787static void ext4_xattr_update_super_block(handle_t *handle,
 788					  struct super_block *sb)
 789{
 790	if (ext4_has_feature_xattr(sb))
 791		return;
 792
 793	BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get_write_access");
 794	if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
 795		ext4_set_feature_xattr(sb);
 796		ext4_handle_dirty_super(handle, sb);
 797	}
 798}
 799
 800int ext4_get_inode_usage(struct inode *inode, qsize_t *usage)
 801{
 802	struct ext4_iloc iloc = { .bh = NULL };
 803	struct buffer_head *bh = NULL;
 804	struct ext4_inode *raw_inode;
 805	struct ext4_xattr_ibody_header *header;
 806	struct ext4_xattr_entry *entry;
 807	qsize_t ea_inode_refs = 0;
 808	void *end;
 809	int ret;
 810
 811	lockdep_assert_held_read(&EXT4_I(inode)->xattr_sem);
 812
 813	if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
 814		ret = ext4_get_inode_loc(inode, &iloc);
 815		if (ret)
 816			goto out;
 817		raw_inode = ext4_raw_inode(&iloc);
 818		header = IHDR(inode, raw_inode);
 819		end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
 820		ret = xattr_check_inode(inode, header, end);
 821		if (ret)
 822			goto out;
 823
 824		for (entry = IFIRST(header); !IS_LAST_ENTRY(entry);
 825		     entry = EXT4_XATTR_NEXT(entry))
 826			if (entry->e_value_inum)
 827				ea_inode_refs++;
 828	}
 829
 830	if (EXT4_I(inode)->i_file_acl) {
 831		bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
 832		if (!bh) {
 833			ret = -EIO;
 834			goto out;
 835		}
 836
 837		ret = ext4_xattr_check_block(inode, bh);
 838		if (ret)
 839			goto out;
 840
 841		for (entry = BFIRST(bh); !IS_LAST_ENTRY(entry);
 842		     entry = EXT4_XATTR_NEXT(entry))
 843			if (entry->e_value_inum)
 844				ea_inode_refs++;
 845	}
 846	*usage = ea_inode_refs + 1;
 847	ret = 0;
 848out:
 849	brelse(iloc.bh);
 850	brelse(bh);
 851	return ret;
 852}
 853
 854static inline size_t round_up_cluster(struct inode *inode, size_t length)
 855{
 856	struct super_block *sb = inode->i_sb;
 857	size_t cluster_size = 1 << (EXT4_SB(sb)->s_cluster_bits +
 858				    inode->i_blkbits);
 859	size_t mask = ~(cluster_size - 1);
 860
 861	return (length + cluster_size - 1) & mask;
 862}
 863
 864static int ext4_xattr_inode_alloc_quota(struct inode *inode, size_t len)
 865{
 866	int err;
 867
 868	err = dquot_alloc_inode(inode);
 869	if (err)
 870		return err;
 871	err = dquot_alloc_space_nodirty(inode, round_up_cluster(inode, len));
 872	if (err)
 873		dquot_free_inode(inode);
 874	return err;
 875}
 876
 877static void ext4_xattr_inode_free_quota(struct inode *parent,
 878					struct inode *ea_inode,
 879					size_t len)
 880{
 881	if (ea_inode &&
 882	    ext4_test_inode_state(ea_inode, EXT4_STATE_LUSTRE_EA_INODE))
 883		return;
 884	dquot_free_space_nodirty(parent, round_up_cluster(parent, len));
 885	dquot_free_inode(parent);
 886}
 887
 888int __ext4_xattr_set_credits(struct super_block *sb, struct inode *inode,
 889			     struct buffer_head *block_bh, size_t value_len,
 890			     bool is_create)
 891{
 892	int credits;
 893	int blocks;
 894
 895	/*
 896	 * 1) Owner inode update
 897	 * 2) Ref count update on old xattr block
 898	 * 3) new xattr block
 899	 * 4) block bitmap update for new xattr block
 900	 * 5) group descriptor for new xattr block
 901	 * 6) block bitmap update for old xattr block
 902	 * 7) group descriptor for old block
 903	 *
 904	 * 6 & 7 can happen if we have two racing threads T_a and T_b
 905	 * which are each trying to set an xattr on inodes I_a and I_b
 906	 * which were both initially sharing an xattr block.
 907	 */
 908	credits = 7;
 909
 910	/* Quota updates. */
 911	credits += EXT4_MAXQUOTAS_TRANS_BLOCKS(sb);
 912
 913	/*
 914	 * In case of inline data, we may push out the data to a block,
 915	 * so we need to reserve credits for this eventuality
 916	 */
 917	if (inode && ext4_has_inline_data(inode))
 918		credits += ext4_writepage_trans_blocks(inode) + 1;
 919
 920	/* We are done if ea_inode feature is not enabled. */
 921	if (!ext4_has_feature_ea_inode(sb))
 922		return credits;
 923
 924	/* New ea_inode, inode map, block bitmap, group descriptor. */
 925	credits += 4;
 926
 927	/* Data blocks. */
 928	blocks = (value_len + sb->s_blocksize - 1) >> sb->s_blocksize_bits;
 929
 930	/* Indirection block or one level of extent tree. */
 931	blocks += 1;
 932
 933	/* Block bitmap and group descriptor updates for each block. */
 934	credits += blocks * 2;
 935
 936	/* Blocks themselves. */
 937	credits += blocks;
 938
 939	if (!is_create) {
 940		/* Dereference ea_inode holding old xattr value.
 941		 * Old ea_inode, inode map, block bitmap, group descriptor.
 942		 */
 943		credits += 4;
 944
 945		/* Data blocks for old ea_inode. */
 946		blocks = XATTR_SIZE_MAX >> sb->s_blocksize_bits;
 947
 948		/* Indirection block or one level of extent tree for old
 949		 * ea_inode.
 950		 */
 951		blocks += 1;
 952
 953		/* Block bitmap and group descriptor updates for each block. */
 954		credits += blocks * 2;
 955	}
 956
 957	/* We may need to clone the existing xattr block in which case we need
 958	 * to increment ref counts for existing ea_inodes referenced by it.
 959	 */
 960	if (block_bh) {
 961		struct ext4_xattr_entry *entry = BFIRST(block_bh);
 962
 963		for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry))
 964			if (entry->e_value_inum)
 965				/* Ref count update on ea_inode. */
 966				credits += 1;
 967	}
 968	return credits;
 969}
 970
 971static int ext4_xattr_ensure_credits(handle_t *handle, struct inode *inode,
 972				     int credits, struct buffer_head *bh,
 973				     bool dirty, bool block_csum)
 974{
 975	int error;
 976
 977	if (!ext4_handle_valid(handle))
 978		return 0;
 979
 980	if (handle->h_buffer_credits >= credits)
 981		return 0;
 982
 983	error = ext4_journal_extend(handle, credits - handle->h_buffer_credits);
 984	if (!error)
 985		return 0;
 986	if (error < 0) {
 987		ext4_warning(inode->i_sb, "Extend journal (error %d)", error);
 988		return error;
 989	}
 990
 991	if (bh && dirty) {
 992		if (block_csum)
 993			ext4_xattr_block_csum_set(inode, bh);
 994		error = ext4_handle_dirty_metadata(handle, NULL, bh);
 995		if (error) {
 996			ext4_warning(inode->i_sb, "Handle metadata (error %d)",
 997				     error);
 998			return error;
 999		}
1000	}
1001
1002	error = ext4_journal_restart(handle, credits);
1003	if (error) {
1004		ext4_warning(inode->i_sb, "Restart journal (error %d)", error);
1005		return error;
1006	}
1007
1008	if (bh) {
1009		error = ext4_journal_get_write_access(handle, bh);
1010		if (error) {
1011			ext4_warning(inode->i_sb,
1012				     "Get write access failed (error %d)",
1013				     error);
1014			return error;
1015		}
1016	}
1017	return 0;
1018}
1019
1020static int ext4_xattr_inode_update_ref(handle_t *handle, struct inode *ea_inode,
1021				       int ref_change)
1022{
1023	struct mb_cache *ea_inode_cache = EA_INODE_CACHE(ea_inode);
1024	struct ext4_iloc iloc;
1025	s64 ref_count;
1026	u32 hash;
1027	int ret;
1028
1029	inode_lock(ea_inode);
1030
1031	ret = ext4_reserve_inode_write(handle, ea_inode, &iloc);
1032	if (ret) {
1033		iloc.bh = NULL;
1034		goto out;
1035	}
1036
1037	ref_count = ext4_xattr_inode_get_ref(ea_inode);
1038	ref_count += ref_change;
1039	ext4_xattr_inode_set_ref(ea_inode, ref_count);
1040
1041	if (ref_change > 0) {
1042		WARN_ONCE(ref_count <= 0, "EA inode %lu ref_count=%lld",
1043			  ea_inode->i_ino, ref_count);
1044
1045		if (ref_count == 1) {
1046			WARN_ONCE(ea_inode->i_nlink, "EA inode %lu i_nlink=%u",
1047				  ea_inode->i_ino, ea_inode->i_nlink);
1048
1049			set_nlink(ea_inode, 1);
1050			ext4_orphan_del(handle, ea_inode);
1051
1052			if (ea_inode_cache) {
1053				hash = ext4_xattr_inode_get_hash(ea_inode);
1054				mb_cache_entry_create(ea_inode_cache,
1055						      GFP_NOFS, hash,
1056						      ea_inode->i_ino,
1057						      true /* reusable */);
1058			}
1059		}
1060	} else {
1061		WARN_ONCE(ref_count < 0, "EA inode %lu ref_count=%lld",
1062			  ea_inode->i_ino, ref_count);
1063
1064		if (ref_count == 0) {
1065			WARN_ONCE(ea_inode->i_nlink != 1,
1066				  "EA inode %lu i_nlink=%u",
1067				  ea_inode->i_ino, ea_inode->i_nlink);
1068
1069			clear_nlink(ea_inode);
1070			ext4_orphan_add(handle, ea_inode);
1071
1072			if (ea_inode_cache) {
1073				hash = ext4_xattr_inode_get_hash(ea_inode);
1074				mb_cache_entry_delete(ea_inode_cache, hash,
1075						      ea_inode->i_ino);
1076			}
1077		}
1078	}
1079
1080	ret = ext4_mark_iloc_dirty(handle, ea_inode, &iloc);
1081	iloc.bh = NULL;
1082	if (ret)
1083		ext4_warning_inode(ea_inode,
1084				   "ext4_mark_iloc_dirty() failed ret=%d", ret);
1085out:
1086	brelse(iloc.bh);
1087	inode_unlock(ea_inode);
1088	return ret;
1089}
1090
1091static int ext4_xattr_inode_inc_ref(handle_t *handle, struct inode *ea_inode)
1092{
1093	return ext4_xattr_inode_update_ref(handle, ea_inode, 1);
1094}
1095
1096static int ext4_xattr_inode_dec_ref(handle_t *handle, struct inode *ea_inode)
1097{
1098	return ext4_xattr_inode_update_ref(handle, ea_inode, -1);
1099}
1100
1101static int ext4_xattr_inode_inc_ref_all(handle_t *handle, struct inode *parent,
1102					struct ext4_xattr_entry *first)
1103{
1104	struct inode *ea_inode;
1105	struct ext4_xattr_entry *entry;
1106	struct ext4_xattr_entry *failed_entry;
1107	unsigned int ea_ino;
1108	int err, saved_err;
1109
1110	for (entry = first; !IS_LAST_ENTRY(entry);
1111	     entry = EXT4_XATTR_NEXT(entry)) {
1112		if (!entry->e_value_inum)
1113			continue;
1114		ea_ino = le32_to_cpu(entry->e_value_inum);
1115		err = ext4_xattr_inode_iget(parent, ea_ino,
1116					    le32_to_cpu(entry->e_hash),
1117					    &ea_inode);
1118		if (err)
1119			goto cleanup;
1120		err = ext4_xattr_inode_inc_ref(handle, ea_inode);
1121		if (err) {
1122			ext4_warning_inode(ea_inode, "inc ref error %d", err);
1123			iput(ea_inode);
1124			goto cleanup;
1125		}
1126		iput(ea_inode);
1127	}
1128	return 0;
1129
1130cleanup:
1131	saved_err = err;
1132	failed_entry = entry;
1133
1134	for (entry = first; entry != failed_entry;
1135	     entry = EXT4_XATTR_NEXT(entry)) {
1136		if (!entry->e_value_inum)
1137			continue;
1138		ea_ino = le32_to_cpu(entry->e_value_inum);
1139		err = ext4_xattr_inode_iget(parent, ea_ino,
1140					    le32_to_cpu(entry->e_hash),
1141					    &ea_inode);
1142		if (err) {
1143			ext4_warning(parent->i_sb,
1144				     "cleanup ea_ino %u iget error %d", ea_ino,
1145				     err);
1146			continue;
1147		}
1148		err = ext4_xattr_inode_dec_ref(handle, ea_inode);
1149		if (err)
1150			ext4_warning_inode(ea_inode, "cleanup dec ref error %d",
1151					   err);
1152		iput(ea_inode);
1153	}
1154	return saved_err;
1155}
1156
1157static void
1158ext4_xattr_inode_dec_ref_all(handle_t *handle, struct inode *parent,
1159			     struct buffer_head *bh,
1160			     struct ext4_xattr_entry *first, bool block_csum,
1161			     struct ext4_xattr_inode_array **ea_inode_array,
1162			     int extra_credits, bool skip_quota)
1163{
1164	struct inode *ea_inode;
1165	struct ext4_xattr_entry *entry;
1166	bool dirty = false;
1167	unsigned int ea_ino;
1168	int err;
1169	int credits;
1170
1171	/* One credit for dec ref on ea_inode, one for orphan list addition, */
1172	credits = 2 + extra_credits;
1173
1174	for (entry = first; !IS_LAST_ENTRY(entry);
1175	     entry = EXT4_XATTR_NEXT(entry)) {
1176		if (!entry->e_value_inum)
1177			continue;
1178		ea_ino = le32_to_cpu(entry->e_value_inum);
1179		err = ext4_xattr_inode_iget(parent, ea_ino,
1180					    le32_to_cpu(entry->e_hash),
1181					    &ea_inode);
1182		if (err)
1183			continue;
1184
1185		err = ext4_expand_inode_array(ea_inode_array, ea_inode);
1186		if (err) {
1187			ext4_warning_inode(ea_inode,
1188					   "Expand inode array err=%d", err);
1189			iput(ea_inode);
1190			continue;
1191		}
1192
1193		err = ext4_xattr_ensure_credits(handle, parent, credits, bh,
1194						dirty, block_csum);
1195		if (err) {
1196			ext4_warning_inode(ea_inode, "Ensure credits err=%d",
1197					   err);
1198			continue;
1199		}
1200
1201		err = ext4_xattr_inode_dec_ref(handle, ea_inode);
1202		if (err) {
1203			ext4_warning_inode(ea_inode, "ea_inode dec ref err=%d",
1204					   err);
1205			continue;
1206		}
1207
1208		if (!skip_quota)
1209			ext4_xattr_inode_free_quota(parent, ea_inode,
1210					      le32_to_cpu(entry->e_value_size));
1211
1212		/*
1213		 * Forget about ea_inode within the same transaction that
1214		 * decrements the ref count. This avoids duplicate decrements in
1215		 * case the rest of the work spills over to subsequent
1216		 * transactions.
1217		 */
1218		entry->e_value_inum = 0;
1219		entry->e_value_size = 0;
1220
1221		dirty = true;
1222	}
1223
1224	if (dirty) {
1225		/*
1226		 * Note that we are deliberately skipping csum calculation for
1227		 * the final update because we do not expect any journal
1228		 * restarts until xattr block is freed.
1229		 */
1230
1231		err = ext4_handle_dirty_metadata(handle, NULL, bh);
1232		if (err)
1233			ext4_warning_inode(parent,
1234					   "handle dirty metadata err=%d", err);
1235	}
1236}
1237
1238/*
1239 * Release the xattr block BH: If the reference count is > 1, decrement it;
1240 * otherwise free the block.
1241 */
1242static void
1243ext4_xattr_release_block(handle_t *handle, struct inode *inode,
1244			 struct buffer_head *bh,
1245			 struct ext4_xattr_inode_array **ea_inode_array,
1246			 int extra_credits)
1247{
1248	struct mb_cache *ea_block_cache = EA_BLOCK_CACHE(inode);
1249	u32 hash, ref;
1250	int error = 0;
1251
1252	BUFFER_TRACE(bh, "get_write_access");
1253	error = ext4_journal_get_write_access(handle, bh);
1254	if (error)
1255		goto out;
1256
1257	lock_buffer(bh);
1258	hash = le32_to_cpu(BHDR(bh)->h_hash);
1259	ref = le32_to_cpu(BHDR(bh)->h_refcount);
1260	if (ref == 1) {
1261		ea_bdebug(bh, "refcount now=0; freeing");
1262		/*
1263		 * This must happen under buffer lock for
1264		 * ext4_xattr_block_set() to reliably detect freed block
1265		 */
1266		if (ea_block_cache)
1267			mb_cache_entry_delete(ea_block_cache, hash,
1268					      bh->b_blocknr);
1269		get_bh(bh);
1270		unlock_buffer(bh);
1271
1272		if (ext4_has_feature_ea_inode(inode->i_sb))
1273			ext4_xattr_inode_dec_ref_all(handle, inode, bh,
1274						     BFIRST(bh),
1275						     true /* block_csum */,
1276						     ea_inode_array,
1277						     extra_credits,
1278						     true /* skip_quota */);
1279		ext4_free_blocks(handle, inode, bh, 0, 1,
1280				 EXT4_FREE_BLOCKS_METADATA |
1281				 EXT4_FREE_BLOCKS_FORGET);
1282	} else {
1283		ref--;
1284		BHDR(bh)->h_refcount = cpu_to_le32(ref);
1285		if (ref == EXT4_XATTR_REFCOUNT_MAX - 1) {
1286			struct mb_cache_entry *ce;
1287
1288			if (ea_block_cache) {
1289				ce = mb_cache_entry_get(ea_block_cache, hash,
1290							bh->b_blocknr);
1291				if (ce) {
1292					ce->e_reusable = 1;
1293					mb_cache_entry_put(ea_block_cache, ce);
1294				}
1295			}
1296		}
1297
1298		ext4_xattr_block_csum_set(inode, bh);
1299		/*
1300		 * Beware of this ugliness: Releasing of xattr block references
1301		 * from different inodes can race and so we have to protect
1302		 * from a race where someone else frees the block (and releases
1303		 * its journal_head) before we are done dirtying the buffer. In
1304		 * nojournal mode this race is harmless and we actually cannot
1305		 * call ext4_handle_dirty_metadata() with locked buffer as
1306		 * that function can call sync_dirty_buffer() so for that case
1307		 * we handle the dirtying after unlocking the buffer.
1308		 */
1309		if (ext4_handle_valid(handle))
1310			error = ext4_handle_dirty_metadata(handle, inode, bh);
1311		unlock_buffer(bh);
1312		if (!ext4_handle_valid(handle))
1313			error = ext4_handle_dirty_metadata(handle, inode, bh);
1314		if (IS_SYNC(inode))
1315			ext4_handle_sync(handle);
1316		dquot_free_block(inode, EXT4_C2B(EXT4_SB(inode->i_sb), 1));
1317		ea_bdebug(bh, "refcount now=%d; releasing",
1318			  le32_to_cpu(BHDR(bh)->h_refcount));
 
 
1319	}
 
1320out:
1321	ext4_std_error(inode->i_sb, error);
1322	return;
1323}
1324
1325/*
1326 * Find the available free space for EAs. This also returns the total number of
1327 * bytes used by EA entries.
1328 */
1329static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
1330				    size_t *min_offs, void *base, int *total)
1331{
1332	for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1333		if (!last->e_value_inum && last->e_value_size) {
 
1334			size_t offs = le16_to_cpu(last->e_value_offs);
1335			if (offs < *min_offs)
1336				*min_offs = offs;
1337		}
1338		if (total)
1339			*total += EXT4_XATTR_LEN(last->e_name_len);
1340	}
1341	return (*min_offs - ((void *)last - base) - sizeof(__u32));
1342}
1343
1344/*
1345 * Write the value of the EA in an inode.
1346 */
1347static int ext4_xattr_inode_write(handle_t *handle, struct inode *ea_inode,
1348				  const void *buf, int bufsize)
1349{
1350	struct buffer_head *bh = NULL;
1351	unsigned long block = 0;
1352	int blocksize = ea_inode->i_sb->s_blocksize;
1353	int max_blocks = (bufsize + blocksize - 1) >> ea_inode->i_blkbits;
1354	int csize, wsize = 0;
1355	int ret = 0;
1356	int retries = 0;
1357
1358retry:
1359	while (ret >= 0 && ret < max_blocks) {
1360		struct ext4_map_blocks map;
1361		map.m_lblk = block += ret;
1362		map.m_len = max_blocks -= ret;
1363
1364		ret = ext4_map_blocks(handle, ea_inode, &map,
1365				      EXT4_GET_BLOCKS_CREATE);
1366		if (ret <= 0) {
1367			ext4_mark_inode_dirty(handle, ea_inode);
1368			if (ret == -ENOSPC &&
1369			    ext4_should_retry_alloc(ea_inode->i_sb, &retries)) {
1370				ret = 0;
1371				goto retry;
1372			}
1373			break;
1374		}
1375	}
1376
1377	if (ret < 0)
1378		return ret;
1379
1380	block = 0;
1381	while (wsize < bufsize) {
1382		if (bh != NULL)
1383			brelse(bh);
1384		csize = (bufsize - wsize) > blocksize ? blocksize :
1385								bufsize - wsize;
1386		bh = ext4_getblk(handle, ea_inode, block, 0);
1387		if (IS_ERR(bh))
1388			return PTR_ERR(bh);
1389		ret = ext4_journal_get_write_access(handle, bh);
1390		if (ret)
1391			goto out;
1392
1393		memcpy(bh->b_data, buf, csize);
1394		set_buffer_uptodate(bh);
1395		ext4_handle_dirty_metadata(handle, ea_inode, bh);
1396
1397		buf += csize;
1398		wsize += csize;
1399		block += 1;
1400	}
1401
1402	inode_lock(ea_inode);
1403	i_size_write(ea_inode, wsize);
1404	ext4_update_i_disksize(ea_inode, wsize);
1405	inode_unlock(ea_inode);
1406
1407	ext4_mark_inode_dirty(handle, ea_inode);
1408
1409out:
1410	brelse(bh);
1411
1412	return ret;
1413}
1414
1415/*
1416 * Create an inode to store the value of a large EA.
1417 */
1418static struct inode *ext4_xattr_inode_create(handle_t *handle,
1419					     struct inode *inode, u32 hash)
1420{
1421	struct inode *ea_inode = NULL;
1422	uid_t owner[2] = { i_uid_read(inode), i_gid_read(inode) };
1423	int err;
1424
1425	/*
1426	 * Let the next inode be the goal, so we try and allocate the EA inode
1427	 * in the same group, or nearby one.
1428	 */
1429	ea_inode = ext4_new_inode(handle, inode->i_sb->s_root->d_inode,
1430				  S_IFREG | 0600, NULL, inode->i_ino + 1, owner,
1431				  EXT4_EA_INODE_FL);
1432	if (!IS_ERR(ea_inode)) {
1433		ea_inode->i_op = &ext4_file_inode_operations;
1434		ea_inode->i_fop = &ext4_file_operations;
1435		ext4_set_aops(ea_inode);
1436		ext4_xattr_inode_set_class(ea_inode);
1437		unlock_new_inode(ea_inode);
1438		ext4_xattr_inode_set_ref(ea_inode, 1);
1439		ext4_xattr_inode_set_hash(ea_inode, hash);
1440		err = ext4_mark_inode_dirty(handle, ea_inode);
1441		if (!err)
1442			err = ext4_inode_attach_jinode(ea_inode);
1443		if (err) {
1444			iput(ea_inode);
1445			return ERR_PTR(err);
1446		}
1447
1448		/*
1449		 * Xattr inodes are shared therefore quota charging is performed
1450		 * at a higher level.
1451		 */
1452		dquot_free_inode(ea_inode);
1453		dquot_drop(ea_inode);
1454		inode_lock(ea_inode);
1455		ea_inode->i_flags |= S_NOQUOTA;
1456		inode_unlock(ea_inode);
1457	}
1458
1459	return ea_inode;
1460}
1461
1462static struct inode *
1463ext4_xattr_inode_cache_find(struct inode *inode, const void *value,
1464			    size_t value_len, u32 hash)
1465{
1466	struct inode *ea_inode;
1467	struct mb_cache_entry *ce;
1468	struct mb_cache *ea_inode_cache = EA_INODE_CACHE(inode);
1469	void *ea_data;
1470
1471	if (!ea_inode_cache)
1472		return NULL;
1473
1474	ce = mb_cache_entry_find_first(ea_inode_cache, hash);
1475	if (!ce)
1476		return NULL;
1477
1478	ea_data = ext4_kvmalloc(value_len, GFP_NOFS);
1479	if (!ea_data) {
1480		mb_cache_entry_put(ea_inode_cache, ce);
1481		return NULL;
1482	}
1483
1484	while (ce) {
1485		ea_inode = ext4_iget(inode->i_sb, ce->e_value);
1486		if (!IS_ERR(ea_inode) &&
1487		    !is_bad_inode(ea_inode) &&
1488		    (EXT4_I(ea_inode)->i_flags & EXT4_EA_INODE_FL) &&
1489		    i_size_read(ea_inode) == value_len &&
1490		    !ext4_xattr_inode_read(ea_inode, ea_data, value_len) &&
1491		    !ext4_xattr_inode_verify_hashes(ea_inode, NULL, ea_data,
1492						    value_len) &&
1493		    !memcmp(value, ea_data, value_len)) {
1494			mb_cache_entry_touch(ea_inode_cache, ce);
1495			mb_cache_entry_put(ea_inode_cache, ce);
1496			kvfree(ea_data);
1497			return ea_inode;
1498		}
1499
1500		if (!IS_ERR(ea_inode))
1501			iput(ea_inode);
1502		ce = mb_cache_entry_find_next(ea_inode_cache, ce);
1503	}
1504	kvfree(ea_data);
1505	return NULL;
1506}
1507
1508/*
1509 * Add value of the EA in an inode.
1510 */
1511static int ext4_xattr_inode_lookup_create(handle_t *handle, struct inode *inode,
1512					  const void *value, size_t value_len,
1513					  struct inode **ret_inode)
1514{
1515	struct inode *ea_inode;
1516	u32 hash;
1517	int err;
1518
1519	hash = ext4_xattr_inode_hash(EXT4_SB(inode->i_sb), value, value_len);
1520	ea_inode = ext4_xattr_inode_cache_find(inode, value, value_len, hash);
1521	if (ea_inode) {
1522		err = ext4_xattr_inode_inc_ref(handle, ea_inode);
1523		if (err) {
1524			iput(ea_inode);
1525			return err;
1526		}
1527
1528		*ret_inode = ea_inode;
1529		return 0;
1530	}
1531
1532	/* Create an inode for the EA value */
1533	ea_inode = ext4_xattr_inode_create(handle, inode, hash);
1534	if (IS_ERR(ea_inode))
1535		return PTR_ERR(ea_inode);
1536
1537	err = ext4_xattr_inode_write(handle, ea_inode, value, value_len);
1538	if (err) {
1539		ext4_xattr_inode_dec_ref(handle, ea_inode);
1540		iput(ea_inode);
1541		return err;
1542	}
1543
1544	if (EA_INODE_CACHE(inode))
1545		mb_cache_entry_create(EA_INODE_CACHE(inode), GFP_NOFS, hash,
1546				      ea_inode->i_ino, true /* reusable */);
1547
1548	*ret_inode = ea_inode;
1549	return 0;
1550}
1551
1552/*
1553 * Reserve min(block_size/8, 1024) bytes for xattr entries/names if ea_inode
1554 * feature is enabled.
1555 */
1556#define EXT4_XATTR_BLOCK_RESERVE(inode)	min(i_blocksize(inode)/8, 1024U)
1557
1558static int ext4_xattr_set_entry(struct ext4_xattr_info *i,
1559				struct ext4_xattr_search *s,
1560				handle_t *handle, struct inode *inode,
1561				bool is_block)
1562{
1563	struct ext4_xattr_entry *last;
1564	struct ext4_xattr_entry *here = s->here;
1565	size_t min_offs = s->end - s->base, name_len = strlen(i->name);
1566	int in_inode = i->in_inode;
1567	struct inode *old_ea_inode = NULL;
1568	struct inode *new_ea_inode = NULL;
1569	size_t old_size, new_size;
1570	int ret;
1571
1572	/* Space used by old and new values. */
1573	old_size = (!s->not_found && !here->e_value_inum) ?
1574			EXT4_XATTR_SIZE(le32_to_cpu(here->e_value_size)) : 0;
1575	new_size = (i->value && !in_inode) ? EXT4_XATTR_SIZE(i->value_len) : 0;
1576
1577	/*
1578	 * Optimization for the simple case when old and new values have the
1579	 * same padded sizes. Not applicable if external inodes are involved.
1580	 */
1581	if (new_size && new_size == old_size) {
1582		size_t offs = le16_to_cpu(here->e_value_offs);
1583		void *val = s->base + offs;
1584
1585		here->e_value_size = cpu_to_le32(i->value_len);
1586		if (i->value == EXT4_ZERO_XATTR_VALUE) {
1587			memset(val, 0, new_size);
1588		} else {
1589			memcpy(val, i->value, i->value_len);
1590			/* Clear padding bytes. */
1591			memset(val + i->value_len, 0, new_size - i->value_len);
1592		}
1593		goto update_hash;
1594	}
1595
1596	/* Compute min_offs and last. */
1597	last = s->first;
1598	for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1599		if (!last->e_value_inum && last->e_value_size) {
1600			size_t offs = le16_to_cpu(last->e_value_offs);
1601			if (offs < min_offs)
1602				min_offs = offs;
1603		}
1604	}
1605
1606	/* Check whether we have enough space. */
1607	if (i->value) {
1608		size_t free;
1609
1610		free = min_offs - ((void *)last - s->base) - sizeof(__u32);
1611		if (!s->not_found)
1612			free += EXT4_XATTR_LEN(name_len) + old_size;
1613
1614		if (free < EXT4_XATTR_LEN(name_len) + new_size) {
1615			ret = -ENOSPC;
1616			goto out;
1617		}
1618
1619		/*
1620		 * If storing the value in an external inode is an option,
1621		 * reserve space for xattr entries/names in the external
1622		 * attribute block so that a long value does not occupy the
1623		 * whole space and prevent futher entries being added.
1624		 */
1625		if (ext4_has_feature_ea_inode(inode->i_sb) &&
1626		    new_size && is_block &&
1627		    (min_offs + old_size - new_size) <
1628					EXT4_XATTR_BLOCK_RESERVE(inode)) {
1629			ret = -ENOSPC;
1630			goto out;
1631		}
 
1632	}
1633
1634	/*
1635	 * Getting access to old and new ea inodes is subject to failures.
1636	 * Finish that work before doing any modifications to the xattr data.
1637	 */
1638	if (!s->not_found && here->e_value_inum) {
1639		ret = ext4_xattr_inode_iget(inode,
1640					    le32_to_cpu(here->e_value_inum),
1641					    le32_to_cpu(here->e_hash),
1642					    &old_ea_inode);
1643		if (ret) {
1644			old_ea_inode = NULL;
1645			goto out;
1646		}
1647	}
1648	if (i->value && in_inode) {
1649		WARN_ON_ONCE(!i->value_len);
1650
1651		ret = ext4_xattr_inode_alloc_quota(inode, i->value_len);
1652		if (ret)
1653			goto out;
1654
1655		ret = ext4_xattr_inode_lookup_create(handle, inode, i->value,
1656						     i->value_len,
1657						     &new_ea_inode);
1658		if (ret) {
1659			new_ea_inode = NULL;
1660			ext4_xattr_inode_free_quota(inode, NULL, i->value_len);
1661			goto out;
1662		}
1663	}
1664
1665	if (old_ea_inode) {
1666		/* We are ready to release ref count on the old_ea_inode. */
1667		ret = ext4_xattr_inode_dec_ref(handle, old_ea_inode);
1668		if (ret) {
1669			/* Release newly required ref count on new_ea_inode. */
1670			if (new_ea_inode) {
1671				int err;
1672
1673				err = ext4_xattr_inode_dec_ref(handle,
1674							       new_ea_inode);
1675				if (err)
1676					ext4_warning_inode(new_ea_inode,
1677						  "dec ref new_ea_inode err=%d",
1678						  err);
1679				ext4_xattr_inode_free_quota(inode, new_ea_inode,
1680							    i->value_len);
1681			}
1682			goto out;
1683		}
1684
1685		ext4_xattr_inode_free_quota(inode, old_ea_inode,
1686					    le32_to_cpu(here->e_value_size));
1687	}
1688
1689	/* No failures allowed past this point. */
1690
1691	if (!s->not_found && here->e_value_offs) {
1692		/* Remove the old value. */
1693		void *first_val = s->base + min_offs;
1694		size_t offs = le16_to_cpu(here->e_value_offs);
1695		void *val = s->base + offs;
1696
1697		memmove(first_val + old_size, first_val, val - first_val);
1698		memset(first_val, 0, old_size);
1699		min_offs += old_size;
1700
1701		/* Adjust all value offsets. */
1702		last = s->first;
1703		while (!IS_LAST_ENTRY(last)) {
1704			size_t o = le16_to_cpu(last->e_value_offs);
1705
1706			if (!last->e_value_inum &&
1707			    last->e_value_size && o < offs)
1708				last->e_value_offs = cpu_to_le16(o + old_size);
1709			last = EXT4_XATTR_NEXT(last);
1710		}
1711	}
1712
1713	if (!i->value) {
1714		/* Remove old name. */
1715		size_t size = EXT4_XATTR_LEN(name_len);
1716
1717		last = ENTRY((void *)last - size);
1718		memmove(here, (void *)here + size,
1719			(void *)last - (void *)here + sizeof(__u32));
1720		memset(last, 0, size);
1721	} else if (s->not_found) {
1722		/* Insert new name. */
1723		size_t size = EXT4_XATTR_LEN(name_len);
1724		size_t rest = (void *)last - (void *)here + sizeof(__u32);
1725
1726		memmove((void *)here + size, here, rest);
1727		memset(here, 0, size);
1728		here->e_name_index = i->name_index;
1729		here->e_name_len = name_len;
1730		memcpy(here->e_name, i->name, name_len);
1731	} else {
1732		/* This is an update, reset value info. */
1733		here->e_value_inum = 0;
1734		here->e_value_offs = 0;
1735		here->e_value_size = 0;
1736	}
 
 
 
 
 
 
 
 
 
 
 
 
1737
1738	if (i->value) {
1739		/* Insert new value. */
1740		if (in_inode) {
1741			here->e_value_inum = cpu_to_le32(new_ea_inode->i_ino);
1742		} else if (i->value_len) {
1743			void *val = s->base + min_offs - new_size;
1744
1745			here->e_value_offs = cpu_to_le16(min_offs - new_size);
1746			if (i->value == EXT4_ZERO_XATTR_VALUE) {
1747				memset(val, 0, new_size);
1748			} else {
1749				memcpy(val, i->value, i->value_len);
1750				/* Clear padding bytes. */
1751				memset(val + i->value_len, 0,
1752				       new_size - i->value_len);
 
1753			}
1754		}
1755		here->e_value_size = cpu_to_le32(i->value_len);
 
 
 
 
 
 
 
1756	}
1757
1758update_hash:
1759	if (i->value) {
1760		__le32 hash = 0;
1761
1762		/* Entry hash calculation. */
1763		if (in_inode) {
1764			__le32 crc32c_hash;
1765
1766			/*
1767			 * Feed crc32c hash instead of the raw value for entry
1768			 * hash calculation. This is to avoid walking
1769			 * potentially long value buffer again.
1770			 */
1771			crc32c_hash = cpu_to_le32(
1772				       ext4_xattr_inode_get_hash(new_ea_inode));
1773			hash = ext4_xattr_hash_entry(here->e_name,
1774						     here->e_name_len,
1775						     &crc32c_hash, 1);
1776		} else if (is_block) {
1777			__le32 *value = s->base + le16_to_cpu(
1778							here->e_value_offs);
1779
1780			hash = ext4_xattr_hash_entry(here->e_name,
1781						     here->e_name_len, value,
1782						     new_size >> 2);
1783		}
1784		here->e_hash = hash;
1785	}
1786
1787	if (is_block)
1788		ext4_xattr_rehash((struct ext4_xattr_header *)s->base);
1789
1790	ret = 0;
1791out:
1792	iput(old_ea_inode);
1793	iput(new_ea_inode);
1794	return ret;
1795}
1796
1797struct ext4_xattr_block_find {
1798	struct ext4_xattr_search s;
1799	struct buffer_head *bh;
1800};
1801
1802static int
1803ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
1804		      struct ext4_xattr_block_find *bs)
1805{
1806	struct super_block *sb = inode->i_sb;
1807	int error;
1808
1809	ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
1810		  i->name_index, i->name, i->value, (long)i->value_len);
1811
1812	if (EXT4_I(inode)->i_file_acl) {
1813		/* The inode already has an extended attribute block. */
1814		bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
1815		error = -EIO;
1816		if (!bs->bh)
1817			goto cleanup;
1818		ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
1819			atomic_read(&(bs->bh->b_count)),
1820			le32_to_cpu(BHDR(bs->bh)->h_refcount));
1821		error = ext4_xattr_check_block(inode, bs->bh);
1822		if (error)
 
 
1823			goto cleanup;
 
1824		/* Find the named attribute. */
1825		bs->s.base = BHDR(bs->bh);
1826		bs->s.first = BFIRST(bs->bh);
1827		bs->s.end = bs->bh->b_data + bs->bh->b_size;
1828		bs->s.here = bs->s.first;
1829		error = xattr_find_entry(inode, &bs->s.here, bs->s.end,
1830					 i->name_index, i->name, 1);
1831		if (error && error != -ENODATA)
1832			goto cleanup;
1833		bs->s.not_found = error;
1834	}
1835	error = 0;
1836
1837cleanup:
1838	return error;
1839}
1840
1841static int
1842ext4_xattr_block_set(handle_t *handle, struct inode *inode,
1843		     struct ext4_xattr_info *i,
1844		     struct ext4_xattr_block_find *bs)
1845{
1846	struct super_block *sb = inode->i_sb;
1847	struct buffer_head *new_bh = NULL;
1848	struct ext4_xattr_search s_copy = bs->s;
1849	struct ext4_xattr_search *s = &s_copy;
1850	struct mb_cache_entry *ce = NULL;
1851	int error = 0;
1852	struct mb_cache *ea_block_cache = EA_BLOCK_CACHE(inode);
1853	struct inode *ea_inode = NULL, *tmp_inode;
1854	size_t old_ea_inode_quota = 0;
1855	unsigned int ea_ino;
1856
1857
1858#define header(x) ((struct ext4_xattr_header *)(x))
1859
 
 
1860	if (s->base) {
1861		BUFFER_TRACE(bs->bh, "get_write_access");
 
1862		error = ext4_journal_get_write_access(handle, bs->bh);
1863		if (error)
1864			goto cleanup;
1865		lock_buffer(bs->bh);
1866
1867		if (header(s->base)->h_refcount == cpu_to_le32(1)) {
1868			__u32 hash = le32_to_cpu(BHDR(bs->bh)->h_hash);
1869
1870			/*
1871			 * This must happen under buffer lock for
1872			 * ext4_xattr_block_set() to reliably detect modified
1873			 * block
1874			 */
1875			if (ea_block_cache)
1876				mb_cache_entry_delete(ea_block_cache, hash,
1877						      bs->bh->b_blocknr);
1878			ea_bdebug(bs->bh, "modifying in-place");
1879			error = ext4_xattr_set_entry(i, s, handle, inode,
1880						     true /* is_block */);
1881			ext4_xattr_block_csum_set(inode, bs->bh);
 
 
 
 
1882			unlock_buffer(bs->bh);
1883			if (error == -EFSCORRUPTED)
1884				goto bad_block;
1885			if (!error)
1886				error = ext4_handle_dirty_metadata(handle,
1887								   inode,
1888								   bs->bh);
1889			if (error)
1890				goto cleanup;
1891			goto inserted;
1892		} else {
1893			int offset = (char *)s->here - bs->bh->b_data;
1894
1895			unlock_buffer(bs->bh);
 
 
 
 
 
1896			ea_bdebug(bs->bh, "cloning");
1897			s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
1898			error = -ENOMEM;
1899			if (s->base == NULL)
1900				goto cleanup;
1901			memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
1902			s->first = ENTRY(header(s->base)+1);
1903			header(s->base)->h_refcount = cpu_to_le32(1);
1904			s->here = ENTRY(s->base + offset);
1905			s->end = s->base + bs->bh->b_size;
1906
1907			/*
1908			 * If existing entry points to an xattr inode, we need
1909			 * to prevent ext4_xattr_set_entry() from decrementing
1910			 * ref count on it because the reference belongs to the
1911			 * original block. In this case, make the entry look
1912			 * like it has an empty value.
1913			 */
1914			if (!s->not_found && s->here->e_value_inum) {
1915				ea_ino = le32_to_cpu(s->here->e_value_inum);
1916				error = ext4_xattr_inode_iget(inode, ea_ino,
1917					      le32_to_cpu(s->here->e_hash),
1918					      &tmp_inode);
1919				if (error)
1920					goto cleanup;
1921
1922				if (!ext4_test_inode_state(tmp_inode,
1923						EXT4_STATE_LUSTRE_EA_INODE)) {
1924					/*
1925					 * Defer quota free call for previous
1926					 * inode until success is guaranteed.
1927					 */
1928					old_ea_inode_quota = le32_to_cpu(
1929							s->here->e_value_size);
1930				}
1931				iput(tmp_inode);
1932
1933				s->here->e_value_inum = 0;
1934				s->here->e_value_size = 0;
1935			}
1936		}
1937	} else {
1938		/* Allocate a buffer where we construct the new block. */
1939		s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
1940		/* assert(header == s->base) */
1941		error = -ENOMEM;
1942		if (s->base == NULL)
1943			goto cleanup;
1944		header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1945		header(s->base)->h_blocks = cpu_to_le32(1);
1946		header(s->base)->h_refcount = cpu_to_le32(1);
1947		s->first = ENTRY(header(s->base)+1);
1948		s->here = ENTRY(header(s->base)+1);
1949		s->end = s->base + sb->s_blocksize;
1950	}
1951
1952	error = ext4_xattr_set_entry(i, s, handle, inode, true /* is_block */);
1953	if (error == -EFSCORRUPTED)
1954		goto bad_block;
1955	if (error)
1956		goto cleanup;
1957
1958	if (i->value && s->here->e_value_inum) {
1959		/*
1960		 * A ref count on ea_inode has been taken as part of the call to
1961		 * ext4_xattr_set_entry() above. We would like to drop this
1962		 * extra ref but we have to wait until the xattr block is
1963		 * initialized and has its own ref count on the ea_inode.
1964		 */
1965		ea_ino = le32_to_cpu(s->here->e_value_inum);
1966		error = ext4_xattr_inode_iget(inode, ea_ino,
1967					      le32_to_cpu(s->here->e_hash),
1968					      &ea_inode);
1969		if (error) {
1970			ea_inode = NULL;
1971			goto cleanup;
1972		}
1973	}
1974
1975inserted:
1976	if (!IS_LAST_ENTRY(s->first)) {
1977		new_bh = ext4_xattr_block_cache_find(inode, header(s->base),
1978						     &ce);
1979		if (new_bh) {
1980			/* We found an identical block in the cache. */
1981			if (new_bh == bs->bh)
1982				ea_bdebug(new_bh, "keeping");
1983			else {
1984				u32 ref;
1985
1986				WARN_ON_ONCE(dquot_initialize_needed(inode));
1987
1988				/* The old block is released after updating
1989				   the inode. */
1990				error = dquot_alloc_block(inode,
1991						EXT4_C2B(EXT4_SB(sb), 1));
1992				if (error)
1993					goto cleanup;
1994				BUFFER_TRACE(new_bh, "get_write_access");
1995				error = ext4_journal_get_write_access(handle,
1996								      new_bh);
1997				if (error)
1998					goto cleanup_dquot;
1999				lock_buffer(new_bh);
2000				/*
2001				 * We have to be careful about races with
2002				 * freeing, rehashing or adding references to
2003				 * xattr block. Once we hold buffer lock xattr
2004				 * block's state is stable so we can check
2005				 * whether the block got freed / rehashed or
2006				 * not.  Since we unhash mbcache entry under
2007				 * buffer lock when freeing / rehashing xattr
2008				 * block, checking whether entry is still
2009				 * hashed is reliable. Same rules hold for
2010				 * e_reusable handling.
2011				 */
2012				if (hlist_bl_unhashed(&ce->e_hash_list) ||
2013				    !ce->e_reusable) {
2014					/*
2015					 * Undo everything and check mbcache
2016					 * again.
2017					 */
2018					unlock_buffer(new_bh);
2019					dquot_free_block(inode,
2020							 EXT4_C2B(EXT4_SB(sb),
2021								  1));
2022					brelse(new_bh);
2023					mb_cache_entry_put(ea_block_cache, ce);
2024					ce = NULL;
2025					new_bh = NULL;
2026					goto inserted;
2027				}
2028				ref = le32_to_cpu(BHDR(new_bh)->h_refcount) + 1;
2029				BHDR(new_bh)->h_refcount = cpu_to_le32(ref);
2030				if (ref >= EXT4_XATTR_REFCOUNT_MAX)
2031					ce->e_reusable = 0;
2032				ea_bdebug(new_bh, "reusing; refcount now=%d",
2033					  ref);
2034				ext4_xattr_block_csum_set(inode, new_bh);
2035				unlock_buffer(new_bh);
2036				error = ext4_handle_dirty_metadata(handle,
2037								   inode,
2038								   new_bh);
2039				if (error)
2040					goto cleanup_dquot;
2041			}
2042			mb_cache_entry_touch(ea_block_cache, ce);
2043			mb_cache_entry_put(ea_block_cache, ce);
2044			ce = NULL;
2045		} else if (bs->bh && s->base == bs->bh->b_data) {
2046			/* We were modifying this block in-place. */
2047			ea_bdebug(bs->bh, "keeping this block");
2048			ext4_xattr_block_cache_insert(ea_block_cache, bs->bh);
2049			new_bh = bs->bh;
2050			get_bh(new_bh);
2051		} else {
2052			/* We need to allocate a new block */
2053			ext4_fsblk_t goal, block;
2054
2055			WARN_ON_ONCE(dquot_initialize_needed(inode));
2056
2057			goal = ext4_group_first_block_no(sb,
2058						EXT4_I(inode)->i_block_group);
2059
2060			/* non-extent files can't have physical blocks past 2^32 */
2061			if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
2062				goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
2063
2064			block = ext4_new_meta_blocks(handle, inode, goal, 0,
2065						     NULL, &error);
2066			if (error)
2067				goto cleanup;
2068
2069			if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
2070				BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
2071
2072			ea_idebug(inode, "creating block %llu",
2073				  (unsigned long long)block);
2074
2075			new_bh = sb_getblk(sb, block);
2076			if (unlikely(!new_bh)) {
2077				error = -ENOMEM;
2078getblk_failed:
2079				ext4_free_blocks(handle, inode, NULL, block, 1,
2080						 EXT4_FREE_BLOCKS_METADATA);
 
2081				goto cleanup;
2082			}
2083			error = ext4_xattr_inode_inc_ref_all(handle, inode,
2084						      ENTRY(header(s->base)+1));
2085			if (error)
2086				goto getblk_failed;
2087			if (ea_inode) {
2088				/* Drop the extra ref on ea_inode. */
2089				error = ext4_xattr_inode_dec_ref(handle,
2090								 ea_inode);
2091				if (error)
2092					ext4_warning_inode(ea_inode,
2093							   "dec ref error=%d",
2094							   error);
2095				iput(ea_inode);
2096				ea_inode = NULL;
2097			}
2098
2099			lock_buffer(new_bh);
2100			error = ext4_journal_get_create_access(handle, new_bh);
2101			if (error) {
2102				unlock_buffer(new_bh);
2103				error = -EIO;
2104				goto getblk_failed;
2105			}
2106			memcpy(new_bh->b_data, s->base, new_bh->b_size);
2107			ext4_xattr_block_csum_set(inode, new_bh);
2108			set_buffer_uptodate(new_bh);
2109			unlock_buffer(new_bh);
2110			ext4_xattr_block_cache_insert(ea_block_cache, new_bh);
2111			error = ext4_handle_dirty_metadata(handle, inode,
2112							   new_bh);
2113			if (error)
2114				goto cleanup;
2115		}
2116	}
2117
2118	if (old_ea_inode_quota)
2119		ext4_xattr_inode_free_quota(inode, NULL, old_ea_inode_quota);
2120
2121	/* Update the inode. */
2122	EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
2123
2124	/* Drop the previous xattr block. */
2125	if (bs->bh && bs->bh != new_bh) {
2126		struct ext4_xattr_inode_array *ea_inode_array = NULL;
2127
2128		ext4_xattr_release_block(handle, inode, bs->bh,
2129					 &ea_inode_array,
2130					 0 /* extra_credits */);
2131		ext4_xattr_inode_array_free(ea_inode_array);
2132	}
2133	error = 0;
2134
2135cleanup:
2136	if (ea_inode) {
2137		int error2;
2138
2139		error2 = ext4_xattr_inode_dec_ref(handle, ea_inode);
2140		if (error2)
2141			ext4_warning_inode(ea_inode, "dec ref error=%d",
2142					   error2);
2143
2144		/* If there was an error, revert the quota charge. */
2145		if (error)
2146			ext4_xattr_inode_free_quota(inode, ea_inode,
2147						    i_size_read(ea_inode));
2148		iput(ea_inode);
2149	}
2150	if (ce)
2151		mb_cache_entry_put(ea_block_cache, ce);
2152	brelse(new_bh);
2153	if (!(bs->bh && s->base == bs->bh->b_data))
2154		kfree(s->base);
2155
2156	return error;
2157
2158cleanup_dquot:
2159	dquot_free_block(inode, EXT4_C2B(EXT4_SB(sb), 1));
2160	goto cleanup;
2161
2162bad_block:
2163	EXT4_ERROR_INODE(inode, "bad block %llu",
2164			 EXT4_I(inode)->i_file_acl);
2165	goto cleanup;
2166
2167#undef header
2168}
2169
2170int ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
2171			  struct ext4_xattr_ibody_find *is)
 
 
 
 
 
 
2172{
2173	struct ext4_xattr_ibody_header *header;
2174	struct ext4_inode *raw_inode;
2175	int error;
2176
2177	if (EXT4_I(inode)->i_extra_isize == 0)
2178		return 0;
2179	raw_inode = ext4_raw_inode(&is->iloc);
2180	header = IHDR(inode, raw_inode);
2181	is->s.base = is->s.first = IFIRST(header);
2182	is->s.here = is->s.first;
2183	is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
2184	if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
2185		error = xattr_check_inode(inode, header, is->s.end);
2186		if (error)
2187			return error;
2188		/* Find the named attribute. */
2189		error = xattr_find_entry(inode, &is->s.here, is->s.end,
2190					 i->name_index, i->name, 0);
 
2191		if (error && error != -ENODATA)
2192			return error;
2193		is->s.not_found = error;
2194	}
2195	return 0;
2196}
2197
2198int ext4_xattr_ibody_inline_set(handle_t *handle, struct inode *inode,
2199				struct ext4_xattr_info *i,
2200				struct ext4_xattr_ibody_find *is)
2201{
2202	struct ext4_xattr_ibody_header *header;
2203	struct ext4_xattr_search *s = &is->s;
2204	int error;
2205
2206	if (EXT4_I(inode)->i_extra_isize == 0)
2207		return -ENOSPC;
2208	error = ext4_xattr_set_entry(i, s, handle, inode, false /* is_block */);
2209	if (error) {
2210		if (error == -ENOSPC &&
2211		    ext4_has_inline_data(inode)) {
2212			error = ext4_try_to_evict_inline_data(handle, inode,
2213					EXT4_XATTR_LEN(strlen(i->name) +
2214					EXT4_XATTR_SIZE(i->value_len)));
2215			if (error)
2216				return error;
2217			error = ext4_xattr_ibody_find(inode, i, is);
2218			if (error)
2219				return error;
2220			error = ext4_xattr_set_entry(i, s, handle, inode,
2221						     false /* is_block */);
2222		}
2223		if (error)
2224			return error;
2225	}
2226	header = IHDR(inode, ext4_raw_inode(&is->iloc));
2227	if (!IS_LAST_ENTRY(s->first)) {
2228		header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
2229		ext4_set_inode_state(inode, EXT4_STATE_XATTR);
2230	} else {
2231		header->h_magic = cpu_to_le32(0);
2232		ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
2233	}
2234	return 0;
2235}
2236
2237static int ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
2238				struct ext4_xattr_info *i,
2239				struct ext4_xattr_ibody_find *is)
2240{
2241	struct ext4_xattr_ibody_header *header;
2242	struct ext4_xattr_search *s = &is->s;
2243	int error;
2244
2245	if (EXT4_I(inode)->i_extra_isize == 0)
2246		return -ENOSPC;
2247	error = ext4_xattr_set_entry(i, s, handle, inode, false /* is_block */);
2248	if (error)
2249		return error;
2250	header = IHDR(inode, ext4_raw_inode(&is->iloc));
2251	if (!IS_LAST_ENTRY(s->first)) {
2252		header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
2253		ext4_set_inode_state(inode, EXT4_STATE_XATTR);
2254	} else {
2255		header->h_magic = cpu_to_le32(0);
2256		ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
2257	}
2258	return 0;
2259}
2260
2261static int ext4_xattr_value_same(struct ext4_xattr_search *s,
2262				 struct ext4_xattr_info *i)
2263{
2264	void *value;
2265
2266	/* When e_value_inum is set the value is stored externally. */
2267	if (s->here->e_value_inum)
2268		return 0;
2269	if (le32_to_cpu(s->here->e_value_size) != i->value_len)
2270		return 0;
2271	value = ((void *)s->base) + le16_to_cpu(s->here->e_value_offs);
2272	return !memcmp(value, i->value, i->value_len);
2273}
2274
2275static struct buffer_head *ext4_xattr_get_block(struct inode *inode)
2276{
2277	struct buffer_head *bh;
2278	int error;
2279
2280	if (!EXT4_I(inode)->i_file_acl)
2281		return NULL;
2282	bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
2283	if (!bh)
2284		return ERR_PTR(-EIO);
2285	error = ext4_xattr_check_block(inode, bh);
2286	if (error)
2287		return ERR_PTR(error);
2288	return bh;
2289}
2290
2291/*
2292 * ext4_xattr_set_handle()
2293 *
2294 * Create, replace or remove an extended attribute for this inode.  Value
2295 * is NULL to remove an existing extended attribute, and non-NULL to
2296 * either replace an existing extended attribute, or create a new extended
2297 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
2298 * specify that an extended attribute must exist and must not exist
2299 * previous to the call, respectively.
2300 *
2301 * Returns 0, or a negative error number on failure.
2302 */
2303int
2304ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
2305		      const char *name, const void *value, size_t value_len,
2306		      int flags)
2307{
2308	struct ext4_xattr_info i = {
2309		.name_index = name_index,
2310		.name = name,
2311		.value = value,
2312		.value_len = value_len,
2313		.in_inode = 0,
2314	};
2315	struct ext4_xattr_ibody_find is = {
2316		.s = { .not_found = -ENODATA, },
2317	};
2318	struct ext4_xattr_block_find bs = {
2319		.s = { .not_found = -ENODATA, },
2320	};
2321	int no_expand;
2322	int error;
2323
2324	if (!name)
2325		return -EINVAL;
2326	if (strlen(name) > 255)
2327		return -ERANGE;
 
 
 
2328
2329	ext4_write_lock_xattr(inode, &no_expand);
2330
2331	/* Check journal credits under write lock. */
2332	if (ext4_handle_valid(handle)) {
2333		struct buffer_head *bh;
2334		int credits;
2335
2336		bh = ext4_xattr_get_block(inode);
2337		if (IS_ERR(bh)) {
2338			error = PTR_ERR(bh);
2339			goto cleanup;
2340		}
2341
2342		credits = __ext4_xattr_set_credits(inode->i_sb, inode, bh,
2343						   value_len,
2344						   flags & XATTR_CREATE);
2345		brelse(bh);
2346
2347		if (!ext4_handle_has_enough_credits(handle, credits)) {
2348			error = -ENOSPC;
2349			goto cleanup;
2350		}
2351	}
2352
2353	error = ext4_reserve_inode_write(handle, inode, &is.iloc);
2354	if (error)
2355		goto cleanup;
2356
2357	if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
2358		struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
2359		memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
2360		ext4_clear_inode_state(inode, EXT4_STATE_NEW);
2361	}
2362
2363	error = ext4_xattr_ibody_find(inode, &i, &is);
2364	if (error)
2365		goto cleanup;
2366	if (is.s.not_found)
2367		error = ext4_xattr_block_find(inode, &i, &bs);
2368	if (error)
2369		goto cleanup;
2370	if (is.s.not_found && bs.s.not_found) {
2371		error = -ENODATA;
2372		if (flags & XATTR_REPLACE)
2373			goto cleanup;
2374		error = 0;
2375		if (!value)
2376			goto cleanup;
2377	} else {
2378		error = -EEXIST;
2379		if (flags & XATTR_CREATE)
2380			goto cleanup;
2381	}
2382
2383	if (!value) {
2384		if (!is.s.not_found)
2385			error = ext4_xattr_ibody_set(handle, inode, &i, &is);
2386		else if (!bs.s.not_found)
2387			error = ext4_xattr_block_set(handle, inode, &i, &bs);
2388	} else {
2389		error = 0;
2390		/* Xattr value did not change? Save us some work and bail out */
2391		if (!is.s.not_found && ext4_xattr_value_same(&is.s, &i))
2392			goto cleanup;
2393		if (!bs.s.not_found && ext4_xattr_value_same(&bs.s, &i))
2394			goto cleanup;
2395
2396		if (ext4_has_feature_ea_inode(inode->i_sb) &&
2397		    (EXT4_XATTR_SIZE(i.value_len) >
2398			EXT4_XATTR_MIN_LARGE_EA_SIZE(inode->i_sb->s_blocksize)))
2399			i.in_inode = 1;
2400retry_inode:
2401		error = ext4_xattr_ibody_set(handle, inode, &i, &is);
2402		if (!error && !bs.s.not_found) {
2403			i.value = NULL;
2404			error = ext4_xattr_block_set(handle, inode, &i, &bs);
2405		} else if (error == -ENOSPC) {
2406			if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
2407				error = ext4_xattr_block_find(inode, &i, &bs);
2408				if (error)
2409					goto cleanup;
2410			}
2411			error = ext4_xattr_block_set(handle, inode, &i, &bs);
2412			if (!error && !is.s.not_found) {
 
 
2413				i.value = NULL;
2414				error = ext4_xattr_ibody_set(handle, inode, &i,
2415							     &is);
2416			} else if (error == -ENOSPC) {
2417				/*
2418				 * Xattr does not fit in the block, store at
2419				 * external inode if possible.
2420				 */
2421				if (ext4_has_feature_ea_inode(inode->i_sb) &&
2422				    !i.in_inode) {
2423					i.in_inode = 1;
2424					goto retry_inode;
2425				}
2426			}
2427		}
2428	}
2429	if (!error) {
2430		ext4_xattr_update_super_block(handle, inode->i_sb);
2431		inode->i_ctime = current_time(inode);
2432		if (!value)
2433			no_expand = 0;
2434		error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
2435		/*
2436		 * The bh is consumed by ext4_mark_iloc_dirty, even with
2437		 * error != 0.
2438		 */
2439		is.iloc.bh = NULL;
2440		if (IS_SYNC(inode))
2441			ext4_handle_sync(handle);
2442	}
2443
2444cleanup:
2445	brelse(is.iloc.bh);
2446	brelse(bs.bh);
2447	ext4_write_unlock_xattr(inode, &no_expand);
 
 
2448	return error;
2449}
2450
2451int ext4_xattr_set_credits(struct inode *inode, size_t value_len,
2452			   bool is_create, int *credits)
2453{
2454	struct buffer_head *bh;
2455	int err;
2456
2457	*credits = 0;
2458
2459	if (!EXT4_SB(inode->i_sb)->s_journal)
2460		return 0;
2461
2462	down_read(&EXT4_I(inode)->xattr_sem);
2463
2464	bh = ext4_xattr_get_block(inode);
2465	if (IS_ERR(bh)) {
2466		err = PTR_ERR(bh);
2467	} else {
2468		*credits = __ext4_xattr_set_credits(inode->i_sb, inode, bh,
2469						    value_len, is_create);
2470		brelse(bh);
2471		err = 0;
2472	}
2473
2474	up_read(&EXT4_I(inode)->xattr_sem);
2475	return err;
2476}
2477
2478/*
2479 * ext4_xattr_set()
2480 *
2481 * Like ext4_xattr_set_handle, but start from an inode. This extended
2482 * attribute modification is a filesystem transaction by itself.
2483 *
2484 * Returns 0, or a negative error number on failure.
2485 */
2486int
2487ext4_xattr_set(struct inode *inode, int name_index, const char *name,
2488	       const void *value, size_t value_len, int flags)
2489{
2490	handle_t *handle;
2491	struct super_block *sb = inode->i_sb;
2492	int error, retries = 0;
2493	int credits;
2494
2495	error = dquot_initialize(inode);
2496	if (error)
2497		return error;
2498
2499retry:
2500	error = ext4_xattr_set_credits(inode, value_len, flags & XATTR_CREATE,
2501				       &credits);
2502	if (error)
2503		return error;
2504
2505	handle = ext4_journal_start(inode, EXT4_HT_XATTR, credits);
2506	if (IS_ERR(handle)) {
2507		error = PTR_ERR(handle);
2508	} else {
2509		int error2;
2510
2511		error = ext4_xattr_set_handle(handle, inode, name_index, name,
2512					      value, value_len, flags);
2513		error2 = ext4_journal_stop(handle);
2514		if (error == -ENOSPC &&
2515		    ext4_should_retry_alloc(sb, &retries))
2516			goto retry;
2517		if (error == 0)
2518			error = error2;
2519	}
2520
2521	return error;
2522}
2523
2524/*
2525 * Shift the EA entries in the inode to create space for the increased
2526 * i_extra_isize.
2527 */
2528static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
2529				     int value_offs_shift, void *to,
2530				     void *from, size_t n)
2531{
2532	struct ext4_xattr_entry *last = entry;
2533	int new_offs;
2534
2535	/* We always shift xattr headers further thus offsets get lower */
2536	BUG_ON(value_offs_shift > 0);
2537
2538	/* Adjust the value offsets of the entries */
2539	for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
2540		if (!last->e_value_inum && last->e_value_size) {
2541			new_offs = le16_to_cpu(last->e_value_offs) +
2542							value_offs_shift;
 
 
2543			last->e_value_offs = cpu_to_le16(new_offs);
2544		}
2545	}
2546	/* Shift the entries by n bytes */
2547	memmove(to, from, n);
2548}
2549
2550/*
2551 * Move xattr pointed to by 'entry' from inode into external xattr block
2552 */
2553static int ext4_xattr_move_to_block(handle_t *handle, struct inode *inode,
2554				    struct ext4_inode *raw_inode,
2555				    struct ext4_xattr_entry *entry)
2556{
2557	struct ext4_xattr_ibody_find *is = NULL;
2558	struct ext4_xattr_block_find *bs = NULL;
2559	char *buffer = NULL, *b_entry_name = NULL;
2560	size_t value_size = le32_to_cpu(entry->e_value_size);
2561	struct ext4_xattr_info i = {
2562		.value = NULL,
2563		.value_len = 0,
2564		.name_index = entry->e_name_index,
2565		.in_inode = !!entry->e_value_inum,
2566	};
2567	struct ext4_xattr_ibody_header *header = IHDR(inode, raw_inode);
2568	int error;
2569
2570	is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
2571	bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
2572	buffer = kmalloc(value_size, GFP_NOFS);
2573	b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
2574	if (!is || !bs || !buffer || !b_entry_name) {
2575		error = -ENOMEM;
2576		goto out;
2577	}
2578
2579	is->s.not_found = -ENODATA;
2580	bs->s.not_found = -ENODATA;
2581	is->iloc.bh = NULL;
2582	bs->bh = NULL;
2583
2584	/* Save the entry name and the entry value */
2585	if (entry->e_value_inum) {
2586		error = ext4_xattr_inode_get(inode, entry, buffer, value_size);
2587		if (error)
2588			goto out;
2589	} else {
2590		size_t value_offs = le16_to_cpu(entry->e_value_offs);
2591		memcpy(buffer, (void *)IFIRST(header) + value_offs, value_size);
2592	}
2593
2594	memcpy(b_entry_name, entry->e_name, entry->e_name_len);
2595	b_entry_name[entry->e_name_len] = '\0';
2596	i.name = b_entry_name;
2597
2598	error = ext4_get_inode_loc(inode, &is->iloc);
2599	if (error)
2600		goto out;
2601
2602	error = ext4_xattr_ibody_find(inode, &i, is);
2603	if (error)
2604		goto out;
2605
2606	/* Remove the chosen entry from the inode */
2607	error = ext4_xattr_ibody_set(handle, inode, &i, is);
2608	if (error)
2609		goto out;
2610
2611	i.value = buffer;
2612	i.value_len = value_size;
2613	error = ext4_xattr_block_find(inode, &i, bs);
2614	if (error)
2615		goto out;
2616
2617	/* Add entry which was removed from the inode into the block */
2618	error = ext4_xattr_block_set(handle, inode, &i, bs);
2619	if (error)
2620		goto out;
2621	error = 0;
2622out:
2623	kfree(b_entry_name);
2624	kfree(buffer);
2625	if (is)
2626		brelse(is->iloc.bh);
2627	kfree(is);
2628	kfree(bs);
2629
2630	return error;
2631}
2632
2633static int ext4_xattr_make_inode_space(handle_t *handle, struct inode *inode,
2634				       struct ext4_inode *raw_inode,
2635				       int isize_diff, size_t ifree,
2636				       size_t bfree, int *total_ino)
2637{
2638	struct ext4_xattr_ibody_header *header = IHDR(inode, raw_inode);
2639	struct ext4_xattr_entry *small_entry;
2640	struct ext4_xattr_entry *entry;
2641	struct ext4_xattr_entry *last;
2642	unsigned int entry_size;	/* EA entry size */
2643	unsigned int total_size;	/* EA entry size + value size */
2644	unsigned int min_total_size;
2645	int error;
2646
2647	while (isize_diff > ifree) {
2648		entry = NULL;
2649		small_entry = NULL;
2650		min_total_size = ~0U;
2651		last = IFIRST(header);
2652		/* Find the entry best suited to be pushed into EA block */
2653		for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
2654			total_size = EXT4_XATTR_LEN(last->e_name_len);
2655			if (!last->e_value_inum)
2656				total_size += EXT4_XATTR_SIZE(
2657					       le32_to_cpu(last->e_value_size));
2658			if (total_size <= bfree &&
2659			    total_size < min_total_size) {
2660				if (total_size + ifree < isize_diff) {
2661					small_entry = last;
2662				} else {
2663					entry = last;
2664					min_total_size = total_size;
2665				}
2666			}
2667		}
2668
2669		if (entry == NULL) {
2670			if (small_entry == NULL)
2671				return -ENOSPC;
2672			entry = small_entry;
2673		}
2674
2675		entry_size = EXT4_XATTR_LEN(entry->e_name_len);
2676		total_size = entry_size;
2677		if (!entry->e_value_inum)
2678			total_size += EXT4_XATTR_SIZE(
2679					      le32_to_cpu(entry->e_value_size));
2680		error = ext4_xattr_move_to_block(handle, inode, raw_inode,
2681						 entry);
2682		if (error)
2683			return error;
2684
2685		*total_ino -= entry_size;
2686		ifree += total_size;
2687		bfree -= total_size;
2688	}
2689
2690	return 0;
2691}
2692
2693/*
2694 * Expand an inode by new_extra_isize bytes when EAs are present.
2695 * Returns 0 on success or negative error number on failure.
2696 */
2697int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
2698			       struct ext4_inode *raw_inode, handle_t *handle)
2699{
2700	struct ext4_xattr_ibody_header *header;
2701	struct buffer_head *bh;
2702	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
2703	static unsigned int mnt_count;
2704	size_t min_offs;
2705	size_t ifree, bfree;
2706	int total_ino;
2707	void *base, *end;
2708	int error = 0, tried_min_extra_isize = 0;
2709	int s_min_extra_isize = le16_to_cpu(sbi->s_es->s_min_extra_isize);
2710	int isize_diff;	/* How much do we need to grow i_extra_isize */
2711
 
2712retry:
2713	isize_diff = new_extra_isize - EXT4_I(inode)->i_extra_isize;
2714	if (EXT4_I(inode)->i_extra_isize >= new_extra_isize)
2715		return 0;
 
2716
2717	header = IHDR(inode, raw_inode);
 
2718
2719	/*
2720	 * Check if enough free space is available in the inode to shift the
2721	 * entries ahead by new_extra_isize.
2722	 */
2723
2724	base = IFIRST(header);
2725	end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
2726	min_offs = end - base;
 
2727	total_ino = sizeof(struct ext4_xattr_ibody_header);
2728
2729	error = xattr_check_inode(inode, header, end);
2730	if (error)
 
 
 
 
 
 
 
 
2731		goto cleanup;
2732
2733	ifree = ext4_xattr_free_space(base, &min_offs, base, &total_ino);
2734	if (ifree >= isize_diff)
2735		goto shift;
2736
2737	/*
2738	 * Enough free space isn't available in the inode, check if
2739	 * EA block can hold new_extra_isize bytes.
2740	 */
2741	if (EXT4_I(inode)->i_file_acl) {
2742		bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
2743		error = -EIO;
2744		if (!bh)
2745			goto cleanup;
2746		error = ext4_xattr_check_block(inode, bh);
2747		if (error)
 
 
2748			goto cleanup;
 
2749		base = BHDR(bh);
 
2750		end = bh->b_data + bh->b_size;
2751		min_offs = end - base;
2752		bfree = ext4_xattr_free_space(BFIRST(bh), &min_offs, base,
2753					      NULL);
2754		brelse(bh);
2755		if (bfree + ifree < isize_diff) {
2756			if (!tried_min_extra_isize && s_min_extra_isize) {
2757				tried_min_extra_isize++;
2758				new_extra_isize = s_min_extra_isize;
 
2759				goto retry;
2760			}
2761			error = -ENOSPC;
2762			goto cleanup;
2763		}
2764	} else {
2765		bfree = inode->i_sb->s_blocksize;
2766	}
2767
2768	error = ext4_xattr_make_inode_space(handle, inode, raw_inode,
2769					    isize_diff, ifree, bfree,
2770					    &total_ino);
2771	if (error) {
2772		if (error == -ENOSPC && !tried_min_extra_isize &&
2773		    s_min_extra_isize) {
2774			tried_min_extra_isize++;
2775			new_extra_isize = s_min_extra_isize;
2776			goto retry;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2777		}
2778		goto cleanup;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2779	}
2780shift:
2781	/* Adjust the offsets and shift the remaining entries ahead */
2782	ext4_xattr_shift_entries(IFIRST(header), EXT4_I(inode)->i_extra_isize
2783			- new_extra_isize, (void *)raw_inode +
2784			EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
2785			(void *)header, total_ino);
2786	EXT4_I(inode)->i_extra_isize = new_extra_isize;
2787
2788cleanup:
2789	if (error && (mnt_count != le16_to_cpu(sbi->s_es->s_mnt_count))) {
2790		ext4_warning(inode->i_sb, "Unable to expand inode %lu. Delete some EAs or run e2fsck.",
2791			     inode->i_ino);
2792		mnt_count = le16_to_cpu(sbi->s_es->s_mnt_count);
2793	}
 
 
 
2794	return error;
2795}
2796
2797#define EIA_INCR 16 /* must be 2^n */
2798#define EIA_MASK (EIA_INCR - 1)
2799
2800/* Add the large xattr @inode into @ea_inode_array for deferred iput().
2801 * If @ea_inode_array is new or full it will be grown and the old
2802 * contents copied over.
2803 */
2804static int
2805ext4_expand_inode_array(struct ext4_xattr_inode_array **ea_inode_array,
2806			struct inode *inode)
2807{
2808	if (*ea_inode_array == NULL) {
2809		/*
2810		 * Start with 15 inodes, so it fits into a power-of-two size.
2811		 * If *ea_inode_array is NULL, this is essentially offsetof()
2812		 */
2813		(*ea_inode_array) =
2814			kmalloc(offsetof(struct ext4_xattr_inode_array,
2815					 inodes[EIA_MASK]),
2816				GFP_NOFS);
2817		if (*ea_inode_array == NULL)
2818			return -ENOMEM;
2819		(*ea_inode_array)->count = 0;
2820	} else if (((*ea_inode_array)->count & EIA_MASK) == EIA_MASK) {
2821		/* expand the array once all 15 + n * 16 slots are full */
2822		struct ext4_xattr_inode_array *new_array = NULL;
2823		int count = (*ea_inode_array)->count;
2824
2825		/* if new_array is NULL, this is essentially offsetof() */
2826		new_array = kmalloc(
2827				offsetof(struct ext4_xattr_inode_array,
2828					 inodes[count + EIA_INCR]),
2829				GFP_NOFS);
2830		if (new_array == NULL)
2831			return -ENOMEM;
2832		memcpy(new_array, *ea_inode_array,
2833		       offsetof(struct ext4_xattr_inode_array, inodes[count]));
2834		kfree(*ea_inode_array);
2835		*ea_inode_array = new_array;
2836	}
2837	(*ea_inode_array)->inodes[(*ea_inode_array)->count++] = inode;
2838	return 0;
2839}
2840
2841/*
2842 * ext4_xattr_delete_inode()
2843 *
2844 * Free extended attribute resources associated with this inode. Traverse
2845 * all entries and decrement reference on any xattr inodes associated with this
2846 * inode. This is called immediately before an inode is freed. We have exclusive
2847 * access to the inode. If an orphan inode is deleted it will also release its
2848 * references on xattr block and xattr inodes.
2849 */
2850int ext4_xattr_delete_inode(handle_t *handle, struct inode *inode,
2851			    struct ext4_xattr_inode_array **ea_inode_array,
2852			    int extra_credits)
2853{
2854	struct buffer_head *bh = NULL;
2855	struct ext4_xattr_ibody_header *header;
2856	struct ext4_iloc iloc = { .bh = NULL };
2857	struct ext4_xattr_entry *entry;
2858	struct inode *ea_inode;
2859	int error;
2860
2861	error = ext4_xattr_ensure_credits(handle, inode, extra_credits,
2862					  NULL /* bh */,
2863					  false /* dirty */,
2864					  false /* block_csum */);
2865	if (error) {
2866		EXT4_ERROR_INODE(inode, "ensure credits (error %d)", error);
2867		goto cleanup;
2868	}
2869
2870	if (ext4_has_feature_ea_inode(inode->i_sb) &&
2871	    ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
2872
2873		error = ext4_get_inode_loc(inode, &iloc);
2874		if (error) {
2875			EXT4_ERROR_INODE(inode, "inode loc (error %d)", error);
2876			goto cleanup;
2877		}
2878
2879		error = ext4_journal_get_write_access(handle, iloc.bh);
2880		if (error) {
2881			EXT4_ERROR_INODE(inode, "write access (error %d)",
2882					 error);
2883			goto cleanup;
2884		}
2885
2886		header = IHDR(inode, ext4_raw_inode(&iloc));
2887		if (header->h_magic == cpu_to_le32(EXT4_XATTR_MAGIC))
2888			ext4_xattr_inode_dec_ref_all(handle, inode, iloc.bh,
2889						     IFIRST(header),
2890						     false /* block_csum */,
2891						     ea_inode_array,
2892						     extra_credits,
2893						     false /* skip_quota */);
2894	}
 
 
2895
2896	if (EXT4_I(inode)->i_file_acl) {
2897		bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
2898		if (!bh) {
2899			EXT4_ERROR_INODE(inode, "block %llu read error",
2900					 EXT4_I(inode)->i_file_acl);
2901			error = -EIO;
2902			goto cleanup;
2903		}
2904		error = ext4_xattr_check_block(inode, bh);
2905		if (error)
2906			goto cleanup;
2907
2908		if (ext4_has_feature_ea_inode(inode->i_sb)) {
2909			for (entry = BFIRST(bh); !IS_LAST_ENTRY(entry);
2910			     entry = EXT4_XATTR_NEXT(entry)) {
2911				if (!entry->e_value_inum)
2912					continue;
2913				error = ext4_xattr_inode_iget(inode,
2914					      le32_to_cpu(entry->e_value_inum),
2915					      le32_to_cpu(entry->e_hash),
2916					      &ea_inode);
2917				if (error)
2918					continue;
2919				ext4_xattr_inode_free_quota(inode, ea_inode,
2920					      le32_to_cpu(entry->e_value_size));
2921				iput(ea_inode);
2922			}
2923
2924		}
2925
2926		ext4_xattr_release_block(handle, inode, bh, ea_inode_array,
2927					 extra_credits);
2928		/*
2929		 * Update i_file_acl value in the same transaction that releases
2930		 * block.
2931		 */
2932		EXT4_I(inode)->i_file_acl = 0;
2933		error = ext4_mark_inode_dirty(handle, inode);
2934		if (error) {
2935			EXT4_ERROR_INODE(inode, "mark inode dirty (error %d)",
2936					 error);
2937			goto cleanup;
2938		}
2939	}
2940	error = 0;
2941cleanup:
2942	brelse(iloc.bh);
2943	brelse(bh);
2944	return error;
2945}
2946
2947void ext4_xattr_inode_array_free(struct ext4_xattr_inode_array *ea_inode_array)
 
 
 
 
 
 
2948{
2949	int idx;
2950
2951	if (ea_inode_array == NULL)
2952		return;
2953
2954	for (idx = 0; idx < ea_inode_array->count; ++idx)
2955		iput(ea_inode_array->inodes[idx]);
2956	kfree(ea_inode_array);
2957}
2958
2959/*
2960 * ext4_xattr_block_cache_insert()
2961 *
2962 * Create a new entry in the extended attribute block cache, and insert
2963 * it unless such an entry is already in the cache.
2964 *
2965 * Returns 0, or a negative error number on failure.
2966 */
2967static void
2968ext4_xattr_block_cache_insert(struct mb_cache *ea_block_cache,
2969			      struct buffer_head *bh)
2970{
2971	struct ext4_xattr_header *header = BHDR(bh);
2972	__u32 hash = le32_to_cpu(header->h_hash);
2973	int reusable = le32_to_cpu(header->h_refcount) <
2974		       EXT4_XATTR_REFCOUNT_MAX;
2975	int error;
2976
2977	if (!ea_block_cache)
 
 
2978		return;
2979	error = mb_cache_entry_create(ea_block_cache, GFP_NOFS, hash,
2980				      bh->b_blocknr, reusable);
2981	if (error) {
2982		if (error == -EBUSY)
 
2983			ea_bdebug(bh, "already in cache");
2984	} else
 
 
2985		ea_bdebug(bh, "inserting [%x]", (int)hash);
 
 
2986}
2987
2988/*
2989 * ext4_xattr_cmp()
2990 *
2991 * Compare two extended attribute blocks for equality.
2992 *
2993 * Returns 0 if the blocks are equal, 1 if they differ, and
2994 * a negative error number on errors.
2995 */
2996static int
2997ext4_xattr_cmp(struct ext4_xattr_header *header1,
2998	       struct ext4_xattr_header *header2)
2999{
3000	struct ext4_xattr_entry *entry1, *entry2;
3001
3002	entry1 = ENTRY(header1+1);
3003	entry2 = ENTRY(header2+1);
3004	while (!IS_LAST_ENTRY(entry1)) {
3005		if (IS_LAST_ENTRY(entry2))
3006			return 1;
3007		if (entry1->e_hash != entry2->e_hash ||
3008		    entry1->e_name_index != entry2->e_name_index ||
3009		    entry1->e_name_len != entry2->e_name_len ||
3010		    entry1->e_value_size != entry2->e_value_size ||
3011		    entry1->e_value_inum != entry2->e_value_inum ||
3012		    memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
3013			return 1;
3014		if (!entry1->e_value_inum &&
3015		    memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
 
3016			   (char *)header2 + le16_to_cpu(entry2->e_value_offs),
3017			   le32_to_cpu(entry1->e_value_size)))
3018			return 1;
3019
3020		entry1 = EXT4_XATTR_NEXT(entry1);
3021		entry2 = EXT4_XATTR_NEXT(entry2);
3022	}
3023	if (!IS_LAST_ENTRY(entry2))
3024		return 1;
3025	return 0;
3026}
3027
3028/*
3029 * ext4_xattr_block_cache_find()
3030 *
3031 * Find an identical extended attribute block.
3032 *
3033 * Returns a pointer to the block found, or NULL if such a block was
3034 * not found or an error occurred.
3035 */
3036static struct buffer_head *
3037ext4_xattr_block_cache_find(struct inode *inode,
3038			    struct ext4_xattr_header *header,
3039			    struct mb_cache_entry **pce)
3040{
3041	__u32 hash = le32_to_cpu(header->h_hash);
3042	struct mb_cache_entry *ce;
3043	struct mb_cache *ea_block_cache = EA_BLOCK_CACHE(inode);
3044
3045	if (!ea_block_cache)
3046		return NULL;
3047	if (!header->h_hash)
3048		return NULL;  /* never share */
3049	ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
3050	ce = mb_cache_entry_find_first(ea_block_cache, hash);
 
 
3051	while (ce) {
3052		struct buffer_head *bh;
3053
3054		bh = sb_bread(inode->i_sb, ce->e_value);
 
 
 
 
 
3055		if (!bh) {
3056			EXT4_ERROR_INODE(inode, "block %lu read error",
3057					 (unsigned long)ce->e_value);
 
 
 
 
 
 
3058		} else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
3059			*pce = ce;
3060			return bh;
3061		}
3062		brelse(bh);
3063		ce = mb_cache_entry_find_next(ea_block_cache, ce);
3064	}
3065	return NULL;
3066}
3067
3068#define NAME_HASH_SHIFT 5
3069#define VALUE_HASH_SHIFT 16
3070
3071/*
3072 * ext4_xattr_hash_entry()
3073 *
3074 * Compute the hash of an extended attribute.
3075 */
3076static __le32 ext4_xattr_hash_entry(char *name, size_t name_len, __le32 *value,
3077				    size_t value_count)
3078{
3079	__u32 hash = 0;
 
 
3080
3081	while (name_len--) {
3082		hash = (hash << NAME_HASH_SHIFT) ^
3083		       (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
3084		       *name++;
3085	}
3086	while (value_count--) {
3087		hash = (hash << VALUE_HASH_SHIFT) ^
3088		       (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
3089		       le32_to_cpu(*value++);
 
 
 
 
 
 
3090	}
3091	return cpu_to_le32(hash);
3092}
3093
3094#undef NAME_HASH_SHIFT
3095#undef VALUE_HASH_SHIFT
3096
3097#define BLOCK_HASH_SHIFT 16
3098
3099/*
3100 * ext4_xattr_rehash()
3101 *
3102 * Re-compute the extended attribute hash value after an entry has changed.
3103 */
3104static void ext4_xattr_rehash(struct ext4_xattr_header *header)
 
3105{
3106	struct ext4_xattr_entry *here;
3107	__u32 hash = 0;
3108
 
3109	here = ENTRY(header+1);
3110	while (!IS_LAST_ENTRY(here)) {
3111		if (!here->e_hash) {
3112			/* Block is not shared if an entry's hash value == 0 */
3113			hash = 0;
3114			break;
3115		}
3116		hash = (hash << BLOCK_HASH_SHIFT) ^
3117		       (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
3118		       le32_to_cpu(here->e_hash);
3119		here = EXT4_XATTR_NEXT(here);
3120	}
3121	header->h_hash = cpu_to_le32(hash);
3122}
3123
3124#undef BLOCK_HASH_SHIFT
3125
3126#define	HASH_BUCKET_BITS	10
3127
3128struct mb_cache *
3129ext4_xattr_create_cache(void)
3130{
3131	return mb_cache_create(HASH_BUCKET_BITS);
 
 
 
3132}
3133
3134void ext4_xattr_destroy_cache(struct mb_cache *cache)
 
3135{
3136	if (cache)
3137		mb_cache_destroy(cache);
 
3138}
3139