blob: 02b39a5deb74033d62deb7c5ccd86d96b6ddf9b9 [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
501 * @chain: chain of indirect blocks (with a missing link - see
502 * ext2_alloc_branch)
503 * @where: location of missing link
504 * @num: number of indirect blocks we are adding
505 * @blks: number of direct blocks we are adding
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 *
Martin J. Bligha686cd82007-10-16 23:30:46 -0700507 * This function fills the missing link and does all housekeeping needed in
508 * inode (->i_blocks, etc.). In case of success we end up with the full
509 * chain to new block and return 0.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700511static void ext2_splice_branch(struct inode *inode,
512 long block, Indirect *where, int num, int blks)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 int i;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700515 struct ext2_block_alloc_info *block_i;
516 ext2_fsblk_t current_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517
Martin J. Bligha686cd82007-10-16 23:30:46 -0700518 block_i = EXT2_I(inode)->i_block_alloc_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519
Martin J. Bligha686cd82007-10-16 23:30:46 -0700520 /* XXX LOCKING probably should have i_meta_lock ?*/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 /* That's it */
522
523 *where->p = where->key;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524
Martin J. Bligha686cd82007-10-16 23:30:46 -0700525 /*
526 * Update the host buffer_head or inode to point to more just allocated
527 * direct blocks blocks
528 */
529 if (num == 0 && blks > 1) {
530 current_block = le32_to_cpu(where->key) + 1;
531 for (i = 1; i < blks; i++)
532 *(where->p + i ) = cpu_to_le32(current_block++);
533 }
534
535 /*
536 * update the most recently allocated logical & physical block
537 * in i_block_alloc_info, to assist find the proper goal block for next
538 * allocation
539 */
540 if (block_i) {
541 block_i->last_alloc_logical_block = block + blks - 1;
542 block_i->last_alloc_physical_block =
543 le32_to_cpu(where[num].key) + blks - 1;
544 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
546 /* We are done with atomic stuff, now do the rest of housekeeping */
547
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 /* had we spliced it onto indirect block? */
549 if (where->bh)
550 mark_buffer_dirty_inode(where->bh, inode);
551
Martin J. Bligha686cd82007-10-16 23:30:46 -0700552 inode->i_ctime = CURRENT_TIME_SEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 mark_inode_dirty(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554}
555
556/*
557 * Allocation strategy is simple: if we have to allocate something, we will
558 * have to go the whole way to leaf. So let's do it before attaching anything
559 * to tree, set linkage between the newborn blocks, write them if sync is
560 * required, recheck the path, free and repeat if check fails, otherwise
561 * set the last missing link (that will protect us from any truncate-generated
562 * removals - all blocks on the path are immune now) and possibly force the
563 * write on the parent block.
564 * That has a nice additional property: no special recovery from the failed
565 * allocations is needed - we simply release blocks and do not touch anything
566 * reachable from inode.
Martin J. Bligha686cd82007-10-16 23:30:46 -0700567 *
568 * `handle' can be NULL if create == 0.
569 *
Martin J. Bligha686cd82007-10-16 23:30:46 -0700570 * return > 0, # of blocks mapped or allocated.
571 * return = 0, if plain lookup failed.
572 * return < 0, error case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700574static int ext2_get_blocks(struct inode *inode,
575 sector_t iblock, unsigned long maxblocks,
576 struct buffer_head *bh_result,
577 int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578{
579 int err = -EIO;
580 int offsets[4];
581 Indirect chain[4];
582 Indirect *partial;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700583 ext2_fsblk_t goal;
584 int indirect_blks;
585 int blocks_to_boundary = 0;
586 int depth;
587 struct ext2_inode_info *ei = EXT2_I(inode);
588 int count = 0;
589 ext2_fsblk_t first_block = 0;
590
591 depth = ext2_block_to_path(inode,iblock,offsets,&blocks_to_boundary);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
593 if (depth == 0)
Martin J. Bligha686cd82007-10-16 23:30:46 -0700594 return (err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595reread:
596 partial = ext2_get_branch(inode, depth, offsets, chain, &err);
597
598 /* Simplest case - block found, no allocation needed */
599 if (!partial) {
Martin J. Bligha686cd82007-10-16 23:30:46 -0700600 first_block = le32_to_cpu(chain[depth - 1].key);
601 clear_buffer_new(bh_result); /* What's this do? */
602 count++;
603 /*map more blocks*/
604 while (count < maxblocks && count <= blocks_to_boundary) {
605 ext2_fsblk_t blk;
606
607 if (!verify_chain(chain, partial)) {
608 /*
609 * Indirect block might be removed by
610 * truncate while we were reading it.
611 * Handling of that case: forget what we've
612 * got now, go to reread.
613 */
614 count = 0;
615 goto changed;
616 }
617 blk = le32_to_cpu(*(chain[depth-1].p + count));
618 if (blk == first_block + count)
619 count++;
620 else
621 break;
622 }
623 goto got_it;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 }
625
626 /* Next simple case - plain lookup or failed read of indirect block */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700627 if (!create || err == -EIO)
628 goto cleanup;
629
630 mutex_lock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631
632 /*
Martin J. Bligha686cd82007-10-16 23:30:46 -0700633 * Okay, we need to do block allocation. Lazily initialize the block
634 * allocation info here if necessary
635 */
636 if (S_ISREG(inode->i_mode) && (!ei->i_block_alloc_info))
637 ext2_init_block_alloc_info(inode);
638
Akinobu Mitafb01bfd2008-02-06 01:40:16 -0800639 goal = ext2_find_goal(inode, iblock, partial);
Martin J. Bligha686cd82007-10-16 23:30:46 -0700640
641 /* the number of blocks need to allocate for [d,t]indirect blocks */
642 indirect_blks = (chain + depth) - partial - 1;
643 /*
644 * Next look up the indirect map to count the totoal number of
645 * direct blocks to allocate for this branch.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 */
Martin J. Bligha686cd82007-10-16 23:30:46 -0700647 count = ext2_blks_to_allocate(partial, indirect_blks,
648 maxblocks, blocks_to_boundary);
649 /*
650 * XXX ???? Block out ext2_truncate while we alter the tree
651 */
652 err = ext2_alloc_branch(inode, indirect_blks, &count, goal,
653 offsets + (partial - chain), partial);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654
Martin J. Bligha686cd82007-10-16 23:30:46 -0700655 if (err) {
656 mutex_unlock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 goto cleanup;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700658 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659
Carsten Otte6d791252005-06-23 22:05:26 -0700660 if (ext2_use_xip(inode->i_sb)) {
661 /*
662 * we need to clear the block
663 */
664 err = ext2_clear_xip_target (inode,
665 le32_to_cpu(chain[depth-1].key));
Martin J. Bligha686cd82007-10-16 23:30:46 -0700666 if (err) {
667 mutex_unlock(&ei->truncate_mutex);
Carsten Otte6d791252005-06-23 22:05:26 -0700668 goto cleanup;
Martin J. Bligha686cd82007-10-16 23:30:46 -0700669 }
Carsten Otte6d791252005-06-23 22:05:26 -0700670 }
671
Martin J. Bligha686cd82007-10-16 23:30:46 -0700672 ext2_splice_branch(inode, iblock, partial, indirect_blks, count);
673 mutex_unlock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 set_buffer_new(bh_result);
Martin J. Bligha686cd82007-10-16 23:30:46 -0700675got_it:
676 map_bh(bh_result, inode->i_sb, le32_to_cpu(chain[depth-1].key));
677 if (count > blocks_to_boundary)
678 set_buffer_boundary(bh_result);
679 err = count;
680 /* Clean up and exit */
681 partial = chain + depth - 1; /* the whole chain */
682cleanup:
683 while (partial > chain) {
684 brelse(partial->bh);
685 partial--;
686 }
687 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688changed:
689 while (partial > chain) {
690 brelse(partial->bh);
691 partial--;
692 }
693 goto reread;
694}
695
Martin J. Bligha686cd82007-10-16 23:30:46 -0700696int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create)
697{
698 unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;
699 int ret = ext2_get_blocks(inode, iblock, max_blocks,
700 bh_result, create);
701 if (ret > 0) {
702 bh_result->b_size = (ret << inode->i_blkbits);
703 ret = 0;
704 }
705 return ret;
706
707}
708
Josef Bacik68c9d702008-10-03 17:32:43 -0400709int ext2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
710 u64 start, u64 len)
711{
712 return generic_block_fiemap(inode, fieinfo, start, len,
713 ext2_get_block);
714}
715
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716static int ext2_writepage(struct page *page, struct writeback_control *wbc)
717{
718 return block_write_full_page(page, ext2_get_block, wbc);
719}
720
721static int ext2_readpage(struct file *file, struct page *page)
722{
723 return mpage_readpage(page, ext2_get_block);
724}
725
726static int
727ext2_readpages(struct file *file, struct address_space *mapping,
728 struct list_head *pages, unsigned nr_pages)
729{
730 return mpage_readpages(mapping, pages, nr_pages, ext2_get_block);
731}
732
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700733int __ext2_write_begin(struct file *file, struct address_space *mapping,
734 loff_t pos, unsigned len, unsigned flags,
735 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736{
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700737 return block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
738 ext2_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739}
740
741static int
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700742ext2_write_begin(struct file *file, struct address_space *mapping,
743 loff_t pos, unsigned len, unsigned flags,
744 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745{
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700746 *pagep = NULL;
747 return __ext2_write_begin(file, mapping, pos, len, flags, pagep,fsdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748}
749
Nick Piggin03158cd2007-10-16 01:25:25 -0700750static int
751ext2_nobh_write_begin(struct file *file, struct address_space *mapping,
752 loff_t pos, unsigned len, unsigned flags,
753 struct page **pagep, void **fsdata)
754{
755 /*
756 * Dir-in-pagecache still uses ext2_write_begin. Would have to rework
757 * directory handling code to pass around offsets rather than struct
758 * pages in order to make this work easily.
759 */
760 return nobh_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
761 ext2_get_block);
762}
763
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764static int ext2_nobh_writepage(struct page *page,
765 struct writeback_control *wbc)
766{
767 return nobh_writepage(page, ext2_get_block, wbc);
768}
769
770static sector_t ext2_bmap(struct address_space *mapping, sector_t block)
771{
772 return generic_block_bmap(mapping,block,ext2_get_block);
773}
774
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775static ssize_t
776ext2_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
777 loff_t offset, unsigned long nr_segs)
778{
779 struct file *file = iocb->ki_filp;
780 struct inode *inode = file->f_mapping->host;
781
782 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
Badari Pulavarty1d8fa7a2006-03-26 01:38:02 -0800783 offset, nr_segs, ext2_get_block, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784}
785
786static int
787ext2_writepages(struct address_space *mapping, struct writeback_control *wbc)
788{
789 return mpage_writepages(mapping, wbc, ext2_get_block);
790}
791
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700792const struct address_space_operations ext2_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 .readpage = ext2_readpage,
794 .readpages = ext2_readpages,
795 .writepage = ext2_writepage,
796 .sync_page = block_sync_page,
Nick Pigginf34fb6e2007-10-16 01:25:04 -0700797 .write_begin = ext2_write_begin,
798 .write_end = generic_write_end,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 .bmap = ext2_bmap,
800 .direct_IO = ext2_direct_IO,
801 .writepages = ext2_writepages,
Christoph Lametere965f962006-02-01 03:05:41 -0800802 .migratepage = buffer_migrate_page,
Hisashi Hifumi8ab22b92008-07-28 15:46:36 -0700803 .is_partially_uptodate = block_is_partially_uptodate,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804};
805
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700806const struct address_space_operations ext2_aops_xip = {
Carsten Otte6d791252005-06-23 22:05:26 -0700807 .bmap = ext2_bmap,
Nick Piggin70688e42008-04-28 02:13:02 -0700808 .get_xip_mem = ext2_get_xip_mem,
Carsten Otte6d791252005-06-23 22:05:26 -0700809};
810
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700811const struct address_space_operations ext2_nobh_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 .readpage = ext2_readpage,
813 .readpages = ext2_readpages,
814 .writepage = ext2_nobh_writepage,
815 .sync_page = block_sync_page,
Nick Piggin03158cd2007-10-16 01:25:25 -0700816 .write_begin = ext2_nobh_write_begin,
817 .write_end = nobh_write_end,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 .bmap = ext2_bmap,
819 .direct_IO = ext2_direct_IO,
820 .writepages = ext2_writepages,
Christoph Lametere965f962006-02-01 03:05:41 -0800821 .migratepage = buffer_migrate_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822};
823
824/*
825 * Probably it should be a library function... search for first non-zero word
826 * or memcmp with zero_page, whatever is better for particular architecture.
827 * Linus?
828 */
829static inline int all_zeroes(__le32 *p, __le32 *q)
830{
831 while (p < q)
832 if (*p++)
833 return 0;
834 return 1;
835}
836
837/**
838 * ext2_find_shared - find the indirect blocks for partial truncation.
839 * @inode: inode in question
840 * @depth: depth of the affected branch
841 * @offsets: offsets of pointers in that branch (see ext2_block_to_path)
842 * @chain: place to store the pointers to partial indirect blocks
843 * @top: place to the (detached) top of branch
844 *
845 * This is a helper function used by ext2_truncate().
846 *
847 * When we do truncate() we may have to clean the ends of several indirect
848 * blocks but leave the blocks themselves alive. Block is partially
849 * truncated if some data below the new i_size is refered from it (and
850 * it is on the path to the first completely truncated data block, indeed).
851 * We have to free the top of that path along with everything to the right
852 * of the path. Since no allocation past the truncation point is possible
853 * until ext2_truncate() finishes, we may safely do the latter, but top
854 * of branch may require special attention - pageout below the truncation
855 * point might try to populate it.
856 *
857 * We atomically detach the top of branch from the tree, store the block
858 * number of its root in *@top, pointers to buffer_heads of partially
859 * truncated blocks - in @chain[].bh and pointers to their last elements
860 * that should not be removed - in @chain[].p. Return value is the pointer
861 * to last filled element of @chain.
862 *
863 * The work left to caller to do the actual freeing of subtrees:
864 * a) free the subtree starting from *@top
865 * b) free the subtrees whose roots are stored in
866 * (@chain[i].p+1 .. end of @chain[i].bh->b_data)
867 * c) free the subtrees growing from the inode past the @chain[0].p
868 * (no partially truncated stuff there).
869 */
870
871static Indirect *ext2_find_shared(struct inode *inode,
872 int depth,
873 int offsets[4],
874 Indirect chain[4],
875 __le32 *top)
876{
877 Indirect *partial, *p;
878 int k, err;
879
880 *top = 0;
881 for (k = depth; k > 1 && !offsets[k-1]; k--)
882 ;
883 partial = ext2_get_branch(inode, k, offsets, chain, &err);
884 if (!partial)
885 partial = chain + k-1;
886 /*
887 * If the branch acquired continuation since we've looked at it -
888 * fine, it should all survive and (new) top doesn't belong to us.
889 */
890 write_lock(&EXT2_I(inode)->i_meta_lock);
891 if (!partial->key && *partial->p) {
892 write_unlock(&EXT2_I(inode)->i_meta_lock);
893 goto no_top;
894 }
895 for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--)
896 ;
897 /*
898 * OK, we've found the last block that must survive. The rest of our
899 * branch should be detached before unlocking. However, if that rest
900 * of branch is all ours and does not grow immediately from the inode
901 * it's easier to cheat and just decrement partial->p.
902 */
903 if (p == chain + k - 1 && p > chain) {
904 p->p--;
905 } else {
906 *top = *p->p;
907 *p->p = 0;
908 }
909 write_unlock(&EXT2_I(inode)->i_meta_lock);
910
911 while(partial > p)
912 {
913 brelse(partial->bh);
914 partial--;
915 }
916no_top:
917 return partial;
918}
919
920/**
921 * ext2_free_data - free a list of data blocks
922 * @inode: inode we are dealing with
923 * @p: array of block numbers
924 * @q: points immediately past the end of array
925 *
926 * We are freeing all blocks refered from that array (numbers are
927 * stored as little-endian 32-bit) and updating @inode->i_blocks
928 * appropriately.
929 */
930static inline void ext2_free_data(struct inode *inode, __le32 *p, __le32 *q)
931{
932 unsigned long block_to_free = 0, count = 0;
933 unsigned long nr;
934
935 for ( ; p < q ; p++) {
936 nr = le32_to_cpu(*p);
937 if (nr) {
938 *p = 0;
939 /* accumulate blocks to free if they're contiguous */
940 if (count == 0)
941 goto free_this;
942 else if (block_to_free == nr - count)
943 count++;
944 else {
945 mark_inode_dirty(inode);
946 ext2_free_blocks (inode, block_to_free, count);
947 free_this:
948 block_to_free = nr;
949 count = 1;
950 }
951 }
952 }
953 if (count > 0) {
954 mark_inode_dirty(inode);
955 ext2_free_blocks (inode, block_to_free, count);
956 }
957}
958
959/**
960 * ext2_free_branches - free an array of branches
961 * @inode: inode we are dealing with
962 * @p: array of block numbers
963 * @q: pointer immediately past the end of array
964 * @depth: depth of the branches to free
965 *
966 * We are freeing all blocks refered from these branches (numbers are
967 * stored as little-endian 32-bit) and updating @inode->i_blocks
968 * appropriately.
969 */
970static void ext2_free_branches(struct inode *inode, __le32 *p, __le32 *q, int depth)
971{
972 struct buffer_head * bh;
973 unsigned long nr;
974
975 if (depth--) {
976 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
977 for ( ; p < q ; p++) {
978 nr = le32_to_cpu(*p);
979 if (!nr)
980 continue;
981 *p = 0;
982 bh = sb_bread(inode->i_sb, nr);
983 /*
984 * A read failure? Report error and clear slot
985 * (should be rare).
986 */
987 if (!bh) {
988 ext2_error(inode->i_sb, "ext2_free_branches",
989 "Read failure, inode=%ld, block=%ld",
990 inode->i_ino, nr);
991 continue;
992 }
993 ext2_free_branches(inode,
994 (__le32*)bh->b_data,
995 (__le32*)bh->b_data + addr_per_block,
996 depth);
997 bforget(bh);
998 ext2_free_blocks(inode, nr, 1);
999 mark_inode_dirty(inode);
1000 }
1001 } else
1002 ext2_free_data(inode, p, q);
1003}
1004
Martin J. Bligha686cd82007-10-16 23:30:46 -07001005void ext2_truncate(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006{
1007 __le32 *i_data = EXT2_I(inode)->i_data;
Martin J. Bligha686cd82007-10-16 23:30:46 -07001008 struct ext2_inode_info *ei = EXT2_I(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
1010 int offsets[4];
1011 Indirect chain[4];
1012 Indirect *partial;
1013 __le32 nr = 0;
1014 int n;
1015 long iblock;
1016 unsigned blocksize;
1017
1018 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1019 S_ISLNK(inode->i_mode)))
1020 return;
1021 if (ext2_inode_is_fast_symlink(inode))
1022 return;
1023 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1024 return;
1025
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 blocksize = inode->i_sb->s_blocksize;
1027 iblock = (inode->i_size + blocksize-1)
1028 >> EXT2_BLOCK_SIZE_BITS(inode->i_sb);
1029
Carsten Otte6d791252005-06-23 22:05:26 -07001030 if (mapping_is_xip(inode->i_mapping))
1031 xip_truncate_page(inode->i_mapping, inode->i_size);
1032 else if (test_opt(inode->i_sb, NOBH))
Nick Piggin03158cd2007-10-16 01:25:25 -07001033 nobh_truncate_page(inode->i_mapping,
1034 inode->i_size, ext2_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 else
1036 block_truncate_page(inode->i_mapping,
1037 inode->i_size, ext2_get_block);
1038
1039 n = ext2_block_to_path(inode, iblock, offsets, NULL);
1040 if (n == 0)
1041 return;
1042
Martin J. Bligha686cd82007-10-16 23:30:46 -07001043 /*
1044 * From here we block out all ext2_get_block() callers who want to
1045 * modify the block allocation tree.
1046 */
1047 mutex_lock(&ei->truncate_mutex);
1048
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 if (n == 1) {
1050 ext2_free_data(inode, i_data+offsets[0],
1051 i_data + EXT2_NDIR_BLOCKS);
1052 goto do_indirects;
1053 }
1054
1055 partial = ext2_find_shared(inode, n, offsets, chain, &nr);
1056 /* Kill the top of shared branch (already detached) */
1057 if (nr) {
1058 if (partial == chain)
1059 mark_inode_dirty(inode);
1060 else
1061 mark_buffer_dirty_inode(partial->bh, inode);
1062 ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
1063 }
1064 /* Clear the ends of indirect blocks on the shared branch */
1065 while (partial > chain) {
1066 ext2_free_branches(inode,
1067 partial->p + 1,
1068 (__le32*)partial->bh->b_data+addr_per_block,
1069 (chain+n-1) - partial);
1070 mark_buffer_dirty_inode(partial->bh, inode);
1071 brelse (partial->bh);
1072 partial--;
1073 }
1074do_indirects:
1075 /* Kill the remaining (whole) subtrees */
1076 switch (offsets[0]) {
1077 default:
1078 nr = i_data[EXT2_IND_BLOCK];
1079 if (nr) {
1080 i_data[EXT2_IND_BLOCK] = 0;
1081 mark_inode_dirty(inode);
1082 ext2_free_branches(inode, &nr, &nr+1, 1);
1083 }
1084 case EXT2_IND_BLOCK:
1085 nr = i_data[EXT2_DIND_BLOCK];
1086 if (nr) {
1087 i_data[EXT2_DIND_BLOCK] = 0;
1088 mark_inode_dirty(inode);
1089 ext2_free_branches(inode, &nr, &nr+1, 2);
1090 }
1091 case EXT2_DIND_BLOCK:
1092 nr = i_data[EXT2_TIND_BLOCK];
1093 if (nr) {
1094 i_data[EXT2_TIND_BLOCK] = 0;
1095 mark_inode_dirty(inode);
1096 ext2_free_branches(inode, &nr, &nr+1, 3);
1097 }
1098 case EXT2_TIND_BLOCK:
1099 ;
1100 }
Martin J. Bligha686cd82007-10-16 23:30:46 -07001101
1102 ext2_discard_reservation(inode);
1103
1104 mutex_unlock(&ei->truncate_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 inode->i_mtime = inode->i_ctime = CURRENT_TIME_SEC;
1106 if (inode_needs_sync(inode)) {
1107 sync_mapping_buffers(inode->i_mapping);
1108 ext2_sync_inode (inode);
1109 } else {
1110 mark_inode_dirty(inode);
1111 }
1112}
1113
1114static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
1115 struct buffer_head **p)
1116{
1117 struct buffer_head * bh;
1118 unsigned long block_group;
1119 unsigned long block;
1120 unsigned long offset;
1121 struct ext2_group_desc * gdp;
1122
1123 *p = NULL;
1124 if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) ||
1125 ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
1126 goto Einval;
1127
1128 block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb);
Eric Sandeenef2fb672007-10-16 23:26:30 -07001129 gdp = ext2_get_group_desc(sb, block_group, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 if (!gdp)
1131 goto Egdp;
1132 /*
1133 * Figure out the offset within the block group inode table
1134 */
1135 offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb);
1136 block = le32_to_cpu(gdp->bg_inode_table) +
1137 (offset >> EXT2_BLOCK_SIZE_BITS(sb));
1138 if (!(bh = sb_bread(sb, block)))
1139 goto Eio;
1140
1141 *p = bh;
1142 offset &= (EXT2_BLOCK_SIZE(sb) - 1);
1143 return (struct ext2_inode *) (bh->b_data + offset);
1144
1145Einval:
1146 ext2_error(sb, "ext2_get_inode", "bad inode number: %lu",
1147 (unsigned long) ino);
1148 return ERR_PTR(-EINVAL);
1149Eio:
1150 ext2_error(sb, "ext2_get_inode",
1151 "unable to read inode block - inode=%lu, block=%lu",
1152 (unsigned long) ino, block);
1153Egdp:
1154 return ERR_PTR(-EIO);
1155}
1156
1157void ext2_set_inode_flags(struct inode *inode)
1158{
1159 unsigned int flags = EXT2_I(inode)->i_flags;
1160
1161 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
1162 if (flags & EXT2_SYNC_FL)
1163 inode->i_flags |= S_SYNC;
1164 if (flags & EXT2_APPEND_FL)
1165 inode->i_flags |= S_APPEND;
1166 if (flags & EXT2_IMMUTABLE_FL)
1167 inode->i_flags |= S_IMMUTABLE;
1168 if (flags & EXT2_NOATIME_FL)
1169 inode->i_flags |= S_NOATIME;
1170 if (flags & EXT2_DIRSYNC_FL)
1171 inode->i_flags |= S_DIRSYNC;
1172}
1173
Jan Kara4f99ed62007-05-08 00:31:04 -07001174/* Propagate flags from i_flags to EXT2_I(inode)->i_flags */
1175void ext2_get_inode_flags(struct ext2_inode_info *ei)
1176{
1177 unsigned int flags = ei->vfs_inode.i_flags;
1178
1179 ei->i_flags &= ~(EXT2_SYNC_FL|EXT2_APPEND_FL|
1180 EXT2_IMMUTABLE_FL|EXT2_NOATIME_FL|EXT2_DIRSYNC_FL);
1181 if (flags & S_SYNC)
1182 ei->i_flags |= EXT2_SYNC_FL;
1183 if (flags & S_APPEND)
1184 ei->i_flags |= EXT2_APPEND_FL;
1185 if (flags & S_IMMUTABLE)
1186 ei->i_flags |= EXT2_IMMUTABLE_FL;
1187 if (flags & S_NOATIME)
1188 ei->i_flags |= EXT2_NOATIME_FL;
1189 if (flags & S_DIRSYNC)
1190 ei->i_flags |= EXT2_DIRSYNC_FL;
1191}
1192
David Howells52fcf702008-02-07 00:15:35 -08001193struct inode *ext2_iget (struct super_block *sb, unsigned long ino)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194{
David Howells52fcf702008-02-07 00:15:35 -08001195 struct ext2_inode_info *ei;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 struct buffer_head * bh;
David Howells52fcf702008-02-07 00:15:35 -08001197 struct ext2_inode *raw_inode;
1198 struct inode *inode;
1199 long ret = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 int n;
1201
David Howells52fcf702008-02-07 00:15:35 -08001202 inode = iget_locked(sb, ino);
1203 if (!inode)
1204 return ERR_PTR(-ENOMEM);
1205 if (!(inode->i_state & I_NEW))
1206 return inode;
1207
1208 ei = EXT2_I(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209#ifdef CONFIG_EXT2_FS_POSIX_ACL
1210 ei->i_acl = EXT2_ACL_NOT_CACHED;
1211 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
1212#endif
Martin J. Bligha686cd82007-10-16 23:30:46 -07001213 ei->i_block_alloc_info = NULL;
1214
David Howells52fcf702008-02-07 00:15:35 -08001215 raw_inode = ext2_get_inode(inode->i_sb, ino, &bh);
1216 if (IS_ERR(raw_inode)) {
1217 ret = PTR_ERR(raw_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 goto bad_inode;
David Howells52fcf702008-02-07 00:15:35 -08001219 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220
1221 inode->i_mode = le16_to_cpu(raw_inode->i_mode);
1222 inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
1223 inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
1224 if (!(test_opt (inode->i_sb, NO_UID32))) {
1225 inode->i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
1226 inode->i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
1227 }
1228 inode->i_nlink = le16_to_cpu(raw_inode->i_links_count);
1229 inode->i_size = le32_to_cpu(raw_inode->i_size);
Markus Rechberger4d7bf112007-05-08 00:23:39 -07001230 inode->i_atime.tv_sec = (signed)le32_to_cpu(raw_inode->i_atime);
1231 inode->i_ctime.tv_sec = (signed)le32_to_cpu(raw_inode->i_ctime);
1232 inode->i_mtime.tv_sec = (signed)le32_to_cpu(raw_inode->i_mtime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0;
1234 ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
1235 /* We now have enough fields to check if the inode was active or not.
1236 * This is needed because nfsd might try to access dead inodes
1237 * the test is that same one that e2fsck uses
1238 * NeilBrown 1999oct15
1239 */
1240 if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) {
1241 /* this inode is deleted */
1242 brelse (bh);
David Howells52fcf702008-02-07 00:15:35 -08001243 ret = -ESTALE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 goto bad_inode;
1245 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 inode->i_blocks = le32_to_cpu(raw_inode->i_blocks);
1247 ei->i_flags = le32_to_cpu(raw_inode->i_flags);
1248 ei->i_faddr = le32_to_cpu(raw_inode->i_faddr);
1249 ei->i_frag_no = raw_inode->i_frag;
1250 ei->i_frag_size = raw_inode->i_fsize;
1251 ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl);
1252 ei->i_dir_acl = 0;
1253 if (S_ISREG(inode->i_mode))
1254 inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32;
1255 else
1256 ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl);
1257 ei->i_dtime = 0;
1258 inode->i_generation = le32_to_cpu(raw_inode->i_generation);
1259 ei->i_state = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb);
1261 ei->i_dir_start_lookup = 0;
1262
1263 /*
1264 * NOTE! The in-memory inode i_data array is in little-endian order
1265 * even on big-endian machines: we do NOT byteswap the block numbers!
1266 */
1267 for (n = 0; n < EXT2_N_BLOCKS; n++)
1268 ei->i_data[n] = raw_inode->i_block[n];
1269
1270 if (S_ISREG(inode->i_mode)) {
1271 inode->i_op = &ext2_file_inode_operations;
Carsten Otte6d791252005-06-23 22:05:26 -07001272 if (ext2_use_xip(inode->i_sb)) {
1273 inode->i_mapping->a_ops = &ext2_aops_xip;
1274 inode->i_fop = &ext2_xip_file_operations;
1275 } else if (test_opt(inode->i_sb, NOBH)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 inode->i_mapping->a_ops = &ext2_nobh_aops;
Carsten Otte6d791252005-06-23 22:05:26 -07001277 inode->i_fop = &ext2_file_operations;
1278 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 inode->i_mapping->a_ops = &ext2_aops;
Carsten Otte6d791252005-06-23 22:05:26 -07001280 inode->i_fop = &ext2_file_operations;
1281 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 } else if (S_ISDIR(inode->i_mode)) {
1283 inode->i_op = &ext2_dir_inode_operations;
1284 inode->i_fop = &ext2_dir_operations;
1285 if (test_opt(inode->i_sb, NOBH))
1286 inode->i_mapping->a_ops = &ext2_nobh_aops;
1287 else
1288 inode->i_mapping->a_ops = &ext2_aops;
1289 } else if (S_ISLNK(inode->i_mode)) {
Duane Griffin8d6d0c42008-12-19 20:47:13 +00001290 if (ext2_inode_is_fast_symlink(inode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 inode->i_op = &ext2_fast_symlink_inode_operations;
Duane Griffin8d6d0c42008-12-19 20:47:13 +00001292 nd_terminate_link(ei->i_data, inode->i_size,
1293 sizeof(ei->i_data) - 1);
1294 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 inode->i_op = &ext2_symlink_inode_operations;
1296 if (test_opt(inode->i_sb, NOBH))
1297 inode->i_mapping->a_ops = &ext2_nobh_aops;
1298 else
1299 inode->i_mapping->a_ops = &ext2_aops;
1300 }
1301 } else {
1302 inode->i_op = &ext2_special_inode_operations;
1303 if (raw_inode->i_block[0])
1304 init_special_inode(inode, inode->i_mode,
1305 old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
1306 else
1307 init_special_inode(inode, inode->i_mode,
1308 new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
1309 }
1310 brelse (bh);
1311 ext2_set_inode_flags(inode);
David Howells52fcf702008-02-07 00:15:35 -08001312 unlock_new_inode(inode);
1313 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314
1315bad_inode:
David Howells52fcf702008-02-07 00:15:35 -08001316 iget_failed(inode);
1317 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318}
1319
1320static int ext2_update_inode(struct inode * inode, int do_sync)
1321{
1322 struct ext2_inode_info *ei = EXT2_I(inode);
1323 struct super_block *sb = inode->i_sb;
1324 ino_t ino = inode->i_ino;
1325 uid_t uid = inode->i_uid;
1326 gid_t gid = inode->i_gid;
1327 struct buffer_head * bh;
1328 struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh);
1329 int n;
1330 int err = 0;
1331
1332 if (IS_ERR(raw_inode))
1333 return -EIO;
1334
1335 /* For fields not not tracking in the in-memory inode,
1336 * initialise them to zero for new inodes. */
1337 if (ei->i_state & EXT2_STATE_NEW)
1338 memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size);
1339
Jan Kara4f99ed62007-05-08 00:31:04 -07001340 ext2_get_inode_flags(ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 raw_inode->i_mode = cpu_to_le16(inode->i_mode);
1342 if (!(test_opt(sb, NO_UID32))) {
1343 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid));
1344 raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid));
1345/*
1346 * Fix up interoperability with old kernels. Otherwise, old inodes get
1347 * re-used with the upper 16 bits of the uid/gid intact
1348 */
1349 if (!ei->i_dtime) {
1350 raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid));
1351 raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid));
1352 } else {
1353 raw_inode->i_uid_high = 0;
1354 raw_inode->i_gid_high = 0;
1355 }
1356 } else {
1357 raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid));
1358 raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid));
1359 raw_inode->i_uid_high = 0;
1360 raw_inode->i_gid_high = 0;
1361 }
1362 raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
1363 raw_inode->i_size = cpu_to_le32(inode->i_size);
1364 raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
1365 raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
1366 raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
1367
1368 raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
1369 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
1370 raw_inode->i_flags = cpu_to_le32(ei->i_flags);
1371 raw_inode->i_faddr = cpu_to_le32(ei->i_faddr);
1372 raw_inode->i_frag = ei->i_frag_no;
1373 raw_inode->i_fsize = ei->i_frag_size;
1374 raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl);
1375 if (!S_ISREG(inode->i_mode))
1376 raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl);
1377 else {
1378 raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32);
1379 if (inode->i_size > 0x7fffffffULL) {
1380 if (!EXT2_HAS_RO_COMPAT_FEATURE(sb,
1381 EXT2_FEATURE_RO_COMPAT_LARGE_FILE) ||
1382 EXT2_SB(sb)->s_es->s_rev_level ==
1383 cpu_to_le32(EXT2_GOOD_OLD_REV)) {
1384 /* If this is the first large file
1385 * created, add a flag to the superblock.
1386 */
1387 lock_kernel();
1388 ext2_update_dynamic_rev(sb);
1389 EXT2_SET_RO_COMPAT_FEATURE(sb,
1390 EXT2_FEATURE_RO_COMPAT_LARGE_FILE);
1391 unlock_kernel();
1392 ext2_write_super(sb);
1393 }
1394 }
1395 }
1396
1397 raw_inode->i_generation = cpu_to_le32(inode->i_generation);
1398 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
1399 if (old_valid_dev(inode->i_rdev)) {
1400 raw_inode->i_block[0] =
1401 cpu_to_le32(old_encode_dev(inode->i_rdev));
1402 raw_inode->i_block[1] = 0;
1403 } else {
1404 raw_inode->i_block[0] = 0;
1405 raw_inode->i_block[1] =
1406 cpu_to_le32(new_encode_dev(inode->i_rdev));
1407 raw_inode->i_block[2] = 0;
1408 }
1409 } else for (n = 0; n < EXT2_N_BLOCKS; n++)
1410 raw_inode->i_block[n] = ei->i_data[n];
1411 mark_buffer_dirty(bh);
1412 if (do_sync) {
1413 sync_dirty_buffer(bh);
1414 if (buffer_req(bh) && !buffer_uptodate(bh)) {
1415 printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1416 sb->s_id, (unsigned long) ino);
1417 err = -EIO;
1418 }
1419 }
1420 ei->i_state &= ~EXT2_STATE_NEW;
1421 brelse (bh);
1422 return err;
1423}
1424
1425int ext2_write_inode(struct inode *inode, int wait)
1426{
1427 return ext2_update_inode(inode, wait);
1428}
1429
1430int ext2_sync_inode(struct inode *inode)
1431{
1432 struct writeback_control wbc = {
1433 .sync_mode = WB_SYNC_ALL,
1434 .nr_to_write = 0, /* sys_fsync did this */
1435 };
1436 return sync_inode(inode, &wbc);
1437}
1438
1439int ext2_setattr(struct dentry *dentry, struct iattr *iattr)
1440{
1441 struct inode *inode = dentry->d_inode;
1442 int error;
1443
1444 error = inode_change_ok(inode, iattr);
1445 if (error)
1446 return error;
1447 if ((iattr->ia_valid & ATTR_UID && iattr->ia_uid != inode->i_uid) ||
1448 (iattr->ia_valid & ATTR_GID && iattr->ia_gid != inode->i_gid)) {
1449 error = DQUOT_TRANSFER(inode, iattr) ? -EDQUOT : 0;
1450 if (error)
1451 return error;
1452 }
1453 error = inode_setattr(inode, iattr);
1454 if (!error && (iattr->ia_valid & ATTR_MODE))
1455 error = ext2_acl_chmod(inode);
1456 return error;
1457}