]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext2/inode.c
Don't dirty the victim in ext2_xattr_delete_inode()
[net-next-2.6.git] / fs / ext2 / inode.c
CommitLineData
1da177e4
LT
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
1da177e4
LT
25#include <linux/time.h>
26#include <linux/highuid.h>
27#include <linux/pagemap.h>
28#include <linux/quotaops.h>
29#include <linux/module.h>
30#include <linux/writeback.h>
31#include <linux/buffer_head.h>
32#include <linux/mpage.h>
68c9d702 33#include <linux/fiemap.h>
8d6d0c4d 34#include <linux/namei.h>
1da177e4
LT
35#include "ext2.h"
36#include "acl.h"
6d79125b 37#include "xip.h"
1da177e4
LT
38
39MODULE_AUTHOR("Remy Card and others");
40MODULE_DESCRIPTION("Second Extended Filesystem");
41MODULE_LICENSE("GPL");
42
a9185b41
CH
43static int __ext2_write_inode(struct inode *inode, int do_sync);
44
1da177e4
LT
45/*
46 * Test whether an inode is a fast symlink.
47 */
48static inline int ext2_inode_is_fast_symlink(struct inode *inode)
49{
50 int ea_blocks = EXT2_I(inode)->i_file_acl ?
51 (inode->i_sb->s_blocksize >> 9) : 0;
52
53 return (S_ISLNK(inode->i_mode) &&
54 inode->i_blocks - ea_blocks == 0);
55}
56
737f2e93 57static void ext2_truncate_blocks(struct inode *inode, loff_t offset);
58
59static void ext2_write_failed(struct address_space *mapping, loff_t to)
60{
61 struct inode *inode = mapping->host;
62
63 if (to > inode->i_size) {
64 truncate_pagecache(inode, to, inode->i_size);
65 ext2_truncate_blocks(inode, inode->i_size);
66 }
67}
68
1da177e4
LT
69/*
70 * Called at the last iput() if i_nlink is zero.
71 */
72void ext2_delete_inode (struct inode * inode)
73{
907f4554 74 if (!is_bad_inode(inode))
871a2931 75 dquot_initialize(inode);
fef26658
MF
76 truncate_inode_pages(&inode->i_data, 0);
77
1da177e4
LT
78 if (is_bad_inode(inode))
79 goto no_delete;
80 EXT2_I(inode)->i_dtime = get_seconds();
81 mark_inode_dirty(inode);
a9185b41 82 __ext2_write_inode(inode, inode_needs_sync(inode));
1da177e4
LT
83
84 inode->i_size = 0;
85 if (inode->i_blocks)
737f2e93 86 ext2_truncate_blocks(inode, 0);
1da177e4
LT
87 ext2_free_inode (inode);
88
89 return;
90no_delete:
91 clear_inode(inode); /* We must guarantee clearing of inode... */
92}
93
1da177e4
LT
94typedef struct {
95 __le32 *p;
96 __le32 key;
97 struct buffer_head *bh;
98} Indirect;
99
100static inline void add_chain(Indirect *p, struct buffer_head *bh, __le32 *v)
101{
102 p->key = *(p->p = v);
103 p->bh = bh;
104}
105
106static inline int verify_chain(Indirect *from, Indirect *to)
107{
108 while (from <= to && from->key == *from->p)
109 from++;
110 return (from > to);
111}
112
113/**
114 * ext2_block_to_path - parse the block number into array of offsets
115 * @inode: inode in question (we are only interested in its superblock)
116 * @i_block: block number to be parsed
117 * @offsets: array to store the offsets in
118 * @boundary: set this non-zero if the referred-to block is likely to be
119 * followed (on disk) by an indirect block.
120 * To store the locations of file's data ext2 uses a data structure common
121 * for UNIX filesystems - tree of pointers anchored in the inode, with
122 * data blocks at leaves and indirect blocks in intermediate nodes.
123 * This function translates the block number into path in that tree -
124 * return value is the path length and @offsets[n] is the offset of
125 * pointer to (n+1)th node in the nth one. If @block is out of range
126 * (negative or too large) warning is printed and zero returned.
127 *
128 * Note: function doesn't find node addresses, so no IO is needed. All
129 * we need to know is the capacity of indirect blocks (taken from the
130 * inode->i_sb).
131 */
132
133/*
134 * Portability note: the last comparison (check that we fit into triple
135 * indirect block) is spelled differently, because otherwise on an
136 * architecture with 32-bit longs and 8Kb pages we might get into trouble
137 * if our filesystem had 8Kb blocks. We might use long long, but that would
138 * kill us on x86. Oh, well, at least the sign propagation does not matter -
139 * i_block would have to be negative in the very beginning, so we would not
140 * get there at all.
141 */
142
143static int ext2_block_to_path(struct inode *inode,
144 long i_block, int offsets[4], int *boundary)
145{
146 int ptrs = EXT2_ADDR_PER_BLOCK(inode->i_sb);
147 int ptrs_bits = EXT2_ADDR_PER_BLOCK_BITS(inode->i_sb);
148 const long direct_blocks = EXT2_NDIR_BLOCKS,
149 indirect_blocks = ptrs,
150 double_blocks = (1 << (ptrs_bits * 2));
151 int n = 0;
152 int final = 0;
153
154 if (i_block < 0) {
2314b07c
AF
155 ext2_msg(inode->i_sb, KERN_WARNING,
156 "warning: %s: block < 0", __func__);
1da177e4
LT
157 } else if (i_block < direct_blocks) {
158 offsets[n++] = i_block;
159 final = direct_blocks;
160 } else if ( (i_block -= direct_blocks) < indirect_blocks) {
161 offsets[n++] = EXT2_IND_BLOCK;
162 offsets[n++] = i_block;
163 final = ptrs;
164 } else if ((i_block -= indirect_blocks) < double_blocks) {
165 offsets[n++] = EXT2_DIND_BLOCK;
166 offsets[n++] = i_block >> ptrs_bits;
167 offsets[n++] = i_block & (ptrs - 1);
168 final = ptrs;
169 } else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) {
170 offsets[n++] = EXT2_TIND_BLOCK;
171 offsets[n++] = i_block >> (ptrs_bits * 2);
172 offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1);
173 offsets[n++] = i_block & (ptrs - 1);
174 final = ptrs;
175 } else {
2314b07c
AF
176 ext2_msg(inode->i_sb, KERN_WARNING,
177 "warning: %s: block is too big", __func__);
1da177e4
LT
178 }
179 if (boundary)
a686cd89
MB
180 *boundary = final - 1 - (i_block & (ptrs - 1));
181
1da177e4
LT
182 return n;
183}
184
185/**
186 * ext2_get_branch - read the chain of indirect blocks leading to data
187 * @inode: inode in question
188 * @depth: depth of the chain (1 - direct pointer, etc.)
189 * @offsets: offsets of pointers in inode/indirect blocks
190 * @chain: place to store the result
191 * @err: here we store the error value
192 *
193 * Function fills the array of triples <key, p, bh> and returns %NULL
194 * if everything went OK or the pointer to the last filled triple
195 * (incomplete one) otherwise. Upon the return chain[i].key contains
196 * the number of (i+1)-th block in the chain (as it is stored in memory,
197 * i.e. little-endian 32-bit), chain[i].p contains the address of that
198 * number (it points into struct inode for i==0 and into the bh->b_data
199 * for i>0) and chain[i].bh points to the buffer_head of i-th indirect
200 * block for i>0 and NULL for i==0. In other words, it holds the block
201 * numbers of the chain, addresses they were taken from (and where we can
202 * verify that chain did not change) and buffer_heads hosting these
203 * numbers.
204 *
205 * Function stops when it stumbles upon zero pointer (absent block)
206 * (pointer to last triple returned, *@err == 0)
207 * or when it gets an IO error reading an indirect block
208 * (ditto, *@err == -EIO)
209 * or when it notices that chain had been changed while it was reading
210 * (ditto, *@err == -EAGAIN)
211 * or when it reads all @depth-1 indirect blocks successfully and finds
212 * the whole chain, all way to the data (returns %NULL, *err == 0).
213 */
214static Indirect *ext2_get_branch(struct inode *inode,
215 int depth,
216 int *offsets,
217 Indirect chain[4],
218 int *err)
219{
220 struct super_block *sb = inode->i_sb;
221 Indirect *p = chain;
222 struct buffer_head *bh;
223
224 *err = 0;
225 /* i_data is not going away, no lock needed */
226 add_chain (chain, NULL, EXT2_I(inode)->i_data + *offsets);
227 if (!p->key)
228 goto no_block;
229 while (--depth) {
230 bh = sb_bread(sb, le32_to_cpu(p->key));
231 if (!bh)
232 goto failure;
233 read_lock(&EXT2_I(inode)->i_meta_lock);
234 if (!verify_chain(chain, p))
235 goto changed;
236 add_chain(++p, bh, (__le32*)bh->b_data + *++offsets);
237 read_unlock(&EXT2_I(inode)->i_meta_lock);
238 if (!p->key)
239 goto no_block;
240 }
241 return NULL;
242
243changed:
244 read_unlock(&EXT2_I(inode)->i_meta_lock);
245 brelse(bh);
246 *err = -EAGAIN;
247 goto no_block;
248failure:
249 *err = -EIO;
250no_block:
251 return p;
252}
253
254/**
255 * ext2_find_near - find a place for allocation with sufficient locality
256 * @inode: owner
257 * @ind: descriptor of indirect block.
258 *
1cc8dcf5 259 * This function returns the preferred place for block allocation.
1da177e4
LT
260 * It is used when heuristic for sequential allocation fails.
261 * Rules are:
262 * + if there is a block to the left of our position - allocate near it.
263 * + if pointer will live in indirect block - allocate near that block.
264 * + if pointer will live in inode - allocate in the same cylinder group.
265 *
266 * In the latter case we colour the starting block by the callers PID to
267 * prevent it from clashing with concurrent allocations for a different inode
268 * in the same block group. The PID is used here so that functionally related
269 * files will be close-by on-disk.
270 *
271 * Caller must make sure that @ind is valid and will stay that way.
272 */
273
4c8b3125 274static ext2_fsblk_t ext2_find_near(struct inode *inode, Indirect *ind)
1da177e4
LT
275{
276 struct ext2_inode_info *ei = EXT2_I(inode);
277 __le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data;
278 __le32 *p;
4c8b3125
AM
279 ext2_fsblk_t bg_start;
280 ext2_fsblk_t colour;
1da177e4
LT
281
282 /* Try to find previous block */
283 for (p = ind->p - 1; p >= start; p--)
284 if (*p)
285 return le32_to_cpu(*p);
286
287 /* No such thing, so let's try location of indirect block */
288 if (ind->bh)
289 return ind->bh->b_blocknr;
290
291 /*
292 * It is going to be refered from inode itself? OK, just put it into
293 * the same cylinder group then.
294 */
24097d12 295 bg_start = ext2_group_first_block_no(inode->i_sb, ei->i_block_group);
1da177e4
LT
296 colour = (current->pid % 16) *
297 (EXT2_BLOCKS_PER_GROUP(inode->i_sb) / 16);
298 return bg_start + colour;
299}
300
301/**
1cc8dcf5 302 * ext2_find_goal - find a preferred place for allocation.
1da177e4
LT
303 * @inode: owner
304 * @block: block we want
1da177e4 305 * @partial: pointer to the last triple within a chain
1da177e4 306 *
a686cd89 307 * Returns preferred place for a block (the goal).
1da177e4
LT
308 */
309
4c8b3125
AM
310static inline ext2_fsblk_t ext2_find_goal(struct inode *inode, long block,
311 Indirect *partial)
1da177e4 312{
a686cd89
MB
313 struct ext2_block_alloc_info *block_i;
314
315 block_i = EXT2_I(inode)->i_block_alloc_info;
316
317 /*
318 * try the heuristic for sequential allocation,
319 * failing that at least try to get decent locality.
320 */
321 if (block_i && (block == block_i->last_alloc_logical_block + 1)
322 && (block_i->last_alloc_physical_block != 0)) {
323 return block_i->last_alloc_physical_block + 1;
324 }
325
326 return ext2_find_near(inode, partial);
327}
328
329/**
330 * ext2_blks_to_allocate: Look up the block map and count the number
331 * of direct blocks need to be allocated for the given branch.
332 *
333 * @branch: chain of indirect blocks
334 * @k: number of blocks need for indirect blocks
335 * @blks: number of data blocks to be mapped.
336 * @blocks_to_boundary: the offset in the indirect block
337 *
338 * return the total number of blocks to be allocate, including the
339 * direct and indirect blocks.
340 */
341static int
342ext2_blks_to_allocate(Indirect * branch, int k, unsigned long blks,
343 int blocks_to_boundary)
344{
345 unsigned long count = 0;
346
347 /*
348 * Simple case, [t,d]Indirect block(s) has not allocated yet
349 * then it's clear blocks on that path have not allocated
350 */
351 if (k > 0) {
352 /* right now don't hanel cross boundary allocation */
353 if (blks < blocks_to_boundary + 1)
354 count += blks;
355 else
356 count += blocks_to_boundary + 1;
357 return count;
1da177e4 358 }
a686cd89
MB
359
360 count++;
361 while (count < blks && count <= blocks_to_boundary
362 && le32_to_cpu(*(branch[0].p + count)) == 0) {
363 count++;
364 }
365 return count;
366}
367
368/**
369 * ext2_alloc_blocks: multiple allocate blocks needed for a branch
370 * @indirect_blks: the number of blocks need to allocate for indirect
371 * blocks
372 *
373 * @new_blocks: on return it will store the new block numbers for
374 * the indirect blocks(if needed) and the first direct block,
375 * @blks: on return it will store the total number of allocated
376 * direct blocks
377 */
378static int ext2_alloc_blocks(struct inode *inode,
379 ext2_fsblk_t goal, int indirect_blks, int blks,
380 ext2_fsblk_t new_blocks[4], int *err)
381{
382 int target, i;
383 unsigned long count = 0;
384 int index = 0;
385 ext2_fsblk_t current_block = 0;
386 int ret = 0;
387
388 /*
389 * Here we try to allocate the requested multiple blocks at once,
390 * on a best-effort basis.
391 * To build a branch, we should allocate blocks for
392 * the indirect blocks(if not allocated yet), and at least
393 * the first direct block of this branch. That's the
394 * minimum number of blocks need to allocate(required)
395 */
396 target = blks + indirect_blks;
397
398 while (1) {
399 count = target;
400 /* allocating blocks for indirect blocks and direct blocks */
401 current_block = ext2_new_blocks(inode,goal,&count,err);
402 if (*err)
403 goto failed_out;
404
405 target -= count;
406 /* allocate blocks for indirect blocks */
407 while (index < indirect_blks && count) {
408 new_blocks[index++] = current_block++;
409 count--;
410 }
411
412 if (count > 0)
413 break;
414 }
415
416 /* save the new block number for the first direct block */
417 new_blocks[index] = current_block;
418
419 /* total number of blocks allocated for direct blocks */
420 ret = count;
421 *err = 0;
422 return ret;
423failed_out:
424 for (i = 0; i <index; i++)
425 ext2_free_blocks(inode, new_blocks[i], 1);
addacc7d
AV
426 if (index)
427 mark_inode_dirty(inode);
a686cd89 428 return ret;
1da177e4
LT
429}
430
431/**
432 * ext2_alloc_branch - allocate and set up a chain of blocks.
433 * @inode: owner
434 * @num: depth of the chain (number of blocks to allocate)
435 * @offsets: offsets (in the blocks) to store the pointers to next.
436 * @branch: place to store the chain in.
437 *
438 * This function allocates @num blocks, zeroes out all but the last one,
439 * links them into chain and (if we are synchronous) writes them to disk.
440 * In other words, it prepares a branch that can be spliced onto the
441 * inode. It stores the information about that chain in the branch[], in
442 * the same format as ext2_get_branch() would do. We are calling it after
443 * we had read the existing part of chain and partial points to the last
444 * triple of that (one with zero ->key). Upon the exit we have the same
445 * picture as after the successful ext2_get_block(), excpet that in one
446 * place chain is disconnected - *branch->p is still zero (we did not
447 * set the last link), but branch->key contains the number that should
448 * be placed into *branch->p to fill that gap.
449 *
450 * If allocation fails we free all blocks we've allocated (and forget
451 * their buffer_heads) and return the error value the from failed
452 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
453 * as described above and return 0.
454 */
455
456static int ext2_alloc_branch(struct inode *inode,
a686cd89
MB
457 int indirect_blks, int *blks, ext2_fsblk_t goal,
458 int *offsets, Indirect *branch)
1da177e4
LT
459{
460 int blocksize = inode->i_sb->s_blocksize;
a686cd89
MB
461 int i, n = 0;
462 int err = 0;
463 struct buffer_head *bh;
464 int num;
465 ext2_fsblk_t new_blocks[4];
466 ext2_fsblk_t current_block;
467
468 num = ext2_alloc_blocks(inode, goal, indirect_blks,
469 *blks, new_blocks, &err);
470 if (err)
471 return err;
472
473 branch[0].key = cpu_to_le32(new_blocks[0]);
474 /*
475 * metadata blocks and data blocks are allocated.
476 */
477 for (n = 1; n <= indirect_blks; n++) {
1da177e4 478 /*
a686cd89
MB
479 * Get buffer_head for parent block, zero it out
480 * and set the pointer to new one, then send
481 * parent to disk.
1da177e4 482 */
a686cd89
MB
483 bh = sb_getblk(inode->i_sb, new_blocks[n-1]);
484 branch[n].bh = bh;
1da177e4
LT
485 lock_buffer(bh);
486 memset(bh->b_data, 0, blocksize);
1da177e4 487 branch[n].p = (__le32 *) bh->b_data + offsets[n];
a686cd89 488 branch[n].key = cpu_to_le32(new_blocks[n]);
1da177e4 489 *branch[n].p = branch[n].key;
a686cd89
MB
490 if ( n == indirect_blks) {
491 current_block = new_blocks[n];
492 /*
493 * End of chain, update the last new metablock of
494 * the chain to point to the new allocated
495 * data blocks numbers
496 */
497 for (i=1; i < num; i++)
498 *(branch[n].p + i) = cpu_to_le32(++current_block);
499 }
1da177e4
LT
500 set_buffer_uptodate(bh);
501 unlock_buffer(bh);
502 mark_buffer_dirty_inode(bh, inode);
503 /* We used to sync bh here if IS_SYNC(inode).
a2a735ad 504 * But we now rely upon generic_write_sync()
1da177e4
LT
505 * and b_inode_buffers. But not for directories.
506 */
507 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
508 sync_dirty_buffer(bh);
1da177e4 509 }
a686cd89 510 *blks = num;
1da177e4
LT
511 return err;
512}
513
514/**
a686cd89
MB
515 * ext2_splice_branch - splice the allocated branch onto inode.
516 * @inode: owner
517 * @block: (logical) number of block we are adding
a686cd89
MB
518 * @where: location of missing link
519 * @num: number of indirect blocks we are adding
520 * @blks: number of direct blocks we are adding
1da177e4 521 *
a686cd89
MB
522 * This function fills the missing link and does all housekeeping needed in
523 * inode (->i_blocks, etc.). In case of success we end up with the full
524 * chain to new block and return 0.
1da177e4 525 */
a686cd89
MB
526static void ext2_splice_branch(struct inode *inode,
527 long block, Indirect *where, int num, int blks)
1da177e4 528{
1da177e4 529 int i;
a686cd89
MB
530 struct ext2_block_alloc_info *block_i;
531 ext2_fsblk_t current_block;
1da177e4 532
a686cd89 533 block_i = EXT2_I(inode)->i_block_alloc_info;
1da177e4 534
a686cd89 535 /* XXX LOCKING probably should have i_meta_lock ?*/
1da177e4
LT
536 /* That's it */
537
538 *where->p = where->key;
1da177e4 539
a686cd89
MB
540 /*
541 * Update the host buffer_head or inode to point to more just allocated
542 * direct blocks blocks
543 */
544 if (num == 0 && blks > 1) {
545 current_block = le32_to_cpu(where->key) + 1;
546 for (i = 1; i < blks; i++)
547 *(where->p + i ) = cpu_to_le32(current_block++);
548 }
1da177e4 549
a686cd89
MB
550 /*
551 * update the most recently allocated logical & physical block
552 * in i_block_alloc_info, to assist find the proper goal block for next
553 * allocation
554 */
555 if (block_i) {
556 block_i->last_alloc_logical_block = block + blks - 1;
557 block_i->last_alloc_physical_block =
558 le32_to_cpu(where[num].key) + blks - 1;
559 }
1da177e4 560
a686cd89 561 /* We are done with atomic stuff, now do the rest of housekeeping */
1da177e4
LT
562
563 /* had we spliced it onto indirect block? */
564 if (where->bh)
565 mark_buffer_dirty_inode(where->bh, inode);
566
a686cd89 567 inode->i_ctime = CURRENT_TIME_SEC;
1da177e4 568 mark_inode_dirty(inode);
1da177e4
LT
569}
570
571/*
572 * Allocation strategy is simple: if we have to allocate something, we will
573 * have to go the whole way to leaf. So let's do it before attaching anything
574 * to tree, set linkage between the newborn blocks, write them if sync is
575 * required, recheck the path, free and repeat if check fails, otherwise
576 * set the last missing link (that will protect us from any truncate-generated
577 * removals - all blocks on the path are immune now) and possibly force the
578 * write on the parent block.
579 * That has a nice additional property: no special recovery from the failed
580 * allocations is needed - we simply release blocks and do not touch anything
581 * reachable from inode.
a686cd89
MB
582 *
583 * `handle' can be NULL if create == 0.
584 *
a686cd89
MB
585 * return > 0, # of blocks mapped or allocated.
586 * return = 0, if plain lookup failed.
587 * return < 0, error case.
1da177e4 588 */
a686cd89
MB
589static int ext2_get_blocks(struct inode *inode,
590 sector_t iblock, unsigned long maxblocks,
591 struct buffer_head *bh_result,
592 int create)
1da177e4
LT
593{
594 int err = -EIO;
595 int offsets[4];
596 Indirect chain[4];
597 Indirect *partial;
a686cd89
MB
598 ext2_fsblk_t goal;
599 int indirect_blks;
600 int blocks_to_boundary = 0;
601 int depth;
602 struct ext2_inode_info *ei = EXT2_I(inode);
603 int count = 0;
604 ext2_fsblk_t first_block = 0;
1da177e4 605
a686cd89 606 depth = ext2_block_to_path(inode,iblock,offsets,&blocks_to_boundary);
1da177e4 607
a686cd89
MB
608 if (depth == 0)
609 return (err);
1da177e4 610
316cb4ef 611 partial = ext2_get_branch(inode, depth, offsets, chain, &err);
1da177e4
LT
612 /* Simplest case - block found, no allocation needed */
613 if (!partial) {
a686cd89
MB
614 first_block = le32_to_cpu(chain[depth - 1].key);
615 clear_buffer_new(bh_result); /* What's this do? */
616 count++;
617 /*map more blocks*/
618 while (count < maxblocks && count <= blocks_to_boundary) {
619 ext2_fsblk_t blk;
620
316cb4ef 621 if (!verify_chain(chain, chain + depth - 1)) {
a686cd89
MB
622 /*
623 * Indirect block might be removed by
624 * truncate while we were reading it.
625 * Handling of that case: forget what we've
626 * got now, go to reread.
627 */
316cb4ef 628 err = -EAGAIN;
a686cd89 629 count = 0;
316cb4ef 630 break;
a686cd89
MB
631 }
632 blk = le32_to_cpu(*(chain[depth-1].p + count));
633 if (blk == first_block + count)
634 count++;
635 else
636 break;
637 }
316cb4ef
JK
638 if (err != -EAGAIN)
639 goto got_it;
1da177e4
LT
640 }
641
642 /* Next simple case - plain lookup or failed read of indirect block */
a686cd89
MB
643 if (!create || err == -EIO)
644 goto cleanup;
645
646 mutex_lock(&ei->truncate_mutex);
316cb4ef
JK
647 /*
648 * If the indirect block is missing while we are reading
649 * the chain(ext3_get_branch() returns -EAGAIN err), or
650 * if the chain has been changed after we grab the semaphore,
651 * (either because another process truncated this branch, or
652 * another get_block allocated this branch) re-grab the chain to see if
653 * the request block has been allocated or not.
654 *
655 * Since we already block the truncate/other get_block
656 * at this point, we will have the current copy of the chain when we
657 * splice the branch into the tree.
658 */
659 if (err == -EAGAIN || !verify_chain(chain, partial)) {
660 while (partial > chain) {
661 brelse(partial->bh);
662 partial--;
663 }
664 partial = ext2_get_branch(inode, depth, offsets, chain, &err);
665 if (!partial) {
666 count++;
667 mutex_unlock(&ei->truncate_mutex);
668 if (err)
669 goto cleanup;
670 clear_buffer_new(bh_result);
671 goto got_it;
672 }
673 }
1da177e4
LT
674
675 /*
a686cd89
MB
676 * Okay, we need to do block allocation. Lazily initialize the block
677 * allocation info here if necessary
678 */
679 if (S_ISREG(inode->i_mode) && (!ei->i_block_alloc_info))
680 ext2_init_block_alloc_info(inode);
1da177e4 681
fb01bfda 682 goal = ext2_find_goal(inode, iblock, partial);
1da177e4 683
a686cd89
MB
684 /* the number of blocks need to allocate for [d,t]indirect blocks */
685 indirect_blks = (chain + depth) - partial - 1;
686 /*
687 * Next look up the indirect map to count the totoal number of
688 * direct blocks to allocate for this branch.
689 */
690 count = ext2_blks_to_allocate(partial, indirect_blks,
691 maxblocks, blocks_to_boundary);
692 /*
693 * XXX ???? Block out ext2_truncate while we alter the tree
694 */
695 err = ext2_alloc_branch(inode, indirect_blks, &count, goal,
696 offsets + (partial - chain), partial);
697
698 if (err) {
699 mutex_unlock(&ei->truncate_mutex);
1da177e4 700 goto cleanup;
a686cd89 701 }
1da177e4 702
6d79125b
CO
703 if (ext2_use_xip(inode->i_sb)) {
704 /*
705 * we need to clear the block
706 */
707 err = ext2_clear_xip_target (inode,
708 le32_to_cpu(chain[depth-1].key));
a686cd89
MB
709 if (err) {
710 mutex_unlock(&ei->truncate_mutex);
6d79125b 711 goto cleanup;
a686cd89 712 }
6d79125b
CO
713 }
714
a686cd89
MB
715 ext2_splice_branch(inode, iblock, partial, indirect_blks, count);
716 mutex_unlock(&ei->truncate_mutex);
1da177e4 717 set_buffer_new(bh_result);
a686cd89
MB
718got_it:
719 map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key));
720 if (count > blocks_to_boundary)
721 set_buffer_boundary(bh_result);
722 err = count;
723 /* Clean up and exit */
724 partial = chain + depth - 1; /* the whole chain */
725cleanup:
726 while (partial > chain) {
727 brelse(partial->bh);
728 partial--;
729 }
730 return err;
1da177e4
LT
731}
732
a686cd89
MB
733int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create)
734{
735 unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;
736 int ret = ext2_get_blocks(inode, iblock, max_blocks,
737 bh_result, create);
738 if (ret > 0) {
739 bh_result->b_size = (ret << inode->i_blkbits);
740 ret = 0;
741 }
742 return ret;
743
744}
745
68c9d702
JB
746int ext2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
747 u64 start, u64 len)
748{
749 return generic_block_fiemap(inode, fieinfo, start, len,
750 ext2_get_block);
751}
752
1da177e4
LT
753static int ext2_writepage(struct page *page, struct writeback_control *wbc)
754{
755 return block_write_full_page(page, ext2_get_block, wbc);
756}
757
758static int ext2_readpage(struct file *file, struct page *page)
759{
760 return mpage_readpage(page, ext2_get_block);
761}
762
763static int
764ext2_readpages(struct file *file, struct address_space *mapping,
765 struct list_head *pages, unsigned nr_pages)
766{
767 return mpage_readpages(mapping, pages, nr_pages, ext2_get_block);
768}
769
1da177e4 770static int
f34fb6ec
NP
771ext2_write_begin(struct file *file, struct address_space *mapping,
772 loff_t pos, unsigned len, unsigned flags,
773 struct page **pagep, void **fsdata)
1da177e4 774{
737f2e93 775 int ret;
776
155130a4
CH
777 ret = block_write_begin(mapping, pos, len, flags, pagep,
778 ext2_get_block);
737f2e93 779 if (ret < 0)
780 ext2_write_failed(mapping, pos + len);
781 return ret;
782}
783
784static int ext2_write_end(struct file *file, struct address_space *mapping,
785 loff_t pos, unsigned len, unsigned copied,
786 struct page *page, void *fsdata)
787{
788 int ret;
789
790 ret = generic_write_end(file, mapping, pos, len, copied, page, fsdata);
791 if (ret < len)
792 ext2_write_failed(mapping, pos + len);
793 return ret;
1da177e4
LT
794}
795
03158cd7
NP
796static int
797ext2_nobh_write_begin(struct file *file, struct address_space *mapping,
798 loff_t pos, unsigned len, unsigned flags,
799 struct page **pagep, void **fsdata)
800{
737f2e93 801 int ret;
802
ea0f04e5
CH
803 ret = nobh_write_begin(mapping, pos, len, flags, pagep, fsdata,
804 ext2_get_block);
737f2e93 805 if (ret < 0)
806 ext2_write_failed(mapping, pos + len);
807 return ret;
03158cd7
NP
808}
809
1da177e4
LT
810static int ext2_nobh_writepage(struct page *page,
811 struct writeback_control *wbc)
812{
813 return nobh_writepage(page, ext2_get_block, wbc);
814}
815
816static sector_t ext2_bmap(struct address_space *mapping, sector_t block)
817{
818 return generic_block_bmap(mapping,block,ext2_get_block);
819}
820
1da177e4
LT
821static ssize_t
822ext2_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
823 loff_t offset, unsigned long nr_segs)
824{
825 struct file *file = iocb->ki_filp;
737f2e93 826 struct address_space *mapping = file->f_mapping;
827 struct inode *inode = mapping->host;
828 ssize_t ret;
829
eafdc7d1 830 ret = blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev,
737f2e93 831 iov, offset, nr_segs, ext2_get_block, NULL);
832 if (ret < 0 && (rw & WRITE))
833 ext2_write_failed(mapping, offset + iov_length(iov, nr_segs));
834 return ret;
1da177e4
LT
835}
836
837static int
838ext2_writepages(struct address_space *mapping, struct writeback_control *wbc)
839{
840 return mpage_writepages(mapping, wbc, ext2_get_block);
841}
842
f5e54d6e 843const struct address_space_operations ext2_aops = {
1da177e4
LT
844 .readpage = ext2_readpage,
845 .readpages = ext2_readpages,
846 .writepage = ext2_writepage,
847 .sync_page = block_sync_page,
f34fb6ec 848 .write_begin = ext2_write_begin,
737f2e93 849 .write_end = ext2_write_end,
1da177e4
LT
850 .bmap = ext2_bmap,
851 .direct_IO = ext2_direct_IO,
852 .writepages = ext2_writepages,
e965f963 853 .migratepage = buffer_migrate_page,
8ab22b9a 854 .is_partially_uptodate = block_is_partially_uptodate,
aa261f54 855 .error_remove_page = generic_error_remove_page,
1da177e4
LT
856};
857
f5e54d6e 858const struct address_space_operations ext2_aops_xip = {
6d79125b 859 .bmap = ext2_bmap,
70688e4d 860 .get_xip_mem = ext2_get_xip_mem,
6d79125b
CO
861};
862
f5e54d6e 863const struct address_space_operations ext2_nobh_aops = {
1da177e4
LT
864 .readpage = ext2_readpage,
865 .readpages = ext2_readpages,
866 .writepage = ext2_nobh_writepage,
867 .sync_page = block_sync_page,
03158cd7
NP
868 .write_begin = ext2_nobh_write_begin,
869 .write_end = nobh_write_end,
1da177e4
LT
870 .bmap = ext2_bmap,
871 .direct_IO = ext2_direct_IO,
872 .writepages = ext2_writepages,
e965f963 873 .migratepage = buffer_migrate_page,
aa261f54 874 .error_remove_page = generic_error_remove_page,
1da177e4
LT
875};
876
877/*
878 * Probably it should be a library function... search for first non-zero word
879 * or memcmp with zero_page, whatever is better for particular architecture.
880 * Linus?
881 */
882static inline int all_zeroes(__le32 *p, __le32 *q)
883{
884 while (p < q)
885 if (*p++)
886 return 0;
887 return 1;
888}
889
890/**
891 * ext2_find_shared - find the indirect blocks for partial truncation.
892 * @inode: inode in question
893 * @depth: depth of the affected branch
894 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
895 * @chain: place to store the pointers to partial indirect blocks
896 * @top: place to the (detached) top of branch
897 *
898 * This is a helper function used by ext2_truncate().
899 *
900 * When we do truncate() we may have to clean the ends of several indirect
901 * blocks but leave the blocks themselves alive. Block is partially
902 * truncated if some data below the new i_size is refered from it (and
903 * it is on the path to the first completely truncated data block, indeed).
904 * We have to free the top of that path along with everything to the right
905 * of the path. Since no allocation past the truncation point is possible
906 * until ext2_truncate() finishes, we may safely do the latter, but top
907 * of branch may require special attention - pageout below the truncation
908 * point might try to populate it.
909 *
910 * We atomically detach the top of branch from the tree, store the block
911 * number of its root in *@top, pointers to buffer_heads of partially
912 * truncated blocks - in @chain[].bh and pointers to their last elements
913 * that should not be removed - in @chain[].p. Return value is the pointer
914 * to last filled element of @chain.
915 *
916 * The work left to caller to do the actual freeing of subtrees:
917 * a) free the subtree starting from *@top
918 * b) free the subtrees whose roots are stored in
919 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
920 * c) free the subtrees growing from the inode past the @chain[0].p
921 * (no partially truncated stuff there).
922 */
923
924static Indirect *ext2_find_shared(struct inode *inode,
925 int depth,
926 int offsets[4],
927 Indirect chain[4],
928 __le32 *top)
929{
930 Indirect *partial, *p;
931 int k, err;
932
933 *top = 0;
934 for (k = depth; k > 1 && !offsets[k-1]; k--)
935 ;
936 partial = ext2_get_branch(inode, k, offsets, chain, &err);
937 if (!partial)
938 partial = chain + k-1;
939 /*
940 * If the branch acquired continuation since we've looked at it -
941 * fine, it should all survive and (new) top doesn't belong to us.
942 */
943 write_lock(&EXT2_I(inode)->i_meta_lock);
944 if (!partial->key && *partial->p) {
945 write_unlock(&EXT2_I(inode)->i_meta_lock);
946 goto no_top;
947 }
948 for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--)
949 ;
950 /*
951 * OK, we've found the last block that must survive. The rest of our
952 * branch should be detached before unlocking. However, if that rest
953 * of branch is all ours and does not grow immediately from the inode
954 * it's easier to cheat and just decrement partial->p.
955 */
956 if (p == chain + k - 1 && p > chain) {
957 p->p--;
958 } else {
959 *top = *p->p;
960 *p->p = 0;
961 }
962 write_unlock(&EXT2_I(inode)->i_meta_lock);
963
964 while(partial > p)
965 {
966 brelse(partial->bh);
967 partial--;
968 }
969no_top:
970 return partial;
971}
972
973/**
974 * ext2_free_data - free a list of data blocks
975 * @inode: inode we are dealing with
976 * @p: array of block numbers
977 * @q: points immediately past the end of array
978 *
979 * We are freeing all blocks refered from that array (numbers are
980 * stored as little-endian 32-bit) and updating @inode->i_blocks
981 * appropriately.
982 */
983static inline void ext2_free_data(struct inode *inode, __le32 *p, __le32 *q)
984{
985 unsigned long block_to_free = 0, count = 0;
986 unsigned long nr;
987
988 for ( ; p < q ; p++) {
989 nr = le32_to_cpu(*p);
990 if (nr) {
991 *p = 0;
992 /* accumulate blocks to free if they're contiguous */
993 if (count == 0)
994 goto free_this;
995 else if (block_to_free == nr - count)
996 count++;
997 else {
1da177e4 998 ext2_free_blocks (inode, block_to_free, count);
addacc7d 999 mark_inode_dirty(inode);
1da177e4
LT
1000 free_this:
1001 block_to_free = nr;
1002 count = 1;
1003 }
1004 }
1005 }
1006 if (count > 0) {
1da177e4 1007 ext2_free_blocks (inode, block_to_free, count);
addacc7d 1008 mark_inode_dirty(inode);
1da177e4
LT
1009 }
1010}
1011
1012/**
1013 * ext2_free_branches - free an array of branches
1014 * @inode: inode we are dealing with
1015 * @p: array of block numbers
1016 * @q: pointer immediately past the end of array
1017 * @depth: depth of the branches to free
1018 *
1019 * We are freeing all blocks refered from these branches (numbers are
1020 * stored as little-endian 32-bit) and updating @inode->i_blocks
1021 * appropriately.
1022 */
1023static void ext2_free_branches(struct inode *inode, __le32 *p, __le32 *q, int depth)
1024{
1025 struct buffer_head * bh;
1026 unsigned long nr;
1027
1028 if (depth--) {
1029 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
1030 for ( ; p < q ; p++) {
1031 nr = le32_to_cpu(*p);
1032 if (!nr)
1033 continue;
1034 *p = 0;
1035 bh = sb_bread(inode->i_sb, nr);
1036 /*
1037 * A read failure? Report error and clear slot
1038 * (should be rare).
1039 */
1040 if (!bh) {
1041 ext2_error(inode->i_sb, "ext2_free_branches",
1042 "Read failure, inode=%ld, block=%ld",
1043 inode->i_ino, nr);
1044 continue;
1045 }
1046 ext2_free_branches(inode,
1047 (__le32*)bh->b_data,
1048 (__le32*)bh->b_data + addr_per_block,
1049 depth);
1050 bforget(bh);
1051 ext2_free_blocks(inode, nr, 1);
1052 mark_inode_dirty(inode);
1053 }
1054 } else
1055 ext2_free_data(inode, p, q);
1056}
1057
737f2e93 1058static void __ext2_truncate_blocks(struct inode *inode, loff_t offset)
1da177e4
LT
1059{
1060 __le32 *i_data = EXT2_I(inode)->i_data;
a686cd89 1061 struct ext2_inode_info *ei = EXT2_I(inode);
1da177e4
LT
1062 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
1063 int offsets[4];
1064 Indirect chain[4];
1065 Indirect *partial;
1066 __le32 nr = 0;
1067 int n;
1068 long iblock;
1069 unsigned blocksize;
1da177e4 1070 blocksize = inode->i_sb->s_blocksize;
737f2e93 1071 iblock = (offset + blocksize-1) >> EXT2_BLOCK_SIZE_BITS(inode->i_sb);
1da177e4
LT
1072
1073 n = ext2_block_to_path(inode, iblock, offsets, NULL);
1074 if (n == 0)
1075 return;
1076
a686cd89
MB
1077 /*
1078 * From here we block out all ext2_get_block() callers who want to
1079 * modify the block allocation tree.
1080 */
1081 mutex_lock(&ei->truncate_mutex);
1082
1da177e4
LT
1083 if (n == 1) {
1084 ext2_free_data(inode, i_data+offsets[0],
1085 i_data + EXT2_NDIR_BLOCKS);
1086 goto do_indirects;
1087 }
1088
1089 partial = ext2_find_shared(inode, n, offsets, chain, &nr);
1090 /* Kill the top of shared branch (already detached) */
1091 if (nr) {
1092 if (partial == chain)
1093 mark_inode_dirty(inode);
1094 else
1095 mark_buffer_dirty_inode(partial->bh, inode);
1096 ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
1097 }
1098 /* Clear the ends of indirect blocks on the shared branch */
1099 while (partial > chain) {
1100 ext2_free_branches(inode,
1101 partial->p + 1,
1102 (__le32*)partial->bh->b_data+addr_per_block,
1103 (chain+n-1) - partial);
1104 mark_buffer_dirty_inode(partial->bh, inode);
1105 brelse (partial->bh);
1106 partial--;
1107 }
1108do_indirects:
1109 /* Kill the remaining (whole) subtrees */
1110 switch (offsets[0]) {
1111 default:
1112 nr = i_data[EXT2_IND_BLOCK];
1113 if (nr) {
1114 i_data[EXT2_IND_BLOCK] = 0;
1115 mark_inode_dirty(inode);
1116 ext2_free_branches(inode, &nr, &nr+1, 1);
1117 }
1118 case EXT2_IND_BLOCK:
1119 nr = i_data[EXT2_DIND_BLOCK];
1120 if (nr) {
1121 i_data[EXT2_DIND_BLOCK] = 0;
1122 mark_inode_dirty(inode);
1123 ext2_free_branches(inode, &nr, &nr+1, 2);
1124 }
1125 case EXT2_DIND_BLOCK:
1126 nr = i_data[EXT2_TIND_BLOCK];
1127 if (nr) {
1128 i_data[EXT2_TIND_BLOCK] = 0;
1129 mark_inode_dirty(inode);
1130 ext2_free_branches(inode, &nr, &nr+1, 3);
1131 }
1132 case EXT2_TIND_BLOCK:
1133 ;
1134 }
a686cd89
MB
1135
1136 ext2_discard_reservation(inode);
1137
1138 mutex_unlock(&ei->truncate_mutex);
737f2e93 1139}
1140
1141static void ext2_truncate_blocks(struct inode *inode, loff_t offset)
1142{
1143 /*
1144 * XXX: it seems like a bug here that we don't allow
1145 * IS_APPEND inode to have blocks-past-i_size trimmed off.
1146 * review and fix this.
1147 *
1148 * Also would be nice to be able to handle IO errors and such,
1149 * but that's probably too much to ask.
1150 */
1151 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1152 S_ISLNK(inode->i_mode)))
1153 return;
1154 if (ext2_inode_is_fast_symlink(inode))
1155 return;
1156 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1157 return;
1158 __ext2_truncate_blocks(inode, offset);
1159}
1160
2c27c65e 1161static int ext2_setsize(struct inode *inode, loff_t newsize)
737f2e93 1162{
737f2e93 1163 int error;
1164
737f2e93 1165 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1166 S_ISLNK(inode->i_mode)))
1167 return -EINVAL;
1168 if (ext2_inode_is_fast_symlink(inode))
1169 return -EINVAL;
1170 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1171 return -EPERM;
1172
1173 if (mapping_is_xip(inode->i_mapping))
1174 error = xip_truncate_page(inode->i_mapping, newsize);
1175 else if (test_opt(inode->i_sb, NOBH))
1176 error = nobh_truncate_page(inode->i_mapping,
1177 newsize, ext2_get_block);
1178 else
1179 error = block_truncate_page(inode->i_mapping,
1180 newsize, ext2_get_block);
1181 if (error)
1182 return error;
1183
2c27c65e 1184 truncate_setsize(inode, newsize);
737f2e93 1185 __ext2_truncate_blocks(inode, newsize);
1186
1da177e4
LT
1187 inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
1188 if (inode_needs_sync(inode)) {
1189 sync_mapping_buffers(inode->i_mapping);
1190 ext2_sync_inode (inode);
1191 } else {
1192 mark_inode_dirty(inode);
1193 }
737f2e93 1194
1195 return 0;
1da177e4
LT
1196}
1197
1198static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
1199 struct buffer_head **p)
1200{
1201 struct buffer_head * bh;
1202 unsigned long block_group;
1203 unsigned long block;
1204 unsigned long offset;
1205 struct ext2_group_desc * gdp;
1206
1207 *p = NULL;
1208 if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) ||
1209 ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
1210 goto Einval;
1211
1212 block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb);
ef2fb679 1213 gdp = ext2_get_group_desc(sb, block_group, NULL);
1da177e4
LT
1214 if (!gdp)
1215 goto Egdp;
1216 /*
1217 * Figure out the offset within the block group inode table
1218 */
1219 offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb);
1220 block = le32_to_cpu(gdp->bg_inode_table) +
1221 (offset >> EXT2_BLOCK_SIZE_BITS(sb));
1222 if (!(bh = sb_bread(sb, block)))
1223 goto Eio;
1224
1225 *p = bh;
1226 offset &= (EXT2_BLOCK_SIZE(sb) - 1);
1227 return (struct ext2_inode *) (bh->b_data + offset);
1228
1229Einval:
1230 ext2_error(sb, "ext2_get_inode", "bad inode number: %lu",
1231 (unsigned long) ino);
1232 return ERR_PTR(-EINVAL);
1233Eio:
1234 ext2_error(sb, "ext2_get_inode",
1235 "unable to read inode block - inode=%lu, block=%lu",
1236 (unsigned long) ino, block);
1237Egdp:
1238 return ERR_PTR(-EIO);
1239}
1240
1241void ext2_set_inode_flags(struct inode *inode)
1242{
1243 unsigned int flags = EXT2_I(inode)->i_flags;
1244
1245 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
1246 if (flags & EXT2_SYNC_FL)
1247 inode->i_flags |= S_SYNC;
1248 if (flags & EXT2_APPEND_FL)
1249 inode->i_flags |= S_APPEND;
1250 if (flags & EXT2_IMMUTABLE_FL)
1251 inode->i_flags |= S_IMMUTABLE;
1252 if (flags & EXT2_NOATIME_FL)
1253 inode->i_flags |= S_NOATIME;
1254 if (flags & EXT2_DIRSYNC_FL)
1255 inode->i_flags |= S_DIRSYNC;
1256}
1257
4f99ed67
JK
1258/* Propagate flags from i_flags to EXT2_I(inode)->i_flags */
1259void ext2_get_inode_flags(struct ext2_inode_info *ei)
1260{
1261 unsigned int flags = ei->vfs_inode.i_flags;
1262
1263 ei->i_flags &= ~(EXT2_SYNC_FL|EXT2_APPEND_FL|
1264 EXT2_IMMUTABLE_FL|EXT2_NOATIME_FL|EXT2_DIRSYNC_FL);
1265 if (flags & S_SYNC)
1266 ei->i_flags |= EXT2_SYNC_FL;
1267 if (flags & S_APPEND)
1268 ei->i_flags |= EXT2_APPEND_FL;
1269 if (flags & S_IMMUTABLE)
1270 ei->i_flags |= EXT2_IMMUTABLE_FL;
1271 if (flags & S_NOATIME)
1272 ei->i_flags |= EXT2_NOATIME_FL;
1273 if (flags & S_DIRSYNC)
1274 ei->i_flags |= EXT2_DIRSYNC_FL;
1275}
1276
52fcf703 1277struct inode *ext2_iget (struct super_block *sb, unsigned long ino)
1da177e4 1278{
52fcf703 1279 struct ext2_inode_info *ei;
1da177e4 1280 struct buffer_head * bh;
52fcf703
DH
1281 struct ext2_inode *raw_inode;
1282 struct inode *inode;
1283 long ret = -EIO;
1da177e4
LT
1284 int n;
1285
52fcf703
DH
1286 inode = iget_locked(sb, ino);
1287 if (!inode)
1288 return ERR_PTR(-ENOMEM);
1289 if (!(inode->i_state & I_NEW))
1290 return inode;
1291
1292 ei = EXT2_I(inode);
a686cd89
MB
1293 ei->i_block_alloc_info = NULL;
1294
52fcf703
DH
1295 raw_inode = ext2_get_inode(inode->i_sb, ino, &bh);
1296 if (IS_ERR(raw_inode)) {
1297 ret = PTR_ERR(raw_inode);
1da177e4 1298 goto bad_inode;
52fcf703 1299 }
1da177e4
LT
1300
1301 inode->i_mode = le16_to_cpu(raw_inode->i_mode);
1302 inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
1303 inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
1304 if (!(test_opt (inode->i_sb, NO_UID32))) {
1305 inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
1306 inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
1307 }
1308 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
1309 inode->i_size = le32_to_cpu(raw_inode->i_size);
4d7bf11d
MR
1310 inode->i_atime.tv_sec = (signed)le32_to_cpu(raw_inode->i_atime);
1311 inode->i_ctime.tv_sec = (signed)le32_to_cpu(raw_inode->i_ctime);
1312 inode->i_mtime.tv_sec = (signed)le32_to_cpu(raw_inode->i_mtime);
1da177e4
LT
1313 inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0;
1314 ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
1315 /* We now have enough fields to check if the inode was active or not.
1316 * This is needed because nfsd might try to access dead inodes
1317 * the test is that same one that e2fsck uses
1318 * NeilBrown 1999oct15
1319 */
1320 if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) {
1321 /* this inode is deleted */
1322 brelse (bh);
52fcf703 1323 ret = -ESTALE;
1da177e4
LT
1324 goto bad_inode;
1325 }
1da177e4
LT
1326 inode->i_blocks = le32_to_cpu(raw_inode->i_blocks);
1327 ei->i_flags = le32_to_cpu(raw_inode->i_flags);
1328 ei->i_faddr = le32_to_cpu(raw_inode->i_faddr);
1329 ei->i_frag_no = raw_inode->i_frag;
1330 ei->i_frag_size = raw_inode->i_fsize;
1331 ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl);
1332 ei->i_dir_acl = 0;
1333 if (S_ISREG(inode->i_mode))
1334 inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32;
1335 else
1336 ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl);
1337 ei->i_dtime = 0;
1338 inode->i_generation = le32_to_cpu(raw_inode->i_generation);
1339 ei->i_state = 0;
1da177e4
LT
1340 ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb);
1341 ei->i_dir_start_lookup = 0;
1342
1343 /*
1344 * NOTE! The in-memory inode i_data array is in little-endian order
1345 * even on big-endian machines: we do NOT byteswap the block numbers!
1346 */
1347 for (n = 0; n < EXT2_N_BLOCKS; n++)
1348 ei->i_data[n] = raw_inode->i_block[n];
1349
1350 if (S_ISREG(inode->i_mode)) {
1351 inode->i_op = &ext2_file_inode_operations;
6d79125b
CO
1352 if (ext2_use_xip(inode->i_sb)) {
1353 inode->i_mapping->a_ops = &ext2_aops_xip;
1354 inode->i_fop = &ext2_xip_file_operations;
1355 } else if (test_opt(inode->i_sb, NOBH)) {
1da177e4 1356 inode->i_mapping->a_ops = &ext2_nobh_aops;
6d79125b
CO
1357 inode->i_fop = &ext2_file_operations;
1358 } else {
1da177e4 1359 inode->i_mapping->a_ops = &ext2_aops;
6d79125b
CO
1360 inode->i_fop = &ext2_file_operations;
1361 }
1da177e4
LT
1362 } else if (S_ISDIR(inode->i_mode)) {
1363 inode->i_op = &ext2_dir_inode_operations;
1364 inode->i_fop = &ext2_dir_operations;
1365 if (test_opt(inode->i_sb, NOBH))
1366 inode->i_mapping->a_ops = &ext2_nobh_aops;
1367 else
1368 inode->i_mapping->a_ops = &ext2_aops;
1369 } else if (S_ISLNK(inode->i_mode)) {
8d6d0c4d 1370 if (ext2_inode_is_fast_symlink(inode)) {
1da177e4 1371 inode->i_op = &ext2_fast_symlink_inode_operations;
8d6d0c4d
DG
1372 nd_terminate_link(ei->i_data, inode->i_size,
1373 sizeof(ei->i_data) - 1);
1374 } else {
1da177e4
LT
1375 inode->i_op = &ext2_symlink_inode_operations;
1376 if (test_opt(inode->i_sb, NOBH))
1377 inode->i_mapping->a_ops = &ext2_nobh_aops;
1378 else
1379 inode->i_mapping->a_ops = &ext2_aops;
1380 }
1381 } else {
1382 inode->i_op = &ext2_special_inode_operations;
1383 if (raw_inode->i_block[0])
1384 init_special_inode(inode, inode->i_mode,
1385 old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
1386 else
1387 init_special_inode(inode, inode->i_mode,
1388 new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
1389 }
1390 brelse (bh);
1391 ext2_set_inode_flags(inode);
52fcf703
DH
1392 unlock_new_inode(inode);
1393 return inode;
1da177e4
LT
1394
1395bad_inode:
52fcf703
DH
1396 iget_failed(inode);
1397 return ERR_PTR(ret);
1da177e4
LT
1398}
1399
a9185b41 1400static int __ext2_write_inode(struct inode *inode, int do_sync)
1da177e4
LT
1401{
1402 struct ext2_inode_info *ei = EXT2_I(inode);
1403 struct super_block *sb = inode->i_sb;
1404 ino_t ino = inode->i_ino;
1405 uid_t uid = inode->i_uid;
1406 gid_t gid = inode->i_gid;
1407 struct buffer_head * bh;
1408 struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh);
1409 int n;
1410 int err = 0;
1411
1412 if (IS_ERR(raw_inode))
1413 return -EIO;
1414
1415 /* For fields not not tracking in the in-memory inode,
1416 * initialise them to zero for new inodes. */
1417 if (ei->i_state & EXT2_STATE_NEW)
1418 memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size);
1419
4f99ed67 1420 ext2_get_inode_flags(ei);
1da177e4
LT
1421 raw_inode->i_mode = cpu_to_le16(inode->i_mode);
1422 if (!(test_opt(sb, NO_UID32))) {
1423 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid));
1424 raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid));
1425/*
1426 * Fix up interoperability with old kernels. Otherwise, old inodes get
1427 * re-used with the upper 16 bits of the uid/gid intact
1428 */
1429 if (!ei->i_dtime) {
1430 raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid));
1431 raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid));
1432 } else {
1433 raw_inode->i_uid_high = 0;
1434 raw_inode->i_gid_high = 0;
1435 }
1436 } else {
1437 raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid));
1438 raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid));
1439 raw_inode->i_uid_high = 0;
1440 raw_inode->i_gid_high = 0;
1441 }
1442 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
1443 raw_inode->i_size = cpu_to_le32(inode->i_size);
1444 raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
1445 raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
1446 raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
1447
1448 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
1449 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
1450 raw_inode->i_flags = cpu_to_le32(ei->i_flags);
1451 raw_inode->i_faddr = cpu_to_le32(ei->i_faddr);
1452 raw_inode->i_frag = ei->i_frag_no;
1453 raw_inode->i_fsize = ei->i_frag_size;
1454 raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl);
1455 if (!S_ISREG(inode->i_mode))
1456 raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl);
1457 else {
1458 raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32);
1459 if (inode->i_size > 0x7fffffffULL) {
1460 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb,
1461 EXT2_FEATURE_RO_COMPAT_LARGE_FILE) ||
1462 EXT2_SB(sb)->s_es->s_rev_level ==
1463 cpu_to_le32(EXT2_GOOD_OLD_REV)) {
1464 /* If this is the first large file
1465 * created, add a flag to the superblock.
1466 */
c15271f4 1467 spin_lock(&EXT2_SB(sb)->s_lock);
1da177e4
LT
1468 ext2_update_dynamic_rev(sb);
1469 EXT2_SET_RO_COMPAT_FEATURE(sb,
1470 EXT2_FEATURE_RO_COMPAT_LARGE_FILE);
c15271f4 1471 spin_unlock(&EXT2_SB(sb)->s_lock);
1da177e4
LT
1472 ext2_write_super(sb);
1473 }
1474 }
1475 }
1476
1477 raw_inode->i_generation = cpu_to_le32(inode->i_generation);
1478 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
1479 if (old_valid_dev(inode->i_rdev)) {
1480 raw_inode->i_block[0] =
1481 cpu_to_le32(old_encode_dev(inode->i_rdev));
1482 raw_inode->i_block[1] = 0;
1483 } else {
1484 raw_inode->i_block[0] = 0;
1485 raw_inode->i_block[1] =
1486 cpu_to_le32(new_encode_dev(inode->i_rdev));
1487 raw_inode->i_block[2] = 0;
1488 }
1489 } else for (n = 0; n < EXT2_N_BLOCKS; n++)
1490 raw_inode->i_block[n] = ei->i_data[n];
1491 mark_buffer_dirty(bh);
1492 if (do_sync) {
1493 sync_dirty_buffer(bh);
1494 if (buffer_req(bh) && !buffer_uptodate(bh)) {
1495 printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1496 sb->s_id, (unsigned long) ino);
1497 err = -EIO;
1498 }
1499 }
1500 ei->i_state &= ~EXT2_STATE_NEW;
1501 brelse (bh);
1502 return err;
1503}
1504
a9185b41
CH
1505int ext2_write_inode(struct inode *inode, struct writeback_control *wbc)
1506{
1507 return __ext2_write_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
1508}
1509
1da177e4
LT
1510int ext2_sync_inode(struct inode *inode)
1511{
1512 struct writeback_control wbc = {
1513 .sync_mode = WB_SYNC_ALL,
1514 .nr_to_write = 0, /* sys_fsync did this */
1515 };
1516 return sync_inode(inode, &wbc);
1517}
1518
1519int ext2_setattr(struct dentry *dentry, struct iattr *iattr)
1520{
1521 struct inode *inode = dentry->d_inode;
1522 int error;
1523
1524 error = inode_change_ok(inode, iattr);
1525 if (error)
1526 return error;
907f4554 1527
12755627 1528 if (is_quota_modification(inode, iattr))
871a2931 1529 dquot_initialize(inode);
1da177e4
LT
1530 if ((iattr->ia_valid & ATTR_UID && iattr->ia_uid != inode->i_uid) ||
1531 (iattr->ia_valid & ATTR_GID && iattr->ia_gid != inode->i_gid)) {
b43fa828 1532 error = dquot_transfer(inode, iattr);
1da177e4
LT
1533 if (error)
1534 return error;
1535 }
af5a30d8 1536 if (iattr->ia_valid & ATTR_SIZE && iattr->ia_size != inode->i_size) {
737f2e93 1537 error = ext2_setsize(inode, iattr->ia_size);
1538 if (error)
1539 return error;
1540 }
6a1a90ad 1541 setattr_copy(inode, iattr);
737f2e93 1542 if (iattr->ia_valid & ATTR_MODE)
1da177e4 1543 error = ext2_acl_chmod(inode);
737f2e93 1544 mark_inode_dirty(inode);
1545
1da177e4
LT
1546 return error;
1547}