blob: 23fff2f87783f7e8856a6ba3679e2e0057d94767 [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>
Josef Bacik68c9d702008-10-03 17:32:43 -040034#include <linux/fiemap.h>
Duane Griffin8d6d0c42008-12-19 20:47:13 +000035#include <linux/namei.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include "ext2.h"
37#include "acl.h"
Carsten Otte6d791252005-06-23 22:05:26 -070038#include "xip.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070039
40MODULE_AUTHOR("Remy Card and others");
41MODULE_DESCRIPTION("Second Extended Filesystem");
42MODULE_LICENSE("GPL");
43
44static int ext2_update_inode(struct inode * inode, int do_sync);
45
46/*
47 * Test whether an inode is a fast symlink.
48 */
49static inline int ext2_inode_is_fast_symlink(struct inode *inode)
50{
51 int ea_blocks = EXT2_I(inode)->i_file_acl ?
52 (inode->i_sb->s_blocksize >> 9) : 0;
53
54 return (S_ISLNK(inode->i_mode) &&
55 inode->i_blocks - ea_blocks == 0);
56}
57
58/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 * Called at the last iput() if i_nlink is zero.
60 */
61void ext2_delete_inode (struct inode * inode)
62{
Mark Fashehfef26652005-09-09 13:01:31 -070063 truncate_inode_pages(&inode->i_data, 0);
64
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 if (is_bad_inode(inode))
66 goto no_delete;
67 EXT2_I(inode)->i_dtime = get_seconds();
68 mark_inode_dirty(inode);
69 ext2_update_inode(inode, inode_needs_sync(inode));
70
71 inode->i_size = 0;
72 if (inode->i_blocks)
73 ext2_truncate (inode);
74 ext2_free_inode (inode);
75
76 return;
77no_delete:
78 clear_inode(inode); /* We must guarantee clearing of inode... */
79}
80
Linus Torvalds1da177e2005-04-16 15:20:36 -070081typedef struct {
82 __le32 *p;
83 __le32 key;
84 struct buffer_head *bh;
85} Indirect;
86
87static inline void add_chain(Indirect *p, struct buffer_head *bh, __le32 *v)
88{
89 p->key = *(p->p = v);
90 p->bh = bh;
91}
92
93static inline int verify_chain(Indirect *from, Indirect *to)
94{
95 while (from <= to && from->key == *from->p)
96 from++;
97 return (from > to);
98}
99
100/**
101 * ext2_block_to_path - parse the block number into array of offsets
102 * @inode: inode in question (we are only interested in its superblock)
103 * @i_block: block number to be parsed
104 * @offsets: array to store the offsets in
105 * @boundary: set this non-zero if the referred-to block is likely to be
106 * followed (on disk) by an indirect block.
107 * To store the locations of file's data ext2 uses a data structure common
108 * for UNIX filesystems - tree of pointers anchored in the inode, with
109 * data blocks at leaves and indirect blocks in intermediate nodes.
110 * This function translates the block number into path in that tree -
111 * return value is the path length and @offsets[n] is the offset of
112 * pointer to (n+1)th node in the nth one. If @block is out of range
113 * (negative or too large) warning is printed and zero returned.
114 *
115 * Note: function doesn't find node addresses, so no IO is needed. All
116 * we need to know is the capacity of indirect blocks (taken from the
117 * inode->i_sb).
118 */
119
120/*
121 * Portability note: the last comparison (check that we fit into triple
122 * indirect block) is spelled differently, because otherwise on an
123 * architecture with 32-bit longs and 8Kb pages we might get into trouble
124 * if our filesystem had 8Kb blocks. We might use long long, but that would
125 * kill us on x86. Oh, well, at least the sign propagation does not matter -
126 * i_block would have to be negative in the very beginning, so we would not
127 * get there at all.
128 */
129
130static int ext2_block_to_path(struct inode *inode,
131 long i_block, int offsets[4], int *boundary)
132{
133 int ptrs = EXT2_ADDR_PER_BLOCK(inode->i_sb);
134 int ptrs_bits = EXT2_ADDR_PER_BLOCK_BITS(inode->i_sb);
135 const long direct_blocks = EXT2_NDIR_BLOCKS,
136 indirect_blocks = ptrs,
137 double_blocks = (1 << (ptrs_bits * 2));
138 int n = 0;
139 int final = 0;
140
141 if (i_block < 0) {
142 ext2_warning (inode->i_sb, "ext2_block_to_path", "block < 0");
143 } else if (i_block < direct_blocks) {
144 offsets[n++] = i_block;
145 final = direct_blocks;
146 } else if ( (i_block -= direct_blocks) < indirect_blocks) {
147 offsets[n++] = EXT2_IND_BLOCK;
148 offsets[n++] = i_block;
149 final = ptrs;
150 } else if ((i_block -= indirect_blocks) < double_blocks) {
151 offsets[n++] = EXT2_DIND_BLOCK;
152 offsets[n++] = i_block >> ptrs_bits;
153 offsets[n++] = i_block & (ptrs - 1);
154 final = ptrs;
155 } else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) {
156 offsets[n++] = EXT2_TIND_BLOCK;
157 offsets[n++] = i_block >> (ptrs_bits * 2);
158 offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1);
159 offsets[n++] = i_block & (ptrs - 1);
160 final = ptrs;
161 } else {
162 ext2_warning (inode->i_sb, "ext2_block_to_path", "block > big");
163 }
164 if (boundary)
Martin J. Bligha686cd82007-10-16 23:30:46 -0700165 *boundary = final - 1 - (i_block & (ptrs - 1));
166
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167 return n;
168}
169
170/**
171 * ext2_get_branch - read the chain of indirect blocks leading to data
172 * @inode: inode in question
173 * @depth: depth of the chain (1 - direct pointer, etc.)
174 * @offsets: offsets of pointers in inode/indirect blocks
175 * @chain: place to store the result
176 * @err: here we store the error value
177 *
178 * Function fills the array of triples <key, p, bh> and returns %NULL
179 * if everything went OK or the pointer to the last filled triple
180 * (incomplete one) otherwise. Upon the return chain[i].key contains
181 * the number of (i+1)-th block in the chain (as it is stored in memory,
182 * i.e. little-endian 32-bit), chain[i].p contains the address of that
183 * number (it points into struct inode for i==0 and into the bh->b_data
184 * for i>0) and chain[i].bh points to the buffer_head of i-th indirect
185 * block for i>0 and NULL for i==0. In other words, it holds the block
186 * numbers of the chain, addresses they were taken from (and where we can
187 * verify that chain did not change) and buffer_heads hosting these
188 * numbers.
189 *
190 * Function stops when it stumbles upon zero pointer (absent block)
191 * (pointer to last triple returned, *@err == 0)
192 * or when it gets an IO error reading an indirect block
193 * (ditto, *@err == -EIO)
194 * or when it notices that chain had been changed while it was reading
195 * (ditto, *@err == -EAGAIN)
196 * or when it reads all @depth-1 indirect blocks successfully and finds
197 * the whole chain, all way to the data (returns %NULL, *err == 0).
198 */
199static Indirect *ext2_get_branch(struct inode *inode,
200 int depth,
201 int *offsets,
202 Indirect chain[4],
203 int *err)
204{
205 struct super_block *sb = inode->i_sb;
206 Indirect *p = chain;
207 struct buffer_head *bh;
208
209 *err = 0;
210 /* i_data is not going away, no lock needed */
211 add_chain (chain, NULL, EXT2_I(inode)->i_data + *offsets);
212 if (!p->key)
213 goto no_block;
214 while (--depth) {
215 bh = sb_bread(sb, le32_to_cpu(p->key));
216 if (!bh)
217 goto failure;
218 read_lock(&EXT2_I(inode)->i_meta_lock);
219 if (!verify_chain(chain, p))
220 goto changed;
221 add_chain(++p, bh, (__le32*)bh->b_data + *++offsets);
222 read_unlock(&EXT2_I(inode)->i_meta_lock);
223 if (!p->key)
224 goto no_block;
225 }
226 return NULL;
227
228changed:
229 read_unlock(&EXT2_I(inode)->i_meta_lock);
230 brelse(bh);
231 *err = -EAGAIN;
232 goto no_block;
233failure:
234 *err = -EIO;
235no_block:
236 return p;
237}
238
239/**
240 * ext2_find_near - find a place for allocation with sufficient locality
241 * @inode: owner
242 * @ind: descriptor of indirect block.
243 *
Benoit Boissinot1cc8dcf2008-04-21 22:45:55 +0000244 * This function returns the preferred place for block allocation.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 * It is used when heuristic for sequential allocation fails.
246 * Rules are:
247 * + if there is a block to the left of our position - allocate near it.
248 * + if pointer will live in indirect block - allocate near that block.
249 * + if pointer will live in inode - allocate in the same cylinder group.
250 *
251 * In the latter case we colour the starting block by the callers PID to
252 * prevent it from clashing with concurrent allocations for a different inode
253 * in the same block group. The PID is used here so that functionally related
254 * files will be close-by on-disk.
255 *
256 * Caller must make sure that @ind is valid and will stay that way.
257 */
258
Akinobu Mita4c8b3122008-04-28 02:16:02 -0700259static ext2_fsblk_t ext2_find_near(struct inode *inode, Indirect *ind)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260{
261 struct ext2_inode_info *ei = EXT2_I(inode);
262 __le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data;
263 __le32 *p;
Akinobu Mita4c8b3122008-04-28 02:16:02 -0700264 ext2_fsblk_t bg_start;
265 ext2_fsblk_t colour;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266
267 /* Try to find previous block */
268 for (p = ind->p - 1; p >= start; p--)
269 if (*p)
270 return le32_to_cpu(*p);
271
272 /* No such thing, so let's try location of indirect block */
273 if (ind->bh)
274 return ind->bh->b_blocknr;
275
276 /*
277 * It is going to be refered from inode itself? OK, just put it into
278 * the same cylinder group then.
279 */
Akinobu Mita24097d12008-04-28 02:16:01 -0700280 bg_start = ext2_group_first_block_no(inode->i_sb, ei->i_block_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 colour = (current->pid % 16) *
282 (EXT2_BLOCKS_PER_GROUP(inode->i_sb) / 16);
283 return bg_start + colour;
284}
285
286/**
Benoit Boissinot1cc8dcf2008-04-21 22:45:55 +0000287 * ext2_find_goal - find a preferred place for allocation.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 * @inode: owner
289 * @block: block we want
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 * @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
Akinobu Mita4c8b3122008-04-28 02:16:02 -0700295static inline ext2_fsblk_t ext2_find_goal(struct inode *inode, long block,
296 Indirect *partial)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297{
Martin J. Bligha686cd82007-10-16 23:30:46 -0700298 struct ext2_block_alloc_info *block_i;
299
300 block_i = EXT2_I(inode)->i_block_alloc_info;
301
302 /*
303 * try the heuristic for sequential allocation,
304 * failing that at least try to get decent locality.
305 */
306 if (block_i && (block == block_i->last_alloc_logical_block + 1)
307 && (block_i->last_alloc_physical_block != 0)) {
308 return block_i->last_alloc_physical_block + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 }
Martin J. Bligha686cd82007-10-16 23:30:46 -0700310
311 return ext2_find_near(inode, partial);
312}
313
314/**
315 * ext2_blks_to_allocate: Look up the block map and count the number
316 * of direct blocks need to be allocated for the given branch.
317 *
318 * @branch: chain of indirect blocks
319 * @k: number of blocks need for indirect blocks
320 * @blks: number of data blocks to be mapped.
321 * @blocks_to_boundary: the offset in the indirect block
322 *
323 * return the total number of blocks to be allocate, including the
324 * direct and indirect blocks.
325 */
326static int
327ext2_blks_to_allocate(Indirect * branch, int k, unsigned long blks,
328 int blocks_to_boundary)
329{
330 unsigned long count = 0;
331
332 /*
333 * Simple case, [t,d]Indirect block(s) has not allocated yet
334 * then it's clear blocks on that path have not allocated
335 */
336 if (k > 0) {
337 /* right now don't hanel cross boundary allocation */
338 if (blks < blocks_to_boundary + 1)
339 count += blks;
340 else
341 count += blocks_to_boundary + 1;
342 return count;
343 }
344
345 count++;
346 while (count < blks && count <= blocks_to_boundary
347 && le32_to_cpu(*(branch[0].p + count)) == 0) {
348 count++;
349 }
350 return count;
351}
352
353/**
354 * ext2_alloc_blocks: multiple allocate blocks needed for a branch
355 * @indirect_blks: the number of blocks need to allocate for indirect
356 * blocks
357 *
358 * @new_blocks: on return it will store the new block numbers for
359 * the indirect blocks(if needed) and the first direct block,
360 * @blks: on return it will store the total number of allocated
361 * direct blocks
362 */
363static int ext2_alloc_blocks(struct inode *inode,
364 ext2_fsblk_t goal, int indirect_blks, int blks,
365 ext2_fsblk_t new_blocks[4], int *err)
366{
367 int target, i;
368 unsigned long count = 0;
369 int index = 0;
370 ext2_fsblk_t current_block = 0;
371 int ret = 0;
372
373 /*
374 * Here we try to allocate the requested multiple blocks at once,
375 * on a best-effort basis.
376 * To build a branch, we should allocate blocks for
377 * the indirect blocks(if not allocated yet), and at least
378 * the first direct block of this branch. That's the
379 * minimum number of blocks need to allocate(required)
380 */
381 target = blks + indirect_blks;
382
383 while (1) {
384 count = target;
385 /* allocating blocks for indirect blocks and direct blocks */
386 current_block = ext2_new_blocks(inode,goal,&count,err);
387 if (*err)
388 goto failed_out;
389
390 target -= count;
391 /* allocate blocks for indirect blocks */
392 while (index < indirect_blks && count) {
393 new_blocks[index++] = current_block++;
394 count--;
395 }
396
397 if (count > 0)
398 break;
399 }
400
401 /* save the new block number for the first direct block */
402 new_blocks[index] = current_block;
403
404 /* total number of blocks allocated for direct blocks */
405 ret = count;
406 *err = 0;
407 return ret;
408failed_out:
409 for (i = 0; i <index; i++)
410 ext2_free_blocks(inode, new_blocks[i], 1);
411 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412}
413
414/**
415 * ext2_alloc_branch - allocate and set up a chain of blocks.
416 * @inode: owner
417 * @num: depth of the chain (number of blocks to allocate)
418 * @offsets: offsets (in the blocks) to store the pointers to next.
419 * @branch: place to store the chain in.
420 *
421 * This function allocates @num blocks, zeroes out all but the last one,
422 * links them into chain and (if we are synchronous) writes them to disk.
423 * In other words, it prepares a branch that can be spliced onto the
424 * inode. It stores the information about that chain in the branch[], in
425 * the same format as ext2_get_branch() would do. We are calling it after
426 * we had read the existing part of chain and partial points to the last
427 * triple of that (one with zero ->key). Upon the exit we have the same
428 * picture as after the successful ext2_get_block(), excpet that in one
429 * place chain is disconnected - *branch->p is still zero (we did not
430 * set the last link), but branch->key contains the number that should
431 * be placed into *branch->p to fill that gap.
432 *
433 * If allocation fails we free all blocks we've allocated (and forget
434 * their buffer_heads) and return the error value the from failed
435 * ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
436 * as described above and return 0.
437 */
438
439static int ext2_alloc_branch(struct inode *inode,
Martin J. Bligha686cd82007-10-16 23:30:46 -0700440 int indirect_blks, int *blks, ext2_fsblk_t goal,
441 int *offsets, Indirect *branch)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442{
443 int blocksize = inode->i_sb->s_blocksize;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700444 int i, n = 0;
445 int err = 0;
446 struct buffer_head *bh;
447 int num;
448 ext2_fsblk_t new_blocks[4];
449 ext2_fsblk_t current_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
Martin J. Bligha686cd82007-10-16 23:30:46 -0700451 num = ext2_alloc_blocks(inode, goal, indirect_blks,
452 *blks, new_blocks, &err);
453 if (err)
454 return err;
455
456 branch[0].key = cpu_to_le32(new_blocks[0]);
457 /*
458 * metadata blocks and data blocks are allocated.
459 */
460 for (n = 1; n <= indirect_blks; n++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 /*
Martin J. Bligha686cd82007-10-16 23:30:46 -0700462 * Get buffer_head for parent block, zero it out
463 * and set the pointer to new one, then send
464 * parent to disk.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700466 bh = sb_getblk(inode->i_sb, new_blocks[n-1]);
467 branch[n].bh = bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 lock_buffer(bh);
469 memset(bh->b_data, 0, blocksize);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 branch[n].p = (__le32 *) bh->b_data + offsets[n];
Martin J. Bligha686cd82007-10-16 23:30:46 -0700471 branch[n].key = cpu_to_le32(new_blocks[n]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 *branch[n].p = branch[n].key;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700473 if ( n == indirect_blks) {
474 current_block = new_blocks[n];
475 /*
476 * End of chain, update the last new metablock of
477 * the chain to point to the new allocated
478 * data blocks numbers
479 */
480 for (i=1; i < num; i++)
481 *(branch[n].p + i) = cpu_to_le32(++current_block);
482 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 set_buffer_uptodate(bh);
484 unlock_buffer(bh);
485 mark_buffer_dirty_inode(bh, inode);
486 /* We used to sync bh here if IS_SYNC(inode).
487 * But we now rely upon generic_osync_inode()
488 * and b_inode_buffers. But not for directories.
489 */
490 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
491 sync_dirty_buffer(bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 }
Martin J. Bligha686cd82007-10-16 23:30:46 -0700493 *blks = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 return err;
495}
496
497/**
Martin J. Bligha686cd82007-10-16 23:30:46 -0700498 * ext2_splice_branch - splice the allocated branch onto inode.
499 * @inode: owner
500 * @block: (logical) number of block we are adding
Martin J. Bligha686cd82007-10-16 23:30:46 -0700501 * @where: location of missing link
502 * @num: number of indirect blocks we are adding
503 * @blks: number of direct blocks we are adding
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 *
Martin J. Bligha686cd82007-10-16 23:30:46 -0700505 * This function fills the missing link and does all housekeeping needed in
506 * inode (->i_blocks, etc.). In case of success we end up with the full
507 * chain to new block and return 0.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700509static void ext2_splice_branch(struct inode *inode,
510 long block, Indirect *where, int num, int blks)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 int i;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700513 struct ext2_block_alloc_info *block_i;
514 ext2_fsblk_t current_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Martin J. Bligha686cd82007-10-16 23:30:46 -0700516 block_i = EXT2_I(inode)->i_block_alloc_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517
Martin J. Bligha686cd82007-10-16 23:30:46 -0700518 /* XXX LOCKING probably should have i_meta_lock ?*/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 /* That's it */
520
521 *where->p = where->key;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522
Martin J. Bligha686cd82007-10-16 23:30:46 -0700523 /*
524 * Update the host buffer_head or inode to point to more just allocated
525 * direct blocks blocks
526 */
527 if (num == 0 && blks > 1) {
528 current_block = le32_to_cpu(where->key) + 1;
529 for (i = 1; i < blks; i++)
530 *(where->p + i ) = cpu_to_le32(current_block++);
531 }
532
533 /*
534 * update the most recently allocated logical & physical block
535 * in i_block_alloc_info, to assist find the proper goal block for next
536 * allocation
537 */
538 if (block_i) {
539 block_i->last_alloc_logical_block = block + blks - 1;
540 block_i->last_alloc_physical_block =
541 le32_to_cpu(where[num].key) + blks - 1;
542 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543
544 /* We are done with atomic stuff, now do the rest of housekeeping */
545
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 /* had we spliced it onto indirect block? */
547 if (where->bh)
548 mark_buffer_dirty_inode(where->bh, inode);
549
Martin J. Bligha686cd82007-10-16 23:30:46 -0700550 inode->i_ctime = CURRENT_TIME_SEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 mark_inode_dirty(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552}
553
554/*
555 * Allocation strategy is simple: if we have to allocate something, we will
556 * have to go the whole way to leaf. So let's do it before attaching anything
557 * to tree, set linkage between the newborn blocks, write them if sync is
558 * required, recheck the path, free and repeat if check fails, otherwise
559 * set the last missing link (that will protect us from any truncate-generated
560 * removals - all blocks on the path are immune now) and possibly force the
561 * write on the parent block.
562 * That has a nice additional property: no special recovery from the failed
563 * allocations is needed - we simply release blocks and do not touch anything
564 * reachable from inode.
Martin J. Bligha686cd82007-10-16 23:30:46 -0700565 *
566 * `handle' can be NULL if create == 0.
567 *
Martin J. Bligha686cd82007-10-16 23:30:46 -0700568 * return > 0, # of blocks mapped or allocated.
569 * return = 0, if plain lookup failed.
570 * return < 0, error case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700572static int ext2_get_blocks(struct inode *inode,
573 sector_t iblock, unsigned long maxblocks,
574 struct buffer_head *bh_result,
575 int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576{
577 int err = -EIO;
578 int offsets[4];
579 Indirect chain[4];
580 Indirect *partial;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700581 ext2_fsblk_t goal;
582 int indirect_blks;
583 int blocks_to_boundary = 0;
584 int depth;
585 struct ext2_inode_info *ei = EXT2_I(inode);
586 int count = 0;
587 ext2_fsblk_t first_block = 0;
588
589 depth = ext2_block_to_path(inode,iblock,offsets,&blocks_to_boundary);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590
591 if (depth == 0)
Martin J. Bligha686cd82007-10-16 23:30:46 -0700592 return (err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593reread:
594 partial = ext2_get_branch(inode, depth, offsets, chain, &err);
595
596 /* Simplest case - block found, no allocation needed */
597 if (!partial) {
Martin J. Bligha686cd82007-10-16 23:30:46 -0700598 first_block = le32_to_cpu(chain[depth - 1].key);
599 clear_buffer_new(bh_result); /* What's this do? */
600 count++;
601 /*map more blocks*/
602 while (count < maxblocks && count <= blocks_to_boundary) {
603 ext2_fsblk_t blk;
604
605 if (!verify_chain(chain, partial)) {
606 /*
607 * Indirect block might be removed by
608 * truncate while we were reading it.
609 * Handling of that case: forget what we've
610 * got now, go to reread.
611 */
612 count = 0;
613 goto changed;
614 }
615 blk = le32_to_cpu(*(chain[depth-1].p + count));
616 if (blk == first_block + count)
617 count++;
618 else
619 break;
620 }
621 goto got_it;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 }
623
624 /* Next simple case - plain lookup or failed read of indirect block */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700625 if (!create || err == -EIO)
626 goto cleanup;
627
628 mutex_lock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
630 /*
Martin J. Bligha686cd82007-10-16 23:30:46 -0700631 * Okay, we need to do block allocation. Lazily initialize the block
632 * allocation info here if necessary
633 */
634 if (S_ISREG(inode->i_mode) && (!ei->i_block_alloc_info))
635 ext2_init_block_alloc_info(inode);
636
Akinobu Mitafb01bfd2008-02-06 01:40:16 -0800637 goal = ext2_find_goal(inode, iblock, partial);
Martin J. Bligha686cd82007-10-16 23:30:46 -0700638
639 /* the number of blocks need to allocate for [d,t]indirect blocks */
640 indirect_blks = (chain + depth) - partial - 1;
641 /*
642 * Next look up the indirect map to count the totoal number of
643 * direct blocks to allocate for this branch.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700645 count = ext2_blks_to_allocate(partial, indirect_blks,
646 maxblocks, blocks_to_boundary);
647 /*
648 * XXX ???? Block out ext2_truncate while we alter the tree
649 */
650 err = ext2_alloc_branch(inode, indirect_blks, &count, goal,
651 offsets + (partial - chain), partial);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
Martin J. Bligha686cd82007-10-16 23:30:46 -0700653 if (err) {
654 mutex_unlock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 goto cleanup;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700656 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657
Carsten Otte6d791252005-06-23 22:05:26 -0700658 if (ext2_use_xip(inode->i_sb)) {
659 /*
660 * we need to clear the block
661 */
662 err = ext2_clear_xip_target (inode,
663 le32_to_cpu(chain[depth-1].key));
Martin J. Bligha686cd82007-10-16 23:30:46 -0700664 if (err) {
665 mutex_unlock(&ei->truncate_mutex);
Carsten Otte6d791252005-06-23 22:05:26 -0700666 goto cleanup;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700667 }
Carsten Otte6d791252005-06-23 22:05:26 -0700668 }
669
Martin J. Bligha686cd82007-10-16 23:30:46 -0700670 ext2_splice_branch(inode, iblock, partial, indirect_blks, count);
671 mutex_unlock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 set_buffer_new(bh_result);
Martin J. Bligha686cd82007-10-16 23:30:46 -0700673got_it:
674 map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key));
675 if (count > blocks_to_boundary)
676 set_buffer_boundary(bh_result);
677 err = count;
678 /* Clean up and exit */
679 partial = chain + depth - 1; /* the whole chain */
680cleanup:
681 while (partial > chain) {
682 brelse(partial->bh);
683 partial--;
684 }
685 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686changed:
687 while (partial > chain) {
688 brelse(partial->bh);
689 partial--;
690 }
691 goto reread;
692}
693
Martin J. Bligha686cd82007-10-16 23:30:46 -0700694int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create)
695{
696 unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;
697 int ret = ext2_get_blocks(inode, iblock, max_blocks,
698 bh_result, create);
699 if (ret > 0) {
700 bh_result->b_size = (ret << inode->i_blkbits);
701 ret = 0;
702 }
703 return ret;
704
705}
706
Josef Bacik68c9d702008-10-03 17:32:43 -0400707int ext2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
708 u64 start, u64 len)
709{
710 return generic_block_fiemap(inode, fieinfo, start, len,
711 ext2_get_block);
712}
713
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714static int ext2_writepage(struct page *page, struct writeback_control *wbc)
715{
716 return block_write_full_page(page, ext2_get_block, wbc);
717}
718
719static int ext2_readpage(struct file *file, struct page *page)
720{
721 return mpage_readpage(page, ext2_get_block);
722}
723
724static int
725ext2_readpages(struct file *file, struct address_space *mapping,
726 struct list_head *pages, unsigned nr_pages)
727{
728 return mpage_readpages(mapping, pages, nr_pages, ext2_get_block);
729}
730
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700731int __ext2_write_begin(struct file *file, struct address_space *mapping,
732 loff_t pos, unsigned len, unsigned flags,
733 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734{
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700735 return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
736 ext2_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737}
738
739static int
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700740ext2_write_begin(struct file *file, struct address_space *mapping,
741 loff_t pos, unsigned len, unsigned flags,
742 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743{
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700744 *pagep = NULL;
745 return __ext2_write_begin(file, mapping, pos, len, flags, pagep,fsdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746}
747
Nick Piggin03158cd2007-10-16 01:25:25 -0700748static int
749ext2_nobh_write_begin(struct file *file, struct address_space *mapping,
750 loff_t pos, unsigned len, unsigned flags,
751 struct page **pagep, void **fsdata)
752{
753 /*
754 * Dir-in-pagecache still uses ext2_write_begin. Would have to rework
755 * directory handling code to pass around offsets rather than struct
756 * pages in order to make this work easily.
757 */
758 return nobh_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
759 ext2_get_block);
760}
761
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762static int ext2_nobh_writepage(struct page *page,
763 struct writeback_control *wbc)
764{
765 return nobh_writepage(page, ext2_get_block, wbc);
766}
767
768static sector_t ext2_bmap(struct address_space *mapping, sector_t block)
769{
770 return generic_block_bmap(mapping,block,ext2_get_block);
771}
772
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773static ssize_t
774ext2_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
775 loff_t offset, unsigned long nr_segs)
776{
777 struct file *file = iocb->ki_filp;
778 struct inode *inode = file->f_mapping->host;
779
780 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
Badari Pulavarty1d8fa7a2006-03-26 01:38:02 -0800781 offset, nr_segs, ext2_get_block, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782}
783
784static int
785ext2_writepages(struct address_space *mapping, struct writeback_control *wbc)
786{
787 return mpage_writepages(mapping, wbc, ext2_get_block);
788}
789
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700790const struct address_space_operations ext2_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 .readpage = ext2_readpage,
792 .readpages = ext2_readpages,
793 .writepage = ext2_writepage,
794 .sync_page = block_sync_page,
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700795 .write_begin = ext2_write_begin,
796 .write_end = generic_write_end,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 .bmap = ext2_bmap,
798 .direct_IO = ext2_direct_IO,
799 .writepages = ext2_writepages,
Christoph Lametere965f962006-02-01 03:05:41 -0800800 .migratepage = buffer_migrate_page,
Hisashi Hifumi8ab22b92008-07-28 15:46:36 -0700801 .is_partially_uptodate = block_is_partially_uptodate,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802};
803
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700804const struct address_space_operations ext2_aops_xip = {
Carsten Otte6d791252005-06-23 22:05:26 -0700805 .bmap = ext2_bmap,
Nick Piggin70688e42008-04-28 02:13:02 -0700806 .get_xip_mem = ext2_get_xip_mem,
Carsten Otte6d791252005-06-23 22:05:26 -0700807};
808
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700809const struct address_space_operations ext2_nobh_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 .readpage = ext2_readpage,
811 .readpages = ext2_readpages,
812 .writepage = ext2_nobh_writepage,
813 .sync_page = block_sync_page,
Nick Piggin03158cd2007-10-16 01:25:25 -0700814 .write_begin = ext2_nobh_write_begin,
815 .write_end = nobh_write_end,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 .bmap = ext2_bmap,
817 .direct_IO = ext2_direct_IO,
818 .writepages = ext2_writepages,
Christoph Lametere965f962006-02-01 03:05:41 -0800819 .migratepage = buffer_migrate_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820};
821
822/*
823 * Probably it should be a library function... search for first non-zero word
824 * or memcmp with zero_page, whatever is better for particular architecture.
825 * Linus?
826 */
827static inline int all_zeroes(__le32 *p, __le32 *q)
828{
829 while (p < q)
830 if (*p++)
831 return 0;
832 return 1;
833}
834
835/**
836 * ext2_find_shared - find the indirect blocks for partial truncation.
837 * @inode: inode in question
838 * @depth: depth of the affected branch
839 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
840 * @chain: place to store the pointers to partial indirect blocks
841 * @top: place to the (detached) top of branch
842 *
843 * This is a helper function used by ext2_truncate().
844 *
845 * When we do truncate() we may have to clean the ends of several indirect
846 * blocks but leave the blocks themselves alive. Block is partially
847 * truncated if some data below the new i_size is refered from it (and
848 * it is on the path to the first completely truncated data block, indeed).
849 * We have to free the top of that path along with everything to the right
850 * of the path. Since no allocation past the truncation point is possible
851 * until ext2_truncate() finishes, we may safely do the latter, but top
852 * of branch may require special attention - pageout below the truncation
853 * point might try to populate it.
854 *
855 * We atomically detach the top of branch from the tree, store the block
856 * number of its root in *@top, pointers to buffer_heads of partially
857 * truncated blocks - in @chain[].bh and pointers to their last elements
858 * that should not be removed - in @chain[].p. Return value is the pointer
859 * to last filled element of @chain.
860 *
861 * The work left to caller to do the actual freeing of subtrees:
862 * a) free the subtree starting from *@top
863 * b) free the subtrees whose roots are stored in
864 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
865 * c) free the subtrees growing from the inode past the @chain[0].p
866 * (no partially truncated stuff there).
867 */
868
869static Indirect *ext2_find_shared(struct inode *inode,
870 int depth,
871 int offsets[4],
872 Indirect chain[4],
873 __le32 *top)
874{
875 Indirect *partial, *p;
876 int k, err;
877
878 *top = 0;
879 for (k = depth; k > 1 && !offsets[k-1]; k--)
880 ;
881 partial = ext2_get_branch(inode, k, offsets, chain, &err);
882 if (!partial)
883 partial = chain + k-1;
884 /*
885 * If the branch acquired continuation since we've looked at it -
886 * fine, it should all survive and (new) top doesn't belong to us.
887 */
888 write_lock(&EXT2_I(inode)->i_meta_lock);
889 if (!partial->key && *partial->p) {
890 write_unlock(&EXT2_I(inode)->i_meta_lock);
891 goto no_top;
892 }
893 for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--)
894 ;
895 /*
896 * OK, we've found the last block that must survive. The rest of our
897 * branch should be detached before unlocking. However, if that rest
898 * of branch is all ours and does not grow immediately from the inode
899 * it's easier to cheat and just decrement partial->p.
900 */
901 if (p == chain + k - 1 && p > chain) {
902 p->p--;
903 } else {
904 *top = *p->p;
905 *p->p = 0;
906 }
907 write_unlock(&EXT2_I(inode)->i_meta_lock);
908
909 while(partial > p)
910 {
911 brelse(partial->bh);
912 partial--;
913 }
914no_top:
915 return partial;
916}
917
918/**
919 * ext2_free_data - free a list of data blocks
920 * @inode: inode we are dealing with
921 * @p: array of block numbers
922 * @q: points immediately past the end of array
923 *
924 * We are freeing all blocks refered from that array (numbers are
925 * stored as little-endian 32-bit) and updating @inode->i_blocks
926 * appropriately.
927 */
928static inline void ext2_free_data(struct inode *inode, __le32 *p, __le32 *q)
929{
930 unsigned long block_to_free = 0, count = 0;
931 unsigned long nr;
932
933 for ( ; p < q ; p++) {
934 nr = le32_to_cpu(*p);
935 if (nr) {
936 *p = 0;
937 /* accumulate blocks to free if they're contiguous */
938 if (count == 0)
939 goto free_this;
940 else if (block_to_free == nr - count)
941 count++;
942 else {
943 mark_inode_dirty(inode);
944 ext2_free_blocks (inode, block_to_free, count);
945 free_this:
946 block_to_free = nr;
947 count = 1;
948 }
949 }
950 }
951 if (count > 0) {
952 mark_inode_dirty(inode);
953 ext2_free_blocks (inode, block_to_free, count);
954 }
955}
956
957/**
958 * ext2_free_branches - free an array of branches
959 * @inode: inode we are dealing with
960 * @p: array of block numbers
961 * @q: pointer immediately past the end of array
962 * @depth: depth of the branches to free
963 *
964 * We are freeing all blocks refered from these branches (numbers are
965 * stored as little-endian 32-bit) and updating @inode->i_blocks
966 * appropriately.
967 */
968static void ext2_free_branches(struct inode *inode, __le32 *p, __le32 *q, int depth)
969{
970 struct buffer_head * bh;
971 unsigned long nr;
972
973 if (depth--) {
974 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
975 for ( ; p < q ; p++) {
976 nr = le32_to_cpu(*p);
977 if (!nr)
978 continue;
979 *p = 0;
980 bh = sb_bread(inode->i_sb, nr);
981 /*
982 * A read failure? Report error and clear slot
983 * (should be rare).
984 */
985 if (!bh) {
986 ext2_error(inode->i_sb, "ext2_free_branches",
987 "Read failure, inode=%ld, block=%ld",
988 inode->i_ino, nr);
989 continue;
990 }
991 ext2_free_branches(inode,
992 (__le32*)bh->b_data,
993 (__le32*)bh->b_data + addr_per_block,
994 depth);
995 bforget(bh);
996 ext2_free_blocks(inode, nr, 1);
997 mark_inode_dirty(inode);
998 }
999 } else
1000 ext2_free_data(inode, p, q);
1001}
1002
Martin J. Bligha686cd82007-10-16 23:30:46 -07001003void ext2_truncate(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004{
1005 __le32 *i_data = EXT2_I(inode)->i_data;
Martin J. Bligha686cd82007-10-16 23:30:46 -07001006 struct ext2_inode_info *ei = EXT2_I(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
1008 int offsets[4];
1009 Indirect chain[4];
1010 Indirect *partial;
1011 __le32 nr = 0;
1012 int n;
1013 long iblock;
1014 unsigned blocksize;
1015
1016 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1017 S_ISLNK(inode->i_mode)))
1018 return;
1019 if (ext2_inode_is_fast_symlink(inode))
1020 return;
1021 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1022 return;
1023
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 blocksize = inode->i_sb->s_blocksize;
1025 iblock = (inode->i_size + blocksize-1)
1026 >> EXT2_BLOCK_SIZE_BITS(inode->i_sb);
1027
Carsten Otte6d791252005-06-23 22:05:26 -07001028 if (mapping_is_xip(inode->i_mapping))
1029 xip_truncate_page(inode->i_mapping, inode->i_size);
1030 else if (test_opt(inode->i_sb, NOBH))
Nick Piggin03158cd2007-10-16 01:25:25 -07001031 nobh_truncate_page(inode->i_mapping,
1032 inode->i_size, ext2_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 else
1034 block_truncate_page(inode->i_mapping,
1035 inode->i_size, ext2_get_block);
1036
1037 n = ext2_block_to_path(inode, iblock, offsets, NULL);
1038 if (n == 0)
1039 return;
1040
Martin J. Bligha686cd82007-10-16 23:30:46 -07001041 /*
1042 * From here we block out all ext2_get_block() callers who want to
1043 * modify the block allocation tree.
1044 */
1045 mutex_lock(&ei->truncate_mutex);
1046
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 if (n == 1) {
1048 ext2_free_data(inode, i_data+offsets[0],
1049 i_data + EXT2_NDIR_BLOCKS);
1050 goto do_indirects;
1051 }
1052
1053 partial = ext2_find_shared(inode, n, offsets, chain, &nr);
1054 /* Kill the top of shared branch (already detached) */
1055 if (nr) {
1056 if (partial == chain)
1057 mark_inode_dirty(inode);
1058 else
1059 mark_buffer_dirty_inode(partial->bh, inode);
1060 ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
1061 }
1062 /* Clear the ends of indirect blocks on the shared branch */
1063 while (partial > chain) {
1064 ext2_free_branches(inode,
1065 partial->p + 1,
1066 (__le32*)partial->bh->b_data+addr_per_block,
1067 (chain+n-1) - partial);
1068 mark_buffer_dirty_inode(partial->bh, inode);
1069 brelse (partial->bh);
1070 partial--;
1071 }
1072do_indirects:
1073 /* Kill the remaining (whole) subtrees */
1074 switch (offsets[0]) {
1075 default:
1076 nr = i_data[EXT2_IND_BLOCK];
1077 if (nr) {
1078 i_data[EXT2_IND_BLOCK] = 0;
1079 mark_inode_dirty(inode);
1080 ext2_free_branches(inode, &nr, &nr+1, 1);
1081 }
1082 case EXT2_IND_BLOCK:
1083 nr = i_data[EXT2_DIND_BLOCK];
1084 if (nr) {
1085 i_data[EXT2_DIND_BLOCK] = 0;
1086 mark_inode_dirty(inode);
1087 ext2_free_branches(inode, &nr, &nr+1, 2);
1088 }
1089 case EXT2_DIND_BLOCK:
1090 nr = i_data[EXT2_TIND_BLOCK];
1091 if (nr) {
1092 i_data[EXT2_TIND_BLOCK] = 0;
1093 mark_inode_dirty(inode);
1094 ext2_free_branches(inode, &nr, &nr+1, 3);
1095 }
1096 case EXT2_TIND_BLOCK:
1097 ;
1098 }
Martin J. Bligha686cd82007-10-16 23:30:46 -07001099
1100 ext2_discard_reservation(inode);
1101
1102 mutex_unlock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
1104 if (inode_needs_sync(inode)) {
1105 sync_mapping_buffers(inode->i_mapping);
1106 ext2_sync_inode (inode);
1107 } else {
1108 mark_inode_dirty(inode);
1109 }
1110}
1111
1112static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
1113 struct buffer_head **p)
1114{
1115 struct buffer_head * bh;
1116 unsigned long block_group;
1117 unsigned long block;
1118 unsigned long offset;
1119 struct ext2_group_desc * gdp;
1120
1121 *p = NULL;
1122 if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) ||
1123 ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
1124 goto Einval;
1125
1126 block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb);
Eric Sandeenef2fb672007-10-16 23:26:30 -07001127 gdp = ext2_get_group_desc(sb, block_group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 if (!gdp)
1129 goto Egdp;
1130 /*
1131 * Figure out the offset within the block group inode table
1132 */
1133 offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb);
1134 block = le32_to_cpu(gdp->bg_inode_table) +
1135 (offset >> EXT2_BLOCK_SIZE_BITS(sb));
1136 if (!(bh = sb_bread(sb, block)))
1137 goto Eio;
1138
1139 *p = bh;
1140 offset &= (EXT2_BLOCK_SIZE(sb) - 1);
1141 return (struct ext2_inode *) (bh->b_data + offset);
1142
1143Einval:
1144 ext2_error(sb, "ext2_get_inode", "bad inode number: %lu",
1145 (unsigned long) ino);
1146 return ERR_PTR(-EINVAL);
1147Eio:
1148 ext2_error(sb, "ext2_get_inode",
1149 "unable to read inode block - inode=%lu, block=%lu",
1150 (unsigned long) ino, block);
1151Egdp:
1152 return ERR_PTR(-EIO);
1153}
1154
1155void ext2_set_inode_flags(struct inode *inode)
1156{
1157 unsigned int flags = EXT2_I(inode)->i_flags;
1158
1159 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
1160 if (flags & EXT2_SYNC_FL)
1161 inode->i_flags |= S_SYNC;
1162 if (flags & EXT2_APPEND_FL)
1163 inode->i_flags |= S_APPEND;
1164 if (flags & EXT2_IMMUTABLE_FL)
1165 inode->i_flags |= S_IMMUTABLE;
1166 if (flags & EXT2_NOATIME_FL)
1167 inode->i_flags |= S_NOATIME;
1168 if (flags & EXT2_DIRSYNC_FL)
1169 inode->i_flags |= S_DIRSYNC;
1170}
1171
Jan Kara4f99ed62007-05-08 00:31:04 -07001172/* Propagate flags from i_flags to EXT2_I(inode)->i_flags */
1173void ext2_get_inode_flags(struct ext2_inode_info *ei)
1174{
1175 unsigned int flags = ei->vfs_inode.i_flags;
1176
1177 ei->i_flags &= ~(EXT2_SYNC_FL|EXT2_APPEND_FL|
1178 EXT2_IMMUTABLE_FL|EXT2_NOATIME_FL|EXT2_DIRSYNC_FL);
1179 if (flags & S_SYNC)
1180 ei->i_flags |= EXT2_SYNC_FL;
1181 if (flags & S_APPEND)
1182 ei->i_flags |= EXT2_APPEND_FL;
1183 if (flags & S_IMMUTABLE)
1184 ei->i_flags |= EXT2_IMMUTABLE_FL;
1185 if (flags & S_NOATIME)
1186 ei->i_flags |= EXT2_NOATIME_FL;
1187 if (flags & S_DIRSYNC)
1188 ei->i_flags |= EXT2_DIRSYNC_FL;
1189}
1190
David Howells52fcf702008-02-07 00:15:35 -08001191struct inode *ext2_iget (struct super_block *sb, unsigned long ino)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192{
David Howells52fcf702008-02-07 00:15:35 -08001193 struct ext2_inode_info *ei;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 struct buffer_head * bh;
David Howells52fcf702008-02-07 00:15:35 -08001195 struct ext2_inode *raw_inode;
1196 struct inode *inode;
1197 long ret = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 int n;
1199
David Howells52fcf702008-02-07 00:15:35 -08001200 inode = iget_locked(sb, ino);
1201 if (!inode)
1202 return ERR_PTR(-ENOMEM);
1203 if (!(inode->i_state & I_NEW))
1204 return inode;
1205
1206 ei = EXT2_I(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207#ifdef CONFIG_EXT2_FS_POSIX_ACL
1208 ei->i_acl = EXT2_ACL_NOT_CACHED;
1209 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
1210#endif
Martin J. Bligha686cd82007-10-16 23:30:46 -07001211 ei->i_block_alloc_info = NULL;
1212
David Howells52fcf702008-02-07 00:15:35 -08001213 raw_inode = ext2_get_inode(inode->i_sb, ino, &bh);
1214 if (IS_ERR(raw_inode)) {
1215 ret = PTR_ERR(raw_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 goto bad_inode;
David Howells52fcf702008-02-07 00:15:35 -08001217 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218
1219 inode->i_mode = le16_to_cpu(raw_inode->i_mode);
1220 inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
1221 inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
1222 if (!(test_opt (inode->i_sb, NO_UID32))) {
1223 inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
1224 inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
1225 }
1226 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
1227 inode->i_size = le32_to_cpu(raw_inode->i_size);
Markus Rechberger4d7bf112007-05-08 00:23:39 -07001228 inode->i_atime.tv_sec = (signed)le32_to_cpu(raw_inode->i_atime);
1229 inode->i_ctime.tv_sec = (signed)le32_to_cpu(raw_inode->i_ctime);
1230 inode->i_mtime.tv_sec = (signed)le32_to_cpu(raw_inode->i_mtime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0;
1232 ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
1233 /* We now have enough fields to check if the inode was active or not.
1234 * This is needed because nfsd might try to access dead inodes
1235 * the test is that same one that e2fsck uses
1236 * NeilBrown 1999oct15
1237 */
1238 if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) {
1239 /* this inode is deleted */
1240 brelse (bh);
David Howells52fcf702008-02-07 00:15:35 -08001241 ret = -ESTALE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 goto bad_inode;
1243 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 inode->i_blocks = le32_to_cpu(raw_inode->i_blocks);
1245 ei->i_flags = le32_to_cpu(raw_inode->i_flags);
1246 ei->i_faddr = le32_to_cpu(raw_inode->i_faddr);
1247 ei->i_frag_no = raw_inode->i_frag;
1248 ei->i_frag_size = raw_inode->i_fsize;
1249 ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl);
1250 ei->i_dir_acl = 0;
1251 if (S_ISREG(inode->i_mode))
1252 inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32;
1253 else
1254 ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl);
1255 ei->i_dtime = 0;
1256 inode->i_generation = le32_to_cpu(raw_inode->i_generation);
1257 ei->i_state = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb);
1259 ei->i_dir_start_lookup = 0;
1260
1261 /*
1262 * NOTE! The in-memory inode i_data array is in little-endian order
1263 * even on big-endian machines: we do NOT byteswap the block numbers!
1264 */
1265 for (n = 0; n < EXT2_N_BLOCKS; n++)
1266 ei->i_data[n] = raw_inode->i_block[n];
1267
1268 if (S_ISREG(inode->i_mode)) {
1269 inode->i_op = &ext2_file_inode_operations;
Carsten Otte6d791252005-06-23 22:05:26 -07001270 if (ext2_use_xip(inode->i_sb)) {
1271 inode->i_mapping->a_ops = &ext2_aops_xip;
1272 inode->i_fop = &ext2_xip_file_operations;
1273 } else if (test_opt(inode->i_sb, NOBH)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 inode->i_mapping->a_ops = &ext2_nobh_aops;
Carsten Otte6d791252005-06-23 22:05:26 -07001275 inode->i_fop = &ext2_file_operations;
1276 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 inode->i_mapping->a_ops = &ext2_aops;
Carsten Otte6d791252005-06-23 22:05:26 -07001278 inode->i_fop = &ext2_file_operations;
1279 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 } else if (S_ISDIR(inode->i_mode)) {
1281 inode->i_op = &ext2_dir_inode_operations;
1282 inode->i_fop = &ext2_dir_operations;
1283 if (test_opt(inode->i_sb, NOBH))
1284 inode->i_mapping->a_ops = &ext2_nobh_aops;
1285 else
1286 inode->i_mapping->a_ops = &ext2_aops;
1287 } else if (S_ISLNK(inode->i_mode)) {
Duane Griffin8d6d0c42008-12-19 20:47:13 +00001288 if (ext2_inode_is_fast_symlink(inode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 inode->i_op = &ext2_fast_symlink_inode_operations;
Duane Griffin8d6d0c42008-12-19 20:47:13 +00001290 nd_terminate_link(ei->i_data, inode->i_size,
1291 sizeof(ei->i_data) - 1);
1292 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 inode->i_op = &ext2_symlink_inode_operations;
1294 if (test_opt(inode->i_sb, NOBH))
1295 inode->i_mapping->a_ops = &ext2_nobh_aops;
1296 else
1297 inode->i_mapping->a_ops = &ext2_aops;
1298 }
1299 } else {
1300 inode->i_op = &ext2_special_inode_operations;
1301 if (raw_inode->i_block[0])
1302 init_special_inode(inode, inode->i_mode,
1303 old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
1304 else
1305 init_special_inode(inode, inode->i_mode,
1306 new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
1307 }
1308 brelse (bh);
1309 ext2_set_inode_flags(inode);
David Howells52fcf702008-02-07 00:15:35 -08001310 unlock_new_inode(inode);
1311 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312
1313bad_inode:
David Howells52fcf702008-02-07 00:15:35 -08001314 iget_failed(inode);
1315 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316}
1317
1318static int ext2_update_inode(struct inode * inode, int do_sync)
1319{
1320 struct ext2_inode_info *ei = EXT2_I(inode);
1321 struct super_block *sb = inode->i_sb;
1322 ino_t ino = inode->i_ino;
1323 uid_t uid = inode->i_uid;
1324 gid_t gid = inode->i_gid;
1325 struct buffer_head * bh;
1326 struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh);
1327 int n;
1328 int err = 0;
1329
1330 if (IS_ERR(raw_inode))
1331 return -EIO;
1332
1333 /* For fields not not tracking in the in-memory inode,
1334 * initialise them to zero for new inodes. */
1335 if (ei->i_state & EXT2_STATE_NEW)
1336 memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size);
1337
Jan Kara4f99ed62007-05-08 00:31:04 -07001338 ext2_get_inode_flags(ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 raw_inode->i_mode = cpu_to_le16(inode->i_mode);
1340 if (!(test_opt(sb, NO_UID32))) {
1341 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid));
1342 raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid));
1343/*
1344 * Fix up interoperability with old kernels. Otherwise, old inodes get
1345 * re-used with the upper 16 bits of the uid/gid intact
1346 */
1347 if (!ei->i_dtime) {
1348 raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid));
1349 raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid));
1350 } else {
1351 raw_inode->i_uid_high = 0;
1352 raw_inode->i_gid_high = 0;
1353 }
1354 } else {
1355 raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid));
1356 raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid));
1357 raw_inode->i_uid_high = 0;
1358 raw_inode->i_gid_high = 0;
1359 }
1360 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
1361 raw_inode->i_size = cpu_to_le32(inode->i_size);
1362 raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
1363 raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
1364 raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
1365
1366 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
1367 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
1368 raw_inode->i_flags = cpu_to_le32(ei->i_flags);
1369 raw_inode->i_faddr = cpu_to_le32(ei->i_faddr);
1370 raw_inode->i_frag = ei->i_frag_no;
1371 raw_inode->i_fsize = ei->i_frag_size;
1372 raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl);
1373 if (!S_ISREG(inode->i_mode))
1374 raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl);
1375 else {
1376 raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32);
1377 if (inode->i_size > 0x7fffffffULL) {
1378 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb,
1379 EXT2_FEATURE_RO_COMPAT_LARGE_FILE) ||
1380 EXT2_SB(sb)->s_es->s_rev_level ==
1381 cpu_to_le32(EXT2_GOOD_OLD_REV)) {
1382 /* If this is the first large file
1383 * created, add a flag to the superblock.
1384 */
1385 lock_kernel();
1386 ext2_update_dynamic_rev(sb);
1387 EXT2_SET_RO_COMPAT_FEATURE(sb,
1388 EXT2_FEATURE_RO_COMPAT_LARGE_FILE);
1389 unlock_kernel();
1390 ext2_write_super(sb);
1391 }
1392 }
1393 }
1394
1395 raw_inode->i_generation = cpu_to_le32(inode->i_generation);
1396 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
1397 if (old_valid_dev(inode->i_rdev)) {
1398 raw_inode->i_block[0] =
1399 cpu_to_le32(old_encode_dev(inode->i_rdev));
1400 raw_inode->i_block[1] = 0;
1401 } else {
1402 raw_inode->i_block[0] = 0;
1403 raw_inode->i_block[1] =
1404 cpu_to_le32(new_encode_dev(inode->i_rdev));
1405 raw_inode->i_block[2] = 0;
1406 }
1407 } else for (n = 0; n < EXT2_N_BLOCKS; n++)
1408 raw_inode->i_block[n] = ei->i_data[n];
1409 mark_buffer_dirty(bh);
1410 if (do_sync) {
1411 sync_dirty_buffer(bh);
1412 if (buffer_req(bh) && !buffer_uptodate(bh)) {
1413 printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1414 sb->s_id, (unsigned long) ino);
1415 err = -EIO;
1416 }
1417 }
1418 ei->i_state &= ~EXT2_STATE_NEW;
1419 brelse (bh);
1420 return err;
1421}
1422
1423int ext2_write_inode(struct inode *inode, int wait)
1424{
1425 return ext2_update_inode(inode, wait);
1426}
1427
1428int ext2_sync_inode(struct inode *inode)
1429{
1430 struct writeback_control wbc = {
1431 .sync_mode = WB_SYNC_ALL,
1432 .nr_to_write = 0, /* sys_fsync did this */
1433 };
1434 return sync_inode(inode, &wbc);
1435}
1436
1437int ext2_setattr(struct dentry *dentry, struct iattr *iattr)
1438{
1439 struct inode *inode = dentry->d_inode;
1440 int error;
1441
1442 error = inode_change_ok(inode, iattr);
1443 if (error)
1444 return error;
1445 if ((iattr->ia_valid & ATTR_UID && iattr->ia_uid != inode->i_uid) ||
1446 (iattr->ia_valid & ATTR_GID && iattr->ia_gid != inode->i_gid)) {
1447 error = DQUOT_TRANSFER(inode, iattr) ? -EDQUOT : 0;
1448 if (error)
1449 return error;
1450 }
1451 error = inode_setattr(inode, iattr);
1452 if (!error && (iattr->ia_valid & ATTR_MODE))
1453 error = ext2_acl_chmod(inode);
1454 return error;
1455}