1/* 2 * linux/fs/ext2/inode.c 3 * 4 * Copyright (C) 1992, 1993, 1994, 1995 5 * Remy Card (card@masi.ibp.fr) 6 * Laboratoire MASI - Institut Blaise Pascal 7 * Universite Pierre et Marie Curie (Paris VI) 8 * 9 * from 10 * 11 * linux/fs/minix/inode.c 12 * 13 * Copyright (C) 1991, 1992 Linus Torvalds 14 * 15 * Goal-directed block allocation by Stephen Tweedie 16 * (sct@dcs.ed.ac.uk), 1993, 1998 17 * Big-endian to little-endian byte-swapping/bitmaps by 18 * David S. Miller (davem@caip.rutgers.edu), 1995 19 * 64-bit file support on 64-bit platforms by Jakub Jelinek 20 * (jj@sunsite.ms.mff.cuni.cz) 21 * 22 * Assorted race fixes, rewrite of ext2_get_block() by Al Viro, 2000 23 */ 24 25#include <linux/time.h> 26#include <linux/highuid.h> 27#include <linux/pagemap.h> 28#include <linux/quotaops.h> 29#include <linux/module.h> 30#include <linux/writeback.h> 31#include <linux/buffer_head.h> 32#include <linux/mpage.h> 33#include <linux/fiemap.h> 34#include <linux/namei.h> 35#include "ext2.h" 36#include "acl.h" 37#include "xip.h" 38 39MODULE_AUTHOR("Remy Card and others"); 40MODULE_DESCRIPTION("Second Extended Filesystem"); 41MODULE_LICENSE("GPL"); 42 43static int __ext2_write_inode(struct inode *inode, int do_sync); 44 45/* 46 * Test whether an inode is a fast symlink. 47 */ 48static inline int ext2_inode_is_fast_symlink(struct inode *inode) 49{ 50 int ea_blocks = EXT2_I(inode)->i_file_acl ? 51 (inode->i_sb->s_blocksize >> 9) : 0; 52 53 return (S_ISLNK(inode->i_mode) && 54 inode->i_blocks - ea_blocks == 0); 55} 56 57static void ext2_truncate_blocks(struct inode *inode, loff_t offset); 58 59static void ext2_write_failed(struct address_space *mapping, loff_t to) 60{ 61 struct inode *inode = mapping->host; 62 63 if (to > inode->i_size) { 64 truncate_pagecache(inode, to, inode->i_size); 65 ext2_truncate_blocks(inode, inode->i_size); 66 } 67} 68 69/* 70 * Called at the last iput() if i_nlink is zero. 71 */ 72void ext2_evict_inode(struct inode * inode) 73{ 74 struct ext2_block_alloc_info *rsv; 75 int want_delete = 0; 76 77 if (!inode->i_nlink && !is_bad_inode(inode)) { 78 want_delete = 1; 79 dquot_initialize(inode); 80 } else { 81 dquot_drop(inode); 82 } 83 84 truncate_inode_pages(&inode->i_data, 0); 85 86 if (want_delete) { 87 /* set dtime */ 88 EXT2_I(inode)->i_dtime = get_seconds(); 89 mark_inode_dirty(inode); 90 __ext2_write_inode(inode, inode_needs_sync(inode)); 91 /* truncate to 0 */ 92 inode->i_size = 0; 93 if (inode->i_blocks) 94 ext2_truncate_blocks(inode, 0); 95 } 96 97 invalidate_inode_buffers(inode); 98 end_writeback(inode); 99 100 ext2_discard_reservation(inode); 101 rsv = EXT2_I(inode)->i_block_alloc_info; 102 EXT2_I(inode)->i_block_alloc_info = NULL; 103 if (unlikely(rsv)) 104 kfree(rsv); 105 106 if (want_delete) 107 ext2_free_inode(inode); 108} 109 110typedef struct { 111 __le32 *p; 112 __le32 key; 113 struct buffer_head *bh; 114} Indirect; 115 116static inline void add_chain(Indirect *p, struct buffer_head *bh, __le32 *v) 117{ 118 p->key = *(p->p = v); 119 p->bh = bh; 120} 121 122static inline int verify_chain(Indirect *from, Indirect *to) 123{ 124 while (from <= to && from->key == *from->p) 125 from++; 126 return (from > to); 127} 128 129/** 130 * ext2_block_to_path - parse the block number into array of offsets 131 * @inode: inode in question (we are only interested in its superblock) 132 * @i_block: block number to be parsed 133 * @offsets: array to store the offsets in 134 * @boundary: set this non-zero if the referred-to block is likely to be 135 * followed (on disk) by an indirect block. 136 * To store the locations of file's data ext2 uses a data structure common 137 * for UNIX filesystems - tree of pointers anchored in the inode, with 138 * data blocks at leaves and indirect blocks in intermediate nodes. 139 * This function translates the block number into path in that tree - 140 * return value is the path length and @offsets[n] is the offset of 141 * pointer to (n+1)th node in the nth one. If @block is out of range 142 * (negative or too large) warning is printed and zero returned. 143 * 144 * Note: function doesn't find node addresses, so no IO is needed. All 145 * we need to know is the capacity of indirect blocks (taken from the 146 * inode->i_sb). 147 */ 148 149/* 150 * Portability note: the last comparison (check that we fit into triple 151 * indirect block) is spelled differently, because otherwise on an 152 * architecture with 32-bit longs and 8Kb pages we might get into trouble 153 * if our filesystem had 8Kb blocks. We might use long long, but that would 154 * kill us on x86. Oh, well, at least the sign propagation does not matter - 155 * i_block would have to be negative in the very beginning, so we would not 156 * get there at all. 157 */ 158 159static int ext2_block_to_path(struct inode *inode, 160 long i_block, int offsets[4], int *boundary) 161{ 162 int ptrs = EXT2_ADDR_PER_BLOCK(inode->i_sb); 163 int ptrs_bits = EXT2_ADDR_PER_BLOCK_BITS(inode->i_sb); 164 const long direct_blocks = EXT2_NDIR_BLOCKS, 165 indirect_blocks = ptrs, 166 double_blocks = (1 << (ptrs_bits * 2)); 167 int n = 0; 168 int final = 0; 169 170 if (i_block < 0) { 171 ext2_msg(inode->i_sb, KERN_WARNING, 172 "warning: %s: block < 0", __func__); 173 } else if (i_block < direct_blocks) { 174 offsets[n++] = i_block; 175 final = direct_blocks; 176 } else if ( (i_block -= direct_blocks) < indirect_blocks) { 177 offsets[n++] = EXT2_IND_BLOCK; 178 offsets[n++] = i_block; 179 final = ptrs; 180 } else if ((i_block -= indirect_blocks) < double_blocks) { 181 offsets[n++] = EXT2_DIND_BLOCK; 182 offsets[n++] = i_block >> ptrs_bits; 183 offsets[n++] = i_block & (ptrs - 1); 184 final = ptrs; 185 } else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) { 186 offsets[n++] = EXT2_TIND_BLOCK; 187 offsets[n++] = i_block >> (ptrs_bits * 2); 188 offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1); 189 offsets[n++] = i_block & (ptrs - 1); 190 final = ptrs; 191 } else { 192 ext2_msg(inode->i_sb, KERN_WARNING, 193 "warning: %s: block is too big", __func__); 194 } 195 if (boundary) 196 *boundary = final - 1 - (i_block & (ptrs - 1)); 197 198 return n; 199} 200 201/** 202 * ext2_get_branch - read the chain of indirect blocks leading to data 203 * @inode: inode in question 204 * @depth: depth of the chain (1 - direct pointer, etc.) 205 * @offsets: offsets of pointers in inode/indirect blocks 206 * @chain: place to store the result 207 * @err: here we store the error value 208 * 209 * Function fills the array of triples <key, p, bh> and returns %NULL 210 * if everything went OK or the pointer to the last filled triple 211 * (incomplete one) otherwise. Upon the return chain[i].key contains 212 * the number of (i+1)-th block in the chain (as it is stored in memory, 213 * i.e. little-endian 32-bit), chain[i].p contains the address of that 214 * number (it points into struct inode for i==0 and into the bh->b_data 215 * for i>0) and chain[i].bh points to the buffer_head of i-th indirect 216 * block for i>0 and NULL for i==0. In other words, it holds the block 217 * numbers of the chain, addresses they were taken from (and where we can 218 * verify that chain did not change) and buffer_heads hosting these 219 * numbers. 220 * 221 * Function stops when it stumbles upon zero pointer (absent block) 222 * (pointer to last triple returned, *@err == 0) 223 * or when it gets an IO error reading an indirect block 224 * (ditto, *@err == -EIO) 225 * or when it notices that chain had been changed while it was reading 226 * (ditto, *@err == -EAGAIN) 227 * or when it reads all @depth-1 indirect blocks successfully and finds 228 * the whole chain, all way to the data (returns %NULL, *err == 0). 229 */ 230static Indirect *ext2_get_branch(struct inode *inode, 231 int depth, 232 int *offsets, 233 Indirect chain[4], 234 int *err) 235{ 236 struct super_block *sb = inode->i_sb; 237 Indirect *p = chain; 238 struct buffer_head *bh; 239 240 *err = 0; 241 /* i_data is not going away, no lock needed */ 242 add_chain (chain, NULL, EXT2_I(inode)->i_data + *offsets); 243 if (!p->key) 244 goto no_block; 245 while (--depth) { 246 bh = sb_bread(sb, le32_to_cpu(p->key)); 247 if (!bh) 248 goto failure; 249 read_lock(&EXT2_I(inode)->i_meta_lock); 250 if (!verify_chain(chain, p)) 251 goto changed; 252 add_chain(++p, bh, (__le32*)bh->b_data + *++offsets); 253 read_unlock(&EXT2_I(inode)->i_meta_lock); 254 if (!p->key) 255 goto no_block; 256 } 257 return NULL; 258 259changed: 260 read_unlock(&EXT2_I(inode)->i_meta_lock); 261 brelse(bh); 262 *err = -EAGAIN; 263 goto no_block; 264failure: 265 *err = -EIO; 266no_block: 267 return p; 268} 269 270/** 271 * ext2_find_near - find a place for allocation with sufficient locality 272 * @inode: owner 273 * @ind: descriptor of indirect block. 274 * 275 * This function returns the preferred place for block allocation. 276 * It is used when heuristic for sequential allocation fails. 277 * Rules are: 278 * + if there is a block to the left of our position - allocate near it. 279 * + if pointer will live in indirect block - allocate near that block. 280 * + if pointer will live in inode - allocate in the same cylinder group. 281 * 282 * In the latter case we colour the starting block by the callers PID to 283 * prevent it from clashing with concurrent allocations for a different inode 284 * in the same block group. The PID is used here so that functionally related 285 * files will be close-by on-disk. 286 * 287 * Caller must make sure that @ind is valid and will stay that way. 288 */ 289 290static ext2_fsblk_t ext2_find_near(struct inode *inode, Indirect *ind) 291{ 292 struct ext2_inode_info *ei = EXT2_I(inode); 293 __le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data; 294 __le32 *p; 295 ext2_fsblk_t bg_start; 296 ext2_fsblk_t colour; 297 298 /* Try to find previous block */ 299 for (p = ind->p - 1; p >= start; p--) 300 if (*p) 301 return le32_to_cpu(*p); 302 303 /* No such thing, so let's try location of indirect block */ 304 if (ind->bh) 305 return ind->bh->b_blocknr; 306 307 /* 308 * It is going to be refered from inode itself? OK, just put it into 309 * the same cylinder group then. 310 */ 311 bg_start = ext2_group_first_block_no(inode->i_sb, ei->i_block_group); 312 colour = (current->pid % 16) * 313 (EXT2_BLOCKS_PER_GROUP(inode->i_sb) / 16); 314 return bg_start + colour; 315} 316 317/** 318 * ext2_find_goal - find a preferred place for allocation. 319 * @inode: owner 320 * @block: block we want 321 * @partial: pointer to the last triple within a chain 322 * 323 * Returns preferred place for a block (the goal). 324 */ 325 326static inline ext2_fsblk_t ext2_find_goal(struct inode *inode, long block, 327 Indirect *partial) 328{ 329 struct ext2_block_alloc_info *block_i; 330 331 block_i = EXT2_I(inode)->i_block_alloc_info; 332 333 /* 334 * try the heuristic for sequential allocation, 335 * failing that at least try to get decent locality. 336 */ 337 if (block_i && (block == block_i->last_alloc_logical_block + 1) 338 && (block_i->last_alloc_physical_block != 0)) { 339 return block_i->last_alloc_physical_block + 1; 340 } 341 342 return ext2_find_near(inode, partial); 343} 344 345/** 346 * ext2_blks_to_allocate: Look up the block map and count the number 347 * of direct blocks need to be allocated for the given branch. 348 * 349 * @branch: chain of indirect blocks 350 * @k: number of blocks need for indirect blocks 351 * @blks: number of data blocks to be mapped. 352 * @blocks_to_boundary: the offset in the indirect block 353 * 354 * return the total number of blocks to be allocate, including the 355 * direct and indirect blocks. 356 */ 357static int 358ext2_blks_to_allocate(Indirect * branch, int k, unsigned long blks, 359 int blocks_to_boundary) 360{ 361 unsigned long count = 0; 362 363 /* 364 * Simple case, [t,d]Indirect block(s) has not allocated yet 365 * then it's clear blocks on that path have not allocated 366 */ 367 if (k > 0) { 368 /* right now don't hanel cross boundary allocation */ 369 if (blks < blocks_to_boundary + 1) 370 count += blks; 371 else 372 count += blocks_to_boundary + 1; 373 return count; 374 } 375 376 count++; 377 while (count < blks && count <= blocks_to_boundary 378 && le32_to_cpu(*(branch[0].p + count)) == 0) { 379 count++; 380 } 381 return count; 382} 383 384/** 385 * ext2_alloc_blocks: multiple allocate blocks needed for a branch 386 * @indirect_blks: the number of blocks need to allocate for indirect 387 * blocks 388 * 389 * @new_blocks: on return it will store the new block numbers for 390 * the indirect blocks(if needed) and the first direct block, 391 * @blks: on return it will store the total number of allocated 392 * direct blocks 393 */ 394static int ext2_alloc_blocks(struct inode *inode, 395 ext2_fsblk_t goal, int indirect_blks, int blks, 396 ext2_fsblk_t new_blocks[4], int *err) 397{ 398 int target, i; 399 unsigned long count = 0; 400 int index = 0; 401 ext2_fsblk_t current_block = 0; 402 int ret = 0; 403 404 /* 405 * Here we try to allocate the requested multiple blocks at once, 406 * on a best-effort basis. 407 * To build a branch, we should allocate blocks for 408 * the indirect blocks(if not allocated yet), and at least 409 * the first direct block of this branch. That's the 410 * minimum number of blocks need to allocate(required) 411 */ 412 target = blks + indirect_blks; 413 414 while (1) { 415 count = target; 416 /* allocating blocks for indirect blocks and direct blocks */ 417 current_block = ext2_new_blocks(inode,goal,&count,err); 418 if (*err) 419 goto failed_out; 420 421 target -= count; 422 /* allocate blocks for indirect blocks */ 423 while (index < indirect_blks && count) { 424 new_blocks[index++] = current_block++; 425 count--; 426 } 427 428 if (count > 0) 429 break; 430 } 431 432 /* save the new block number for the first direct block */ 433 new_blocks[index] = current_block; 434 435 /* total number of blocks allocated for direct blocks */ 436 ret = count; 437 *err = 0; 438 return ret; 439failed_out: 440 for (i = 0; i <index; i++) 441 ext2_free_blocks(inode, new_blocks[i], 1); 442 if (index) 443 mark_inode_dirty(inode); 444 return ret; 445} 446 447/** 448 * ext2_alloc_branch - allocate and set up a chain of blocks. 449 * @inode: owner 450 * @num: depth of the chain (number of blocks to allocate) 451 * @offsets: offsets (in the blocks) to store the pointers to next. 452 * @branch: place to store the chain in. 453 * 454 * This function allocates @num blocks, zeroes out all but the last one, 455 * links them into chain and (if we are synchronous) writes them to disk. 456 * In other words, it prepares a branch that can be spliced onto the 457 * inode. It stores the information about that chain in the branch[], in 458 * the same format as ext2_get_branch() would do. We are calling it after 459 * we had read the existing part of chain and partial points to the last 460 * triple of that (one with zero ->key). Upon the exit we have the same 461 * picture as after the successful ext2_get_block(), excpet that in one 462 * place chain is disconnected - *branch->p is still zero (we did not 463 * set the last link), but branch->key contains the number that should 464 * be placed into *branch->p to fill that gap. 465 * 466 * If allocation fails we free all blocks we've allocated (and forget 467 * their buffer_heads) and return the error value the from failed 468 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain 469 * as described above and return 0. 470 */ 471 472static int ext2_alloc_branch(struct inode *inode, 473 int indirect_blks, int *blks, ext2_fsblk_t goal, 474 int *offsets, Indirect *branch) 475{ 476 int blocksize = inode->i_sb->s_blocksize; 477 int i, n = 0; 478 int err = 0; 479 struct buffer_head *bh; 480 int num; 481 ext2_fsblk_t new_blocks[4]; 482 ext2_fsblk_t current_block; 483 484 num = ext2_alloc_blocks(inode, goal, indirect_blks, 485 *blks, new_blocks, &err); 486 if (err) 487 return err; 488 489 branch[0].key = cpu_to_le32(new_blocks[0]); 490 /* 491 * metadata blocks and data blocks are allocated. 492 */ 493 for (n = 1; n <= indirect_blks; n++) { 494 /* 495 * Get buffer_head for parent block, zero it out 496 * and set the pointer to new one, then send 497 * parent to disk. 498 */ 499 bh = sb_getblk(inode->i_sb, new_blocks[n-1]); 500 branch[n].bh = bh; 501 lock_buffer(bh); 502 memset(bh->b_data, 0, blocksize); 503 branch[n].p = (__le32 *) bh->b_data + offsets[n]; 504 branch[n].key = cpu_to_le32(new_blocks[n]); 505 *branch[n].p = branch[n].key; 506 if ( n == indirect_blks) { 507 current_block = new_blocks[n]; 508 /* 509 * End of chain, update the last new metablock of 510 * the chain to point to the new allocated 511 * data blocks numbers 512 */ 513 for (i=1; i < num; i++) 514 *(branch[n].p + i) = cpu_to_le32(++current_block); 515 } 516 set_buffer_uptodate(bh); 517 unlock_buffer(bh); 518 mark_buffer_dirty_inode(bh, inode); 519 /* We used to sync bh here if IS_SYNC(inode). 520 * But we now rely upon generic_write_sync() 521 * and b_inode_buffers. But not for directories. 522 */ 523 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) 524 sync_dirty_buffer(bh); 525 } 526 *blks = num; 527 return err; 528} 529 530/** 531 * ext2_splice_branch - splice the allocated branch onto inode. 532 * @inode: owner 533 * @block: (logical) number of block we are adding 534 * @where: location of missing link 535 * @num: number of indirect blocks we are adding 536 * @blks: number of direct blocks we are adding 537 * 538 * This function fills the missing link and does all housekeeping needed in 539 * inode (->i_blocks, etc.). In case of success we end up with the full 540 * chain to new block and return 0. 541 */ 542static void ext2_splice_branch(struct inode *inode, 543 long block, Indirect *where, int num, int blks) 544{ 545 int i; 546 struct ext2_block_alloc_info *block_i; 547 ext2_fsblk_t current_block; 548 549 block_i = EXT2_I(inode)->i_block_alloc_info; 550 551 /* That's it */ 552 553 *where->p = where->key; 554 555 /* 556 * Update the host buffer_head or inode to point to more just allocated 557 * direct blocks blocks 558 */ 559 if (num == 0 && blks > 1) { 560 current_block = le32_to_cpu(where->key) + 1; 561 for (i = 1; i < blks; i++) 562 *(where->p + i ) = cpu_to_le32(current_block++); 563 } 564 565 /* 566 * update the most recently allocated logical & physical block 567 * in i_block_alloc_info, to assist find the proper goal block for next 568 * allocation 569 */ 570 if (block_i) { 571 block_i->last_alloc_logical_block = block + blks - 1; 572 block_i->last_alloc_physical_block = 573 le32_to_cpu(where[num].key) + blks - 1; 574 } 575 576 /* We are done with atomic stuff, now do the rest of housekeeping */ 577 578 /* had we spliced it onto indirect block? */ 579 if (where->bh) 580 mark_buffer_dirty_inode(where->bh, inode); 581 582 inode->i_ctime = CURRENT_TIME_SEC; 583 mark_inode_dirty(inode); 584} 585 586/* 587 * Allocation strategy is simple: if we have to allocate something, we will 588 * have to go the whole way to leaf. So let's do it before attaching anything 589 * to tree, set linkage between the newborn blocks, write them if sync is 590 * required, recheck the path, free and repeat if check fails, otherwise 591 * set the last missing link (that will protect us from any truncate-generated 592 * removals - all blocks on the path are immune now) and possibly force the 593 * write on the parent block. 594 * That has a nice additional property: no special recovery from the failed 595 * allocations is needed - we simply release blocks and do not touch anything 596 * reachable from inode. 597 * 598 * `handle' can be NULL if create == 0. 599 * 600 * return > 0, # of blocks mapped or allocated. 601 * return = 0, if plain lookup failed. 602 * return < 0, error case. 603 */ 604static int ext2_get_blocks(struct inode *inode, 605 sector_t iblock, unsigned long maxblocks, 606 struct buffer_head *bh_result, 607 int create) 608{ 609 int err = -EIO; 610 int offsets[4]; 611 Indirect chain[4]; 612 Indirect *partial; 613 ext2_fsblk_t goal; 614 int indirect_blks; 615 int blocks_to_boundary = 0; 616 int depth; 617 struct ext2_inode_info *ei = EXT2_I(inode); 618 int count = 0; 619 ext2_fsblk_t first_block = 0; 620 621 depth = ext2_block_to_path(inode,iblock,offsets,&blocks_to_boundary); 622 623 if (depth == 0) 624 return (err); 625 626 partial = ext2_get_branch(inode, depth, offsets, chain, &err); 627 /* Simplest case - block found, no allocation needed */ 628 if (!partial) { 629 first_block = le32_to_cpu(chain[depth - 1].key); 630 clear_buffer_new(bh_result); /* What's this do? */ 631 count++; 632 /*map more blocks*/ 633 while (count < maxblocks && count <= blocks_to_boundary) { 634 ext2_fsblk_t blk; 635 636 if (!verify_chain(chain, chain + depth - 1)) { 637 /* 638 * Indirect block might be removed by 639 * truncate while we were reading it. 640 * Handling of that case: forget what we've 641 * got now, go to reread. 642 */ 643 err = -EAGAIN; 644 count = 0; 645 break; 646 } 647 blk = le32_to_cpu(*(chain[depth-1].p + count)); 648 if (blk == first_block + count) 649 count++; 650 else 651 break; 652 } 653 if (err != -EAGAIN) 654 goto got_it; 655 } 656 657 /* Next simple case - plain lookup or failed read of indirect block */ 658 if (!create || err == -EIO) 659 goto cleanup; 660 661 mutex_lock(&ei->truncate_mutex); 662 /* 663 * If the indirect block is missing while we are reading 664 * the chain(ext3_get_branch() returns -EAGAIN err), or 665 * if the chain has been changed after we grab the semaphore, 666 * (either because another process truncated this branch, or 667 * another get_block allocated this branch) re-grab the chain to see if 668 * the request block has been allocated or not. 669 * 670 * Since we already block the truncate/other get_block 671 * at this point, we will have the current copy of the chain when we 672 * splice the branch into the tree. 673 */ 674 if (err == -EAGAIN || !verify_chain(chain, partial)) { 675 while (partial > chain) { 676 brelse(partial->bh); 677 partial--; 678 } 679 partial = ext2_get_branch(inode, depth, offsets, chain, &err); 680 if (!partial) { 681 count++; 682 mutex_unlock(&ei->truncate_mutex); 683 if (err) 684 goto cleanup; 685 clear_buffer_new(bh_result); 686 goto got_it; 687 } 688 } 689 690 /* 691 * Okay, we need to do block allocation. Lazily initialize the block 692 * allocation info here if necessary 693 */ 694 if (S_ISREG(inode->i_mode) && (!ei->i_block_alloc_info)) 695 ext2_init_block_alloc_info(inode); 696 697 goal = ext2_find_goal(inode, iblock, partial); 698 699 /* the number of blocks need to allocate for [d,t]indirect blocks */ 700 indirect_blks = (chain + depth) - partial - 1; 701 /* 702 * Next look up the indirect map to count the totoal number of 703 * direct blocks to allocate for this branch. 704 */ 705 count = ext2_blks_to_allocate(partial, indirect_blks, 706 maxblocks, blocks_to_boundary); 707 err = ext2_alloc_branch(inode, indirect_blks, &count, goal, 708 offsets + (partial - chain), partial); 709 710 if (err) { 711 mutex_unlock(&ei->truncate_mutex); 712 goto cleanup; 713 } 714 715 if (ext2_use_xip(inode->i_sb)) { 716 /* 717 * we need to clear the block 718 */ 719 err = ext2_clear_xip_target (inode, 720 le32_to_cpu(chain[depth-1].key)); 721 if (err) { 722 mutex_unlock(&ei->truncate_mutex); 723 goto cleanup; 724 } 725 } 726 727 ext2_splice_branch(inode, iblock, partial, indirect_blks, count); 728 mutex_unlock(&ei->truncate_mutex); 729 set_buffer_new(bh_result); 730got_it: 731 map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key)); 732 if (count > blocks_to_boundary) 733 set_buffer_boundary(bh_result); 734 err = count; 735 /* Clean up and exit */ 736 partial = chain + depth - 1; /* the whole chain */ 737cleanup: 738 while (partial > chain) { 739 brelse(partial->bh); 740 partial--; 741 } 742 return err; 743} 744 745int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create) 746{ 747 unsigned max_blocks = bh_result->b_size >> inode->i_blkbits; 748 int ret = ext2_get_blocks(inode, iblock, max_blocks, 749 bh_result, create); 750 if (ret > 0) { 751 bh_result->b_size = (ret << inode->i_blkbits); 752 ret = 0; 753 } 754 return ret; 755 756} 757 758int ext2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 759 u64 start, u64 len) 760{ 761 return generic_block_fiemap(inode, fieinfo, start, len, 762 ext2_get_block); 763} 764 765static int ext2_writepage(struct page *page, struct writeback_control *wbc) 766{ 767 return block_write_full_page(page, ext2_get_block, wbc); 768} 769 770static int ext2_readpage(struct file *file, struct page *page) 771{ 772 return mpage_readpage(page, ext2_get_block); 773} 774 775static int 776ext2_readpages(struct file *file, struct address_space *mapping, 777 struct list_head *pages, unsigned nr_pages) 778{ 779 return mpage_readpages(mapping, pages, nr_pages, ext2_get_block); 780} 781 782static int 783ext2_write_begin(struct file *file, struct address_space *mapping, 784 loff_t pos, unsigned len, unsigned flags, 785 struct page **pagep, void **fsdata) 786{ 787 int ret; 788 789 ret = block_write_begin(mapping, pos, len, flags, pagep, 790 ext2_get_block); 791 if (ret < 0) 792 ext2_write_failed(mapping, pos + len); 793 return ret; 794} 795 796static int ext2_write_end(struct file *file, struct address_space *mapping, 797 loff_t pos, unsigned len, unsigned copied, 798 struct page *page, void *fsdata) 799{ 800 int ret; 801 802 ret = generic_write_end(file, mapping, pos, len, copied, page, fsdata); 803 if (ret < len) 804 ext2_write_failed(mapping, pos + len); 805 return ret; 806} 807 808static int 809ext2_nobh_write_begin(struct file *file, struct address_space *mapping, 810 loff_t pos, unsigned len, unsigned flags, 811 struct page **pagep, void **fsdata) 812{ 813 int ret; 814 815 ret = nobh_write_begin(mapping, pos, len, flags, pagep, fsdata, 816 ext2_get_block); 817 if (ret < 0) 818 ext2_write_failed(mapping, pos + len); 819 return ret; 820} 821 822static int ext2_nobh_writepage(struct page *page, 823 struct writeback_control *wbc) 824{ 825 return nobh_writepage(page, ext2_get_block, wbc); 826} 827 828static sector_t ext2_bmap(struct address_space *mapping, sector_t block) 829{ 830 return generic_block_bmap(mapping,block,ext2_get_block); 831} 832 833static ssize_t 834ext2_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, 835 loff_t offset, unsigned long nr_segs) 836{ 837 struct file *file = iocb->ki_filp; 838 struct address_space *mapping = file->f_mapping; 839 struct inode *inode = mapping->host; 840 ssize_t ret; 841 842 ret = blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, 843 iov, offset, nr_segs, ext2_get_block, NULL); 844 if (ret < 0 && (rw & WRITE)) 845 ext2_write_failed(mapping, offset + iov_length(iov, nr_segs)); 846 return ret; 847} 848 849static int 850ext2_writepages(struct address_space *mapping, struct writeback_control *wbc) 851{ 852 return mpage_writepages(mapping, wbc, ext2_get_block); 853} 854 855const struct address_space_operations ext2_aops = { 856 .readpage = ext2_readpage, 857 .readpages = ext2_readpages, 858 .writepage = ext2_writepage, 859 .sync_page = block_sync_page, 860 .write_begin = ext2_write_begin, 861 .write_end = ext2_write_end, 862 .bmap = ext2_bmap, 863 .direct_IO = ext2_direct_IO, 864 .writepages = ext2_writepages, 865 .migratepage = buffer_migrate_page, 866 .is_partially_uptodate = block_is_partially_uptodate, 867 .error_remove_page = generic_error_remove_page, 868}; 869 870const struct address_space_operations ext2_aops_xip = { 871 .bmap = ext2_bmap, 872 .get_xip_mem = ext2_get_xip_mem, 873}; 874 875const struct address_space_operations ext2_nobh_aops = { 876 .readpage = ext2_readpage, 877 .readpages = ext2_readpages, 878 .writepage = ext2_nobh_writepage, 879 .sync_page = block_sync_page, 880 .write_begin = ext2_nobh_write_begin, 881 .write_end = nobh_write_end, 882 .bmap = ext2_bmap, 883 .direct_IO = ext2_direct_IO, 884 .writepages = ext2_writepages, 885 .migratepage = buffer_migrate_page, 886 .error_remove_page = generic_error_remove_page, 887}; 888 889/* 890 * Probably it should be a library function... search for first non-zero word 891 * or memcmp with zero_page, whatever is better for particular architecture. 892 * Linus? 893 */ 894static inline int all_zeroes(__le32 *p, __le32 *q) 895{ 896 while (p < q) 897 if (*p++) 898 return 0; 899 return 1; 900} 901 902/** 903 * ext2_find_shared - find the indirect blocks for partial truncation. 904 * @inode: inode in question 905 * @depth: depth of the affected branch 906 * @offsets: offsets of pointers in that branch (see ext2_block_to_path) 907 * @chain: place to store the pointers to partial indirect blocks 908 * @top: place to the (detached) top of branch 909 * 910 * This is a helper function used by ext2_truncate(). 911 * 912 * When we do truncate() we may have to clean the ends of several indirect 913 * blocks but leave the blocks themselves alive. Block is partially 914 * truncated if some data below the new i_size is refered from it (and 915 * it is on the path to the first completely truncated data block, indeed). 916 * We have to free the top of that path along with everything to the right 917 * of the path. Since no allocation past the truncation point is possible 918 * until ext2_truncate() finishes, we may safely do the latter, but top 919 * of branch may require special attention - pageout below the truncation 920 * point might try to populate it. 921 * 922 * We atomically detach the top of branch from the tree, store the block 923 * number of its root in *@top, pointers to buffer_heads of partially 924 * truncated blocks - in @chain[].bh and pointers to their last elements 925 * that should not be removed - in @chain[].p. Return value is the pointer 926 * to last filled element of @chain. 927 * 928 * The work left to caller to do the actual freeing of subtrees: 929 * a) free the subtree starting from *@top 930 * b) free the subtrees whose roots are stored in 931 * (@chain[i].p+1 .. end of @chain[i].bh->b_data) 932 * c) free the subtrees growing from the inode past the @chain[0].p 933 * (no partially truncated stuff there). 934 */ 935 936static Indirect *ext2_find_shared(struct inode *inode, 937 int depth, 938 int offsets[4], 939 Indirect chain[4], 940 __le32 *top) 941{ 942 Indirect *partial, *p; 943 int k, err; 944 945 *top = 0; 946 for (k = depth; k > 1 && !offsets[k-1]; k--) 947 ; 948 partial = ext2_get_branch(inode, k, offsets, chain, &err); 949 if (!partial) 950 partial = chain + k-1; 951 /* 952 * If the branch acquired continuation since we've looked at it - 953 * fine, it should all survive and (new) top doesn't belong to us. 954 */ 955 write_lock(&EXT2_I(inode)->i_meta_lock); 956 if (!partial->key && *partial->p) { 957 write_unlock(&EXT2_I(inode)->i_meta_lock); 958 goto no_top; 959 } 960 for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--) 961 ; 962 /* 963 * OK, we've found the last block that must survive. The rest of our 964 * branch should be detached before unlocking. However, if that rest 965 * of branch is all ours and does not grow immediately from the inode 966 * it's easier to cheat and just decrement partial->p. 967 */ 968 if (p == chain + k - 1 && p > chain) { 969 p->p--; 970 } else { 971 *top = *p->p; 972 *p->p = 0; 973 } 974 write_unlock(&EXT2_I(inode)->i_meta_lock); 975 976 while(partial > p) 977 { 978 brelse(partial->bh); 979 partial--; 980 } 981no_top: 982 return partial; 983} 984 985/** 986 * ext2_free_data - free a list of data blocks 987 * @inode: inode we are dealing with 988 * @p: array of block numbers 989 * @q: points immediately past the end of array 990 * 991 * We are freeing all blocks refered from that array (numbers are 992 * stored as little-endian 32-bit) and updating @inode->i_blocks 993 * appropriately. 994 */ 995static inline void ext2_free_data(struct inode *inode, __le32 *p, __le32 *q) 996{ 997 unsigned long block_to_free = 0, count = 0; 998 unsigned long nr; 999 1000 for ( ; p < q ; p++) { 1001 nr = le32_to_cpu(*p); 1002 if (nr) { 1003 *p = 0; 1004 /* accumulate blocks to free if they're contiguous */ 1005 if (count == 0) 1006 goto free_this; 1007 else if (block_to_free == nr - count) 1008 count++; 1009 else { 1010 ext2_free_blocks (inode, block_to_free, count); 1011 mark_inode_dirty(inode); 1012 free_this: 1013 block_to_free = nr; 1014 count = 1; 1015 } 1016 } 1017 } 1018 if (count > 0) { 1019 ext2_free_blocks (inode, block_to_free, count); 1020 mark_inode_dirty(inode); 1021 } 1022} 1023 1024/** 1025 * ext2_free_branches - free an array of branches 1026 * @inode: inode we are dealing with 1027 * @p: array of block numbers 1028 * @q: pointer immediately past the end of array 1029 * @depth: depth of the branches to free 1030 * 1031 * We are freeing all blocks refered from these branches (numbers are 1032 * stored as little-endian 32-bit) and updating @inode->i_blocks 1033 * appropriately. 1034 */ 1035static void ext2_free_branches(struct inode *inode, __le32 *p, __le32 *q, int depth) 1036{ 1037 struct buffer_head * bh; 1038 unsigned long nr; 1039 1040 if (depth--) { 1041 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb); 1042 for ( ; p < q ; p++) { 1043 nr = le32_to_cpu(*p); 1044 if (!nr) 1045 continue; 1046 *p = 0; 1047 bh = sb_bread(inode->i_sb, nr); 1048 /* 1049 * A read failure? Report error and clear slot 1050 * (should be rare). 1051 */ 1052 if (!bh) { 1053 ext2_error(inode->i_sb, "ext2_free_branches", 1054 "Read failure, inode=%ld, block=%ld", 1055 inode->i_ino, nr); 1056 continue; 1057 } 1058 ext2_free_branches(inode, 1059 (__le32*)bh->b_data, 1060 (__le32*)bh->b_data + addr_per_block, 1061 depth); 1062 bforget(bh); 1063 ext2_free_blocks(inode, nr, 1); 1064 mark_inode_dirty(inode); 1065 } 1066 } else 1067 ext2_free_data(inode, p, q); 1068} 1069 1070static void __ext2_truncate_blocks(struct inode *inode, loff_t offset) 1071{ 1072 __le32 *i_data = EXT2_I(inode)->i_data; 1073 struct ext2_inode_info *ei = EXT2_I(inode); 1074 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb); 1075 int offsets[4]; 1076 Indirect chain[4]; 1077 Indirect *partial; 1078 __le32 nr = 0; 1079 int n; 1080 long iblock; 1081 unsigned blocksize; 1082 blocksize = inode->i_sb->s_blocksize; 1083 iblock = (offset + blocksize-1) >> EXT2_BLOCK_SIZE_BITS(inode->i_sb); 1084 1085 n = ext2_block_to_path(inode, iblock, offsets, NULL); 1086 if (n == 0) 1087 return; 1088 1089 /* 1090 * From here we block out all ext2_get_block() callers who want to 1091 * modify the block allocation tree. 1092 */ 1093 mutex_lock(&ei->truncate_mutex); 1094 1095 if (n == 1) { 1096 ext2_free_data(inode, i_data+offsets[0], 1097 i_data + EXT2_NDIR_BLOCKS); 1098 goto do_indirects; 1099 } 1100 1101 partial = ext2_find_shared(inode, n, offsets, chain, &nr); 1102 /* Kill the top of shared branch (already detached) */ 1103 if (nr) { 1104 if (partial == chain) 1105 mark_inode_dirty(inode); 1106 else 1107 mark_buffer_dirty_inode(partial->bh, inode); 1108 ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial); 1109 } 1110 /* Clear the ends of indirect blocks on the shared branch */ 1111 while (partial > chain) { 1112 ext2_free_branches(inode, 1113 partial->p + 1, 1114 (__le32*)partial->bh->b_data+addr_per_block, 1115 (chain+n-1) - partial); 1116 mark_buffer_dirty_inode(partial->bh, inode); 1117 brelse (partial->bh); 1118 partial--; 1119 } 1120do_indirects: 1121 /* Kill the remaining (whole) subtrees */ 1122 switch (offsets[0]) { 1123 default: 1124 nr = i_data[EXT2_IND_BLOCK]; 1125 if (nr) { 1126 i_data[EXT2_IND_BLOCK] = 0; 1127 mark_inode_dirty(inode); 1128 ext2_free_branches(inode, &nr, &nr+1, 1); 1129 } 1130 case EXT2_IND_BLOCK: 1131 nr = i_data[EXT2_DIND_BLOCK]; 1132 if (nr) { 1133 i_data[EXT2_DIND_BLOCK] = 0; 1134 mark_inode_dirty(inode); 1135 ext2_free_branches(inode, &nr, &nr+1, 2); 1136 } 1137 case EXT2_DIND_BLOCK: 1138 nr = i_data[EXT2_TIND_BLOCK]; 1139 if (nr) { 1140 i_data[EXT2_TIND_BLOCK] = 0; 1141 mark_inode_dirty(inode); 1142 ext2_free_branches(inode, &nr, &nr+1, 3); 1143 } 1144 case EXT2_TIND_BLOCK: 1145 ; 1146 } 1147 1148 ext2_discard_reservation(inode); 1149 1150 mutex_unlock(&ei->truncate_mutex); 1151} 1152 1153static void ext2_truncate_blocks(struct inode *inode, loff_t offset) 1154{ 1155 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 1156 S_ISLNK(inode->i_mode))) 1157 return; 1158 if (ext2_inode_is_fast_symlink(inode)) 1159 return; 1160 if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) 1161 return; 1162 __ext2_truncate_blocks(inode, offset); 1163} 1164 1165static int ext2_setsize(struct inode *inode, loff_t newsize) 1166{ 1167 int error; 1168 1169 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 1170 S_ISLNK(inode->i_mode))) 1171 return -EINVAL; 1172 if (ext2_inode_is_fast_symlink(inode)) 1173 return -EINVAL; 1174 if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) 1175 return -EPERM; 1176 1177 if (mapping_is_xip(inode->i_mapping)) 1178 error = xip_truncate_page(inode->i_mapping, newsize); 1179 else if (test_opt(inode->i_sb, NOBH)) 1180 error = nobh_truncate_page(inode->i_mapping, 1181 newsize, ext2_get_block); 1182 else 1183 error = block_truncate_page(inode->i_mapping, 1184 newsize, ext2_get_block); 1185 if (error) 1186 return error; 1187 1188 truncate_setsize(inode, newsize); 1189 __ext2_truncate_blocks(inode, newsize); 1190 1191 inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC; 1192 if (inode_needs_sync(inode)) { 1193 sync_mapping_buffers(inode->i_mapping); 1194 ext2_sync_inode (inode); 1195 } else { 1196 mark_inode_dirty(inode); 1197 } 1198 1199 return 0; 1200} 1201 1202static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino, 1203 struct buffer_head **p) 1204{ 1205 struct buffer_head * bh; 1206 unsigned long block_group; 1207 unsigned long block; 1208 unsigned long offset; 1209 struct ext2_group_desc * gdp; 1210 1211 *p = NULL; 1212 if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) || 1213 ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count)) 1214 goto Einval; 1215 1216 block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb); 1217 gdp = ext2_get_group_desc(sb, block_group, NULL); 1218 if (!gdp) 1219 goto Egdp; 1220 /* 1221 * Figure out the offset within the block group inode table 1222 */ 1223 offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb); 1224 block = le32_to_cpu(gdp->bg_inode_table) + 1225 (offset >> EXT2_BLOCK_SIZE_BITS(sb)); 1226 if (!(bh = sb_bread(sb, block))) 1227 goto Eio; 1228 1229 *p = bh; 1230 offset &= (EXT2_BLOCK_SIZE(sb) - 1); 1231 return (struct ext2_inode *) (bh->b_data + offset); 1232 1233Einval: 1234 ext2_error(sb, "ext2_get_inode", "bad inode number: %lu", 1235 (unsigned long) ino); 1236 return ERR_PTR(-EINVAL); 1237Eio: 1238 ext2_error(sb, "ext2_get_inode", 1239 "unable to read inode block - inode=%lu, block=%lu", 1240 (unsigned long) ino, block); 1241Egdp: 1242 return ERR_PTR(-EIO); 1243} 1244 1245void ext2_set_inode_flags(struct inode *inode) 1246{ 1247 unsigned int flags = EXT2_I(inode)->i_flags; 1248 1249 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); 1250 if (flags & EXT2_SYNC_FL) 1251 inode->i_flags |= S_SYNC; 1252 if (flags & EXT2_APPEND_FL) 1253 inode->i_flags |= S_APPEND; 1254 if (flags & EXT2_IMMUTABLE_FL) 1255 inode->i_flags |= S_IMMUTABLE; 1256 if (flags & EXT2_NOATIME_FL) 1257 inode->i_flags |= S_NOATIME; 1258 if (flags & EXT2_DIRSYNC_FL) 1259 inode->i_flags |= S_DIRSYNC; 1260} 1261 1262/* Propagate flags from i_flags to EXT2_I(inode)->i_flags */ 1263void ext2_get_inode_flags(struct ext2_inode_info *ei) 1264{ 1265 unsigned int flags = ei->vfs_inode.i_flags; 1266 1267 ei->i_flags &= ~(EXT2_SYNC_FL|EXT2_APPEND_FL| 1268 EXT2_IMMUTABLE_FL|EXT2_NOATIME_FL|EXT2_DIRSYNC_FL); 1269 if (flags & S_SYNC) 1270 ei->i_flags |= EXT2_SYNC_FL; 1271 if (flags & S_APPEND) 1272 ei->i_flags |= EXT2_APPEND_FL; 1273 if (flags & S_IMMUTABLE) 1274 ei->i_flags |= EXT2_IMMUTABLE_FL; 1275 if (flags & S_NOATIME) 1276 ei->i_flags |= EXT2_NOATIME_FL; 1277 if (flags & S_DIRSYNC) 1278 ei->i_flags |= EXT2_DIRSYNC_FL; 1279} 1280 1281struct inode *ext2_iget (struct super_block *sb, unsigned long ino) 1282{ 1283 struct ext2_inode_info *ei; 1284 struct buffer_head * bh; 1285 struct ext2_inode *raw_inode; 1286 struct inode *inode; 1287 long ret = -EIO; 1288 int n; 1289 1290 inode = iget_locked(sb, ino); 1291 if (!inode) 1292 return ERR_PTR(-ENOMEM); 1293 if (!(inode->i_state & I_NEW)) 1294 return inode; 1295 1296 ei = EXT2_I(inode); 1297 ei->i_block_alloc_info = NULL; 1298 1299 raw_inode = ext2_get_inode(inode->i_sb, ino, &bh); 1300 if (IS_ERR(raw_inode)) { 1301 ret = PTR_ERR(raw_inode); 1302 goto bad_inode; 1303 } 1304 1305 inode->i_mode = le16_to_cpu(raw_inode->i_mode); 1306 inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low); 1307 inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low); 1308 if (!(test_opt (inode->i_sb, NO_UID32))) { 1309 inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16; 1310 inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16; 1311 } 1312 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count); 1313 inode->i_size = le32_to_cpu(raw_inode->i_size); 1314 inode->i_atime.tv_sec = (signed)le32_to_cpu(raw_inode->i_atime); 1315 inode->i_ctime.tv_sec = (signed)le32_to_cpu(raw_inode->i_ctime); 1316 inode->i_mtime.tv_sec = (signed)le32_to_cpu(raw_inode->i_mtime); 1317 inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0; 1318 ei->i_dtime = le32_to_cpu(raw_inode->i_dtime); 1319 /* We now have enough fields to check if the inode was active or not. 1320 * This is needed because nfsd might try to access dead inodes 1321 * the test is that same one that e2fsck uses 1322 * NeilBrown 1999oct15 1323 */ 1324 if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) { 1325 /* this inode is deleted */ 1326 brelse (bh); 1327 ret = -ESTALE; 1328 goto bad_inode; 1329 } 1330 inode->i_blocks = le32_to_cpu(raw_inode->i_blocks); 1331 ei->i_flags = le32_to_cpu(raw_inode->i_flags); 1332 ei->i_faddr = le32_to_cpu(raw_inode->i_faddr); 1333 ei->i_frag_no = raw_inode->i_frag; 1334 ei->i_frag_size = raw_inode->i_fsize; 1335 ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl); 1336 ei->i_dir_acl = 0; 1337 if (S_ISREG(inode->i_mode)) 1338 inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32; 1339 else 1340 ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl); 1341 ei->i_dtime = 0; 1342 inode->i_generation = le32_to_cpu(raw_inode->i_generation); 1343 ei->i_state = 0; 1344 ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb); 1345 ei->i_dir_start_lookup = 0; 1346 1347 /* 1348 * NOTE! The in-memory inode i_data array is in little-endian order 1349 * even on big-endian machines: we do NOT byteswap the block numbers! 1350 */ 1351 for (n = 0; n < EXT2_N_BLOCKS; n++) 1352 ei->i_data[n] = raw_inode->i_block[n]; 1353 1354 if (S_ISREG(inode->i_mode)) { 1355 inode->i_op = &ext2_file_inode_operations; 1356 if (ext2_use_xip(inode->i_sb)) { 1357 inode->i_mapping->a_ops = &ext2_aops_xip; 1358 inode->i_fop = &ext2_xip_file_operations; 1359 } else if (test_opt(inode->i_sb, NOBH)) { 1360 inode->i_mapping->a_ops = &ext2_nobh_aops; 1361 inode->i_fop = &ext2_file_operations; 1362 } else { 1363 inode->i_mapping->a_ops = &ext2_aops; 1364 inode->i_fop = &ext2_file_operations; 1365 } 1366 } else if (S_ISDIR(inode->i_mode)) { 1367 inode->i_op = &ext2_dir_inode_operations; 1368 inode->i_fop = &ext2_dir_operations; 1369 if (test_opt(inode->i_sb, NOBH)) 1370 inode->i_mapping->a_ops = &ext2_nobh_aops; 1371 else 1372 inode->i_mapping->a_ops = &ext2_aops; 1373 } else if (S_ISLNK(inode->i_mode)) { 1374 if (ext2_inode_is_fast_symlink(inode)) { 1375 inode->i_op = &ext2_fast_symlink_inode_operations; 1376 nd_terminate_link(ei->i_data, inode->i_size, 1377 sizeof(ei->i_data) - 1); 1378 } else { 1379 inode->i_op = &ext2_symlink_inode_operations; 1380 if (test_opt(inode->i_sb, NOBH)) 1381 inode->i_mapping->a_ops = &ext2_nobh_aops; 1382 else 1383 inode->i_mapping->a_ops = &ext2_aops; 1384 } 1385 } else { 1386 inode->i_op = &ext2_special_inode_operations; 1387 if (raw_inode->i_block[0]) 1388 init_special_inode(inode, inode->i_mode, 1389 old_decode_dev(le32_to_cpu(raw_inode->i_block[0]))); 1390 else 1391 init_special_inode(inode, inode->i_mode, 1392 new_decode_dev(le32_to_cpu(raw_inode->i_block[1]))); 1393 } 1394 brelse (bh); 1395 ext2_set_inode_flags(inode); 1396 unlock_new_inode(inode); 1397 return inode; 1398 1399bad_inode: 1400 iget_failed(inode); 1401 return ERR_PTR(ret); 1402} 1403 1404static int __ext2_write_inode(struct inode *inode, int do_sync) 1405{ 1406 struct ext2_inode_info *ei = EXT2_I(inode); 1407 struct super_block *sb = inode->i_sb; 1408 ino_t ino = inode->i_ino; 1409 uid_t uid = inode->i_uid; 1410 gid_t gid = inode->i_gid; 1411 struct buffer_head * bh; 1412 struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh); 1413 int n; 1414 int err = 0; 1415 1416 if (IS_ERR(raw_inode)) 1417 return -EIO; 1418 1419 /* For fields not not tracking in the in-memory inode, 1420 * initialise them to zero for new inodes. */ 1421 if (ei->i_state & EXT2_STATE_NEW) 1422 memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size); 1423 1424 ext2_get_inode_flags(ei); 1425 raw_inode->i_mode = cpu_to_le16(inode->i_mode); 1426 if (!(test_opt(sb, NO_UID32))) { 1427 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid)); 1428 raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid)); 1429/* 1430 * Fix up interoperability with old kernels. Otherwise, old inodes get 1431 * re-used with the upper 16 bits of the uid/gid intact 1432 */ 1433 if (!ei->i_dtime) { 1434 raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid)); 1435 raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid)); 1436 } else { 1437 raw_inode->i_uid_high = 0; 1438 raw_inode->i_gid_high = 0; 1439 } 1440 } else { 1441 raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid)); 1442 raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid)); 1443 raw_inode->i_uid_high = 0; 1444 raw_inode->i_gid_high = 0; 1445 } 1446 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink); 1447 raw_inode->i_size = cpu_to_le32(inode->i_size); 1448 raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec); 1449 raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec); 1450 raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec); 1451 1452 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks); 1453 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime); 1454 raw_inode->i_flags = cpu_to_le32(ei->i_flags); 1455 raw_inode->i_faddr = cpu_to_le32(ei->i_faddr); 1456 raw_inode->i_frag = ei->i_frag_no; 1457 raw_inode->i_fsize = ei->i_frag_size; 1458 raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl); 1459 if (!S_ISREG(inode->i_mode)) 1460 raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl); 1461 else { 1462 raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32); 1463 if (inode->i_size > 0x7fffffffULL) { 1464 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb, 1465 EXT2_FEATURE_RO_COMPAT_LARGE_FILE) || 1466 EXT2_SB(sb)->s_es->s_rev_level == 1467 cpu_to_le32(EXT2_GOOD_OLD_REV)) { 1468 /* If this is the first large file 1469 * created, add a flag to the superblock. 1470 */ 1471 spin_lock(&EXT2_SB(sb)->s_lock); 1472 ext2_update_dynamic_rev(sb); 1473 EXT2_SET_RO_COMPAT_FEATURE(sb, 1474 EXT2_FEATURE_RO_COMPAT_LARGE_FILE); 1475 spin_unlock(&EXT2_SB(sb)->s_lock); 1476 ext2_write_super(sb); 1477 } 1478 } 1479 } 1480 1481 raw_inode->i_generation = cpu_to_le32(inode->i_generation); 1482 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) { 1483 if (old_valid_dev(inode->i_rdev)) { 1484 raw_inode->i_block[0] = 1485 cpu_to_le32(old_encode_dev(inode->i_rdev)); 1486 raw_inode->i_block[1] = 0; 1487 } else { 1488 raw_inode->i_block[0] = 0; 1489 raw_inode->i_block[1] = 1490 cpu_to_le32(new_encode_dev(inode->i_rdev)); 1491 raw_inode->i_block[2] = 0; 1492 } 1493 } else for (n = 0; n < EXT2_N_BLOCKS; n++) 1494 raw_inode->i_block[n] = ei->i_data[n]; 1495 mark_buffer_dirty(bh); 1496 if (do_sync) { 1497 sync_dirty_buffer(bh); 1498 if (buffer_req(bh) && !buffer_uptodate(bh)) { 1499 printk ("IO error syncing ext2 inode [%s:%08lx]\n", 1500 sb->s_id, (unsigned long) ino); 1501 err = -EIO; 1502 } 1503 } 1504 ei->i_state &= ~EXT2_STATE_NEW; 1505 brelse (bh); 1506 return err; 1507} 1508 1509int ext2_write_inode(struct inode *inode, struct writeback_control *wbc) 1510{ 1511 return __ext2_write_inode(inode, wbc->sync_mode == WB_SYNC_ALL); 1512} 1513 1514int ext2_sync_inode(struct inode *inode) 1515{ 1516 struct writeback_control wbc = { 1517 .sync_mode = WB_SYNC_ALL, 1518 .nr_to_write = 0, /* sys_fsync did this */ 1519 }; 1520 return sync_inode(inode, &wbc); 1521} 1522 1523int ext2_setattr(struct dentry *dentry, struct iattr *iattr) 1524{ 1525 struct inode *inode = dentry->d_inode; 1526 int error; 1527 1528 error = inode_change_ok(inode, iattr); 1529 if (error) 1530 return error; 1531 1532 if (is_quota_modification(inode, iattr)) 1533 dquot_initialize(inode); 1534 if ((iattr->ia_valid & ATTR_UID && iattr->ia_uid != inode->i_uid) || 1535 (iattr->ia_valid & ATTR_GID && iattr->ia_gid != inode->i_gid)) { 1536 error = dquot_transfer(inode, iattr); 1537 if (error) 1538 return error; 1539 } 1540 if (iattr->ia_valid & ATTR_SIZE && iattr->ia_size != inode->i_size) { 1541 error = ext2_setsize(inode, iattr->ia_size); 1542 if (error) 1543 return error; 1544 } 1545 setattr_copy(inode, iattr); 1546 if (iattr->ia_valid & ATTR_MODE) 1547 error = ext2_acl_chmod(inode); 1548 mark_inode_dirty(inode); 1549 1550 return error; 1551} 1552