1/* 2 * dir.c - NILFS directory entry operations 3 * 4 * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation. 5 * 6 * This program is free software; you can redistribute it and/or modify 7 * it under the terms of the GNU General Public License as published by 8 * the Free Software Foundation; either version 2 of the License, or 9 * (at your option) any later version. 10 * 11 * This program is distributed in the hope that it will be useful, 12 * but WITHOUT ANY WARRANTY; without even the implied warranty of 13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 14 * GNU General Public License for more details. 15 * 16 * You should have received a copy of the GNU General Public License 17 * along with this program; if not, write to the Free Software 18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA 19 * 20 * Modified for NILFS by Amagai Yoshiji <amagai@osrg.net> 21 */ 22/* 23 * linux/fs/ext2/dir.c 24 * 25 * Copyright (C) 1992, 1993, 1994, 1995 26 * Remy Card (card@masi.ibp.fr) 27 * Laboratoire MASI - Institut Blaise Pascal 28 * Universite Pierre et Marie Curie (Paris VI) 29 * 30 * from 31 * 32 * linux/fs/minix/dir.c 33 * 34 * Copyright (C) 1991, 1992 Linus Torvalds 35 * 36 * ext2 directory handling functions 37 * 38 * Big-endian to little-endian byte-swapping/bitmaps by 39 * David S. Miller (davem@caip.rutgers.edu), 1995 40 * 41 * All code that works with directory layout had been switched to pagecache 42 * and moved here. AV 43 */ 44 45#include <linux/pagemap.h> 46#include "nilfs.h" 47#include "page.h" 48 49/* 50 * nilfs uses block-sized chunks. Arguably, sector-sized ones would be 51 * more robust, but we have what we have 52 */ 53static inline unsigned nilfs_chunk_size(struct inode *inode) 54{ 55 return inode->i_sb->s_blocksize; 56} 57 58static inline void nilfs_put_page(struct page *page) 59{ 60 kunmap(page); 61 page_cache_release(page); 62} 63 64static inline unsigned long dir_pages(struct inode *inode) 65{ 66 return (inode->i_size+PAGE_CACHE_SIZE-1)>>PAGE_CACHE_SHIFT; 67} 68 69/* 70 * Return the offset into page `page_nr' of the last valid 71 * byte in that page, plus one. 72 */ 73static unsigned nilfs_last_byte(struct inode *inode, unsigned long page_nr) 74{ 75 unsigned last_byte = inode->i_size; 76 77 last_byte -= page_nr << PAGE_CACHE_SHIFT; 78 if (last_byte > PAGE_CACHE_SIZE) 79 last_byte = PAGE_CACHE_SIZE; 80 return last_byte; 81} 82 83static int nilfs_prepare_chunk(struct page *page, unsigned from, unsigned to) 84{ 85 loff_t pos = page_offset(page) + from; 86 return __block_write_begin(page, pos, to - from, nilfs_get_block); 87} 88 89static void nilfs_commit_chunk(struct page *page, 90 struct address_space *mapping, 91 unsigned from, unsigned to) 92{ 93 struct inode *dir = mapping->host; 94 loff_t pos = page_offset(page) + from; 95 unsigned len = to - from; 96 unsigned nr_dirty, copied; 97 int err; 98 99 nr_dirty = nilfs_page_count_clean_buffers(page, from, to); 100 copied = block_write_end(NULL, mapping, pos, len, len, page, NULL); 101 if (pos + copied > dir->i_size) 102 i_size_write(dir, pos + copied); 103 if (IS_DIRSYNC(dir)) 104 nilfs_set_transaction_flag(NILFS_TI_SYNC); 105 err = nilfs_set_file_dirty(dir, nr_dirty); 106 WARN_ON(err); /* do not happen */ 107 unlock_page(page); 108} 109 110static void nilfs_check_page(struct page *page) 111{ 112 struct inode *dir = page->mapping->host; 113 struct super_block *sb = dir->i_sb; 114 unsigned chunk_size = nilfs_chunk_size(dir); 115 char *kaddr = page_address(page); 116 unsigned offs, rec_len; 117 unsigned limit = PAGE_CACHE_SIZE; 118 struct nilfs_dir_entry *p; 119 char *error; 120 121 if ((dir->i_size >> PAGE_CACHE_SHIFT) == page->index) { 122 limit = dir->i_size & ~PAGE_CACHE_MASK; 123 if (limit & (chunk_size - 1)) 124 goto Ebadsize; 125 if (!limit) 126 goto out; 127 } 128 for (offs = 0; offs <= limit - NILFS_DIR_REC_LEN(1); offs += rec_len) { 129 p = (struct nilfs_dir_entry *)(kaddr + offs); 130 rec_len = nilfs_rec_len_from_disk(p->rec_len); 131 132 if (rec_len < NILFS_DIR_REC_LEN(1)) 133 goto Eshort; 134 if (rec_len & 3) 135 goto Ealign; 136 if (rec_len < NILFS_DIR_REC_LEN(p->name_len)) 137 goto Enamelen; 138 if (((offs + rec_len - 1) ^ offs) & ~(chunk_size-1)) 139 goto Espan; 140 } 141 if (offs != limit) 142 goto Eend; 143out: 144 SetPageChecked(page); 145 return; 146 147 /* Too bad, we had an error */ 148 149Ebadsize: 150 nilfs_error(sb, "nilfs_check_page", 151 "size of directory #%lu is not a multiple of chunk size", 152 dir->i_ino 153 ); 154 goto fail; 155Eshort: 156 error = "rec_len is smaller than minimal"; 157 goto bad_entry; 158Ealign: 159 error = "unaligned directory entry"; 160 goto bad_entry; 161Enamelen: 162 error = "rec_len is too small for name_len"; 163 goto bad_entry; 164Espan: 165 error = "directory entry across blocks"; 166bad_entry: 167 nilfs_error(sb, "nilfs_check_page", "bad entry in directory #%lu: %s - " 168 "offset=%lu, inode=%lu, rec_len=%d, name_len=%d", 169 dir->i_ino, error, (page->index<<PAGE_CACHE_SHIFT)+offs, 170 (unsigned long) le64_to_cpu(p->inode), 171 rec_len, p->name_len); 172 goto fail; 173Eend: 174 p = (struct nilfs_dir_entry *)(kaddr + offs); 175 nilfs_error(sb, "nilfs_check_page", 176 "entry in directory #%lu spans the page boundary" 177 "offset=%lu, inode=%lu", 178 dir->i_ino, (page->index<<PAGE_CACHE_SHIFT)+offs, 179 (unsigned long) le64_to_cpu(p->inode)); 180fail: 181 SetPageChecked(page); 182 SetPageError(page); 183} 184 185static struct page *nilfs_get_page(struct inode *dir, unsigned long n) 186{ 187 struct address_space *mapping = dir->i_mapping; 188 struct page *page = read_mapping_page(mapping, n, NULL); 189 190 if (!IS_ERR(page)) { 191 kmap(page); 192 if (!PageChecked(page)) 193 nilfs_check_page(page); 194 if (PageError(page)) 195 goto fail; 196 } 197 return page; 198 199fail: 200 nilfs_put_page(page); 201 return ERR_PTR(-EIO); 202} 203 204/* 205 * NOTE! unlike strncmp, nilfs_match returns 1 for success, 0 for failure. 206 * 207 * len <= NILFS_NAME_LEN and de != NULL are guaranteed by caller. 208 */ 209static int 210nilfs_match(int len, const unsigned char *name, struct nilfs_dir_entry *de) 211{ 212 if (len != de->name_len) 213 return 0; 214 if (!de->inode) 215 return 0; 216 return !memcmp(name, de->name, len); 217} 218 219/* 220 * p is at least 6 bytes before the end of page 221 */ 222static struct nilfs_dir_entry *nilfs_next_entry(struct nilfs_dir_entry *p) 223{ 224 return (struct nilfs_dir_entry *)((char *)p + 225 nilfs_rec_len_from_disk(p->rec_len)); 226} 227 228static unsigned char 229nilfs_filetype_table[NILFS_FT_MAX] = { 230 [NILFS_FT_UNKNOWN] = DT_UNKNOWN, 231 [NILFS_FT_REG_FILE] = DT_REG, 232 [NILFS_FT_DIR] = DT_DIR, 233 [NILFS_FT_CHRDEV] = DT_CHR, 234 [NILFS_FT_BLKDEV] = DT_BLK, 235 [NILFS_FT_FIFO] = DT_FIFO, 236 [NILFS_FT_SOCK] = DT_SOCK, 237 [NILFS_FT_SYMLINK] = DT_LNK, 238}; 239 240#define S_SHIFT 12 241static unsigned char 242nilfs_type_by_mode[S_IFMT >> S_SHIFT] = { 243 [S_IFREG >> S_SHIFT] = NILFS_FT_REG_FILE, 244 [S_IFDIR >> S_SHIFT] = NILFS_FT_DIR, 245 [S_IFCHR >> S_SHIFT] = NILFS_FT_CHRDEV, 246 [S_IFBLK >> S_SHIFT] = NILFS_FT_BLKDEV, 247 [S_IFIFO >> S_SHIFT] = NILFS_FT_FIFO, 248 [S_IFSOCK >> S_SHIFT] = NILFS_FT_SOCK, 249 [S_IFLNK >> S_SHIFT] = NILFS_FT_SYMLINK, 250}; 251 252static void nilfs_set_de_type(struct nilfs_dir_entry *de, struct inode *inode) 253{ 254 umode_t mode = inode->i_mode; 255 256 de->file_type = nilfs_type_by_mode[(mode & S_IFMT)>>S_SHIFT]; 257} 258 259static int nilfs_readdir(struct file *file, struct dir_context *ctx) 260{ 261 loff_t pos = ctx->pos; 262 struct inode *inode = file_inode(file); 263 struct super_block *sb = inode->i_sb; 264 unsigned int offset = pos & ~PAGE_CACHE_MASK; 265 unsigned long n = pos >> PAGE_CACHE_SHIFT; 266 unsigned long npages = dir_pages(inode); 267/* unsigned chunk_mask = ~(nilfs_chunk_size(inode)-1); */ 268 269 if (pos > inode->i_size - NILFS_DIR_REC_LEN(1)) 270 return 0; 271 272 for ( ; n < npages; n++, offset = 0) { 273 char *kaddr, *limit; 274 struct nilfs_dir_entry *de; 275 struct page *page = nilfs_get_page(inode, n); 276 277 if (IS_ERR(page)) { 278 nilfs_error(sb, __func__, "bad page in #%lu", 279 inode->i_ino); 280 ctx->pos += PAGE_CACHE_SIZE - offset; 281 return -EIO; 282 } 283 kaddr = page_address(page); 284 de = (struct nilfs_dir_entry *)(kaddr + offset); 285 limit = kaddr + nilfs_last_byte(inode, n) - 286 NILFS_DIR_REC_LEN(1); 287 for ( ; (char *)de <= limit; de = nilfs_next_entry(de)) { 288 if (de->rec_len == 0) { 289 nilfs_error(sb, __func__, 290 "zero-length directory entry"); 291 nilfs_put_page(page); 292 return -EIO; 293 } 294 if (de->inode) { 295 unsigned char t; 296 297 if (de->file_type < NILFS_FT_MAX) 298 t = nilfs_filetype_table[de->file_type]; 299 else 300 t = DT_UNKNOWN; 301 302 if (!dir_emit(ctx, de->name, de->name_len, 303 le64_to_cpu(de->inode), t)) { 304 nilfs_put_page(page); 305 return 0; 306 } 307 } 308 ctx->pos += nilfs_rec_len_from_disk(de->rec_len); 309 } 310 nilfs_put_page(page); 311 } 312 return 0; 313} 314 315/* 316 * nilfs_find_entry() 317 * 318 * finds an entry in the specified directory with the wanted name. It 319 * returns the page in which the entry was found, and the entry itself 320 * (as a parameter - res_dir). Page is returned mapped and unlocked. 321 * Entry is guaranteed to be valid. 322 */ 323struct nilfs_dir_entry * 324nilfs_find_entry(struct inode *dir, const struct qstr *qstr, 325 struct page **res_page) 326{ 327 const unsigned char *name = qstr->name; 328 int namelen = qstr->len; 329 unsigned reclen = NILFS_DIR_REC_LEN(namelen); 330 unsigned long start, n; 331 unsigned long npages = dir_pages(dir); 332 struct page *page = NULL; 333 struct nilfs_inode_info *ei = NILFS_I(dir); 334 struct nilfs_dir_entry *de; 335 336 if (npages == 0) 337 goto out; 338 339 /* OFFSET_CACHE */ 340 *res_page = NULL; 341 342 start = ei->i_dir_start_lookup; 343 if (start >= npages) 344 start = 0; 345 n = start; 346 do { 347 char *kaddr; 348 page = nilfs_get_page(dir, n); 349 if (!IS_ERR(page)) { 350 kaddr = page_address(page); 351 de = (struct nilfs_dir_entry *)kaddr; 352 kaddr += nilfs_last_byte(dir, n) - reclen; 353 while ((char *) de <= kaddr) { 354 if (de->rec_len == 0) { 355 nilfs_error(dir->i_sb, __func__, 356 "zero-length directory entry"); 357 nilfs_put_page(page); 358 goto out; 359 } 360 if (nilfs_match(namelen, name, de)) 361 goto found; 362 de = nilfs_next_entry(de); 363 } 364 nilfs_put_page(page); 365 } 366 if (++n >= npages) 367 n = 0; 368 /* next page is past the blocks we've got */ 369 if (unlikely(n > (dir->i_blocks >> (PAGE_CACHE_SHIFT - 9)))) { 370 nilfs_error(dir->i_sb, __func__, 371 "dir %lu size %lld exceeds block count %llu", 372 dir->i_ino, dir->i_size, 373 (unsigned long long)dir->i_blocks); 374 goto out; 375 } 376 } while (n != start); 377out: 378 return NULL; 379 380found: 381 *res_page = page; 382 ei->i_dir_start_lookup = n; 383 return de; 384} 385 386struct nilfs_dir_entry *nilfs_dotdot(struct inode *dir, struct page **p) 387{ 388 struct page *page = nilfs_get_page(dir, 0); 389 struct nilfs_dir_entry *de = NULL; 390 391 if (!IS_ERR(page)) { 392 de = nilfs_next_entry( 393 (struct nilfs_dir_entry *)page_address(page)); 394 *p = page; 395 } 396 return de; 397} 398 399ino_t nilfs_inode_by_name(struct inode *dir, const struct qstr *qstr) 400{ 401 ino_t res = 0; 402 struct nilfs_dir_entry *de; 403 struct page *page; 404 405 de = nilfs_find_entry(dir, qstr, &page); 406 if (de) { 407 res = le64_to_cpu(de->inode); 408 kunmap(page); 409 page_cache_release(page); 410 } 411 return res; 412} 413 414/* Releases the page */ 415void nilfs_set_link(struct inode *dir, struct nilfs_dir_entry *de, 416 struct page *page, struct inode *inode) 417{ 418 unsigned from = (char *) de - (char *) page_address(page); 419 unsigned to = from + nilfs_rec_len_from_disk(de->rec_len); 420 struct address_space *mapping = page->mapping; 421 int err; 422 423 lock_page(page); 424 err = nilfs_prepare_chunk(page, from, to); 425 BUG_ON(err); 426 de->inode = cpu_to_le64(inode->i_ino); 427 nilfs_set_de_type(de, inode); 428 nilfs_commit_chunk(page, mapping, from, to); 429 nilfs_put_page(page); 430 dir->i_mtime = dir->i_ctime = CURRENT_TIME; 431} 432 433/* 434 * Parent is locked. 435 */ 436int nilfs_add_link(struct dentry *dentry, struct inode *inode) 437{ 438 struct inode *dir = d_inode(dentry->d_parent); 439 const unsigned char *name = dentry->d_name.name; 440 int namelen = dentry->d_name.len; 441 unsigned chunk_size = nilfs_chunk_size(dir); 442 unsigned reclen = NILFS_DIR_REC_LEN(namelen); 443 unsigned short rec_len, name_len; 444 struct page *page = NULL; 445 struct nilfs_dir_entry *de; 446 unsigned long npages = dir_pages(dir); 447 unsigned long n; 448 char *kaddr; 449 unsigned from, to; 450 int err; 451 452 /* 453 * We take care of directory expansion in the same loop. 454 * This code plays outside i_size, so it locks the page 455 * to protect that region. 456 */ 457 for (n = 0; n <= npages; n++) { 458 char *dir_end; 459 460 page = nilfs_get_page(dir, n); 461 err = PTR_ERR(page); 462 if (IS_ERR(page)) 463 goto out; 464 lock_page(page); 465 kaddr = page_address(page); 466 dir_end = kaddr + nilfs_last_byte(dir, n); 467 de = (struct nilfs_dir_entry *)kaddr; 468 kaddr += PAGE_CACHE_SIZE - reclen; 469 while ((char *)de <= kaddr) { 470 if ((char *)de == dir_end) { 471 /* We hit i_size */ 472 name_len = 0; 473 rec_len = chunk_size; 474 de->rec_len = nilfs_rec_len_to_disk(chunk_size); 475 de->inode = 0; 476 goto got_it; 477 } 478 if (de->rec_len == 0) { 479 nilfs_error(dir->i_sb, __func__, 480 "zero-length directory entry"); 481 err = -EIO; 482 goto out_unlock; 483 } 484 err = -EEXIST; 485 if (nilfs_match(namelen, name, de)) 486 goto out_unlock; 487 name_len = NILFS_DIR_REC_LEN(de->name_len); 488 rec_len = nilfs_rec_len_from_disk(de->rec_len); 489 if (!de->inode && rec_len >= reclen) 490 goto got_it; 491 if (rec_len >= name_len + reclen) 492 goto got_it; 493 de = (struct nilfs_dir_entry *)((char *)de + rec_len); 494 } 495 unlock_page(page); 496 nilfs_put_page(page); 497 } 498 BUG(); 499 return -EINVAL; 500 501got_it: 502 from = (char *)de - (char *)page_address(page); 503 to = from + rec_len; 504 err = nilfs_prepare_chunk(page, from, to); 505 if (err) 506 goto out_unlock; 507 if (de->inode) { 508 struct nilfs_dir_entry *de1; 509 510 de1 = (struct nilfs_dir_entry *)((char *)de + name_len); 511 de1->rec_len = nilfs_rec_len_to_disk(rec_len - name_len); 512 de->rec_len = nilfs_rec_len_to_disk(name_len); 513 de = de1; 514 } 515 de->name_len = namelen; 516 memcpy(de->name, name, namelen); 517 de->inode = cpu_to_le64(inode->i_ino); 518 nilfs_set_de_type(de, inode); 519 nilfs_commit_chunk(page, page->mapping, from, to); 520 dir->i_mtime = dir->i_ctime = CURRENT_TIME; 521 nilfs_mark_inode_dirty(dir); 522 /* OFFSET_CACHE */ 523out_put: 524 nilfs_put_page(page); 525out: 526 return err; 527out_unlock: 528 unlock_page(page); 529 goto out_put; 530} 531 532/* 533 * nilfs_delete_entry deletes a directory entry by merging it with the 534 * previous entry. Page is up-to-date. Releases the page. 535 */ 536int nilfs_delete_entry(struct nilfs_dir_entry *dir, struct page *page) 537{ 538 struct address_space *mapping = page->mapping; 539 struct inode *inode = mapping->host; 540 char *kaddr = page_address(page); 541 unsigned from = ((char *)dir - kaddr) & ~(nilfs_chunk_size(inode) - 1); 542 unsigned to = ((char *)dir - kaddr) + 543 nilfs_rec_len_from_disk(dir->rec_len); 544 struct nilfs_dir_entry *pde = NULL; 545 struct nilfs_dir_entry *de = (struct nilfs_dir_entry *)(kaddr + from); 546 int err; 547 548 while ((char *)de < (char *)dir) { 549 if (de->rec_len == 0) { 550 nilfs_error(inode->i_sb, __func__, 551 "zero-length directory entry"); 552 err = -EIO; 553 goto out; 554 } 555 pde = de; 556 de = nilfs_next_entry(de); 557 } 558 if (pde) 559 from = (char *)pde - (char *)page_address(page); 560 lock_page(page); 561 err = nilfs_prepare_chunk(page, from, to); 562 BUG_ON(err); 563 if (pde) 564 pde->rec_len = nilfs_rec_len_to_disk(to - from); 565 dir->inode = 0; 566 nilfs_commit_chunk(page, mapping, from, to); 567 inode->i_ctime = inode->i_mtime = CURRENT_TIME; 568out: 569 nilfs_put_page(page); 570 return err; 571} 572 573/* 574 * Set the first fragment of directory. 575 */ 576int nilfs_make_empty(struct inode *inode, struct inode *parent) 577{ 578 struct address_space *mapping = inode->i_mapping; 579 struct page *page = grab_cache_page(mapping, 0); 580 unsigned chunk_size = nilfs_chunk_size(inode); 581 struct nilfs_dir_entry *de; 582 int err; 583 void *kaddr; 584 585 if (!page) 586 return -ENOMEM; 587 588 err = nilfs_prepare_chunk(page, 0, chunk_size); 589 if (unlikely(err)) { 590 unlock_page(page); 591 goto fail; 592 } 593 kaddr = kmap_atomic(page); 594 memset(kaddr, 0, chunk_size); 595 de = (struct nilfs_dir_entry *)kaddr; 596 de->name_len = 1; 597 de->rec_len = nilfs_rec_len_to_disk(NILFS_DIR_REC_LEN(1)); 598 memcpy(de->name, ".\0\0", 4); 599 de->inode = cpu_to_le64(inode->i_ino); 600 nilfs_set_de_type(de, inode); 601 602 de = (struct nilfs_dir_entry *)(kaddr + NILFS_DIR_REC_LEN(1)); 603 de->name_len = 2; 604 de->rec_len = nilfs_rec_len_to_disk(chunk_size - NILFS_DIR_REC_LEN(1)); 605 de->inode = cpu_to_le64(parent->i_ino); 606 memcpy(de->name, "..\0", 4); 607 nilfs_set_de_type(de, inode); 608 kunmap_atomic(kaddr); 609 nilfs_commit_chunk(page, mapping, 0, chunk_size); 610fail: 611 page_cache_release(page); 612 return err; 613} 614 615/* 616 * routine to check that the specified directory is empty (for rmdir) 617 */ 618int nilfs_empty_dir(struct inode *inode) 619{ 620 struct page *page = NULL; 621 unsigned long i, npages = dir_pages(inode); 622 623 for (i = 0; i < npages; i++) { 624 char *kaddr; 625 struct nilfs_dir_entry *de; 626 627 page = nilfs_get_page(inode, i); 628 if (IS_ERR(page)) 629 continue; 630 631 kaddr = page_address(page); 632 de = (struct nilfs_dir_entry *)kaddr; 633 kaddr += nilfs_last_byte(inode, i) - NILFS_DIR_REC_LEN(1); 634 635 while ((char *)de <= kaddr) { 636 if (de->rec_len == 0) { 637 nilfs_error(inode->i_sb, __func__, 638 "zero-length directory entry " 639 "(kaddr=%p, de=%p)\n", kaddr, de); 640 goto not_empty; 641 } 642 if (de->inode != 0) { 643 /* check for . and .. */ 644 if (de->name[0] != '.') 645 goto not_empty; 646 if (de->name_len > 2) 647 goto not_empty; 648 if (de->name_len < 2) { 649 if (de->inode != 650 cpu_to_le64(inode->i_ino)) 651 goto not_empty; 652 } else if (de->name[1] != '.') 653 goto not_empty; 654 } 655 de = nilfs_next_entry(de); 656 } 657 nilfs_put_page(page); 658 } 659 return 1; 660 661not_empty: 662 nilfs_put_page(page); 663 return 0; 664} 665 666const struct file_operations nilfs_dir_operations = { 667 .llseek = generic_file_llseek, 668 .read = generic_read_dir, 669 .iterate = nilfs_readdir, 670 .unlocked_ioctl = nilfs_ioctl, 671#ifdef CONFIG_COMPAT 672 .compat_ioctl = nilfs_compat_ioctl, 673#endif /* CONFIG_COMPAT */ 674 .fsync = nilfs_sync_file, 675 676}; 677