Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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/smp_lock.h> |
| 26 | #include <linux/time.h> |
| 27 | #include <linux/highuid.h> |
| 28 | #include <linux/pagemap.h> |
| 29 | #include <linux/quotaops.h> |
| 30 | #include <linux/module.h> |
| 31 | #include <linux/writeback.h> |
| 32 | #include <linux/buffer_head.h> |
| 33 | #include <linux/mpage.h> |
| 34 | #include "ext2.h" |
| 35 | #include "acl.h" |
Carsten Otte | 6d79125 | 2005-06-23 22:05:26 -0700 | [diff] [blame] | 36 | #include "xip.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 37 | |
| 38 | MODULE_AUTHOR("Remy Card and others"); |
| 39 | MODULE_DESCRIPTION("Second Extended Filesystem"); |
| 40 | MODULE_LICENSE("GPL"); |
| 41 | |
| 42 | static int ext2_update_inode(struct inode * inode, int do_sync); |
| 43 | |
| 44 | /* |
| 45 | * Test whether an inode is a fast symlink. |
| 46 | */ |
| 47 | static inline int ext2_inode_is_fast_symlink(struct inode *inode) |
| 48 | { |
| 49 | int ea_blocks = EXT2_I(inode)->i_file_acl ? |
| 50 | (inode->i_sb->s_blocksize >> 9) : 0; |
| 51 | |
| 52 | return (S_ISLNK(inode->i_mode) && |
| 53 | inode->i_blocks - ea_blocks == 0); |
| 54 | } |
| 55 | |
| 56 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 57 | * Called at the last iput() if i_nlink is zero. |
| 58 | */ |
| 59 | void ext2_delete_inode (struct inode * inode) |
| 60 | { |
Mark Fasheh | fef2665 | 2005-09-09 13:01:31 -0700 | [diff] [blame] | 61 | truncate_inode_pages(&inode->i_data, 0); |
| 62 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 63 | if (is_bad_inode(inode)) |
| 64 | goto no_delete; |
| 65 | EXT2_I(inode)->i_dtime = get_seconds(); |
| 66 | mark_inode_dirty(inode); |
| 67 | ext2_update_inode(inode, inode_needs_sync(inode)); |
| 68 | |
| 69 | inode->i_size = 0; |
| 70 | if (inode->i_blocks) |
| 71 | ext2_truncate (inode); |
| 72 | ext2_free_inode (inode); |
| 73 | |
| 74 | return; |
| 75 | no_delete: |
| 76 | clear_inode(inode); /* We must guarantee clearing of inode... */ |
| 77 | } |
| 78 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 79 | typedef struct { |
| 80 | __le32 *p; |
| 81 | __le32 key; |
| 82 | struct buffer_head *bh; |
| 83 | } Indirect; |
| 84 | |
| 85 | static inline void add_chain(Indirect *p, struct buffer_head *bh, __le32 *v) |
| 86 | { |
| 87 | p->key = *(p->p = v); |
| 88 | p->bh = bh; |
| 89 | } |
| 90 | |
| 91 | static inline int verify_chain(Indirect *from, Indirect *to) |
| 92 | { |
| 93 | while (from <= to && from->key == *from->p) |
| 94 | from++; |
| 95 | return (from > to); |
| 96 | } |
| 97 | |
| 98 | /** |
| 99 | * ext2_block_to_path - parse the block number into array of offsets |
| 100 | * @inode: inode in question (we are only interested in its superblock) |
| 101 | * @i_block: block number to be parsed |
| 102 | * @offsets: array to store the offsets in |
| 103 | * @boundary: set this non-zero if the referred-to block is likely to be |
| 104 | * followed (on disk) by an indirect block. |
| 105 | * To store the locations of file's data ext2 uses a data structure common |
| 106 | * for UNIX filesystems - tree of pointers anchored in the inode, with |
| 107 | * data blocks at leaves and indirect blocks in intermediate nodes. |
| 108 | * This function translates the block number into path in that tree - |
| 109 | * return value is the path length and @offsets[n] is the offset of |
| 110 | * pointer to (n+1)th node in the nth one. If @block is out of range |
| 111 | * (negative or too large) warning is printed and zero returned. |
| 112 | * |
| 113 | * Note: function doesn't find node addresses, so no IO is needed. All |
| 114 | * we need to know is the capacity of indirect blocks (taken from the |
| 115 | * inode->i_sb). |
| 116 | */ |
| 117 | |
| 118 | /* |
| 119 | * Portability note: the last comparison (check that we fit into triple |
| 120 | * indirect block) is spelled differently, because otherwise on an |
| 121 | * architecture with 32-bit longs and 8Kb pages we might get into trouble |
| 122 | * if our filesystem had 8Kb blocks. We might use long long, but that would |
| 123 | * kill us on x86. Oh, well, at least the sign propagation does not matter - |
| 124 | * i_block would have to be negative in the very beginning, so we would not |
| 125 | * get there at all. |
| 126 | */ |
| 127 | |
| 128 | static int ext2_block_to_path(struct inode *inode, |
| 129 | long i_block, int offsets[4], int *boundary) |
| 130 | { |
| 131 | int ptrs = EXT2_ADDR_PER_BLOCK(inode->i_sb); |
| 132 | int ptrs_bits = EXT2_ADDR_PER_BLOCK_BITS(inode->i_sb); |
| 133 | const long direct_blocks = EXT2_NDIR_BLOCKS, |
| 134 | indirect_blocks = ptrs, |
| 135 | double_blocks = (1 << (ptrs_bits * 2)); |
| 136 | int n = 0; |
| 137 | int final = 0; |
| 138 | |
| 139 | if (i_block < 0) { |
| 140 | ext2_warning (inode->i_sb, "ext2_block_to_path", "block < 0"); |
| 141 | } else if (i_block < direct_blocks) { |
| 142 | offsets[n++] = i_block; |
| 143 | final = direct_blocks; |
| 144 | } else if ( (i_block -= direct_blocks) < indirect_blocks) { |
| 145 | offsets[n++] = EXT2_IND_BLOCK; |
| 146 | offsets[n++] = i_block; |
| 147 | final = ptrs; |
| 148 | } else if ((i_block -= indirect_blocks) < double_blocks) { |
| 149 | offsets[n++] = EXT2_DIND_BLOCK; |
| 150 | offsets[n++] = i_block >> ptrs_bits; |
| 151 | offsets[n++] = i_block & (ptrs - 1); |
| 152 | final = ptrs; |
| 153 | } else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) { |
| 154 | offsets[n++] = EXT2_TIND_BLOCK; |
| 155 | offsets[n++] = i_block >> (ptrs_bits * 2); |
| 156 | offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1); |
| 157 | offsets[n++] = i_block & (ptrs - 1); |
| 158 | final = ptrs; |
| 159 | } else { |
| 160 | ext2_warning (inode->i_sb, "ext2_block_to_path", "block > big"); |
| 161 | } |
| 162 | if (boundary) |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 163 | *boundary = final - 1 - (i_block & (ptrs - 1)); |
| 164 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 165 | return n; |
| 166 | } |
| 167 | |
| 168 | /** |
| 169 | * ext2_get_branch - read the chain of indirect blocks leading to data |
| 170 | * @inode: inode in question |
| 171 | * @depth: depth of the chain (1 - direct pointer, etc.) |
| 172 | * @offsets: offsets of pointers in inode/indirect blocks |
| 173 | * @chain: place to store the result |
| 174 | * @err: here we store the error value |
| 175 | * |
| 176 | * Function fills the array of triples <key, p, bh> and returns %NULL |
| 177 | * if everything went OK or the pointer to the last filled triple |
| 178 | * (incomplete one) otherwise. Upon the return chain[i].key contains |
| 179 | * the number of (i+1)-th block in the chain (as it is stored in memory, |
| 180 | * i.e. little-endian 32-bit), chain[i].p contains the address of that |
| 181 | * number (it points into struct inode for i==0 and into the bh->b_data |
| 182 | * for i>0) and chain[i].bh points to the buffer_head of i-th indirect |
| 183 | * block for i>0 and NULL for i==0. In other words, it holds the block |
| 184 | * numbers of the chain, addresses they were taken from (and where we can |
| 185 | * verify that chain did not change) and buffer_heads hosting these |
| 186 | * numbers. |
| 187 | * |
| 188 | * Function stops when it stumbles upon zero pointer (absent block) |
| 189 | * (pointer to last triple returned, *@err == 0) |
| 190 | * or when it gets an IO error reading an indirect block |
| 191 | * (ditto, *@err == -EIO) |
| 192 | * or when it notices that chain had been changed while it was reading |
| 193 | * (ditto, *@err == -EAGAIN) |
| 194 | * or when it reads all @depth-1 indirect blocks successfully and finds |
| 195 | * the whole chain, all way to the data (returns %NULL, *err == 0). |
| 196 | */ |
| 197 | static Indirect *ext2_get_branch(struct inode *inode, |
| 198 | int depth, |
| 199 | int *offsets, |
| 200 | Indirect chain[4], |
| 201 | int *err) |
| 202 | { |
| 203 | struct super_block *sb = inode->i_sb; |
| 204 | Indirect *p = chain; |
| 205 | struct buffer_head *bh; |
| 206 | |
| 207 | *err = 0; |
| 208 | /* i_data is not going away, no lock needed */ |
| 209 | add_chain (chain, NULL, EXT2_I(inode)->i_data + *offsets); |
| 210 | if (!p->key) |
| 211 | goto no_block; |
| 212 | while (--depth) { |
| 213 | bh = sb_bread(sb, le32_to_cpu(p->key)); |
| 214 | if (!bh) |
| 215 | goto failure; |
| 216 | read_lock(&EXT2_I(inode)->i_meta_lock); |
| 217 | if (!verify_chain(chain, p)) |
| 218 | goto changed; |
| 219 | add_chain(++p, bh, (__le32*)bh->b_data + *++offsets); |
| 220 | read_unlock(&EXT2_I(inode)->i_meta_lock); |
| 221 | if (!p->key) |
| 222 | goto no_block; |
| 223 | } |
| 224 | return NULL; |
| 225 | |
| 226 | changed: |
| 227 | read_unlock(&EXT2_I(inode)->i_meta_lock); |
| 228 | brelse(bh); |
| 229 | *err = -EAGAIN; |
| 230 | goto no_block; |
| 231 | failure: |
| 232 | *err = -EIO; |
| 233 | no_block: |
| 234 | return p; |
| 235 | } |
| 236 | |
| 237 | /** |
| 238 | * ext2_find_near - find a place for allocation with sufficient locality |
| 239 | * @inode: owner |
| 240 | * @ind: descriptor of indirect block. |
| 241 | * |
Benoit Boissinot | 1cc8dcf | 2008-04-21 22:45:55 +0000 | [diff] [blame] | 242 | * This function returns the preferred place for block allocation. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 243 | * It is used when heuristic for sequential allocation fails. |
| 244 | * Rules are: |
| 245 | * + if there is a block to the left of our position - allocate near it. |
| 246 | * + if pointer will live in indirect block - allocate near that block. |
| 247 | * + if pointer will live in inode - allocate in the same cylinder group. |
| 248 | * |
| 249 | * In the latter case we colour the starting block by the callers PID to |
| 250 | * prevent it from clashing with concurrent allocations for a different inode |
| 251 | * in the same block group. The PID is used here so that functionally related |
| 252 | * files will be close-by on-disk. |
| 253 | * |
| 254 | * Caller must make sure that @ind is valid and will stay that way. |
| 255 | */ |
| 256 | |
| 257 | static unsigned long ext2_find_near(struct inode *inode, Indirect *ind) |
| 258 | { |
| 259 | struct ext2_inode_info *ei = EXT2_I(inode); |
| 260 | __le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data; |
| 261 | __le32 *p; |
| 262 | unsigned long bg_start; |
| 263 | unsigned long colour; |
| 264 | |
| 265 | /* Try to find previous block */ |
| 266 | for (p = ind->p - 1; p >= start; p--) |
| 267 | if (*p) |
| 268 | return le32_to_cpu(*p); |
| 269 | |
| 270 | /* No such thing, so let's try location of indirect block */ |
| 271 | if (ind->bh) |
| 272 | return ind->bh->b_blocknr; |
| 273 | |
| 274 | /* |
| 275 | * It is going to be refered from inode itself? OK, just put it into |
| 276 | * the same cylinder group then. |
| 277 | */ |
Akinobu Mita | 24097d1 | 2008-04-28 02:16:01 -0700 | [diff] [blame^] | 278 | bg_start = ext2_group_first_block_no(inode->i_sb, ei->i_block_group); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 279 | colour = (current->pid % 16) * |
| 280 | (EXT2_BLOCKS_PER_GROUP(inode->i_sb) / 16); |
| 281 | return bg_start + colour; |
| 282 | } |
| 283 | |
| 284 | /** |
Benoit Boissinot | 1cc8dcf | 2008-04-21 22:45:55 +0000 | [diff] [blame] | 285 | * ext2_find_goal - find a preferred place for allocation. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 286 | * @inode: owner |
| 287 | * @block: block we want |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 288 | * @partial: pointer to the last triple within a chain |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 289 | * |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 290 | * Returns preferred place for a block (the goal). |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 291 | */ |
| 292 | |
Akinobu Mita | fb01bfd | 2008-02-06 01:40:16 -0800 | [diff] [blame] | 293 | static inline int ext2_find_goal(struct inode *inode, long block, |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 294 | Indirect *partial) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 295 | { |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 296 | struct ext2_block_alloc_info *block_i; |
| 297 | |
| 298 | block_i = EXT2_I(inode)->i_block_alloc_info; |
| 299 | |
| 300 | /* |
| 301 | * try the heuristic for sequential allocation, |
| 302 | * failing that at least try to get decent locality. |
| 303 | */ |
| 304 | if (block_i && (block == block_i->last_alloc_logical_block + 1) |
| 305 | && (block_i->last_alloc_physical_block != 0)) { |
| 306 | return block_i->last_alloc_physical_block + 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 307 | } |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 308 | |
| 309 | return ext2_find_near(inode, partial); |
| 310 | } |
| 311 | |
| 312 | /** |
| 313 | * ext2_blks_to_allocate: Look up the block map and count the number |
| 314 | * of direct blocks need to be allocated for the given branch. |
| 315 | * |
| 316 | * @branch: chain of indirect blocks |
| 317 | * @k: number of blocks need for indirect blocks |
| 318 | * @blks: number of data blocks to be mapped. |
| 319 | * @blocks_to_boundary: the offset in the indirect block |
| 320 | * |
| 321 | * return the total number of blocks to be allocate, including the |
| 322 | * direct and indirect blocks. |
| 323 | */ |
| 324 | static int |
| 325 | ext2_blks_to_allocate(Indirect * branch, int k, unsigned long blks, |
| 326 | int blocks_to_boundary) |
| 327 | { |
| 328 | unsigned long count = 0; |
| 329 | |
| 330 | /* |
| 331 | * Simple case, [t,d]Indirect block(s) has not allocated yet |
| 332 | * then it's clear blocks on that path have not allocated |
| 333 | */ |
| 334 | if (k > 0) { |
| 335 | /* right now don't hanel cross boundary allocation */ |
| 336 | if (blks < blocks_to_boundary + 1) |
| 337 | count += blks; |
| 338 | else |
| 339 | count += blocks_to_boundary + 1; |
| 340 | return count; |
| 341 | } |
| 342 | |
| 343 | count++; |
| 344 | while (count < blks && count <= blocks_to_boundary |
| 345 | && le32_to_cpu(*(branch[0].p + count)) == 0) { |
| 346 | count++; |
| 347 | } |
| 348 | return count; |
| 349 | } |
| 350 | |
| 351 | /** |
| 352 | * ext2_alloc_blocks: multiple allocate blocks needed for a branch |
| 353 | * @indirect_blks: the number of blocks need to allocate for indirect |
| 354 | * blocks |
| 355 | * |
| 356 | * @new_blocks: on return it will store the new block numbers for |
| 357 | * the indirect blocks(if needed) and the first direct block, |
| 358 | * @blks: on return it will store the total number of allocated |
| 359 | * direct blocks |
| 360 | */ |
| 361 | static int ext2_alloc_blocks(struct inode *inode, |
| 362 | ext2_fsblk_t goal, int indirect_blks, int blks, |
| 363 | ext2_fsblk_t new_blocks[4], int *err) |
| 364 | { |
| 365 | int target, i; |
| 366 | unsigned long count = 0; |
| 367 | int index = 0; |
| 368 | ext2_fsblk_t current_block = 0; |
| 369 | int ret = 0; |
| 370 | |
| 371 | /* |
| 372 | * Here we try to allocate the requested multiple blocks at once, |
| 373 | * on a best-effort basis. |
| 374 | * To build a branch, we should allocate blocks for |
| 375 | * the indirect blocks(if not allocated yet), and at least |
| 376 | * the first direct block of this branch. That's the |
| 377 | * minimum number of blocks need to allocate(required) |
| 378 | */ |
| 379 | target = blks + indirect_blks; |
| 380 | |
| 381 | while (1) { |
| 382 | count = target; |
| 383 | /* allocating blocks for indirect blocks and direct blocks */ |
| 384 | current_block = ext2_new_blocks(inode,goal,&count,err); |
| 385 | if (*err) |
| 386 | goto failed_out; |
| 387 | |
| 388 | target -= count; |
| 389 | /* allocate blocks for indirect blocks */ |
| 390 | while (index < indirect_blks && count) { |
| 391 | new_blocks[index++] = current_block++; |
| 392 | count--; |
| 393 | } |
| 394 | |
| 395 | if (count > 0) |
| 396 | break; |
| 397 | } |
| 398 | |
| 399 | /* save the new block number for the first direct block */ |
| 400 | new_blocks[index] = current_block; |
| 401 | |
| 402 | /* total number of blocks allocated for direct blocks */ |
| 403 | ret = count; |
| 404 | *err = 0; |
| 405 | return ret; |
| 406 | failed_out: |
| 407 | for (i = 0; i <index; i++) |
| 408 | ext2_free_blocks(inode, new_blocks[i], 1); |
| 409 | return ret; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 410 | } |
| 411 | |
| 412 | /** |
| 413 | * ext2_alloc_branch - allocate and set up a chain of blocks. |
| 414 | * @inode: owner |
| 415 | * @num: depth of the chain (number of blocks to allocate) |
| 416 | * @offsets: offsets (in the blocks) to store the pointers to next. |
| 417 | * @branch: place to store the chain in. |
| 418 | * |
| 419 | * This function allocates @num blocks, zeroes out all but the last one, |
| 420 | * links them into chain and (if we are synchronous) writes them to disk. |
| 421 | * In other words, it prepares a branch that can be spliced onto the |
| 422 | * inode. It stores the information about that chain in the branch[], in |
| 423 | * the same format as ext2_get_branch() would do. We are calling it after |
| 424 | * we had read the existing part of chain and partial points to the last |
| 425 | * triple of that (one with zero ->key). Upon the exit we have the same |
| 426 | * picture as after the successful ext2_get_block(), excpet that in one |
| 427 | * place chain is disconnected - *branch->p is still zero (we did not |
| 428 | * set the last link), but branch->key contains the number that should |
| 429 | * be placed into *branch->p to fill that gap. |
| 430 | * |
| 431 | * If allocation fails we free all blocks we've allocated (and forget |
| 432 | * their buffer_heads) and return the error value the from failed |
| 433 | * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain |
| 434 | * as described above and return 0. |
| 435 | */ |
| 436 | |
| 437 | static int ext2_alloc_branch(struct inode *inode, |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 438 | int indirect_blks, int *blks, ext2_fsblk_t goal, |
| 439 | int *offsets, Indirect *branch) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 440 | { |
| 441 | int blocksize = inode->i_sb->s_blocksize; |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 442 | int i, n = 0; |
| 443 | int err = 0; |
| 444 | struct buffer_head *bh; |
| 445 | int num; |
| 446 | ext2_fsblk_t new_blocks[4]; |
| 447 | ext2_fsblk_t current_block; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 448 | |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 449 | num = ext2_alloc_blocks(inode, goal, indirect_blks, |
| 450 | *blks, new_blocks, &err); |
| 451 | if (err) |
| 452 | return err; |
| 453 | |
| 454 | branch[0].key = cpu_to_le32(new_blocks[0]); |
| 455 | /* |
| 456 | * metadata blocks and data blocks are allocated. |
| 457 | */ |
| 458 | for (n = 1; n <= indirect_blks; n++) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 459 | /* |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 460 | * Get buffer_head for parent block, zero it out |
| 461 | * and set the pointer to new one, then send |
| 462 | * parent to disk. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 463 | */ |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 464 | bh = sb_getblk(inode->i_sb, new_blocks[n-1]); |
| 465 | branch[n].bh = bh; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 466 | lock_buffer(bh); |
| 467 | memset(bh->b_data, 0, blocksize); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 468 | branch[n].p = (__le32 *) bh->b_data + offsets[n]; |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 469 | branch[n].key = cpu_to_le32(new_blocks[n]); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 470 | *branch[n].p = branch[n].key; |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 471 | if ( n == indirect_blks) { |
| 472 | current_block = new_blocks[n]; |
| 473 | /* |
| 474 | * End of chain, update the last new metablock of |
| 475 | * the chain to point to the new allocated |
| 476 | * data blocks numbers |
| 477 | */ |
| 478 | for (i=1; i < num; i++) |
| 479 | *(branch[n].p + i) = cpu_to_le32(++current_block); |
| 480 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 481 | set_buffer_uptodate(bh); |
| 482 | unlock_buffer(bh); |
| 483 | mark_buffer_dirty_inode(bh, inode); |
| 484 | /* We used to sync bh here if IS_SYNC(inode). |
| 485 | * But we now rely upon generic_osync_inode() |
| 486 | * and b_inode_buffers. But not for directories. |
| 487 | */ |
| 488 | if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode)) |
| 489 | sync_dirty_buffer(bh); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 490 | } |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 491 | *blks = num; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 492 | return err; |
| 493 | } |
| 494 | |
| 495 | /** |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 496 | * ext2_splice_branch - splice the allocated branch onto inode. |
| 497 | * @inode: owner |
| 498 | * @block: (logical) number of block we are adding |
| 499 | * @chain: chain of indirect blocks (with a missing link - see |
| 500 | * ext2_alloc_branch) |
| 501 | * @where: location of missing link |
| 502 | * @num: number of indirect blocks we are adding |
| 503 | * @blks: number of direct blocks we are adding |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 504 | * |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 505 | * This function fills the missing link and does all housekeeping needed in |
| 506 | * inode (->i_blocks, etc.). In case of success we end up with the full |
| 507 | * chain to new block and return 0. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 508 | */ |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 509 | static void ext2_splice_branch(struct inode *inode, |
| 510 | long block, Indirect *where, int num, int blks) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 511 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 512 | int i; |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 513 | struct ext2_block_alloc_info *block_i; |
| 514 | ext2_fsblk_t current_block; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 515 | |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 516 | block_i = EXT2_I(inode)->i_block_alloc_info; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 517 | |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 518 | /* XXX LOCKING probably should have i_meta_lock ?*/ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 519 | /* That's it */ |
| 520 | |
| 521 | *where->p = where->key; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 522 | |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 523 | /* |
| 524 | * Update the host buffer_head or inode to point to more just allocated |
| 525 | * direct blocks blocks |
| 526 | */ |
| 527 | if (num == 0 && blks > 1) { |
| 528 | current_block = le32_to_cpu(where->key) + 1; |
| 529 | for (i = 1; i < blks; i++) |
| 530 | *(where->p + i ) = cpu_to_le32(current_block++); |
| 531 | } |
| 532 | |
| 533 | /* |
| 534 | * update the most recently allocated logical & physical block |
| 535 | * in i_block_alloc_info, to assist find the proper goal block for next |
| 536 | * allocation |
| 537 | */ |
| 538 | if (block_i) { |
| 539 | block_i->last_alloc_logical_block = block + blks - 1; |
| 540 | block_i->last_alloc_physical_block = |
| 541 | le32_to_cpu(where[num].key) + blks - 1; |
| 542 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 543 | |
| 544 | /* We are done with atomic stuff, now do the rest of housekeeping */ |
| 545 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 546 | /* had we spliced it onto indirect block? */ |
| 547 | if (where->bh) |
| 548 | mark_buffer_dirty_inode(where->bh, inode); |
| 549 | |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 550 | inode->i_ctime = CURRENT_TIME_SEC; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 551 | mark_inode_dirty(inode); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 552 | } |
| 553 | |
| 554 | /* |
| 555 | * Allocation strategy is simple: if we have to allocate something, we will |
| 556 | * have to go the whole way to leaf. So let's do it before attaching anything |
| 557 | * to tree, set linkage between the newborn blocks, write them if sync is |
| 558 | * required, recheck the path, free and repeat if check fails, otherwise |
| 559 | * set the last missing link (that will protect us from any truncate-generated |
| 560 | * removals - all blocks on the path are immune now) and possibly force the |
| 561 | * write on the parent block. |
| 562 | * That has a nice additional property: no special recovery from the failed |
| 563 | * allocations is needed - we simply release blocks and do not touch anything |
| 564 | * reachable from inode. |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 565 | * |
| 566 | * `handle' can be NULL if create == 0. |
| 567 | * |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 568 | * return > 0, # of blocks mapped or allocated. |
| 569 | * return = 0, if plain lookup failed. |
| 570 | * return < 0, error case. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 571 | */ |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 572 | static int ext2_get_blocks(struct inode *inode, |
| 573 | sector_t iblock, unsigned long maxblocks, |
| 574 | struct buffer_head *bh_result, |
| 575 | int create) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 576 | { |
| 577 | int err = -EIO; |
| 578 | int offsets[4]; |
| 579 | Indirect chain[4]; |
| 580 | Indirect *partial; |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 581 | ext2_fsblk_t goal; |
| 582 | int indirect_blks; |
| 583 | int blocks_to_boundary = 0; |
| 584 | int depth; |
| 585 | struct ext2_inode_info *ei = EXT2_I(inode); |
| 586 | int count = 0; |
| 587 | ext2_fsblk_t first_block = 0; |
| 588 | |
| 589 | depth = ext2_block_to_path(inode,iblock,offsets,&blocks_to_boundary); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 590 | |
| 591 | if (depth == 0) |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 592 | return (err); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 593 | reread: |
| 594 | partial = ext2_get_branch(inode, depth, offsets, chain, &err); |
| 595 | |
| 596 | /* Simplest case - block found, no allocation needed */ |
| 597 | if (!partial) { |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 598 | first_block = le32_to_cpu(chain[depth - 1].key); |
| 599 | clear_buffer_new(bh_result); /* What's this do? */ |
| 600 | count++; |
| 601 | /*map more blocks*/ |
| 602 | while (count < maxblocks && count <= blocks_to_boundary) { |
| 603 | ext2_fsblk_t blk; |
| 604 | |
| 605 | if (!verify_chain(chain, partial)) { |
| 606 | /* |
| 607 | * Indirect block might be removed by |
| 608 | * truncate while we were reading it. |
| 609 | * Handling of that case: forget what we've |
| 610 | * got now, go to reread. |
| 611 | */ |
| 612 | count = 0; |
| 613 | goto changed; |
| 614 | } |
| 615 | blk = le32_to_cpu(*(chain[depth-1].p + count)); |
| 616 | if (blk == first_block + count) |
| 617 | count++; |
| 618 | else |
| 619 | break; |
| 620 | } |
| 621 | goto got_it; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 622 | } |
| 623 | |
| 624 | /* Next simple case - plain lookup or failed read of indirect block */ |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 625 | if (!create || err == -EIO) |
| 626 | goto cleanup; |
| 627 | |
| 628 | mutex_lock(&ei->truncate_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 629 | |
| 630 | /* |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 631 | * Okay, we need to do block allocation. Lazily initialize the block |
| 632 | * allocation info here if necessary |
| 633 | */ |
| 634 | if (S_ISREG(inode->i_mode) && (!ei->i_block_alloc_info)) |
| 635 | ext2_init_block_alloc_info(inode); |
| 636 | |
Akinobu Mita | fb01bfd | 2008-02-06 01:40:16 -0800 | [diff] [blame] | 637 | goal = ext2_find_goal(inode, iblock, partial); |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 638 | |
| 639 | /* the number of blocks need to allocate for [d,t]indirect blocks */ |
| 640 | indirect_blks = (chain + depth) - partial - 1; |
| 641 | /* |
| 642 | * Next look up the indirect map to count the totoal number of |
| 643 | * direct blocks to allocate for this branch. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 644 | */ |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 645 | count = ext2_blks_to_allocate(partial, indirect_blks, |
| 646 | maxblocks, blocks_to_boundary); |
| 647 | /* |
| 648 | * XXX ???? Block out ext2_truncate while we alter the tree |
| 649 | */ |
| 650 | err = ext2_alloc_branch(inode, indirect_blks, &count, goal, |
| 651 | offsets + (partial - chain), partial); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 652 | |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 653 | if (err) { |
| 654 | mutex_unlock(&ei->truncate_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 655 | goto cleanup; |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 656 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 657 | |
Carsten Otte | 6d79125 | 2005-06-23 22:05:26 -0700 | [diff] [blame] | 658 | if (ext2_use_xip(inode->i_sb)) { |
| 659 | /* |
| 660 | * we need to clear the block |
| 661 | */ |
| 662 | err = ext2_clear_xip_target (inode, |
| 663 | le32_to_cpu(chain[depth-1].key)); |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 664 | if (err) { |
| 665 | mutex_unlock(&ei->truncate_mutex); |
Carsten Otte | 6d79125 | 2005-06-23 22:05:26 -0700 | [diff] [blame] | 666 | goto cleanup; |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 667 | } |
Carsten Otte | 6d79125 | 2005-06-23 22:05:26 -0700 | [diff] [blame] | 668 | } |
| 669 | |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 670 | ext2_splice_branch(inode, iblock, partial, indirect_blks, count); |
| 671 | mutex_unlock(&ei->truncate_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 672 | set_buffer_new(bh_result); |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 673 | got_it: |
| 674 | map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key)); |
| 675 | if (count > blocks_to_boundary) |
| 676 | set_buffer_boundary(bh_result); |
| 677 | err = count; |
| 678 | /* Clean up and exit */ |
| 679 | partial = chain + depth - 1; /* the whole chain */ |
| 680 | cleanup: |
| 681 | while (partial > chain) { |
| 682 | brelse(partial->bh); |
| 683 | partial--; |
| 684 | } |
| 685 | return err; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 686 | changed: |
| 687 | while (partial > chain) { |
| 688 | brelse(partial->bh); |
| 689 | partial--; |
| 690 | } |
| 691 | goto reread; |
| 692 | } |
| 693 | |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 694 | int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create) |
| 695 | { |
| 696 | unsigned max_blocks = bh_result->b_size >> inode->i_blkbits; |
| 697 | int ret = ext2_get_blocks(inode, iblock, max_blocks, |
| 698 | bh_result, create); |
| 699 | if (ret > 0) { |
| 700 | bh_result->b_size = (ret << inode->i_blkbits); |
| 701 | ret = 0; |
| 702 | } |
| 703 | return ret; |
| 704 | |
| 705 | } |
| 706 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 707 | static int ext2_writepage(struct page *page, struct writeback_control *wbc) |
| 708 | { |
| 709 | return block_write_full_page(page, ext2_get_block, wbc); |
| 710 | } |
| 711 | |
| 712 | static int ext2_readpage(struct file *file, struct page *page) |
| 713 | { |
| 714 | return mpage_readpage(page, ext2_get_block); |
| 715 | } |
| 716 | |
| 717 | static int |
| 718 | ext2_readpages(struct file *file, struct address_space *mapping, |
| 719 | struct list_head *pages, unsigned nr_pages) |
| 720 | { |
| 721 | return mpage_readpages(mapping, pages, nr_pages, ext2_get_block); |
| 722 | } |
| 723 | |
Nick Piggin | f34fb6e | 2007-10-16 01:25:04 -0700 | [diff] [blame] | 724 | int __ext2_write_begin(struct file *file, struct address_space *mapping, |
| 725 | loff_t pos, unsigned len, unsigned flags, |
| 726 | struct page **pagep, void **fsdata) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 727 | { |
Nick Piggin | f34fb6e | 2007-10-16 01:25:04 -0700 | [diff] [blame] | 728 | return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata, |
| 729 | ext2_get_block); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 730 | } |
| 731 | |
| 732 | static int |
Nick Piggin | f34fb6e | 2007-10-16 01:25:04 -0700 | [diff] [blame] | 733 | ext2_write_begin(struct file *file, struct address_space *mapping, |
| 734 | loff_t pos, unsigned len, unsigned flags, |
| 735 | struct page **pagep, void **fsdata) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 736 | { |
Nick Piggin | f34fb6e | 2007-10-16 01:25:04 -0700 | [diff] [blame] | 737 | *pagep = NULL; |
| 738 | return __ext2_write_begin(file, mapping, pos, len, flags, pagep,fsdata); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 739 | } |
| 740 | |
Nick Piggin | 03158cd | 2007-10-16 01:25:25 -0700 | [diff] [blame] | 741 | static int |
| 742 | ext2_nobh_write_begin(struct file *file, struct address_space *mapping, |
| 743 | loff_t pos, unsigned len, unsigned flags, |
| 744 | struct page **pagep, void **fsdata) |
| 745 | { |
| 746 | /* |
| 747 | * Dir-in-pagecache still uses ext2_write_begin. Would have to rework |
| 748 | * directory handling code to pass around offsets rather than struct |
| 749 | * pages in order to make this work easily. |
| 750 | */ |
| 751 | return nobh_write_begin(file, mapping, pos, len, flags, pagep, fsdata, |
| 752 | ext2_get_block); |
| 753 | } |
| 754 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 755 | static int ext2_nobh_writepage(struct page *page, |
| 756 | struct writeback_control *wbc) |
| 757 | { |
| 758 | return nobh_writepage(page, ext2_get_block, wbc); |
| 759 | } |
| 760 | |
| 761 | static sector_t ext2_bmap(struct address_space *mapping, sector_t block) |
| 762 | { |
| 763 | return generic_block_bmap(mapping,block,ext2_get_block); |
| 764 | } |
| 765 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 766 | static ssize_t |
| 767 | ext2_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, |
| 768 | loff_t offset, unsigned long nr_segs) |
| 769 | { |
| 770 | struct file *file = iocb->ki_filp; |
| 771 | struct inode *inode = file->f_mapping->host; |
| 772 | |
| 773 | return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov, |
Badari Pulavarty | 1d8fa7a | 2006-03-26 01:38:02 -0800 | [diff] [blame] | 774 | offset, nr_segs, ext2_get_block, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 775 | } |
| 776 | |
| 777 | static int |
| 778 | ext2_writepages(struct address_space *mapping, struct writeback_control *wbc) |
| 779 | { |
| 780 | return mpage_writepages(mapping, wbc, ext2_get_block); |
| 781 | } |
| 782 | |
Christoph Hellwig | f5e54d6 | 2006-06-28 04:26:44 -0700 | [diff] [blame] | 783 | const struct address_space_operations ext2_aops = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 784 | .readpage = ext2_readpage, |
| 785 | .readpages = ext2_readpages, |
| 786 | .writepage = ext2_writepage, |
| 787 | .sync_page = block_sync_page, |
Nick Piggin | f34fb6e | 2007-10-16 01:25:04 -0700 | [diff] [blame] | 788 | .write_begin = ext2_write_begin, |
| 789 | .write_end = generic_write_end, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 790 | .bmap = ext2_bmap, |
| 791 | .direct_IO = ext2_direct_IO, |
| 792 | .writepages = ext2_writepages, |
Christoph Lameter | e965f96 | 2006-02-01 03:05:41 -0800 | [diff] [blame] | 793 | .migratepage = buffer_migrate_page, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 794 | }; |
| 795 | |
Christoph Hellwig | f5e54d6 | 2006-06-28 04:26:44 -0700 | [diff] [blame] | 796 | const struct address_space_operations ext2_aops_xip = { |
Carsten Otte | 6d79125 | 2005-06-23 22:05:26 -0700 | [diff] [blame] | 797 | .bmap = ext2_bmap, |
Nick Piggin | 70688e4 | 2008-04-28 02:13:02 -0700 | [diff] [blame] | 798 | .get_xip_mem = ext2_get_xip_mem, |
Carsten Otte | 6d79125 | 2005-06-23 22:05:26 -0700 | [diff] [blame] | 799 | }; |
| 800 | |
Christoph Hellwig | f5e54d6 | 2006-06-28 04:26:44 -0700 | [diff] [blame] | 801 | const struct address_space_operations ext2_nobh_aops = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 802 | .readpage = ext2_readpage, |
| 803 | .readpages = ext2_readpages, |
| 804 | .writepage = ext2_nobh_writepage, |
| 805 | .sync_page = block_sync_page, |
Nick Piggin | 03158cd | 2007-10-16 01:25:25 -0700 | [diff] [blame] | 806 | .write_begin = ext2_nobh_write_begin, |
| 807 | .write_end = nobh_write_end, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 808 | .bmap = ext2_bmap, |
| 809 | .direct_IO = ext2_direct_IO, |
| 810 | .writepages = ext2_writepages, |
Christoph Lameter | e965f96 | 2006-02-01 03:05:41 -0800 | [diff] [blame] | 811 | .migratepage = buffer_migrate_page, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 812 | }; |
| 813 | |
| 814 | /* |
| 815 | * Probably it should be a library function... search for first non-zero word |
| 816 | * or memcmp with zero_page, whatever is better for particular architecture. |
| 817 | * Linus? |
| 818 | */ |
| 819 | static inline int all_zeroes(__le32 *p, __le32 *q) |
| 820 | { |
| 821 | while (p < q) |
| 822 | if (*p++) |
| 823 | return 0; |
| 824 | return 1; |
| 825 | } |
| 826 | |
| 827 | /** |
| 828 | * ext2_find_shared - find the indirect blocks for partial truncation. |
| 829 | * @inode: inode in question |
| 830 | * @depth: depth of the affected branch |
| 831 | * @offsets: offsets of pointers in that branch (see ext2_block_to_path) |
| 832 | * @chain: place to store the pointers to partial indirect blocks |
| 833 | * @top: place to the (detached) top of branch |
| 834 | * |
| 835 | * This is a helper function used by ext2_truncate(). |
| 836 | * |
| 837 | * When we do truncate() we may have to clean the ends of several indirect |
| 838 | * blocks but leave the blocks themselves alive. Block is partially |
| 839 | * truncated if some data below the new i_size is refered from it (and |
| 840 | * it is on the path to the first completely truncated data block, indeed). |
| 841 | * We have to free the top of that path along with everything to the right |
| 842 | * of the path. Since no allocation past the truncation point is possible |
| 843 | * until ext2_truncate() finishes, we may safely do the latter, but top |
| 844 | * of branch may require special attention - pageout below the truncation |
| 845 | * point might try to populate it. |
| 846 | * |
| 847 | * We atomically detach the top of branch from the tree, store the block |
| 848 | * number of its root in *@top, pointers to buffer_heads of partially |
| 849 | * truncated blocks - in @chain[].bh and pointers to their last elements |
| 850 | * that should not be removed - in @chain[].p. Return value is the pointer |
| 851 | * to last filled element of @chain. |
| 852 | * |
| 853 | * The work left to caller to do the actual freeing of subtrees: |
| 854 | * a) free the subtree starting from *@top |
| 855 | * b) free the subtrees whose roots are stored in |
| 856 | * (@chain[i].p+1 .. end of @chain[i].bh->b_data) |
| 857 | * c) free the subtrees growing from the inode past the @chain[0].p |
| 858 | * (no partially truncated stuff there). |
| 859 | */ |
| 860 | |
| 861 | static Indirect *ext2_find_shared(struct inode *inode, |
| 862 | int depth, |
| 863 | int offsets[4], |
| 864 | Indirect chain[4], |
| 865 | __le32 *top) |
| 866 | { |
| 867 | Indirect *partial, *p; |
| 868 | int k, err; |
| 869 | |
| 870 | *top = 0; |
| 871 | for (k = depth; k > 1 && !offsets[k-1]; k--) |
| 872 | ; |
| 873 | partial = ext2_get_branch(inode, k, offsets, chain, &err); |
| 874 | if (!partial) |
| 875 | partial = chain + k-1; |
| 876 | /* |
| 877 | * If the branch acquired continuation since we've looked at it - |
| 878 | * fine, it should all survive and (new) top doesn't belong to us. |
| 879 | */ |
| 880 | write_lock(&EXT2_I(inode)->i_meta_lock); |
| 881 | if (!partial->key && *partial->p) { |
| 882 | write_unlock(&EXT2_I(inode)->i_meta_lock); |
| 883 | goto no_top; |
| 884 | } |
| 885 | for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--) |
| 886 | ; |
| 887 | /* |
| 888 | * OK, we've found the last block that must survive. The rest of our |
| 889 | * branch should be detached before unlocking. However, if that rest |
| 890 | * of branch is all ours and does not grow immediately from the inode |
| 891 | * it's easier to cheat and just decrement partial->p. |
| 892 | */ |
| 893 | if (p == chain + k - 1 && p > chain) { |
| 894 | p->p--; |
| 895 | } else { |
| 896 | *top = *p->p; |
| 897 | *p->p = 0; |
| 898 | } |
| 899 | write_unlock(&EXT2_I(inode)->i_meta_lock); |
| 900 | |
| 901 | while(partial > p) |
| 902 | { |
| 903 | brelse(partial->bh); |
| 904 | partial--; |
| 905 | } |
| 906 | no_top: |
| 907 | return partial; |
| 908 | } |
| 909 | |
| 910 | /** |
| 911 | * ext2_free_data - free a list of data blocks |
| 912 | * @inode: inode we are dealing with |
| 913 | * @p: array of block numbers |
| 914 | * @q: points immediately past the end of array |
| 915 | * |
| 916 | * We are freeing all blocks refered from that array (numbers are |
| 917 | * stored as little-endian 32-bit) and updating @inode->i_blocks |
| 918 | * appropriately. |
| 919 | */ |
| 920 | static inline void ext2_free_data(struct inode *inode, __le32 *p, __le32 *q) |
| 921 | { |
| 922 | unsigned long block_to_free = 0, count = 0; |
| 923 | unsigned long nr; |
| 924 | |
| 925 | for ( ; p < q ; p++) { |
| 926 | nr = le32_to_cpu(*p); |
| 927 | if (nr) { |
| 928 | *p = 0; |
| 929 | /* accumulate blocks to free if they're contiguous */ |
| 930 | if (count == 0) |
| 931 | goto free_this; |
| 932 | else if (block_to_free == nr - count) |
| 933 | count++; |
| 934 | else { |
| 935 | mark_inode_dirty(inode); |
| 936 | ext2_free_blocks (inode, block_to_free, count); |
| 937 | free_this: |
| 938 | block_to_free = nr; |
| 939 | count = 1; |
| 940 | } |
| 941 | } |
| 942 | } |
| 943 | if (count > 0) { |
| 944 | mark_inode_dirty(inode); |
| 945 | ext2_free_blocks (inode, block_to_free, count); |
| 946 | } |
| 947 | } |
| 948 | |
| 949 | /** |
| 950 | * ext2_free_branches - free an array of branches |
| 951 | * @inode: inode we are dealing with |
| 952 | * @p: array of block numbers |
| 953 | * @q: pointer immediately past the end of array |
| 954 | * @depth: depth of the branches to free |
| 955 | * |
| 956 | * We are freeing all blocks refered from these branches (numbers are |
| 957 | * stored as little-endian 32-bit) and updating @inode->i_blocks |
| 958 | * appropriately. |
| 959 | */ |
| 960 | static void ext2_free_branches(struct inode *inode, __le32 *p, __le32 *q, int depth) |
| 961 | { |
| 962 | struct buffer_head * bh; |
| 963 | unsigned long nr; |
| 964 | |
| 965 | if (depth--) { |
| 966 | int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb); |
| 967 | for ( ; p < q ; p++) { |
| 968 | nr = le32_to_cpu(*p); |
| 969 | if (!nr) |
| 970 | continue; |
| 971 | *p = 0; |
| 972 | bh = sb_bread(inode->i_sb, nr); |
| 973 | /* |
| 974 | * A read failure? Report error and clear slot |
| 975 | * (should be rare). |
| 976 | */ |
| 977 | if (!bh) { |
| 978 | ext2_error(inode->i_sb, "ext2_free_branches", |
| 979 | "Read failure, inode=%ld, block=%ld", |
| 980 | inode->i_ino, nr); |
| 981 | continue; |
| 982 | } |
| 983 | ext2_free_branches(inode, |
| 984 | (__le32*)bh->b_data, |
| 985 | (__le32*)bh->b_data + addr_per_block, |
| 986 | depth); |
| 987 | bforget(bh); |
| 988 | ext2_free_blocks(inode, nr, 1); |
| 989 | mark_inode_dirty(inode); |
| 990 | } |
| 991 | } else |
| 992 | ext2_free_data(inode, p, q); |
| 993 | } |
| 994 | |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 995 | void ext2_truncate(struct inode *inode) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 996 | { |
| 997 | __le32 *i_data = EXT2_I(inode)->i_data; |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 998 | struct ext2_inode_info *ei = EXT2_I(inode); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 999 | int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb); |
| 1000 | int offsets[4]; |
| 1001 | Indirect chain[4]; |
| 1002 | Indirect *partial; |
| 1003 | __le32 nr = 0; |
| 1004 | int n; |
| 1005 | long iblock; |
| 1006 | unsigned blocksize; |
| 1007 | |
| 1008 | if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || |
| 1009 | S_ISLNK(inode->i_mode))) |
| 1010 | return; |
| 1011 | if (ext2_inode_is_fast_symlink(inode)) |
| 1012 | return; |
| 1013 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) |
| 1014 | return; |
| 1015 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1016 | blocksize = inode->i_sb->s_blocksize; |
| 1017 | iblock = (inode->i_size + blocksize-1) |
| 1018 | >> EXT2_BLOCK_SIZE_BITS(inode->i_sb); |
| 1019 | |
Carsten Otte | 6d79125 | 2005-06-23 22:05:26 -0700 | [diff] [blame] | 1020 | if (mapping_is_xip(inode->i_mapping)) |
| 1021 | xip_truncate_page(inode->i_mapping, inode->i_size); |
| 1022 | else if (test_opt(inode->i_sb, NOBH)) |
Nick Piggin | 03158cd | 2007-10-16 01:25:25 -0700 | [diff] [blame] | 1023 | nobh_truncate_page(inode->i_mapping, |
| 1024 | inode->i_size, ext2_get_block); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1025 | else |
| 1026 | block_truncate_page(inode->i_mapping, |
| 1027 | inode->i_size, ext2_get_block); |
| 1028 | |
| 1029 | n = ext2_block_to_path(inode, iblock, offsets, NULL); |
| 1030 | if (n == 0) |
| 1031 | return; |
| 1032 | |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 1033 | /* |
| 1034 | * From here we block out all ext2_get_block() callers who want to |
| 1035 | * modify the block allocation tree. |
| 1036 | */ |
| 1037 | mutex_lock(&ei->truncate_mutex); |
| 1038 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1039 | if (n == 1) { |
| 1040 | ext2_free_data(inode, i_data+offsets[0], |
| 1041 | i_data + EXT2_NDIR_BLOCKS); |
| 1042 | goto do_indirects; |
| 1043 | } |
| 1044 | |
| 1045 | partial = ext2_find_shared(inode, n, offsets, chain, &nr); |
| 1046 | /* Kill the top of shared branch (already detached) */ |
| 1047 | if (nr) { |
| 1048 | if (partial == chain) |
| 1049 | mark_inode_dirty(inode); |
| 1050 | else |
| 1051 | mark_buffer_dirty_inode(partial->bh, inode); |
| 1052 | ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial); |
| 1053 | } |
| 1054 | /* Clear the ends of indirect blocks on the shared branch */ |
| 1055 | while (partial > chain) { |
| 1056 | ext2_free_branches(inode, |
| 1057 | partial->p + 1, |
| 1058 | (__le32*)partial->bh->b_data+addr_per_block, |
| 1059 | (chain+n-1) - partial); |
| 1060 | mark_buffer_dirty_inode(partial->bh, inode); |
| 1061 | brelse (partial->bh); |
| 1062 | partial--; |
| 1063 | } |
| 1064 | do_indirects: |
| 1065 | /* Kill the remaining (whole) subtrees */ |
| 1066 | switch (offsets[0]) { |
| 1067 | default: |
| 1068 | nr = i_data[EXT2_IND_BLOCK]; |
| 1069 | if (nr) { |
| 1070 | i_data[EXT2_IND_BLOCK] = 0; |
| 1071 | mark_inode_dirty(inode); |
| 1072 | ext2_free_branches(inode, &nr, &nr+1, 1); |
| 1073 | } |
| 1074 | case EXT2_IND_BLOCK: |
| 1075 | nr = i_data[EXT2_DIND_BLOCK]; |
| 1076 | if (nr) { |
| 1077 | i_data[EXT2_DIND_BLOCK] = 0; |
| 1078 | mark_inode_dirty(inode); |
| 1079 | ext2_free_branches(inode, &nr, &nr+1, 2); |
| 1080 | } |
| 1081 | case EXT2_DIND_BLOCK: |
| 1082 | nr = i_data[EXT2_TIND_BLOCK]; |
| 1083 | if (nr) { |
| 1084 | i_data[EXT2_TIND_BLOCK] = 0; |
| 1085 | mark_inode_dirty(inode); |
| 1086 | ext2_free_branches(inode, &nr, &nr+1, 3); |
| 1087 | } |
| 1088 | case EXT2_TIND_BLOCK: |
| 1089 | ; |
| 1090 | } |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 1091 | |
| 1092 | ext2_discard_reservation(inode); |
| 1093 | |
| 1094 | mutex_unlock(&ei->truncate_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1095 | inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC; |
| 1096 | if (inode_needs_sync(inode)) { |
| 1097 | sync_mapping_buffers(inode->i_mapping); |
| 1098 | ext2_sync_inode (inode); |
| 1099 | } else { |
| 1100 | mark_inode_dirty(inode); |
| 1101 | } |
| 1102 | } |
| 1103 | |
| 1104 | static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino, |
| 1105 | struct buffer_head **p) |
| 1106 | { |
| 1107 | struct buffer_head * bh; |
| 1108 | unsigned long block_group; |
| 1109 | unsigned long block; |
| 1110 | unsigned long offset; |
| 1111 | struct ext2_group_desc * gdp; |
| 1112 | |
| 1113 | *p = NULL; |
| 1114 | if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) || |
| 1115 | ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count)) |
| 1116 | goto Einval; |
| 1117 | |
| 1118 | block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb); |
Eric Sandeen | ef2fb67 | 2007-10-16 23:26:30 -0700 | [diff] [blame] | 1119 | gdp = ext2_get_group_desc(sb, block_group, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1120 | if (!gdp) |
| 1121 | goto Egdp; |
| 1122 | /* |
| 1123 | * Figure out the offset within the block group inode table |
| 1124 | */ |
| 1125 | offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb); |
| 1126 | block = le32_to_cpu(gdp->bg_inode_table) + |
| 1127 | (offset >> EXT2_BLOCK_SIZE_BITS(sb)); |
| 1128 | if (!(bh = sb_bread(sb, block))) |
| 1129 | goto Eio; |
| 1130 | |
| 1131 | *p = bh; |
| 1132 | offset &= (EXT2_BLOCK_SIZE(sb) - 1); |
| 1133 | return (struct ext2_inode *) (bh->b_data + offset); |
| 1134 | |
| 1135 | Einval: |
| 1136 | ext2_error(sb, "ext2_get_inode", "bad inode number: %lu", |
| 1137 | (unsigned long) ino); |
| 1138 | return ERR_PTR(-EINVAL); |
| 1139 | Eio: |
| 1140 | ext2_error(sb, "ext2_get_inode", |
| 1141 | "unable to read inode block - inode=%lu, block=%lu", |
| 1142 | (unsigned long) ino, block); |
| 1143 | Egdp: |
| 1144 | return ERR_PTR(-EIO); |
| 1145 | } |
| 1146 | |
| 1147 | void ext2_set_inode_flags(struct inode *inode) |
| 1148 | { |
| 1149 | unsigned int flags = EXT2_I(inode)->i_flags; |
| 1150 | |
| 1151 | inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); |
| 1152 | if (flags & EXT2_SYNC_FL) |
| 1153 | inode->i_flags |= S_SYNC; |
| 1154 | if (flags & EXT2_APPEND_FL) |
| 1155 | inode->i_flags |= S_APPEND; |
| 1156 | if (flags & EXT2_IMMUTABLE_FL) |
| 1157 | inode->i_flags |= S_IMMUTABLE; |
| 1158 | if (flags & EXT2_NOATIME_FL) |
| 1159 | inode->i_flags |= S_NOATIME; |
| 1160 | if (flags & EXT2_DIRSYNC_FL) |
| 1161 | inode->i_flags |= S_DIRSYNC; |
| 1162 | } |
| 1163 | |
Jan Kara | 4f99ed6 | 2007-05-08 00:31:04 -0700 | [diff] [blame] | 1164 | /* Propagate flags from i_flags to EXT2_I(inode)->i_flags */ |
| 1165 | void ext2_get_inode_flags(struct ext2_inode_info *ei) |
| 1166 | { |
| 1167 | unsigned int flags = ei->vfs_inode.i_flags; |
| 1168 | |
| 1169 | ei->i_flags &= ~(EXT2_SYNC_FL|EXT2_APPEND_FL| |
| 1170 | EXT2_IMMUTABLE_FL|EXT2_NOATIME_FL|EXT2_DIRSYNC_FL); |
| 1171 | if (flags & S_SYNC) |
| 1172 | ei->i_flags |= EXT2_SYNC_FL; |
| 1173 | if (flags & S_APPEND) |
| 1174 | ei->i_flags |= EXT2_APPEND_FL; |
| 1175 | if (flags & S_IMMUTABLE) |
| 1176 | ei->i_flags |= EXT2_IMMUTABLE_FL; |
| 1177 | if (flags & S_NOATIME) |
| 1178 | ei->i_flags |= EXT2_NOATIME_FL; |
| 1179 | if (flags & S_DIRSYNC) |
| 1180 | ei->i_flags |= EXT2_DIRSYNC_FL; |
| 1181 | } |
| 1182 | |
David Howells | 52fcf70 | 2008-02-07 00:15:35 -0800 | [diff] [blame] | 1183 | struct inode *ext2_iget (struct super_block *sb, unsigned long ino) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1184 | { |
David Howells | 52fcf70 | 2008-02-07 00:15:35 -0800 | [diff] [blame] | 1185 | struct ext2_inode_info *ei; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1186 | struct buffer_head * bh; |
David Howells | 52fcf70 | 2008-02-07 00:15:35 -0800 | [diff] [blame] | 1187 | struct ext2_inode *raw_inode; |
| 1188 | struct inode *inode; |
| 1189 | long ret = -EIO; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1190 | int n; |
| 1191 | |
David Howells | 52fcf70 | 2008-02-07 00:15:35 -0800 | [diff] [blame] | 1192 | inode = iget_locked(sb, ino); |
| 1193 | if (!inode) |
| 1194 | return ERR_PTR(-ENOMEM); |
| 1195 | if (!(inode->i_state & I_NEW)) |
| 1196 | return inode; |
| 1197 | |
| 1198 | ei = EXT2_I(inode); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1199 | #ifdef CONFIG_EXT2_FS_POSIX_ACL |
| 1200 | ei->i_acl = EXT2_ACL_NOT_CACHED; |
| 1201 | ei->i_default_acl = EXT2_ACL_NOT_CACHED; |
| 1202 | #endif |
Martin J. Bligh | a686cd8 | 2007-10-16 23:30:46 -0700 | [diff] [blame] | 1203 | ei->i_block_alloc_info = NULL; |
| 1204 | |
David Howells | 52fcf70 | 2008-02-07 00:15:35 -0800 | [diff] [blame] | 1205 | raw_inode = ext2_get_inode(inode->i_sb, ino, &bh); |
| 1206 | if (IS_ERR(raw_inode)) { |
| 1207 | ret = PTR_ERR(raw_inode); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1208 | goto bad_inode; |
David Howells | 52fcf70 | 2008-02-07 00:15:35 -0800 | [diff] [blame] | 1209 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1210 | |
| 1211 | inode->i_mode = le16_to_cpu(raw_inode->i_mode); |
| 1212 | inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low); |
| 1213 | inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low); |
| 1214 | if (!(test_opt (inode->i_sb, NO_UID32))) { |
| 1215 | inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16; |
| 1216 | inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16; |
| 1217 | } |
| 1218 | inode->i_nlink = le16_to_cpu(raw_inode->i_links_count); |
| 1219 | inode->i_size = le32_to_cpu(raw_inode->i_size); |
Markus Rechberger | 4d7bf11 | 2007-05-08 00:23:39 -0700 | [diff] [blame] | 1220 | inode->i_atime.tv_sec = (signed)le32_to_cpu(raw_inode->i_atime); |
| 1221 | inode->i_ctime.tv_sec = (signed)le32_to_cpu(raw_inode->i_ctime); |
| 1222 | inode->i_mtime.tv_sec = (signed)le32_to_cpu(raw_inode->i_mtime); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1223 | inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0; |
| 1224 | ei->i_dtime = le32_to_cpu(raw_inode->i_dtime); |
| 1225 | /* We now have enough fields to check if the inode was active or not. |
| 1226 | * This is needed because nfsd might try to access dead inodes |
| 1227 | * the test is that same one that e2fsck uses |
| 1228 | * NeilBrown 1999oct15 |
| 1229 | */ |
| 1230 | if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) { |
| 1231 | /* this inode is deleted */ |
| 1232 | brelse (bh); |
David Howells | 52fcf70 | 2008-02-07 00:15:35 -0800 | [diff] [blame] | 1233 | ret = -ESTALE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1234 | goto bad_inode; |
| 1235 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1236 | inode->i_blocks = le32_to_cpu(raw_inode->i_blocks); |
| 1237 | ei->i_flags = le32_to_cpu(raw_inode->i_flags); |
| 1238 | ei->i_faddr = le32_to_cpu(raw_inode->i_faddr); |
| 1239 | ei->i_frag_no = raw_inode->i_frag; |
| 1240 | ei->i_frag_size = raw_inode->i_fsize; |
| 1241 | ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl); |
| 1242 | ei->i_dir_acl = 0; |
| 1243 | if (S_ISREG(inode->i_mode)) |
| 1244 | inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32; |
| 1245 | else |
| 1246 | ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl); |
| 1247 | ei->i_dtime = 0; |
| 1248 | inode->i_generation = le32_to_cpu(raw_inode->i_generation); |
| 1249 | ei->i_state = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1250 | ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb); |
| 1251 | ei->i_dir_start_lookup = 0; |
| 1252 | |
| 1253 | /* |
| 1254 | * NOTE! The in-memory inode i_data array is in little-endian order |
| 1255 | * even on big-endian machines: we do NOT byteswap the block numbers! |
| 1256 | */ |
| 1257 | for (n = 0; n < EXT2_N_BLOCKS; n++) |
| 1258 | ei->i_data[n] = raw_inode->i_block[n]; |
| 1259 | |
| 1260 | if (S_ISREG(inode->i_mode)) { |
| 1261 | inode->i_op = &ext2_file_inode_operations; |
Carsten Otte | 6d79125 | 2005-06-23 22:05:26 -0700 | [diff] [blame] | 1262 | if (ext2_use_xip(inode->i_sb)) { |
| 1263 | inode->i_mapping->a_ops = &ext2_aops_xip; |
| 1264 | inode->i_fop = &ext2_xip_file_operations; |
| 1265 | } else if (test_opt(inode->i_sb, NOBH)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1266 | inode->i_mapping->a_ops = &ext2_nobh_aops; |
Carsten Otte | 6d79125 | 2005-06-23 22:05:26 -0700 | [diff] [blame] | 1267 | inode->i_fop = &ext2_file_operations; |
| 1268 | } else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1269 | inode->i_mapping->a_ops = &ext2_aops; |
Carsten Otte | 6d79125 | 2005-06-23 22:05:26 -0700 | [diff] [blame] | 1270 | inode->i_fop = &ext2_file_operations; |
| 1271 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1272 | } else if (S_ISDIR(inode->i_mode)) { |
| 1273 | inode->i_op = &ext2_dir_inode_operations; |
| 1274 | inode->i_fop = &ext2_dir_operations; |
| 1275 | if (test_opt(inode->i_sb, NOBH)) |
| 1276 | inode->i_mapping->a_ops = &ext2_nobh_aops; |
| 1277 | else |
| 1278 | inode->i_mapping->a_ops = &ext2_aops; |
| 1279 | } else if (S_ISLNK(inode->i_mode)) { |
| 1280 | if (ext2_inode_is_fast_symlink(inode)) |
| 1281 | inode->i_op = &ext2_fast_symlink_inode_operations; |
| 1282 | else { |
| 1283 | inode->i_op = &ext2_symlink_inode_operations; |
| 1284 | if (test_opt(inode->i_sb, NOBH)) |
| 1285 | inode->i_mapping->a_ops = &ext2_nobh_aops; |
| 1286 | else |
| 1287 | inode->i_mapping->a_ops = &ext2_aops; |
| 1288 | } |
| 1289 | } else { |
| 1290 | inode->i_op = &ext2_special_inode_operations; |
| 1291 | if (raw_inode->i_block[0]) |
| 1292 | init_special_inode(inode, inode->i_mode, |
| 1293 | old_decode_dev(le32_to_cpu(raw_inode->i_block[0]))); |
| 1294 | else |
| 1295 | init_special_inode(inode, inode->i_mode, |
| 1296 | new_decode_dev(le32_to_cpu(raw_inode->i_block[1]))); |
| 1297 | } |
| 1298 | brelse (bh); |
| 1299 | ext2_set_inode_flags(inode); |
David Howells | 52fcf70 | 2008-02-07 00:15:35 -0800 | [diff] [blame] | 1300 | unlock_new_inode(inode); |
| 1301 | return inode; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1302 | |
| 1303 | bad_inode: |
David Howells | 52fcf70 | 2008-02-07 00:15:35 -0800 | [diff] [blame] | 1304 | iget_failed(inode); |
| 1305 | return ERR_PTR(ret); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1306 | } |
| 1307 | |
| 1308 | static int ext2_update_inode(struct inode * inode, int do_sync) |
| 1309 | { |
| 1310 | struct ext2_inode_info *ei = EXT2_I(inode); |
| 1311 | struct super_block *sb = inode->i_sb; |
| 1312 | ino_t ino = inode->i_ino; |
| 1313 | uid_t uid = inode->i_uid; |
| 1314 | gid_t gid = inode->i_gid; |
| 1315 | struct buffer_head * bh; |
| 1316 | struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh); |
| 1317 | int n; |
| 1318 | int err = 0; |
| 1319 | |
| 1320 | if (IS_ERR(raw_inode)) |
| 1321 | return -EIO; |
| 1322 | |
| 1323 | /* For fields not not tracking in the in-memory inode, |
| 1324 | * initialise them to zero for new inodes. */ |
| 1325 | if (ei->i_state & EXT2_STATE_NEW) |
| 1326 | memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size); |
| 1327 | |
Jan Kara | 4f99ed6 | 2007-05-08 00:31:04 -0700 | [diff] [blame] | 1328 | ext2_get_inode_flags(ei); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1329 | raw_inode->i_mode = cpu_to_le16(inode->i_mode); |
| 1330 | if (!(test_opt(sb, NO_UID32))) { |
| 1331 | raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid)); |
| 1332 | raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid)); |
| 1333 | /* |
| 1334 | * Fix up interoperability with old kernels. Otherwise, old inodes get |
| 1335 | * re-used with the upper 16 bits of the uid/gid intact |
| 1336 | */ |
| 1337 | if (!ei->i_dtime) { |
| 1338 | raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid)); |
| 1339 | raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid)); |
| 1340 | } else { |
| 1341 | raw_inode->i_uid_high = 0; |
| 1342 | raw_inode->i_gid_high = 0; |
| 1343 | } |
| 1344 | } else { |
| 1345 | raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid)); |
| 1346 | raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid)); |
| 1347 | raw_inode->i_uid_high = 0; |
| 1348 | raw_inode->i_gid_high = 0; |
| 1349 | } |
| 1350 | raw_inode->i_links_count = cpu_to_le16(inode->i_nlink); |
| 1351 | raw_inode->i_size = cpu_to_le32(inode->i_size); |
| 1352 | raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec); |
| 1353 | raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec); |
| 1354 | raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec); |
| 1355 | |
| 1356 | raw_inode->i_blocks = cpu_to_le32(inode->i_blocks); |
| 1357 | raw_inode->i_dtime = cpu_to_le32(ei->i_dtime); |
| 1358 | raw_inode->i_flags = cpu_to_le32(ei->i_flags); |
| 1359 | raw_inode->i_faddr = cpu_to_le32(ei->i_faddr); |
| 1360 | raw_inode->i_frag = ei->i_frag_no; |
| 1361 | raw_inode->i_fsize = ei->i_frag_size; |
| 1362 | raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl); |
| 1363 | if (!S_ISREG(inode->i_mode)) |
| 1364 | raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl); |
| 1365 | else { |
| 1366 | raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32); |
| 1367 | if (inode->i_size > 0x7fffffffULL) { |
| 1368 | if (!EXT2_HAS_RO_COMPAT_FEATURE(sb, |
| 1369 | EXT2_FEATURE_RO_COMPAT_LARGE_FILE) || |
| 1370 | EXT2_SB(sb)->s_es->s_rev_level == |
| 1371 | cpu_to_le32(EXT2_GOOD_OLD_REV)) { |
| 1372 | /* If this is the first large file |
| 1373 | * created, add a flag to the superblock. |
| 1374 | */ |
| 1375 | lock_kernel(); |
| 1376 | ext2_update_dynamic_rev(sb); |
| 1377 | EXT2_SET_RO_COMPAT_FEATURE(sb, |
| 1378 | EXT2_FEATURE_RO_COMPAT_LARGE_FILE); |
| 1379 | unlock_kernel(); |
| 1380 | ext2_write_super(sb); |
| 1381 | } |
| 1382 | } |
| 1383 | } |
| 1384 | |
| 1385 | raw_inode->i_generation = cpu_to_le32(inode->i_generation); |
| 1386 | if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) { |
| 1387 | if (old_valid_dev(inode->i_rdev)) { |
| 1388 | raw_inode->i_block[0] = |
| 1389 | cpu_to_le32(old_encode_dev(inode->i_rdev)); |
| 1390 | raw_inode->i_block[1] = 0; |
| 1391 | } else { |
| 1392 | raw_inode->i_block[0] = 0; |
| 1393 | raw_inode->i_block[1] = |
| 1394 | cpu_to_le32(new_encode_dev(inode->i_rdev)); |
| 1395 | raw_inode->i_block[2] = 0; |
| 1396 | } |
| 1397 | } else for (n = 0; n < EXT2_N_BLOCKS; n++) |
| 1398 | raw_inode->i_block[n] = ei->i_data[n]; |
| 1399 | mark_buffer_dirty(bh); |
| 1400 | if (do_sync) { |
| 1401 | sync_dirty_buffer(bh); |
| 1402 | if (buffer_req(bh) && !buffer_uptodate(bh)) { |
| 1403 | printk ("IO error syncing ext2 inode [%s:%08lx]\n", |
| 1404 | sb->s_id, (unsigned long) ino); |
| 1405 | err = -EIO; |
| 1406 | } |
| 1407 | } |
| 1408 | ei->i_state &= ~EXT2_STATE_NEW; |
| 1409 | brelse (bh); |
| 1410 | return err; |
| 1411 | } |
| 1412 | |
| 1413 | int ext2_write_inode(struct inode *inode, int wait) |
| 1414 | { |
| 1415 | return ext2_update_inode(inode, wait); |
| 1416 | } |
| 1417 | |
| 1418 | int ext2_sync_inode(struct inode *inode) |
| 1419 | { |
| 1420 | struct writeback_control wbc = { |
| 1421 | .sync_mode = WB_SYNC_ALL, |
| 1422 | .nr_to_write = 0, /* sys_fsync did this */ |
| 1423 | }; |
| 1424 | return sync_inode(inode, &wbc); |
| 1425 | } |
| 1426 | |
| 1427 | int ext2_setattr(struct dentry *dentry, struct iattr *iattr) |
| 1428 | { |
| 1429 | struct inode *inode = dentry->d_inode; |
| 1430 | int error; |
| 1431 | |
| 1432 | error = inode_change_ok(inode, iattr); |
| 1433 | if (error) |
| 1434 | return error; |
| 1435 | if ((iattr->ia_valid & ATTR_UID && iattr->ia_uid != inode->i_uid) || |
| 1436 | (iattr->ia_valid & ATTR_GID && iattr->ia_gid != inode->i_gid)) { |
| 1437 | error = DQUOT_TRANSFER(inode, iattr) ? -EDQUOT : 0; |
| 1438 | if (error) |
| 1439 | return error; |
| 1440 | } |
| 1441 | error = inode_setattr(inode, iattr); |
| 1442 | if (!error && (iattr->ia_valid & ATTR_MODE)) |
| 1443 | error = ext2_acl_chmod(inode); |
| 1444 | return error; |
| 1445 | } |