Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux
1
fork

Configure Feed

Select the types of activity you want to include in your feed.

at v2.6.26-rc7 1327 lines 35 kB view raw
1/* 2 * linux/fs/ext3/xattr.c 3 * 4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de> 5 * 6 * Fix by Harrison Xing <harrison@mountainviewdata.com>. 7 * Ext3 code with a lot of help from Eric Jarman <ejarman@acm.org>. 8 * Extended attributes for symlinks and special files added per 9 * suggestion of Luka Renko <luka.renko@hermes.si>. 10 * xattr consolidation Copyright (c) 2004 James Morris <jmorris@redhat.com>, 11 * Red Hat Inc. 12 * ea-in-inode support by Alex Tomas <alex@clusterfs.com> aka bzzz 13 * and Andreas Gruenbacher <agruen@suse.de>. 14 */ 15 16/* 17 * Extended attributes are stored directly in inodes (on file systems with 18 * inodes bigger than 128 bytes) and on additional disk blocks. The i_file_acl 19 * field contains the block number if an inode uses an additional block. All 20 * attributes must fit in the inode and one additional block. Blocks that 21 * contain the identical set of attributes may be shared among several inodes. 22 * Identical blocks are detected by keeping a cache of blocks that have 23 * recently been accessed. 24 * 25 * The attributes in inodes and on blocks have a different header; the entries 26 * are stored in the same format: 27 * 28 * +------------------+ 29 * | header | 30 * | entry 1 | | 31 * | entry 2 | | growing downwards 32 * | entry 3 | v 33 * | four null bytes | 34 * | . . . | 35 * | value 1 | ^ 36 * | value 3 | | growing upwards 37 * | value 2 | | 38 * +------------------+ 39 * 40 * The header is followed by multiple entry descriptors. In disk blocks, the 41 * entry descriptors are kept sorted. In inodes, they are unsorted. The 42 * attribute values are aligned to the end of the block in no specific order. 43 * 44 * Locking strategy 45 * ---------------- 46 * EXT3_I(inode)->i_file_acl is protected by EXT3_I(inode)->xattr_sem. 47 * EA blocks are only changed if they are exclusive to an inode, so 48 * holding xattr_sem also means that nothing but the EA block's reference 49 * count can change. Multiple writers to the same block are synchronized 50 * by the buffer lock. 51 */ 52 53#include <linux/init.h> 54#include <linux/fs.h> 55#include <linux/slab.h> 56#include <linux/ext3_jbd.h> 57#include <linux/ext3_fs.h> 58#include <linux/mbcache.h> 59#include <linux/quotaops.h> 60#include <linux/rwsem.h> 61#include "xattr.h" 62#include "acl.h" 63 64#define BHDR(bh) ((struct ext3_xattr_header *)((bh)->b_data)) 65#define ENTRY(ptr) ((struct ext3_xattr_entry *)(ptr)) 66#define BFIRST(bh) ENTRY(BHDR(bh)+1) 67#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0) 68 69#define IHDR(inode, raw_inode) \ 70 ((struct ext3_xattr_ibody_header *) \ 71 ((void *)raw_inode + \ 72 EXT3_GOOD_OLD_INODE_SIZE + \ 73 EXT3_I(inode)->i_extra_isize)) 74#define IFIRST(hdr) ((struct ext3_xattr_entry *)((hdr)+1)) 75 76#ifdef EXT3_XATTR_DEBUG 77# define ea_idebug(inode, f...) do { \ 78 printk(KERN_DEBUG "inode %s:%lu: ", \ 79 inode->i_sb->s_id, inode->i_ino); \ 80 printk(f); \ 81 printk("\n"); \ 82 } while (0) 83# define ea_bdebug(bh, f...) do { \ 84 char b[BDEVNAME_SIZE]; \ 85 printk(KERN_DEBUG "block %s:%lu: ", \ 86 bdevname(bh->b_bdev, b), \ 87 (unsigned long) bh->b_blocknr); \ 88 printk(f); \ 89 printk("\n"); \ 90 } while (0) 91#else 92# define ea_idebug(f...) 93# define ea_bdebug(f...) 94#endif 95 96static void ext3_xattr_cache_insert(struct buffer_head *); 97static struct buffer_head *ext3_xattr_cache_find(struct inode *, 98 struct ext3_xattr_header *, 99 struct mb_cache_entry **); 100static void ext3_xattr_rehash(struct ext3_xattr_header *, 101 struct ext3_xattr_entry *); 102static int ext3_xattr_list(struct inode *inode, char *buffer, 103 size_t buffer_size); 104 105static struct mb_cache *ext3_xattr_cache; 106 107static struct xattr_handler *ext3_xattr_handler_map[] = { 108 [EXT3_XATTR_INDEX_USER] = &ext3_xattr_user_handler, 109#ifdef CONFIG_EXT3_FS_POSIX_ACL 110 [EXT3_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext3_xattr_acl_access_handler, 111 [EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext3_xattr_acl_default_handler, 112#endif 113 [EXT3_XATTR_INDEX_TRUSTED] = &ext3_xattr_trusted_handler, 114#ifdef CONFIG_EXT3_FS_SECURITY 115 [EXT3_XATTR_INDEX_SECURITY] = &ext3_xattr_security_handler, 116#endif 117}; 118 119struct xattr_handler *ext3_xattr_handlers[] = { 120 &ext3_xattr_user_handler, 121 &ext3_xattr_trusted_handler, 122#ifdef CONFIG_EXT3_FS_POSIX_ACL 123 &ext3_xattr_acl_access_handler, 124 &ext3_xattr_acl_default_handler, 125#endif 126#ifdef CONFIG_EXT3_FS_SECURITY 127 &ext3_xattr_security_handler, 128#endif 129 NULL 130}; 131 132static inline struct xattr_handler * 133ext3_xattr_handler(int name_index) 134{ 135 struct xattr_handler *handler = NULL; 136 137 if (name_index > 0 && name_index < ARRAY_SIZE(ext3_xattr_handler_map)) 138 handler = ext3_xattr_handler_map[name_index]; 139 return handler; 140} 141 142/* 143 * Inode operation listxattr() 144 * 145 * dentry->d_inode->i_mutex: don't care 146 */ 147ssize_t 148ext3_listxattr(struct dentry *dentry, char *buffer, size_t size) 149{ 150 return ext3_xattr_list(dentry->d_inode, buffer, size); 151} 152 153static int 154ext3_xattr_check_names(struct ext3_xattr_entry *entry, void *end) 155{ 156 while (!IS_LAST_ENTRY(entry)) { 157 struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(entry); 158 if ((void *)next >= end) 159 return -EIO; 160 entry = next; 161 } 162 return 0; 163} 164 165static inline int 166ext3_xattr_check_block(struct buffer_head *bh) 167{ 168 int error; 169 170 if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || 171 BHDR(bh)->h_blocks != cpu_to_le32(1)) 172 return -EIO; 173 error = ext3_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size); 174 return error; 175} 176 177static inline int 178ext3_xattr_check_entry(struct ext3_xattr_entry *entry, size_t size) 179{ 180 size_t value_size = le32_to_cpu(entry->e_value_size); 181 182 if (entry->e_value_block != 0 || value_size > size || 183 le16_to_cpu(entry->e_value_offs) + value_size > size) 184 return -EIO; 185 return 0; 186} 187 188static int 189ext3_xattr_find_entry(struct ext3_xattr_entry **pentry, int name_index, 190 const char *name, size_t size, int sorted) 191{ 192 struct ext3_xattr_entry *entry; 193 size_t name_len; 194 int cmp = 1; 195 196 if (name == NULL) 197 return -EINVAL; 198 name_len = strlen(name); 199 entry = *pentry; 200 for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) { 201 cmp = name_index - entry->e_name_index; 202 if (!cmp) 203 cmp = name_len - entry->e_name_len; 204 if (!cmp) 205 cmp = memcmp(name, entry->e_name, name_len); 206 if (cmp <= 0 && (sorted || cmp == 0)) 207 break; 208 } 209 *pentry = entry; 210 if (!cmp && ext3_xattr_check_entry(entry, size)) 211 return -EIO; 212 return cmp ? -ENODATA : 0; 213} 214 215static int 216ext3_xattr_block_get(struct inode *inode, int name_index, const char *name, 217 void *buffer, size_t buffer_size) 218{ 219 struct buffer_head *bh = NULL; 220 struct ext3_xattr_entry *entry; 221 size_t size; 222 int error; 223 224 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld", 225 name_index, name, buffer, (long)buffer_size); 226 227 error = -ENODATA; 228 if (!EXT3_I(inode)->i_file_acl) 229 goto cleanup; 230 ea_idebug(inode, "reading block %u", EXT3_I(inode)->i_file_acl); 231 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl); 232 if (!bh) 233 goto cleanup; 234 ea_bdebug(bh, "b_count=%d, refcount=%d", 235 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount)); 236 if (ext3_xattr_check_block(bh)) { 237bad_block: ext3_error(inode->i_sb, __func__, 238 "inode %lu: bad block "E3FSBLK, inode->i_ino, 239 EXT3_I(inode)->i_file_acl); 240 error = -EIO; 241 goto cleanup; 242 } 243 ext3_xattr_cache_insert(bh); 244 entry = BFIRST(bh); 245 error = ext3_xattr_find_entry(&entry, name_index, name, bh->b_size, 1); 246 if (error == -EIO) 247 goto bad_block; 248 if (error) 249 goto cleanup; 250 size = le32_to_cpu(entry->e_value_size); 251 if (buffer) { 252 error = -ERANGE; 253 if (size > buffer_size) 254 goto cleanup; 255 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs), 256 size); 257 } 258 error = size; 259 260cleanup: 261 brelse(bh); 262 return error; 263} 264 265static int 266ext3_xattr_ibody_get(struct inode *inode, int name_index, const char *name, 267 void *buffer, size_t buffer_size) 268{ 269 struct ext3_xattr_ibody_header *header; 270 struct ext3_xattr_entry *entry; 271 struct ext3_inode *raw_inode; 272 struct ext3_iloc iloc; 273 size_t size; 274 void *end; 275 int error; 276 277 if (!(EXT3_I(inode)->i_state & EXT3_STATE_XATTR)) 278 return -ENODATA; 279 error = ext3_get_inode_loc(inode, &iloc); 280 if (error) 281 return error; 282 raw_inode = ext3_raw_inode(&iloc); 283 header = IHDR(inode, raw_inode); 284 entry = IFIRST(header); 285 end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size; 286 error = ext3_xattr_check_names(entry, end); 287 if (error) 288 goto cleanup; 289 error = ext3_xattr_find_entry(&entry, name_index, name, 290 end - (void *)entry, 0); 291 if (error) 292 goto cleanup; 293 size = le32_to_cpu(entry->e_value_size); 294 if (buffer) { 295 error = -ERANGE; 296 if (size > buffer_size) 297 goto cleanup; 298 memcpy(buffer, (void *)IFIRST(header) + 299 le16_to_cpu(entry->e_value_offs), size); 300 } 301 error = size; 302 303cleanup: 304 brelse(iloc.bh); 305 return error; 306} 307 308/* 309 * ext3_xattr_get() 310 * 311 * Copy an extended attribute into the buffer 312 * provided, or compute the buffer size required. 313 * Buffer is NULL to compute the size of the buffer required. 314 * 315 * Returns a negative error number on failure, or the number of bytes 316 * used / required on success. 317 */ 318int 319ext3_xattr_get(struct inode *inode, int name_index, const char *name, 320 void *buffer, size_t buffer_size) 321{ 322 int error; 323 324 down_read(&EXT3_I(inode)->xattr_sem); 325 error = ext3_xattr_ibody_get(inode, name_index, name, buffer, 326 buffer_size); 327 if (error == -ENODATA) 328 error = ext3_xattr_block_get(inode, name_index, name, buffer, 329 buffer_size); 330 up_read(&EXT3_I(inode)->xattr_sem); 331 return error; 332} 333 334static int 335ext3_xattr_list_entries(struct inode *inode, struct ext3_xattr_entry *entry, 336 char *buffer, size_t buffer_size) 337{ 338 size_t rest = buffer_size; 339 340 for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) { 341 struct xattr_handler *handler = 342 ext3_xattr_handler(entry->e_name_index); 343 344 if (handler) { 345 size_t size = handler->list(inode, buffer, rest, 346 entry->e_name, 347 entry->e_name_len); 348 if (buffer) { 349 if (size > rest) 350 return -ERANGE; 351 buffer += size; 352 } 353 rest -= size; 354 } 355 } 356 return buffer_size - rest; 357} 358 359static int 360ext3_xattr_block_list(struct inode *inode, char *buffer, size_t buffer_size) 361{ 362 struct buffer_head *bh = NULL; 363 int error; 364 365 ea_idebug(inode, "buffer=%p, buffer_size=%ld", 366 buffer, (long)buffer_size); 367 368 error = 0; 369 if (!EXT3_I(inode)->i_file_acl) 370 goto cleanup; 371 ea_idebug(inode, "reading block %u", EXT3_I(inode)->i_file_acl); 372 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl); 373 error = -EIO; 374 if (!bh) 375 goto cleanup; 376 ea_bdebug(bh, "b_count=%d, refcount=%d", 377 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount)); 378 if (ext3_xattr_check_block(bh)) { 379 ext3_error(inode->i_sb, __func__, 380 "inode %lu: bad block "E3FSBLK, inode->i_ino, 381 EXT3_I(inode)->i_file_acl); 382 error = -EIO; 383 goto cleanup; 384 } 385 ext3_xattr_cache_insert(bh); 386 error = ext3_xattr_list_entries(inode, BFIRST(bh), buffer, buffer_size); 387 388cleanup: 389 brelse(bh); 390 391 return error; 392} 393 394static int 395ext3_xattr_ibody_list(struct inode *inode, char *buffer, size_t buffer_size) 396{ 397 struct ext3_xattr_ibody_header *header; 398 struct ext3_inode *raw_inode; 399 struct ext3_iloc iloc; 400 void *end; 401 int error; 402 403 if (!(EXT3_I(inode)->i_state & EXT3_STATE_XATTR)) 404 return 0; 405 error = ext3_get_inode_loc(inode, &iloc); 406 if (error) 407 return error; 408 raw_inode = ext3_raw_inode(&iloc); 409 header = IHDR(inode, raw_inode); 410 end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size; 411 error = ext3_xattr_check_names(IFIRST(header), end); 412 if (error) 413 goto cleanup; 414 error = ext3_xattr_list_entries(inode, IFIRST(header), 415 buffer, buffer_size); 416 417cleanup: 418 brelse(iloc.bh); 419 return error; 420} 421 422/* 423 * ext3_xattr_list() 424 * 425 * Copy a list of attribute names into the buffer 426 * provided, or compute the buffer size required. 427 * Buffer is NULL to compute the size of the buffer required. 428 * 429 * Returns a negative error number on failure, or the number of bytes 430 * used / required on success. 431 */ 432static int 433ext3_xattr_list(struct inode *inode, char *buffer, size_t buffer_size) 434{ 435 int i_error, b_error; 436 437 down_read(&EXT3_I(inode)->xattr_sem); 438 i_error = ext3_xattr_ibody_list(inode, buffer, buffer_size); 439 if (i_error < 0) { 440 b_error = 0; 441 } else { 442 if (buffer) { 443 buffer += i_error; 444 buffer_size -= i_error; 445 } 446 b_error = ext3_xattr_block_list(inode, buffer, buffer_size); 447 if (b_error < 0) 448 i_error = 0; 449 } 450 up_read(&EXT3_I(inode)->xattr_sem); 451 return i_error + b_error; 452} 453 454/* 455 * If the EXT3_FEATURE_COMPAT_EXT_ATTR feature of this file system is 456 * not set, set it. 457 */ 458static void ext3_xattr_update_super_block(handle_t *handle, 459 struct super_block *sb) 460{ 461 if (EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR)) 462 return; 463 464 if (ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh) == 0) { 465 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR); 466 sb->s_dirt = 1; 467 ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); 468 } 469} 470 471/* 472 * Release the xattr block BH: If the reference count is > 1, decrement 473 * it; otherwise free the block. 474 */ 475static void 476ext3_xattr_release_block(handle_t *handle, struct inode *inode, 477 struct buffer_head *bh) 478{ 479 struct mb_cache_entry *ce = NULL; 480 int error = 0; 481 482 ce = mb_cache_entry_get(ext3_xattr_cache, bh->b_bdev, bh->b_blocknr); 483 error = ext3_journal_get_write_access(handle, bh); 484 if (error) 485 goto out; 486 487 lock_buffer(bh); 488 489 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) { 490 ea_bdebug(bh, "refcount now=0; freeing"); 491 if (ce) 492 mb_cache_entry_free(ce); 493 ext3_free_blocks(handle, inode, bh->b_blocknr, 1); 494 get_bh(bh); 495 ext3_forget(handle, 1, inode, bh, bh->b_blocknr); 496 } else { 497 le32_add_cpu(&BHDR(bh)->h_refcount, -1); 498 error = ext3_journal_dirty_metadata(handle, bh); 499 if (IS_SYNC(inode)) 500 handle->h_sync = 1; 501 DQUOT_FREE_BLOCK(inode, 1); 502 ea_bdebug(bh, "refcount now=%d; releasing", 503 le32_to_cpu(BHDR(bh)->h_refcount)); 504 if (ce) 505 mb_cache_entry_release(ce); 506 } 507 unlock_buffer(bh); 508out: 509 ext3_std_error(inode->i_sb, error); 510 return; 511} 512 513struct ext3_xattr_info { 514 int name_index; 515 const char *name; 516 const void *value; 517 size_t value_len; 518}; 519 520struct ext3_xattr_search { 521 struct ext3_xattr_entry *first; 522 void *base; 523 void *end; 524 struct ext3_xattr_entry *here; 525 int not_found; 526}; 527 528static int 529ext3_xattr_set_entry(struct ext3_xattr_info *i, struct ext3_xattr_search *s) 530{ 531 struct ext3_xattr_entry *last; 532 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name); 533 534 /* Compute min_offs and last. */ 535 last = s->first; 536 for (; !IS_LAST_ENTRY(last); last = EXT3_XATTR_NEXT(last)) { 537 if (!last->e_value_block && last->e_value_size) { 538 size_t offs = le16_to_cpu(last->e_value_offs); 539 if (offs < min_offs) 540 min_offs = offs; 541 } 542 } 543 free = min_offs - ((void *)last - s->base) - sizeof(__u32); 544 if (!s->not_found) { 545 if (!s->here->e_value_block && s->here->e_value_size) { 546 size_t size = le32_to_cpu(s->here->e_value_size); 547 free += EXT3_XATTR_SIZE(size); 548 } 549 free += EXT3_XATTR_LEN(name_len); 550 } 551 if (i->value) { 552 if (free < EXT3_XATTR_SIZE(i->value_len) || 553 free < EXT3_XATTR_LEN(name_len) + 554 EXT3_XATTR_SIZE(i->value_len)) 555 return -ENOSPC; 556 } 557 558 if (i->value && s->not_found) { 559 /* Insert the new name. */ 560 size_t size = EXT3_XATTR_LEN(name_len); 561 size_t rest = (void *)last - (void *)s->here + sizeof(__u32); 562 memmove((void *)s->here + size, s->here, rest); 563 memset(s->here, 0, size); 564 s->here->e_name_index = i->name_index; 565 s->here->e_name_len = name_len; 566 memcpy(s->here->e_name, i->name, name_len); 567 } else { 568 if (!s->here->e_value_block && s->here->e_value_size) { 569 void *first_val = s->base + min_offs; 570 size_t offs = le16_to_cpu(s->here->e_value_offs); 571 void *val = s->base + offs; 572 size_t size = EXT3_XATTR_SIZE( 573 le32_to_cpu(s->here->e_value_size)); 574 575 if (i->value && size == EXT3_XATTR_SIZE(i->value_len)) { 576 /* The old and the new value have the same 577 size. Just replace. */ 578 s->here->e_value_size = 579 cpu_to_le32(i->value_len); 580 memset(val + size - EXT3_XATTR_PAD, 0, 581 EXT3_XATTR_PAD); /* Clear pad bytes. */ 582 memcpy(val, i->value, i->value_len); 583 return 0; 584 } 585 586 /* Remove the old value. */ 587 memmove(first_val + size, first_val, val - first_val); 588 memset(first_val, 0, size); 589 s->here->e_value_size = 0; 590 s->here->e_value_offs = 0; 591 min_offs += size; 592 593 /* Adjust all value offsets. */ 594 last = s->first; 595 while (!IS_LAST_ENTRY(last)) { 596 size_t o = le16_to_cpu(last->e_value_offs); 597 if (!last->e_value_block && 598 last->e_value_size && o < offs) 599 last->e_value_offs = 600 cpu_to_le16(o + size); 601 last = EXT3_XATTR_NEXT(last); 602 } 603 } 604 if (!i->value) { 605 /* Remove the old name. */ 606 size_t size = EXT3_XATTR_LEN(name_len); 607 last = ENTRY((void *)last - size); 608 memmove(s->here, (void *)s->here + size, 609 (void *)last - (void *)s->here + sizeof(__u32)); 610 memset(last, 0, size); 611 } 612 } 613 614 if (i->value) { 615 /* Insert the new value. */ 616 s->here->e_value_size = cpu_to_le32(i->value_len); 617 if (i->value_len) { 618 size_t size = EXT3_XATTR_SIZE(i->value_len); 619 void *val = s->base + min_offs - size; 620 s->here->e_value_offs = cpu_to_le16(min_offs - size); 621 memset(val + size - EXT3_XATTR_PAD, 0, 622 EXT3_XATTR_PAD); /* Clear the pad bytes. */ 623 memcpy(val, i->value, i->value_len); 624 } 625 } 626 return 0; 627} 628 629struct ext3_xattr_block_find { 630 struct ext3_xattr_search s; 631 struct buffer_head *bh; 632}; 633 634static int 635ext3_xattr_block_find(struct inode *inode, struct ext3_xattr_info *i, 636 struct ext3_xattr_block_find *bs) 637{ 638 struct super_block *sb = inode->i_sb; 639 int error; 640 641 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld", 642 i->name_index, i->name, i->value, (long)i->value_len); 643 644 if (EXT3_I(inode)->i_file_acl) { 645 /* The inode already has an extended attribute block. */ 646 bs->bh = sb_bread(sb, EXT3_I(inode)->i_file_acl); 647 error = -EIO; 648 if (!bs->bh) 649 goto cleanup; 650 ea_bdebug(bs->bh, "b_count=%d, refcount=%d", 651 atomic_read(&(bs->bh->b_count)), 652 le32_to_cpu(BHDR(bs->bh)->h_refcount)); 653 if (ext3_xattr_check_block(bs->bh)) { 654 ext3_error(sb, __func__, 655 "inode %lu: bad block "E3FSBLK, inode->i_ino, 656 EXT3_I(inode)->i_file_acl); 657 error = -EIO; 658 goto cleanup; 659 } 660 /* Find the named attribute. */ 661 bs->s.base = BHDR(bs->bh); 662 bs->s.first = BFIRST(bs->bh); 663 bs->s.end = bs->bh->b_data + bs->bh->b_size; 664 bs->s.here = bs->s.first; 665 error = ext3_xattr_find_entry(&bs->s.here, i->name_index, 666 i->name, bs->bh->b_size, 1); 667 if (error && error != -ENODATA) 668 goto cleanup; 669 bs->s.not_found = error; 670 } 671 error = 0; 672 673cleanup: 674 return error; 675} 676 677static int 678ext3_xattr_block_set(handle_t *handle, struct inode *inode, 679 struct ext3_xattr_info *i, 680 struct ext3_xattr_block_find *bs) 681{ 682 struct super_block *sb = inode->i_sb; 683 struct buffer_head *new_bh = NULL; 684 struct ext3_xattr_search *s = &bs->s; 685 struct mb_cache_entry *ce = NULL; 686 int error = 0; 687 688#define header(x) ((struct ext3_xattr_header *)(x)) 689 690 if (i->value && i->value_len > sb->s_blocksize) 691 return -ENOSPC; 692 if (s->base) { 693 ce = mb_cache_entry_get(ext3_xattr_cache, bs->bh->b_bdev, 694 bs->bh->b_blocknr); 695 error = ext3_journal_get_write_access(handle, bs->bh); 696 if (error) 697 goto cleanup; 698 lock_buffer(bs->bh); 699 700 if (header(s->base)->h_refcount == cpu_to_le32(1)) { 701 if (ce) { 702 mb_cache_entry_free(ce); 703 ce = NULL; 704 } 705 ea_bdebug(bs->bh, "modifying in-place"); 706 error = ext3_xattr_set_entry(i, s); 707 if (!error) { 708 if (!IS_LAST_ENTRY(s->first)) 709 ext3_xattr_rehash(header(s->base), 710 s->here); 711 ext3_xattr_cache_insert(bs->bh); 712 } 713 unlock_buffer(bs->bh); 714 if (error == -EIO) 715 goto bad_block; 716 if (!error) 717 error = ext3_journal_dirty_metadata(handle, 718 bs->bh); 719 if (error) 720 goto cleanup; 721 goto inserted; 722 } else { 723 int offset = (char *)s->here - bs->bh->b_data; 724 725 unlock_buffer(bs->bh); 726 journal_release_buffer(handle, bs->bh); 727 728 if (ce) { 729 mb_cache_entry_release(ce); 730 ce = NULL; 731 } 732 ea_bdebug(bs->bh, "cloning"); 733 s->base = kmalloc(bs->bh->b_size, GFP_NOFS); 734 error = -ENOMEM; 735 if (s->base == NULL) 736 goto cleanup; 737 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size); 738 s->first = ENTRY(header(s->base)+1); 739 header(s->base)->h_refcount = cpu_to_le32(1); 740 s->here = ENTRY(s->base + offset); 741 s->end = s->base + bs->bh->b_size; 742 } 743 } else { 744 /* Allocate a buffer where we construct the new block. */ 745 s->base = kzalloc(sb->s_blocksize, GFP_NOFS); 746 /* assert(header == s->base) */ 747 error = -ENOMEM; 748 if (s->base == NULL) 749 goto cleanup; 750 header(s->base)->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC); 751 header(s->base)->h_blocks = cpu_to_le32(1); 752 header(s->base)->h_refcount = cpu_to_le32(1); 753 s->first = ENTRY(header(s->base)+1); 754 s->here = ENTRY(header(s->base)+1); 755 s->end = s->base + sb->s_blocksize; 756 } 757 758 error = ext3_xattr_set_entry(i, s); 759 if (error == -EIO) 760 goto bad_block; 761 if (error) 762 goto cleanup; 763 if (!IS_LAST_ENTRY(s->first)) 764 ext3_xattr_rehash(header(s->base), s->here); 765 766inserted: 767 if (!IS_LAST_ENTRY(s->first)) { 768 new_bh = ext3_xattr_cache_find(inode, header(s->base), &ce); 769 if (new_bh) { 770 /* We found an identical block in the cache. */ 771 if (new_bh == bs->bh) 772 ea_bdebug(new_bh, "keeping"); 773 else { 774 /* The old block is released after updating 775 the inode. */ 776 error = -EDQUOT; 777 if (DQUOT_ALLOC_BLOCK(inode, 1)) 778 goto cleanup; 779 error = ext3_journal_get_write_access(handle, 780 new_bh); 781 if (error) 782 goto cleanup_dquot; 783 lock_buffer(new_bh); 784 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1); 785 ea_bdebug(new_bh, "reusing; refcount now=%d", 786 le32_to_cpu(BHDR(new_bh)->h_refcount)); 787 unlock_buffer(new_bh); 788 error = ext3_journal_dirty_metadata(handle, 789 new_bh); 790 if (error) 791 goto cleanup_dquot; 792 } 793 mb_cache_entry_release(ce); 794 ce = NULL; 795 } else if (bs->bh && s->base == bs->bh->b_data) { 796 /* We were modifying this block in-place. */ 797 ea_bdebug(bs->bh, "keeping this block"); 798 new_bh = bs->bh; 799 get_bh(new_bh); 800 } else { 801 /* We need to allocate a new block */ 802 ext3_fsblk_t goal = ext3_group_first_block_no(sb, 803 EXT3_I(inode)->i_block_group); 804 ext3_fsblk_t block = ext3_new_block(handle, inode, 805 goal, &error); 806 if (error) 807 goto cleanup; 808 ea_idebug(inode, "creating block %d", block); 809 810 new_bh = sb_getblk(sb, block); 811 if (!new_bh) { 812getblk_failed: 813 ext3_free_blocks(handle, inode, block, 1); 814 error = -EIO; 815 goto cleanup; 816 } 817 lock_buffer(new_bh); 818 error = ext3_journal_get_create_access(handle, new_bh); 819 if (error) { 820 unlock_buffer(new_bh); 821 goto getblk_failed; 822 } 823 memcpy(new_bh->b_data, s->base, new_bh->b_size); 824 set_buffer_uptodate(new_bh); 825 unlock_buffer(new_bh); 826 ext3_xattr_cache_insert(new_bh); 827 error = ext3_journal_dirty_metadata(handle, new_bh); 828 if (error) 829 goto cleanup; 830 } 831 } 832 833 /* Update the inode. */ 834 EXT3_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0; 835 836 /* Drop the previous xattr block. */ 837 if (bs->bh && bs->bh != new_bh) 838 ext3_xattr_release_block(handle, inode, bs->bh); 839 error = 0; 840 841cleanup: 842 if (ce) 843 mb_cache_entry_release(ce); 844 brelse(new_bh); 845 if (!(bs->bh && s->base == bs->bh->b_data)) 846 kfree(s->base); 847 848 return error; 849 850cleanup_dquot: 851 DQUOT_FREE_BLOCK(inode, 1); 852 goto cleanup; 853 854bad_block: 855 ext3_error(inode->i_sb, __func__, 856 "inode %lu: bad block "E3FSBLK, inode->i_ino, 857 EXT3_I(inode)->i_file_acl); 858 goto cleanup; 859 860#undef header 861} 862 863struct ext3_xattr_ibody_find { 864 struct ext3_xattr_search s; 865 struct ext3_iloc iloc; 866}; 867 868static int 869ext3_xattr_ibody_find(struct inode *inode, struct ext3_xattr_info *i, 870 struct ext3_xattr_ibody_find *is) 871{ 872 struct ext3_xattr_ibody_header *header; 873 struct ext3_inode *raw_inode; 874 int error; 875 876 if (EXT3_I(inode)->i_extra_isize == 0) 877 return 0; 878 raw_inode = ext3_raw_inode(&is->iloc); 879 header = IHDR(inode, raw_inode); 880 is->s.base = is->s.first = IFIRST(header); 881 is->s.here = is->s.first; 882 is->s.end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size; 883 if (EXT3_I(inode)->i_state & EXT3_STATE_XATTR) { 884 error = ext3_xattr_check_names(IFIRST(header), is->s.end); 885 if (error) 886 return error; 887 /* Find the named attribute. */ 888 error = ext3_xattr_find_entry(&is->s.here, i->name_index, 889 i->name, is->s.end - 890 (void *)is->s.base, 0); 891 if (error && error != -ENODATA) 892 return error; 893 is->s.not_found = error; 894 } 895 return 0; 896} 897 898static int 899ext3_xattr_ibody_set(handle_t *handle, struct inode *inode, 900 struct ext3_xattr_info *i, 901 struct ext3_xattr_ibody_find *is) 902{ 903 struct ext3_xattr_ibody_header *header; 904 struct ext3_xattr_search *s = &is->s; 905 int error; 906 907 if (EXT3_I(inode)->i_extra_isize == 0) 908 return -ENOSPC; 909 error = ext3_xattr_set_entry(i, s); 910 if (error) 911 return error; 912 header = IHDR(inode, ext3_raw_inode(&is->iloc)); 913 if (!IS_LAST_ENTRY(s->first)) { 914 header->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC); 915 EXT3_I(inode)->i_state |= EXT3_STATE_XATTR; 916 } else { 917 header->h_magic = cpu_to_le32(0); 918 EXT3_I(inode)->i_state &= ~EXT3_STATE_XATTR; 919 } 920 return 0; 921} 922 923/* 924 * ext3_xattr_set_handle() 925 * 926 * Create, replace or remove an extended attribute for this inode. Buffer 927 * is NULL to remove an existing extended attribute, and non-NULL to 928 * either replace an existing extended attribute, or create a new extended 929 * attribute. The flags XATTR_REPLACE and XATTR_CREATE 930 * specify that an extended attribute must exist and must not exist 931 * previous to the call, respectively. 932 * 933 * Returns 0, or a negative error number on failure. 934 */ 935int 936ext3_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index, 937 const char *name, const void *value, size_t value_len, 938 int flags) 939{ 940 struct ext3_xattr_info i = { 941 .name_index = name_index, 942 .name = name, 943 .value = value, 944 .value_len = value_len, 945 946 }; 947 struct ext3_xattr_ibody_find is = { 948 .s = { .not_found = -ENODATA, }, 949 }; 950 struct ext3_xattr_block_find bs = { 951 .s = { .not_found = -ENODATA, }, 952 }; 953 int error; 954 955 if (!name) 956 return -EINVAL; 957 if (strlen(name) > 255) 958 return -ERANGE; 959 down_write(&EXT3_I(inode)->xattr_sem); 960 error = ext3_get_inode_loc(inode, &is.iloc); 961 if (error) 962 goto cleanup; 963 964 if (EXT3_I(inode)->i_state & EXT3_STATE_NEW) { 965 struct ext3_inode *raw_inode = ext3_raw_inode(&is.iloc); 966 memset(raw_inode, 0, EXT3_SB(inode->i_sb)->s_inode_size); 967 EXT3_I(inode)->i_state &= ~EXT3_STATE_NEW; 968 } 969 970 error = ext3_xattr_ibody_find(inode, &i, &is); 971 if (error) 972 goto cleanup; 973 if (is.s.not_found) 974 error = ext3_xattr_block_find(inode, &i, &bs); 975 if (error) 976 goto cleanup; 977 if (is.s.not_found && bs.s.not_found) { 978 error = -ENODATA; 979 if (flags & XATTR_REPLACE) 980 goto cleanup; 981 error = 0; 982 if (!value) 983 goto cleanup; 984 } else { 985 error = -EEXIST; 986 if (flags & XATTR_CREATE) 987 goto cleanup; 988 } 989 error = ext3_journal_get_write_access(handle, is.iloc.bh); 990 if (error) 991 goto cleanup; 992 if (!value) { 993 if (!is.s.not_found) 994 error = ext3_xattr_ibody_set(handle, inode, &i, &is); 995 else if (!bs.s.not_found) 996 error = ext3_xattr_block_set(handle, inode, &i, &bs); 997 } else { 998 error = ext3_xattr_ibody_set(handle, inode, &i, &is); 999 if (!error && !bs.s.not_found) { 1000 i.value = NULL; 1001 error = ext3_xattr_block_set(handle, inode, &i, &bs); 1002 } else if (error == -ENOSPC) { 1003 if (EXT3_I(inode)->i_file_acl && !bs.s.base) { 1004 error = ext3_xattr_block_find(inode, &i, &bs); 1005 if (error) 1006 goto cleanup; 1007 } 1008 error = ext3_xattr_block_set(handle, inode, &i, &bs); 1009 if (error) 1010 goto cleanup; 1011 if (!is.s.not_found) { 1012 i.value = NULL; 1013 error = ext3_xattr_ibody_set(handle, inode, &i, 1014 &is); 1015 } 1016 } 1017 } 1018 if (!error) { 1019 ext3_xattr_update_super_block(handle, inode->i_sb); 1020 inode->i_ctime = CURRENT_TIME_SEC; 1021 error = ext3_mark_iloc_dirty(handle, inode, &is.iloc); 1022 /* 1023 * The bh is consumed by ext3_mark_iloc_dirty, even with 1024 * error != 0. 1025 */ 1026 is.iloc.bh = NULL; 1027 if (IS_SYNC(inode)) 1028 handle->h_sync = 1; 1029 } 1030 1031cleanup: 1032 brelse(is.iloc.bh); 1033 brelse(bs.bh); 1034 up_write(&EXT3_I(inode)->xattr_sem); 1035 return error; 1036} 1037 1038/* 1039 * ext3_xattr_set() 1040 * 1041 * Like ext3_xattr_set_handle, but start from an inode. This extended 1042 * attribute modification is a filesystem transaction by itself. 1043 * 1044 * Returns 0, or a negative error number on failure. 1045 */ 1046int 1047ext3_xattr_set(struct inode *inode, int name_index, const char *name, 1048 const void *value, size_t value_len, int flags) 1049{ 1050 handle_t *handle; 1051 int error, retries = 0; 1052 1053retry: 1054 handle = ext3_journal_start(inode, EXT3_DATA_TRANS_BLOCKS(inode->i_sb)); 1055 if (IS_ERR(handle)) { 1056 error = PTR_ERR(handle); 1057 } else { 1058 int error2; 1059 1060 error = ext3_xattr_set_handle(handle, inode, name_index, name, 1061 value, value_len, flags); 1062 error2 = ext3_journal_stop(handle); 1063 if (error == -ENOSPC && 1064 ext3_should_retry_alloc(inode->i_sb, &retries)) 1065 goto retry; 1066 if (error == 0) 1067 error = error2; 1068 } 1069 1070 return error; 1071} 1072 1073/* 1074 * ext3_xattr_delete_inode() 1075 * 1076 * Free extended attribute resources associated with this inode. This 1077 * is called immediately before an inode is freed. We have exclusive 1078 * access to the inode. 1079 */ 1080void 1081ext3_xattr_delete_inode(handle_t *handle, struct inode *inode) 1082{ 1083 struct buffer_head *bh = NULL; 1084 1085 if (!EXT3_I(inode)->i_file_acl) 1086 goto cleanup; 1087 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl); 1088 if (!bh) { 1089 ext3_error(inode->i_sb, __func__, 1090 "inode %lu: block "E3FSBLK" read error", inode->i_ino, 1091 EXT3_I(inode)->i_file_acl); 1092 goto cleanup; 1093 } 1094 if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || 1095 BHDR(bh)->h_blocks != cpu_to_le32(1)) { 1096 ext3_error(inode->i_sb, __func__, 1097 "inode %lu: bad block "E3FSBLK, inode->i_ino, 1098 EXT3_I(inode)->i_file_acl); 1099 goto cleanup; 1100 } 1101 ext3_xattr_release_block(handle, inode, bh); 1102 EXT3_I(inode)->i_file_acl = 0; 1103 1104cleanup: 1105 brelse(bh); 1106} 1107 1108/* 1109 * ext3_xattr_put_super() 1110 * 1111 * This is called when a file system is unmounted. 1112 */ 1113void 1114ext3_xattr_put_super(struct super_block *sb) 1115{ 1116 mb_cache_shrink(sb->s_bdev); 1117} 1118 1119/* 1120 * ext3_xattr_cache_insert() 1121 * 1122 * Create a new entry in the extended attribute cache, and insert 1123 * it unless such an entry is already in the cache. 1124 * 1125 * Returns 0, or a negative error number on failure. 1126 */ 1127static void 1128ext3_xattr_cache_insert(struct buffer_head *bh) 1129{ 1130 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash); 1131 struct mb_cache_entry *ce; 1132 int error; 1133 1134 ce = mb_cache_entry_alloc(ext3_xattr_cache, GFP_NOFS); 1135 if (!ce) { 1136 ea_bdebug(bh, "out of memory"); 1137 return; 1138 } 1139 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash); 1140 if (error) { 1141 mb_cache_entry_free(ce); 1142 if (error == -EBUSY) { 1143 ea_bdebug(bh, "already in cache"); 1144 error = 0; 1145 } 1146 } else { 1147 ea_bdebug(bh, "inserting [%x]", (int)hash); 1148 mb_cache_entry_release(ce); 1149 } 1150} 1151 1152/* 1153 * ext3_xattr_cmp() 1154 * 1155 * Compare two extended attribute blocks for equality. 1156 * 1157 * Returns 0 if the blocks are equal, 1 if they differ, and 1158 * a negative error number on errors. 1159 */ 1160static int 1161ext3_xattr_cmp(struct ext3_xattr_header *header1, 1162 struct ext3_xattr_header *header2) 1163{ 1164 struct ext3_xattr_entry *entry1, *entry2; 1165 1166 entry1 = ENTRY(header1+1); 1167 entry2 = ENTRY(header2+1); 1168 while (!IS_LAST_ENTRY(entry1)) { 1169 if (IS_LAST_ENTRY(entry2)) 1170 return 1; 1171 if (entry1->e_hash != entry2->e_hash || 1172 entry1->e_name_index != entry2->e_name_index || 1173 entry1->e_name_len != entry2->e_name_len || 1174 entry1->e_value_size != entry2->e_value_size || 1175 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len)) 1176 return 1; 1177 if (entry1->e_value_block != 0 || entry2->e_value_block != 0) 1178 return -EIO; 1179 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs), 1180 (char *)header2 + le16_to_cpu(entry2->e_value_offs), 1181 le32_to_cpu(entry1->e_value_size))) 1182 return 1; 1183 1184 entry1 = EXT3_XATTR_NEXT(entry1); 1185 entry2 = EXT3_XATTR_NEXT(entry2); 1186 } 1187 if (!IS_LAST_ENTRY(entry2)) 1188 return 1; 1189 return 0; 1190} 1191 1192/* 1193 * ext3_xattr_cache_find() 1194 * 1195 * Find an identical extended attribute block. 1196 * 1197 * Returns a pointer to the block found, or NULL if such a block was 1198 * not found or an error occurred. 1199 */ 1200static struct buffer_head * 1201ext3_xattr_cache_find(struct inode *inode, struct ext3_xattr_header *header, 1202 struct mb_cache_entry **pce) 1203{ 1204 __u32 hash = le32_to_cpu(header->h_hash); 1205 struct mb_cache_entry *ce; 1206 1207 if (!header->h_hash) 1208 return NULL; /* never share */ 1209 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash); 1210again: 1211 ce = mb_cache_entry_find_first(ext3_xattr_cache, 0, 1212 inode->i_sb->s_bdev, hash); 1213 while (ce) { 1214 struct buffer_head *bh; 1215 1216 if (IS_ERR(ce)) { 1217 if (PTR_ERR(ce) == -EAGAIN) 1218 goto again; 1219 break; 1220 } 1221 bh = sb_bread(inode->i_sb, ce->e_block); 1222 if (!bh) { 1223 ext3_error(inode->i_sb, __func__, 1224 "inode %lu: block %lu read error", 1225 inode->i_ino, (unsigned long) ce->e_block); 1226 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >= 1227 EXT3_XATTR_REFCOUNT_MAX) { 1228 ea_idebug(inode, "block %lu refcount %d>=%d", 1229 (unsigned long) ce->e_block, 1230 le32_to_cpu(BHDR(bh)->h_refcount), 1231 EXT3_XATTR_REFCOUNT_MAX); 1232 } else if (ext3_xattr_cmp(header, BHDR(bh)) == 0) { 1233 *pce = ce; 1234 return bh; 1235 } 1236 brelse(bh); 1237 ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash); 1238 } 1239 return NULL; 1240} 1241 1242#define NAME_HASH_SHIFT 5 1243#define VALUE_HASH_SHIFT 16 1244 1245/* 1246 * ext3_xattr_hash_entry() 1247 * 1248 * Compute the hash of an extended attribute. 1249 */ 1250static inline void ext3_xattr_hash_entry(struct ext3_xattr_header *header, 1251 struct ext3_xattr_entry *entry) 1252{ 1253 __u32 hash = 0; 1254 char *name = entry->e_name; 1255 int n; 1256 1257 for (n=0; n < entry->e_name_len; n++) { 1258 hash = (hash << NAME_HASH_SHIFT) ^ 1259 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^ 1260 *name++; 1261 } 1262 1263 if (entry->e_value_block == 0 && entry->e_value_size != 0) { 1264 __le32 *value = (__le32 *)((char *)header + 1265 le16_to_cpu(entry->e_value_offs)); 1266 for (n = (le32_to_cpu(entry->e_value_size) + 1267 EXT3_XATTR_ROUND) >> EXT3_XATTR_PAD_BITS; n; n--) { 1268 hash = (hash << VALUE_HASH_SHIFT) ^ 1269 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^ 1270 le32_to_cpu(*value++); 1271 } 1272 } 1273 entry->e_hash = cpu_to_le32(hash); 1274} 1275 1276#undef NAME_HASH_SHIFT 1277#undef VALUE_HASH_SHIFT 1278 1279#define BLOCK_HASH_SHIFT 16 1280 1281/* 1282 * ext3_xattr_rehash() 1283 * 1284 * Re-compute the extended attribute hash value after an entry has changed. 1285 */ 1286static void ext3_xattr_rehash(struct ext3_xattr_header *header, 1287 struct ext3_xattr_entry *entry) 1288{ 1289 struct ext3_xattr_entry *here; 1290 __u32 hash = 0; 1291 1292 ext3_xattr_hash_entry(header, entry); 1293 here = ENTRY(header+1); 1294 while (!IS_LAST_ENTRY(here)) { 1295 if (!here->e_hash) { 1296 /* Block is not shared if an entry's hash value == 0 */ 1297 hash = 0; 1298 break; 1299 } 1300 hash = (hash << BLOCK_HASH_SHIFT) ^ 1301 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^ 1302 le32_to_cpu(here->e_hash); 1303 here = EXT3_XATTR_NEXT(here); 1304 } 1305 header->h_hash = cpu_to_le32(hash); 1306} 1307 1308#undef BLOCK_HASH_SHIFT 1309 1310int __init 1311init_ext3_xattr(void) 1312{ 1313 ext3_xattr_cache = mb_cache_create("ext3_xattr", NULL, 1314 sizeof(struct mb_cache_entry) + 1315 sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6); 1316 if (!ext3_xattr_cache) 1317 return -ENOMEM; 1318 return 0; 1319} 1320 1321void 1322exit_ext3_xattr(void) 1323{ 1324 if (ext3_xattr_cache) 1325 mb_cache_destroy(ext3_xattr_cache); 1326 ext3_xattr_cache = NULL; 1327}