blob: b1ab32ab5a779381b1b46d29c61bd5fe2eddac44 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 Otte6d791252005-06-23 22:05:26 -070036#include "xip.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
38MODULE_AUTHOR("Remy Card and others");
39MODULE_DESCRIPTION("Second Extended Filesystem");
40MODULE_LICENSE("GPL");
41
42static int ext2_update_inode(struct inode * inode, int do_sync);
43
44/*
45 * Test whether an inode is a fast symlink.
46 */
47static 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 Torvalds1da177e2005-04-16 15:20:36 -070057 * Called at the last iput() if i_nlink is zero.
58 */
59void ext2_delete_inode (struct inode * inode)
60{
Mark Fashehfef26652005-09-09 13:01:31 -070061 truncate_inode_pages(&inode->i_data, 0);
62
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 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;
75no_delete:
76 clear_inode(inode); /* We must guarantee clearing of inode... */
77}
78
Linus Torvalds1da177e2005-04-16 15:20:36 -070079typedef struct {
80 __le32 *p;
81 __le32 key;
82 struct buffer_head *bh;
83} Indirect;
84
85static 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
91static 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
128static 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. Bligha686cd82007-10-16 23:30:46 -0700163 *boundary = final - 1 - (i_block & (ptrs - 1));
164
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 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 */
197static 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
226changed:
227 read_unlock(&EXT2_I(inode)->i_meta_lock);
228 brelse(bh);
229 *err = -EAGAIN;
230 goto no_block;
231failure:
232 *err = -EIO;
233no_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 *
242 * This function returns the prefered place for block allocation.
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
257static 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 */
278 bg_start = (ei->i_block_group * EXT2_BLOCKS_PER_GROUP(inode->i_sb)) +
279 le32_to_cpu(EXT2_SB(inode->i_sb)->s_es->s_first_data_block);
280 colour = (current->pid % 16) *
281 (EXT2_BLOCKS_PER_GROUP(inode->i_sb) / 16);
282 return bg_start + colour;
283}
284
285/**
286 * ext2_find_goal - find a prefered place for allocation.
287 * @inode: owner
288 * @block: block we want
289 * @chain: chain of indirect blocks
290 * @partial: pointer to the last triple within a chain
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 *
Martin J. Bligha686cd82007-10-16 23:30:46 -0700292 * Returns preferred place for a block (the goal).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 */
294
295static inline int ext2_find_goal(struct inode *inode,
296 long block,
297 Indirect chain[4],
Martin J. Bligha686cd82007-10-16 23:30:46 -0700298 Indirect *partial)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299{
Martin J. Bligha686cd82007-10-16 23:30:46 -0700300 struct ext2_block_alloc_info *block_i;
301
302 block_i = EXT2_I(inode)->i_block_alloc_info;
303
304 /*
305 * try the heuristic for sequential allocation,
306 * failing that at least try to get decent locality.
307 */
308 if (block_i && (block == block_i->last_alloc_logical_block + 1)
309 && (block_i->last_alloc_physical_block != 0)) {
310 return block_i->last_alloc_physical_block + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 }
Martin J. Bligha686cd82007-10-16 23:30:46 -0700312
313 return ext2_find_near(inode, partial);
314}
315
316/**
317 * ext2_blks_to_allocate: Look up the block map and count the number
318 * of direct blocks need to be allocated for the given branch.
319 *
320 * @branch: chain of indirect blocks
321 * @k: number of blocks need for indirect blocks
322 * @blks: number of data blocks to be mapped.
323 * @blocks_to_boundary: the offset in the indirect block
324 *
325 * return the total number of blocks to be allocate, including the
326 * direct and indirect blocks.
327 */
328static int
329ext2_blks_to_allocate(Indirect * branch, int k, unsigned long blks,
330 int blocks_to_boundary)
331{
332 unsigned long count = 0;
333
334 /*
335 * Simple case, [t,d]Indirect block(s) has not allocated yet
336 * then it's clear blocks on that path have not allocated
337 */
338 if (k > 0) {
339 /* right now don't hanel cross boundary allocation */
340 if (blks < blocks_to_boundary + 1)
341 count += blks;
342 else
343 count += blocks_to_boundary + 1;
344 return count;
345 }
346
347 count++;
348 while (count < blks && count <= blocks_to_boundary
349 && le32_to_cpu(*(branch[0].p + count)) == 0) {
350 count++;
351 }
352 return count;
353}
354
355/**
356 * ext2_alloc_blocks: multiple allocate blocks needed for a branch
357 * @indirect_blks: the number of blocks need to allocate for indirect
358 * blocks
359 *
360 * @new_blocks: on return it will store the new block numbers for
361 * the indirect blocks(if needed) and the first direct block,
362 * @blks: on return it will store the total number of allocated
363 * direct blocks
364 */
365static int ext2_alloc_blocks(struct inode *inode,
366 ext2_fsblk_t goal, int indirect_blks, int blks,
367 ext2_fsblk_t new_blocks[4], int *err)
368{
369 int target, i;
370 unsigned long count = 0;
371 int index = 0;
372 ext2_fsblk_t current_block = 0;
373 int ret = 0;
374
375 /*
376 * Here we try to allocate the requested multiple blocks at once,
377 * on a best-effort basis.
378 * To build a branch, we should allocate blocks for
379 * the indirect blocks(if not allocated yet), and at least
380 * the first direct block of this branch. That's the
381 * minimum number of blocks need to allocate(required)
382 */
383 target = blks + indirect_blks;
384
385 while (1) {
386 count = target;
387 /* allocating blocks for indirect blocks and direct blocks */
388 current_block = ext2_new_blocks(inode,goal,&count,err);
389 if (*err)
390 goto failed_out;
391
392 target -= count;
393 /* allocate blocks for indirect blocks */
394 while (index < indirect_blks && count) {
395 new_blocks[index++] = current_block++;
396 count--;
397 }
398
399 if (count > 0)
400 break;
401 }
402
403 /* save the new block number for the first direct block */
404 new_blocks[index] = current_block;
405
406 /* total number of blocks allocated for direct blocks */
407 ret = count;
408 *err = 0;
409 return ret;
410failed_out:
411 for (i = 0; i <index; i++)
412 ext2_free_blocks(inode, new_blocks[i], 1);
413 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414}
415
416/**
417 * ext2_alloc_branch - allocate and set up a chain of blocks.
418 * @inode: owner
419 * @num: depth of the chain (number of blocks to allocate)
420 * @offsets: offsets (in the blocks) to store the pointers to next.
421 * @branch: place to store the chain in.
422 *
423 * This function allocates @num blocks, zeroes out all but the last one,
424 * links them into chain and (if we are synchronous) writes them to disk.
425 * In other words, it prepares a branch that can be spliced onto the
426 * inode. It stores the information about that chain in the branch[], in
427 * the same format as ext2_get_branch() would do. We are calling it after
428 * we had read the existing part of chain and partial points to the last
429 * triple of that (one with zero ->key). Upon the exit we have the same
430 * picture as after the successful ext2_get_block(), excpet that in one
431 * place chain is disconnected - *branch->p is still zero (we did not
432 * set the last link), but branch->key contains the number that should
433 * be placed into *branch->p to fill that gap.
434 *
435 * If allocation fails we free all blocks we've allocated (and forget
436 * their buffer_heads) and return the error value the from failed
437 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
438 * as described above and return 0.
439 */
440
441static int ext2_alloc_branch(struct inode *inode,
Martin J. Bligha686cd82007-10-16 23:30:46 -0700442 int indirect_blks, int *blks, ext2_fsblk_t goal,
443 int *offsets, Indirect *branch)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444{
445 int blocksize = inode->i_sb->s_blocksize;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700446 int i, n = 0;
447 int err = 0;
448 struct buffer_head *bh;
449 int num;
450 ext2_fsblk_t new_blocks[4];
451 ext2_fsblk_t current_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452
Martin J. Bligha686cd82007-10-16 23:30:46 -0700453 num = ext2_alloc_blocks(inode, goal, indirect_blks,
454 *blks, new_blocks, &err);
455 if (err)
456 return err;
457
458 branch[0].key = cpu_to_le32(new_blocks[0]);
459 /*
460 * metadata blocks and data blocks are allocated.
461 */
462 for (n = 1; n <= indirect_blks; n++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 /*
Martin J. Bligha686cd82007-10-16 23:30:46 -0700464 * Get buffer_head for parent block, zero it out
465 * and set the pointer to new one, then send
466 * parent to disk.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700468 bh = sb_getblk(inode->i_sb, new_blocks[n-1]);
469 branch[n].bh = bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 lock_buffer(bh);
471 memset(bh->b_data, 0, blocksize);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 branch[n].p = (__le32 *) bh->b_data + offsets[n];
Martin J. Bligha686cd82007-10-16 23:30:46 -0700473 branch[n].key = cpu_to_le32(new_blocks[n]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 *branch[n].p = branch[n].key;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700475 if ( n == indirect_blks) {
476 current_block = new_blocks[n];
477 /*
478 * End of chain, update the last new metablock of
479 * the chain to point to the new allocated
480 * data blocks numbers
481 */
482 for (i=1; i < num; i++)
483 *(branch[n].p + i) = cpu_to_le32(++current_block);
484 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 set_buffer_uptodate(bh);
486 unlock_buffer(bh);
487 mark_buffer_dirty_inode(bh, inode);
488 /* We used to sync bh here if IS_SYNC(inode).
489 * But we now rely upon generic_osync_inode()
490 * and b_inode_buffers. But not for directories.
491 */
492 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
493 sync_dirty_buffer(bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 }
Martin J. Bligha686cd82007-10-16 23:30:46 -0700495 *blks = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 return err;
497}
498
499/**
Martin J. Bligha686cd82007-10-16 23:30:46 -0700500 * ext2_splice_branch - splice the allocated branch onto inode.
501 * @inode: owner
502 * @block: (logical) number of block we are adding
503 * @chain: chain of indirect blocks (with a missing link - see
504 * ext2_alloc_branch)
505 * @where: location of missing link
506 * @num: number of indirect blocks we are adding
507 * @blks: number of direct blocks we are adding
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 *
Martin J. Bligha686cd82007-10-16 23:30:46 -0700509 * This function fills the missing link and does all housekeeping needed in
510 * inode (->i_blocks, etc.). In case of success we end up with the full
511 * chain to new block and return 0.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700513static void ext2_splice_branch(struct inode *inode,
514 long block, Indirect *where, int num, int blks)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 int i;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700517 struct ext2_block_alloc_info *block_i;
518 ext2_fsblk_t current_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519
Martin J. Bligha686cd82007-10-16 23:30:46 -0700520 block_i = EXT2_I(inode)->i_block_alloc_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521
Martin J. Bligha686cd82007-10-16 23:30:46 -0700522 /* XXX LOCKING probably should have i_meta_lock ?*/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 /* That's it */
524
525 *where->p = where->key;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526
Martin J. Bligha686cd82007-10-16 23:30:46 -0700527 /*
528 * Update the host buffer_head or inode to point to more just allocated
529 * direct blocks blocks
530 */
531 if (num == 0 && blks > 1) {
532 current_block = le32_to_cpu(where->key) + 1;
533 for (i = 1; i < blks; i++)
534 *(where->p + i ) = cpu_to_le32(current_block++);
535 }
536
537 /*
538 * update the most recently allocated logical & physical block
539 * in i_block_alloc_info, to assist find the proper goal block for next
540 * allocation
541 */
542 if (block_i) {
543 block_i->last_alloc_logical_block = block + blks - 1;
544 block_i->last_alloc_physical_block =
545 le32_to_cpu(where[num].key) + blks - 1;
546 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547
548 /* We are done with atomic stuff, now do the rest of housekeeping */
549
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 /* had we spliced it onto indirect block? */
551 if (where->bh)
552 mark_buffer_dirty_inode(where->bh, inode);
553
Martin J. Bligha686cd82007-10-16 23:30:46 -0700554 inode->i_ctime = CURRENT_TIME_SEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 mark_inode_dirty(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556}
557
558/*
559 * Allocation strategy is simple: if we have to allocate something, we will
560 * have to go the whole way to leaf. So let's do it before attaching anything
561 * to tree, set linkage between the newborn blocks, write them if sync is
562 * required, recheck the path, free and repeat if check fails, otherwise
563 * set the last missing link (that will protect us from any truncate-generated
564 * removals - all blocks on the path are immune now) and possibly force the
565 * write on the parent block.
566 * That has a nice additional property: no special recovery from the failed
567 * allocations is needed - we simply release blocks and do not touch anything
568 * reachable from inode.
Martin J. Bligha686cd82007-10-16 23:30:46 -0700569 *
570 * `handle' can be NULL if create == 0.
571 *
572 * The BKL may not be held on entry here. Be sure to take it early.
573 * return > 0, # of blocks mapped or allocated.
574 * return = 0, if plain lookup failed.
575 * return < 0, error case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700577static int ext2_get_blocks(struct inode *inode,
578 sector_t iblock, unsigned long maxblocks,
579 struct buffer_head *bh_result,
580 int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581{
582 int err = -EIO;
583 int offsets[4];
584 Indirect chain[4];
585 Indirect *partial;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700586 ext2_fsblk_t goal;
587 int indirect_blks;
588 int blocks_to_boundary = 0;
589 int depth;
590 struct ext2_inode_info *ei = EXT2_I(inode);
591 int count = 0;
592 ext2_fsblk_t first_block = 0;
593
594 depth = ext2_block_to_path(inode,iblock,offsets,&blocks_to_boundary);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595
596 if (depth == 0)
Martin J. Bligha686cd82007-10-16 23:30:46 -0700597 return (err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598reread:
599 partial = ext2_get_branch(inode, depth, offsets, chain, &err);
600
601 /* Simplest case - block found, no allocation needed */
602 if (!partial) {
Martin J. Bligha686cd82007-10-16 23:30:46 -0700603 first_block = le32_to_cpu(chain[depth - 1].key);
604 clear_buffer_new(bh_result); /* What's this do? */
605 count++;
606 /*map more blocks*/
607 while (count < maxblocks && count <= blocks_to_boundary) {
608 ext2_fsblk_t blk;
609
610 if (!verify_chain(chain, partial)) {
611 /*
612 * Indirect block might be removed by
613 * truncate while we were reading it.
614 * Handling of that case: forget what we've
615 * got now, go to reread.
616 */
617 count = 0;
618 goto changed;
619 }
620 blk = le32_to_cpu(*(chain[depth-1].p + count));
621 if (blk == first_block + count)
622 count++;
623 else
624 break;
625 }
626 goto got_it;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 }
628
629 /* Next simple case - plain lookup or failed read of indirect block */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700630 if (!create || err == -EIO)
631 goto cleanup;
632
633 mutex_lock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
635 /*
Martin J. Bligha686cd82007-10-16 23:30:46 -0700636 * Okay, we need to do block allocation. Lazily initialize the block
637 * allocation info here if necessary
638 */
639 if (S_ISREG(inode->i_mode) && (!ei->i_block_alloc_info))
640 ext2_init_block_alloc_info(inode);
641
642 goal = ext2_find_goal(inode, iblock, chain, partial);
643
644 /* the number of blocks need to allocate for [d,t]indirect blocks */
645 indirect_blks = (chain + depth) - partial - 1;
646 /*
647 * Next look up the indirect map to count the totoal number of
648 * direct blocks to allocate for this branch.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700650 count = ext2_blks_to_allocate(partial, indirect_blks,
651 maxblocks, blocks_to_boundary);
652 /*
653 * XXX ???? Block out ext2_truncate while we alter the tree
654 */
655 err = ext2_alloc_branch(inode, indirect_blks, &count, goal,
656 offsets + (partial - chain), partial);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657
Martin J. Bligha686cd82007-10-16 23:30:46 -0700658 if (err) {
659 mutex_unlock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 goto cleanup;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700661 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662
Carsten Otte6d791252005-06-23 22:05:26 -0700663 if (ext2_use_xip(inode->i_sb)) {
664 /*
665 * we need to clear the block
666 */
667 err = ext2_clear_xip_target (inode,
668 le32_to_cpu(chain[depth-1].key));
Martin J. Bligha686cd82007-10-16 23:30:46 -0700669 if (err) {
670 mutex_unlock(&ei->truncate_mutex);
Carsten Otte6d791252005-06-23 22:05:26 -0700671 goto cleanup;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700672 }
Carsten Otte6d791252005-06-23 22:05:26 -0700673 }
674
Martin J. Bligha686cd82007-10-16 23:30:46 -0700675 ext2_splice_branch(inode, iblock, partial, indirect_blks, count);
676 mutex_unlock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 set_buffer_new(bh_result);
Martin J. Bligha686cd82007-10-16 23:30:46 -0700678got_it:
679 map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key));
680 if (count > blocks_to_boundary)
681 set_buffer_boundary(bh_result);
682 err = count;
683 /* Clean up and exit */
684 partial = chain + depth - 1; /* the whole chain */
685cleanup:
686 while (partial > chain) {
687 brelse(partial->bh);
688 partial--;
689 }
690 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691changed:
692 while (partial > chain) {
693 brelse(partial->bh);
694 partial--;
695 }
696 goto reread;
697}
698
Martin J. Bligha686cd82007-10-16 23:30:46 -0700699int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create)
700{
701 unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;
702 int ret = ext2_get_blocks(inode, iblock, max_blocks,
703 bh_result, create);
704 if (ret > 0) {
705 bh_result->b_size = (ret << inode->i_blkbits);
706 ret = 0;
707 }
708 return ret;
709
710}
711
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712static int ext2_writepage(struct page *page, struct writeback_control *wbc)
713{
714 return block_write_full_page(page, ext2_get_block, wbc);
715}
716
717static int ext2_readpage(struct file *file, struct page *page)
718{
719 return mpage_readpage(page, ext2_get_block);
720}
721
722static int
723ext2_readpages(struct file *file, struct address_space *mapping,
724 struct list_head *pages, unsigned nr_pages)
725{
726 return mpage_readpages(mapping, pages, nr_pages, ext2_get_block);
727}
728
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700729int __ext2_write_begin(struct file *file, struct address_space *mapping,
730 loff_t pos, unsigned len, unsigned flags,
731 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732{
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700733 return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
734 ext2_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735}
736
737static int
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700738ext2_write_begin(struct file *file, struct address_space *mapping,
739 loff_t pos, unsigned len, unsigned flags,
740 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741{
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700742 *pagep = NULL;
743 return __ext2_write_begin(file, mapping, pos, len, flags, pagep,fsdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744}
745
Nick Piggin03158cd2007-10-16 01:25:25 -0700746static int
747ext2_nobh_write_begin(struct file *file, struct address_space *mapping,
748 loff_t pos, unsigned len, unsigned flags,
749 struct page **pagep, void **fsdata)
750{
751 /*
752 * Dir-in-pagecache still uses ext2_write_begin. Would have to rework
753 * directory handling code to pass around offsets rather than struct
754 * pages in order to make this work easily.
755 */
756 return nobh_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
757 ext2_get_block);
758}
759
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760static int ext2_nobh_writepage(struct page *page,
761 struct writeback_control *wbc)
762{
763 return nobh_writepage(page, ext2_get_block, wbc);
764}
765
766static sector_t ext2_bmap(struct address_space *mapping, sector_t block)
767{
768 return generic_block_bmap(mapping,block,ext2_get_block);
769}
770
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771static ssize_t
772ext2_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
773 loff_t offset, unsigned long nr_segs)
774{
775 struct file *file = iocb->ki_filp;
776 struct inode *inode = file->f_mapping->host;
777
778 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
Badari Pulavarty1d8fa7a2006-03-26 01:38:02 -0800779 offset, nr_segs, ext2_get_block, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780}
781
782static int
783ext2_writepages(struct address_space *mapping, struct writeback_control *wbc)
784{
785 return mpage_writepages(mapping, wbc, ext2_get_block);
786}
787
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700788const struct address_space_operations ext2_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 .readpage = ext2_readpage,
790 .readpages = ext2_readpages,
791 .writepage = ext2_writepage,
792 .sync_page = block_sync_page,
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700793 .write_begin = ext2_write_begin,
794 .write_end = generic_write_end,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 .bmap = ext2_bmap,
796 .direct_IO = ext2_direct_IO,
797 .writepages = ext2_writepages,
Christoph Lametere965f962006-02-01 03:05:41 -0800798 .migratepage = buffer_migrate_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799};
800
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700801const struct address_space_operations ext2_aops_xip = {
Carsten Otte6d791252005-06-23 22:05:26 -0700802 .bmap = ext2_bmap,
803 .get_xip_page = ext2_get_xip_page,
804};
805
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700806const struct address_space_operations ext2_nobh_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 .readpage = ext2_readpage,
808 .readpages = ext2_readpages,
809 .writepage = ext2_nobh_writepage,
810 .sync_page = block_sync_page,
Nick Piggin03158cd2007-10-16 01:25:25 -0700811 .write_begin = ext2_nobh_write_begin,
812 .write_end = nobh_write_end,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 .bmap = ext2_bmap,
814 .direct_IO = ext2_direct_IO,
815 .writepages = ext2_writepages,
Christoph Lametere965f962006-02-01 03:05:41 -0800816 .migratepage = buffer_migrate_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817};
818
819/*
820 * Probably it should be a library function... search for first non-zero word
821 * or memcmp with zero_page, whatever is better for particular architecture.
822 * Linus?
823 */
824static inline int all_zeroes(__le32 *p, __le32 *q)
825{
826 while (p < q)
827 if (*p++)
828 return 0;
829 return 1;
830}
831
832/**
833 * ext2_find_shared - find the indirect blocks for partial truncation.
834 * @inode: inode in question
835 * @depth: depth of the affected branch
836 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
837 * @chain: place to store the pointers to partial indirect blocks
838 * @top: place to the (detached) top of branch
839 *
840 * This is a helper function used by ext2_truncate().
841 *
842 * When we do truncate() we may have to clean the ends of several indirect
843 * blocks but leave the blocks themselves alive. Block is partially
844 * truncated if some data below the new i_size is refered from it (and
845 * it is on the path to the first completely truncated data block, indeed).
846 * We have to free the top of that path along with everything to the right
847 * of the path. Since no allocation past the truncation point is possible
848 * until ext2_truncate() finishes, we may safely do the latter, but top
849 * of branch may require special attention - pageout below the truncation
850 * point might try to populate it.
851 *
852 * We atomically detach the top of branch from the tree, store the block
853 * number of its root in *@top, pointers to buffer_heads of partially
854 * truncated blocks - in @chain[].bh and pointers to their last elements
855 * that should not be removed - in @chain[].p. Return value is the pointer
856 * to last filled element of @chain.
857 *
858 * The work left to caller to do the actual freeing of subtrees:
859 * a) free the subtree starting from *@top
860 * b) free the subtrees whose roots are stored in
861 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
862 * c) free the subtrees growing from the inode past the @chain[0].p
863 * (no partially truncated stuff there).
864 */
865
866static Indirect *ext2_find_shared(struct inode *inode,
867 int depth,
868 int offsets[4],
869 Indirect chain[4],
870 __le32 *top)
871{
872 Indirect *partial, *p;
873 int k, err;
874
875 *top = 0;
876 for (k = depth; k > 1 && !offsets[k-1]; k--)
877 ;
878 partial = ext2_get_branch(inode, k, offsets, chain, &err);
879 if (!partial)
880 partial = chain + k-1;
881 /*
882 * If the branch acquired continuation since we've looked at it -
883 * fine, it should all survive and (new) top doesn't belong to us.
884 */
885 write_lock(&EXT2_I(inode)->i_meta_lock);
886 if (!partial->key && *partial->p) {
887 write_unlock(&EXT2_I(inode)->i_meta_lock);
888 goto no_top;
889 }
890 for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--)
891 ;
892 /*
893 * OK, we've found the last block that must survive. The rest of our
894 * branch should be detached before unlocking. However, if that rest
895 * of branch is all ours and does not grow immediately from the inode
896 * it's easier to cheat and just decrement partial->p.
897 */
898 if (p == chain + k - 1 && p > chain) {
899 p->p--;
900 } else {
901 *top = *p->p;
902 *p->p = 0;
903 }
904 write_unlock(&EXT2_I(inode)->i_meta_lock);
905
906 while(partial > p)
907 {
908 brelse(partial->bh);
909 partial--;
910 }
911no_top:
912 return partial;
913}
914
915/**
916 * ext2_free_data - free a list of data blocks
917 * @inode: inode we are dealing with
918 * @p: array of block numbers
919 * @q: points immediately past the end of array
920 *
921 * We are freeing all blocks refered from that array (numbers are
922 * stored as little-endian 32-bit) and updating @inode->i_blocks
923 * appropriately.
924 */
925static inline void ext2_free_data(struct inode *inode, __le32 *p, __le32 *q)
926{
927 unsigned long block_to_free = 0, count = 0;
928 unsigned long nr;
929
930 for ( ; p < q ; p++) {
931 nr = le32_to_cpu(*p);
932 if (nr) {
933 *p = 0;
934 /* accumulate blocks to free if they're contiguous */
935 if (count == 0)
936 goto free_this;
937 else if (block_to_free == nr - count)
938 count++;
939 else {
940 mark_inode_dirty(inode);
941 ext2_free_blocks (inode, block_to_free, count);
942 free_this:
943 block_to_free = nr;
944 count = 1;
945 }
946 }
947 }
948 if (count > 0) {
949 mark_inode_dirty(inode);
950 ext2_free_blocks (inode, block_to_free, count);
951 }
952}
953
954/**
955 * ext2_free_branches - free an array of branches
956 * @inode: inode we are dealing with
957 * @p: array of block numbers
958 * @q: pointer immediately past the end of array
959 * @depth: depth of the branches to free
960 *
961 * We are freeing all blocks refered from these branches (numbers are
962 * stored as little-endian 32-bit) and updating @inode->i_blocks
963 * appropriately.
964 */
965static void ext2_free_branches(struct inode *inode, __le32 *p, __le32 *q, int depth)
966{
967 struct buffer_head * bh;
968 unsigned long nr;
969
970 if (depth--) {
971 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
972 for ( ; p < q ; p++) {
973 nr = le32_to_cpu(*p);
974 if (!nr)
975 continue;
976 *p = 0;
977 bh = sb_bread(inode->i_sb, nr);
978 /*
979 * A read failure? Report error and clear slot
980 * (should be rare).
981 */
982 if (!bh) {
983 ext2_error(inode->i_sb, "ext2_free_branches",
984 "Read failure, inode=%ld, block=%ld",
985 inode->i_ino, nr);
986 continue;
987 }
988 ext2_free_branches(inode,
989 (__le32*)bh->b_data,
990 (__le32*)bh->b_data + addr_per_block,
991 depth);
992 bforget(bh);
993 ext2_free_blocks(inode, nr, 1);
994 mark_inode_dirty(inode);
995 }
996 } else
997 ext2_free_data(inode, p, q);
998}
999
Martin J. Bligha686cd82007-10-16 23:30:46 -07001000void ext2_truncate(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001{
1002 __le32 *i_data = EXT2_I(inode)->i_data;
Martin J. Bligha686cd82007-10-16 23:30:46 -07001003 struct ext2_inode_info *ei = EXT2_I(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
1005 int offsets[4];
1006 Indirect chain[4];
1007 Indirect *partial;
1008 __le32 nr = 0;
1009 int n;
1010 long iblock;
1011 unsigned blocksize;
1012
1013 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1014 S_ISLNK(inode->i_mode)))
1015 return;
1016 if (ext2_inode_is_fast_symlink(inode))
1017 return;
1018 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1019 return;
1020
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 blocksize = inode->i_sb->s_blocksize;
1022 iblock = (inode->i_size + blocksize-1)
1023 >> EXT2_BLOCK_SIZE_BITS(inode->i_sb);
1024
Carsten Otte6d791252005-06-23 22:05:26 -07001025 if (mapping_is_xip(inode->i_mapping))
1026 xip_truncate_page(inode->i_mapping, inode->i_size);
1027 else if (test_opt(inode->i_sb, NOBH))
Nick Piggin03158cd2007-10-16 01:25:25 -07001028 nobh_truncate_page(inode->i_mapping,
1029 inode->i_size, ext2_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 else
1031 block_truncate_page(inode->i_mapping,
1032 inode->i_size, ext2_get_block);
1033
1034 n = ext2_block_to_path(inode, iblock, offsets, NULL);
1035 if (n == 0)
1036 return;
1037
Martin J. Bligha686cd82007-10-16 23:30:46 -07001038 /*
1039 * From here we block out all ext2_get_block() callers who want to
1040 * modify the block allocation tree.
1041 */
1042 mutex_lock(&ei->truncate_mutex);
1043
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 if (n == 1) {
1045 ext2_free_data(inode, i_data+offsets[0],
1046 i_data + EXT2_NDIR_BLOCKS);
1047 goto do_indirects;
1048 }
1049
1050 partial = ext2_find_shared(inode, n, offsets, chain, &nr);
1051 /* Kill the top of shared branch (already detached) */
1052 if (nr) {
1053 if (partial == chain)
1054 mark_inode_dirty(inode);
1055 else
1056 mark_buffer_dirty_inode(partial->bh, inode);
1057 ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
1058 }
1059 /* Clear the ends of indirect blocks on the shared branch */
1060 while (partial > chain) {
1061 ext2_free_branches(inode,
1062 partial->p + 1,
1063 (__le32*)partial->bh->b_data+addr_per_block,
1064 (chain+n-1) - partial);
1065 mark_buffer_dirty_inode(partial->bh, inode);
1066 brelse (partial->bh);
1067 partial--;
1068 }
1069do_indirects:
1070 /* Kill the remaining (whole) subtrees */
1071 switch (offsets[0]) {
1072 default:
1073 nr = i_data[EXT2_IND_BLOCK];
1074 if (nr) {
1075 i_data[EXT2_IND_BLOCK] = 0;
1076 mark_inode_dirty(inode);
1077 ext2_free_branches(inode, &nr, &nr+1, 1);
1078 }
1079 case EXT2_IND_BLOCK:
1080 nr = i_data[EXT2_DIND_BLOCK];
1081 if (nr) {
1082 i_data[EXT2_DIND_BLOCK] = 0;
1083 mark_inode_dirty(inode);
1084 ext2_free_branches(inode, &nr, &nr+1, 2);
1085 }
1086 case EXT2_DIND_BLOCK:
1087 nr = i_data[EXT2_TIND_BLOCK];
1088 if (nr) {
1089 i_data[EXT2_TIND_BLOCK] = 0;
1090 mark_inode_dirty(inode);
1091 ext2_free_branches(inode, &nr, &nr+1, 3);
1092 }
1093 case EXT2_TIND_BLOCK:
1094 ;
1095 }
Martin J. Bligha686cd82007-10-16 23:30:46 -07001096
1097 ext2_discard_reservation(inode);
1098
1099 mutex_unlock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
1101 if (inode_needs_sync(inode)) {
1102 sync_mapping_buffers(inode->i_mapping);
1103 ext2_sync_inode (inode);
1104 } else {
1105 mark_inode_dirty(inode);
1106 }
1107}
1108
1109static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
1110 struct buffer_head **p)
1111{
1112 struct buffer_head * bh;
1113 unsigned long block_group;
1114 unsigned long block;
1115 unsigned long offset;
1116 struct ext2_group_desc * gdp;
1117
1118 *p = NULL;
1119 if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) ||
1120 ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
1121 goto Einval;
1122
1123 block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb);
Eric Sandeenef2fb672007-10-16 23:26:30 -07001124 gdp = ext2_get_group_desc(sb, block_group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 if (!gdp)
1126 goto Egdp;
1127 /*
1128 * Figure out the offset within the block group inode table
1129 */
1130 offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb);
1131 block = le32_to_cpu(gdp->bg_inode_table) +
1132 (offset >> EXT2_BLOCK_SIZE_BITS(sb));
1133 if (!(bh = sb_bread(sb, block)))
1134 goto Eio;
1135
1136 *p = bh;
1137 offset &= (EXT2_BLOCK_SIZE(sb) - 1);
1138 return (struct ext2_inode *) (bh->b_data + offset);
1139
1140Einval:
1141 ext2_error(sb, "ext2_get_inode", "bad inode number: %lu",
1142 (unsigned long) ino);
1143 return ERR_PTR(-EINVAL);
1144Eio:
1145 ext2_error(sb, "ext2_get_inode",
1146 "unable to read inode block - inode=%lu, block=%lu",
1147 (unsigned long) ino, block);
1148Egdp:
1149 return ERR_PTR(-EIO);
1150}
1151
1152void ext2_set_inode_flags(struct inode *inode)
1153{
1154 unsigned int flags = EXT2_I(inode)->i_flags;
1155
1156 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
1157 if (flags & EXT2_SYNC_FL)
1158 inode->i_flags |= S_SYNC;
1159 if (flags & EXT2_APPEND_FL)
1160 inode->i_flags |= S_APPEND;
1161 if (flags & EXT2_IMMUTABLE_FL)
1162 inode->i_flags |= S_IMMUTABLE;
1163 if (flags & EXT2_NOATIME_FL)
1164 inode->i_flags |= S_NOATIME;
1165 if (flags & EXT2_DIRSYNC_FL)
1166 inode->i_flags |= S_DIRSYNC;
1167}
1168
Jan Kara4f99ed62007-05-08 00:31:04 -07001169/* Propagate flags from i_flags to EXT2_I(inode)->i_flags */
1170void ext2_get_inode_flags(struct ext2_inode_info *ei)
1171{
1172 unsigned int flags = ei->vfs_inode.i_flags;
1173
1174 ei->i_flags &= ~(EXT2_SYNC_FL|EXT2_APPEND_FL|
1175 EXT2_IMMUTABLE_FL|EXT2_NOATIME_FL|EXT2_DIRSYNC_FL);
1176 if (flags & S_SYNC)
1177 ei->i_flags |= EXT2_SYNC_FL;
1178 if (flags & S_APPEND)
1179 ei->i_flags |= EXT2_APPEND_FL;
1180 if (flags & S_IMMUTABLE)
1181 ei->i_flags |= EXT2_IMMUTABLE_FL;
1182 if (flags & S_NOATIME)
1183 ei->i_flags |= EXT2_NOATIME_FL;
1184 if (flags & S_DIRSYNC)
1185 ei->i_flags |= EXT2_DIRSYNC_FL;
1186}
1187
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188void ext2_read_inode (struct inode * inode)
1189{
1190 struct ext2_inode_info *ei = EXT2_I(inode);
1191 ino_t ino = inode->i_ino;
1192 struct buffer_head * bh;
1193 struct ext2_inode * raw_inode = ext2_get_inode(inode->i_sb, ino, &bh);
1194 int n;
1195
1196#ifdef CONFIG_EXT2_FS_POSIX_ACL
1197 ei->i_acl = EXT2_ACL_NOT_CACHED;
1198 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
1199#endif
Martin J. Bligha686cd82007-10-16 23:30:46 -07001200 ei->i_block_alloc_info = NULL;
1201
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 if (IS_ERR(raw_inode))
1203 goto bad_inode;
1204
1205 inode->i_mode = le16_to_cpu(raw_inode->i_mode);
1206 inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
1207 inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
1208 if (!(test_opt (inode->i_sb, NO_UID32))) {
1209 inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
1210 inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
1211 }
1212 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
1213 inode->i_size = le32_to_cpu(raw_inode->i_size);
Markus Rechberger4d7bf112007-05-08 00:23:39 -07001214 inode->i_atime.tv_sec = (signed)le32_to_cpu(raw_inode->i_atime);
1215 inode->i_ctime.tv_sec = (signed)le32_to_cpu(raw_inode->i_ctime);
1216 inode->i_mtime.tv_sec = (signed)le32_to_cpu(raw_inode->i_mtime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0;
1218 ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
1219 /* We now have enough fields to check if the inode was active or not.
1220 * This is needed because nfsd might try to access dead inodes
1221 * the test is that same one that e2fsck uses
1222 * NeilBrown 1999oct15
1223 */
1224 if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) {
1225 /* this inode is deleted */
1226 brelse (bh);
1227 goto bad_inode;
1228 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 inode->i_blocks = le32_to_cpu(raw_inode->i_blocks);
1230 ei->i_flags = le32_to_cpu(raw_inode->i_flags);
1231 ei->i_faddr = le32_to_cpu(raw_inode->i_faddr);
1232 ei->i_frag_no = raw_inode->i_frag;
1233 ei->i_frag_size = raw_inode->i_fsize;
1234 ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl);
1235 ei->i_dir_acl = 0;
1236 if (S_ISREG(inode->i_mode))
1237 inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32;
1238 else
1239 ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl);
1240 ei->i_dtime = 0;
1241 inode->i_generation = le32_to_cpu(raw_inode->i_generation);
1242 ei->i_state = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb);
1244 ei->i_dir_start_lookup = 0;
1245
1246 /*
1247 * NOTE! The in-memory inode i_data array is in little-endian order
1248 * even on big-endian machines: we do NOT byteswap the block numbers!
1249 */
1250 for (n = 0; n < EXT2_N_BLOCKS; n++)
1251 ei->i_data[n] = raw_inode->i_block[n];
1252
1253 if (S_ISREG(inode->i_mode)) {
1254 inode->i_op = &ext2_file_inode_operations;
Carsten Otte6d791252005-06-23 22:05:26 -07001255 if (ext2_use_xip(inode->i_sb)) {
1256 inode->i_mapping->a_ops = &ext2_aops_xip;
1257 inode->i_fop = &ext2_xip_file_operations;
1258 } else if (test_opt(inode->i_sb, NOBH)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259 inode->i_mapping->a_ops = &ext2_nobh_aops;
Carsten Otte6d791252005-06-23 22:05:26 -07001260 inode->i_fop = &ext2_file_operations;
1261 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 inode->i_mapping->a_ops = &ext2_aops;
Carsten Otte6d791252005-06-23 22:05:26 -07001263 inode->i_fop = &ext2_file_operations;
1264 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 } else if (S_ISDIR(inode->i_mode)) {
1266 inode->i_op = &ext2_dir_inode_operations;
1267 inode->i_fop = &ext2_dir_operations;
1268 if (test_opt(inode->i_sb, NOBH))
1269 inode->i_mapping->a_ops = &ext2_nobh_aops;
1270 else
1271 inode->i_mapping->a_ops = &ext2_aops;
1272 } else if (S_ISLNK(inode->i_mode)) {
1273 if (ext2_inode_is_fast_symlink(inode))
1274 inode->i_op = &ext2_fast_symlink_inode_operations;
1275 else {
1276 inode->i_op = &ext2_symlink_inode_operations;
1277 if (test_opt(inode->i_sb, NOBH))
1278 inode->i_mapping->a_ops = &ext2_nobh_aops;
1279 else
1280 inode->i_mapping->a_ops = &ext2_aops;
1281 }
1282 } else {
1283 inode->i_op = &ext2_special_inode_operations;
1284 if (raw_inode->i_block[0])
1285 init_special_inode(inode, inode->i_mode,
1286 old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
1287 else
1288 init_special_inode(inode, inode->i_mode,
1289 new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
1290 }
1291 brelse (bh);
1292 ext2_set_inode_flags(inode);
1293 return;
1294
1295bad_inode:
1296 make_bad_inode(inode);
1297 return;
1298}
1299
1300static int ext2_update_inode(struct inode * inode, int do_sync)
1301{
1302 struct ext2_inode_info *ei = EXT2_I(inode);
1303 struct super_block *sb = inode->i_sb;
1304 ino_t ino = inode->i_ino;
1305 uid_t uid = inode->i_uid;
1306 gid_t gid = inode->i_gid;
1307 struct buffer_head * bh;
1308 struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh);
1309 int n;
1310 int err = 0;
1311
1312 if (IS_ERR(raw_inode))
1313 return -EIO;
1314
1315 /* For fields not not tracking in the in-memory inode,
1316 * initialise them to zero for new inodes. */
1317 if (ei->i_state & EXT2_STATE_NEW)
1318 memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size);
1319
Jan Kara4f99ed62007-05-08 00:31:04 -07001320 ext2_get_inode_flags(ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 raw_inode->i_mode = cpu_to_le16(inode->i_mode);
1322 if (!(test_opt(sb, NO_UID32))) {
1323 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid));
1324 raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid));
1325/*
1326 * Fix up interoperability with old kernels. Otherwise, old inodes get
1327 * re-used with the upper 16 bits of the uid/gid intact
1328 */
1329 if (!ei->i_dtime) {
1330 raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid));
1331 raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid));
1332 } else {
1333 raw_inode->i_uid_high = 0;
1334 raw_inode->i_gid_high = 0;
1335 }
1336 } else {
1337 raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid));
1338 raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid));
1339 raw_inode->i_uid_high = 0;
1340 raw_inode->i_gid_high = 0;
1341 }
1342 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
1343 raw_inode->i_size = cpu_to_le32(inode->i_size);
1344 raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
1345 raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
1346 raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
1347
1348 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
1349 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
1350 raw_inode->i_flags = cpu_to_le32(ei->i_flags);
1351 raw_inode->i_faddr = cpu_to_le32(ei->i_faddr);
1352 raw_inode->i_frag = ei->i_frag_no;
1353 raw_inode->i_fsize = ei->i_frag_size;
1354 raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl);
1355 if (!S_ISREG(inode->i_mode))
1356 raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl);
1357 else {
1358 raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32);
1359 if (inode->i_size > 0x7fffffffULL) {
1360 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb,
1361 EXT2_FEATURE_RO_COMPAT_LARGE_FILE) ||
1362 EXT2_SB(sb)->s_es->s_rev_level ==
1363 cpu_to_le32(EXT2_GOOD_OLD_REV)) {
1364 /* If this is the first large file
1365 * created, add a flag to the superblock.
1366 */
1367 lock_kernel();
1368 ext2_update_dynamic_rev(sb);
1369 EXT2_SET_RO_COMPAT_FEATURE(sb,
1370 EXT2_FEATURE_RO_COMPAT_LARGE_FILE);
1371 unlock_kernel();
1372 ext2_write_super(sb);
1373 }
1374 }
1375 }
1376
1377 raw_inode->i_generation = cpu_to_le32(inode->i_generation);
1378 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
1379 if (old_valid_dev(inode->i_rdev)) {
1380 raw_inode->i_block[0] =
1381 cpu_to_le32(old_encode_dev(inode->i_rdev));
1382 raw_inode->i_block[1] = 0;
1383 } else {
1384 raw_inode->i_block[0] = 0;
1385 raw_inode->i_block[1] =
1386 cpu_to_le32(new_encode_dev(inode->i_rdev));
1387 raw_inode->i_block[2] = 0;
1388 }
1389 } else for (n = 0; n < EXT2_N_BLOCKS; n++)
1390 raw_inode->i_block[n] = ei->i_data[n];
1391 mark_buffer_dirty(bh);
1392 if (do_sync) {
1393 sync_dirty_buffer(bh);
1394 if (buffer_req(bh) && !buffer_uptodate(bh)) {
1395 printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1396 sb->s_id, (unsigned long) ino);
1397 err = -EIO;
1398 }
1399 }
1400 ei->i_state &= ~EXT2_STATE_NEW;
1401 brelse (bh);
1402 return err;
1403}
1404
1405int ext2_write_inode(struct inode *inode, int wait)
1406{
1407 return ext2_update_inode(inode, wait);
1408}
1409
1410int ext2_sync_inode(struct inode *inode)
1411{
1412 struct writeback_control wbc = {
1413 .sync_mode = WB_SYNC_ALL,
1414 .nr_to_write = 0, /* sys_fsync did this */
1415 };
1416 return sync_inode(inode, &wbc);
1417}
1418
1419int ext2_setattr(struct dentry *dentry, struct iattr *iattr)
1420{
1421 struct inode *inode = dentry->d_inode;
1422 int error;
1423
1424 error = inode_change_ok(inode, iattr);
1425 if (error)
1426 return error;
1427 if ((iattr->ia_valid & ATTR_UID && iattr->ia_uid != inode->i_uid) ||
1428 (iattr->ia_valid & ATTR_GID && iattr->ia_gid != inode->i_gid)) {
1429 error = DQUOT_TRANSFER(inode, iattr) ? -EDQUOT : 0;
1430 if (error)
1431 return error;
1432 }
1433 error = inode_setattr(inode, iattr);
1434 if (!error && (iattr->ia_valid & ATTR_MODE))
1435 error = ext2_acl_chmod(inode);
1436 return error;
1437}