]>
Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
3a8a9a10 | 3 | * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved. |
b3b94faa DT |
4 | * |
5 | * This copyrighted material is made available to anyone wishing to use, | |
6 | * modify, copy, or redistribute it subject to the terms and conditions | |
7 | * of the GNU General Public License v.2. | |
8 | */ | |
9 | ||
10 | #include <linux/sched.h> | |
11 | #include <linux/slab.h> | |
12 | #include <linux/spinlock.h> | |
13 | #include <linux/completion.h> | |
14 | #include <linux/buffer_head.h> | |
5c676f6d | 15 | #include <linux/gfs2_ondisk.h> |
71b86f56 | 16 | #include <linux/crc32.h> |
b3b94faa DT |
17 | |
18 | #include "gfs2.h" | |
5c676f6d SW |
19 | #include "lm_interface.h" |
20 | #include "incore.h" | |
b3b94faa DT |
21 | #include "bmap.h" |
22 | #include "glock.h" | |
23 | #include "inode.h" | |
b3b94faa DT |
24 | #include "meta_io.h" |
25 | #include "page.h" | |
26 | #include "quota.h" | |
27 | #include "rgrp.h" | |
28 | #include "trans.h" | |
18ec7d5c | 29 | #include "dir.h" |
5c676f6d | 30 | #include "util.h" |
b3b94faa DT |
31 | |
32 | /* This doesn't need to be that large as max 64 bit pointers in a 4k | |
33 | * block is 512, so __u16 is fine for that. It saves stack space to | |
34 | * keep it small. | |
35 | */ | |
36 | struct metapath { | |
37 | __u16 mp_list[GFS2_MAX_META_HEIGHT]; | |
38 | }; | |
39 | ||
40 | typedef int (*block_call_t) (struct gfs2_inode *ip, struct buffer_head *dibh, | |
41 | struct buffer_head *bh, uint64_t *top, | |
42 | uint64_t *bottom, unsigned int height, | |
43 | void *data); | |
44 | ||
45 | struct strip_mine { | |
46 | int sm_first; | |
47 | unsigned int sm_height; | |
48 | }; | |
49 | ||
b3b94faa DT |
50 | /** |
51 | * gfs2_unstuff_dinode - Unstuff a dinode when the data has grown too big | |
52 | * @ip: The GFS2 inode to unstuff | |
53 | * @unstuffer: the routine that handles unstuffing a non-zero length file | |
54 | * @private: private data for the unstuffer | |
55 | * | |
56 | * This routine unstuffs a dinode and returns it to a "normal" state such | |
57 | * that the height can be grown in the traditional way. | |
58 | * | |
59 | * Returns: errno | |
60 | */ | |
61 | ||
62 | int gfs2_unstuff_dinode(struct gfs2_inode *ip, gfs2_unstuffer_t unstuffer, | |
63 | void *private) | |
64 | { | |
65 | struct buffer_head *bh, *dibh; | |
66 | uint64_t block = 0; | |
18ec7d5c | 67 | int isdir = gfs2_is_dir(ip); |
b3b94faa DT |
68 | int error; |
69 | ||
70 | down_write(&ip->i_rw_mutex); | |
71 | ||
72 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
73 | if (error) | |
74 | goto out; | |
75 | ||
76 | if (ip->i_di.di_size) { | |
77 | /* Get a free block, fill it with the stuffed data, | |
78 | and write it out to disk */ | |
79 | ||
18ec7d5c | 80 | if (isdir) { |
b3b94faa DT |
81 | block = gfs2_alloc_meta(ip); |
82 | ||
61e085a8 | 83 | error = gfs2_dir_get_new_buffer(ip, block, &bh); |
b3b94faa DT |
84 | if (error) |
85 | goto out_brelse; | |
86 | gfs2_buffer_copy_tail(bh, | |
87 | sizeof(struct gfs2_meta_header), | |
88 | dibh, sizeof(struct gfs2_dinode)); | |
89 | brelse(bh); | |
90 | } else { | |
91 | block = gfs2_alloc_data(ip); | |
92 | ||
93 | error = unstuffer(ip, dibh, block, private); | |
94 | if (error) | |
95 | goto out_brelse; | |
96 | } | |
97 | } | |
98 | ||
99 | /* Set up the pointer to the new block */ | |
100 | ||
d4e9c4c3 | 101 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
b3b94faa DT |
102 | |
103 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode)); | |
104 | ||
105 | if (ip->i_di.di_size) { | |
568f4c96 SW |
106 | *(uint64_t *)(dibh->b_data + sizeof(struct gfs2_dinode)) = |
107 | cpu_to_be64(block); | |
b3b94faa DT |
108 | ip->i_di.di_blocks++; |
109 | } | |
110 | ||
111 | ip->i_di.di_height = 1; | |
112 | ||
113 | gfs2_dinode_out(&ip->i_di, dibh->b_data); | |
114 | ||
115 | out_brelse: | |
116 | brelse(dibh); | |
117 | ||
118 | out: | |
119 | up_write(&ip->i_rw_mutex); | |
120 | ||
121 | return error; | |
122 | } | |
123 | ||
124 | /** | |
125 | * calc_tree_height - Calculate the height of a metadata tree | |
126 | * @ip: The GFS2 inode | |
127 | * @size: The proposed size of the file | |
128 | * | |
129 | * Work out how tall a metadata tree needs to be in order to accommodate a | |
130 | * file of a particular size. If size is less than the current size of | |
131 | * the inode, then the current size of the inode is used instead of the | |
132 | * supplied one. | |
133 | * | |
134 | * Returns: the height the tree should be | |
135 | */ | |
136 | ||
137 | static unsigned int calc_tree_height(struct gfs2_inode *ip, uint64_t size) | |
138 | { | |
feaa7bba | 139 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
140 | uint64_t *arr; |
141 | unsigned int max, height; | |
142 | ||
143 | if (ip->i_di.di_size > size) | |
144 | size = ip->i_di.di_size; | |
145 | ||
18ec7d5c | 146 | if (gfs2_is_dir(ip)) { |
b3b94faa DT |
147 | arr = sdp->sd_jheightsize; |
148 | max = sdp->sd_max_jheight; | |
149 | } else { | |
150 | arr = sdp->sd_heightsize; | |
151 | max = sdp->sd_max_height; | |
152 | } | |
153 | ||
154 | for (height = 0; height < max; height++) | |
155 | if (arr[height] >= size) | |
156 | break; | |
157 | ||
158 | return height; | |
159 | } | |
160 | ||
161 | /** | |
162 | * build_height - Build a metadata tree of the requested height | |
163 | * @ip: The GFS2 inode | |
164 | * @height: The height to build to | |
165 | * | |
b3b94faa DT |
166 | * |
167 | * Returns: errno | |
168 | */ | |
169 | ||
e90c01e1 | 170 | static int build_height(struct inode *inode, unsigned height) |
b3b94faa | 171 | { |
feaa7bba | 172 | struct gfs2_inode *ip = GFS2_I(inode); |
e90c01e1 SW |
173 | unsigned new_height = height - ip->i_di.di_height; |
174 | struct buffer_head *dibh; | |
175 | struct buffer_head *blocks[GFS2_MAX_META_HEIGHT]; | |
b3b94faa | 176 | int error; |
e90c01e1 SW |
177 | u64 *bp; |
178 | u64 bn; | |
179 | unsigned n; | |
b3b94faa | 180 | |
e90c01e1 SW |
181 | if (height <= ip->i_di.di_height) |
182 | return 0; | |
b3b94faa | 183 | |
e90c01e1 SW |
184 | error = gfs2_meta_inode_buffer(ip, &dibh); |
185 | if (error) | |
186 | return error; | |
b3b94faa | 187 | |
e90c01e1 SW |
188 | for(n = 0; n < new_height; n++) { |
189 | bn = gfs2_alloc_meta(ip); | |
190 | blocks[n] = gfs2_meta_new(ip->i_gl, bn); | |
191 | gfs2_trans_add_bh(ip->i_gl, blocks[n], 1); | |
192 | } | |
193 | ||
194 | n = 0; | |
195 | bn = blocks[0]->b_blocknr; | |
196 | if (new_height > 1) { | |
197 | for(; n < new_height-1; n++) { | |
198 | gfs2_metatype_set(blocks[n], GFS2_METATYPE_IN, | |
b3b94faa | 199 | GFS2_FORMAT_IN); |
e90c01e1 SW |
200 | gfs2_buffer_clear_tail(blocks[n], |
201 | sizeof(struct gfs2_meta_header)); | |
202 | bp = (u64 *)(blocks[n]->b_data + | |
203 | sizeof(struct gfs2_meta_header)); | |
204 | *bp = cpu_to_be64(blocks[n+1]->b_blocknr); | |
205 | brelse(blocks[n]); | |
206 | blocks[n] = NULL; | |
b3b94faa | 207 | } |
b3b94faa | 208 | } |
e90c01e1 SW |
209 | gfs2_metatype_set(blocks[n], GFS2_METATYPE_IN, GFS2_FORMAT_IN); |
210 | gfs2_buffer_copy_tail(blocks[n], sizeof(struct gfs2_meta_header), | |
211 | dibh, sizeof(struct gfs2_dinode)); | |
212 | brelse(blocks[n]); | |
213 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
214 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode)); | |
215 | bp = (u64 *)(dibh->b_data + sizeof(struct gfs2_dinode)); | |
216 | *bp = cpu_to_be64(bn); | |
217 | ip->i_di.di_height += new_height; | |
218 | ip->i_di.di_blocks += new_height; | |
219 | gfs2_dinode_out(&ip->i_di, dibh->b_data); | |
220 | brelse(dibh); | |
221 | return error; | |
b3b94faa DT |
222 | } |
223 | ||
224 | /** | |
225 | * find_metapath - Find path through the metadata tree | |
226 | * @ip: The inode pointer | |
227 | * @mp: The metapath to return the result in | |
228 | * @block: The disk block to look up | |
229 | * | |
230 | * This routine returns a struct metapath structure that defines a path | |
231 | * through the metadata of inode "ip" to get to block "block". | |
232 | * | |
233 | * Example: | |
234 | * Given: "ip" is a height 3 file, "offset" is 101342453, and this is a | |
235 | * filesystem with a blocksize of 4096. | |
236 | * | |
237 | * find_metapath() would return a struct metapath structure set to: | |
238 | * mp_offset = 101342453, mp_height = 3, mp_list[0] = 0, mp_list[1] = 48, | |
239 | * and mp_list[2] = 165. | |
240 | * | |
241 | * That means that in order to get to the block containing the byte at | |
242 | * offset 101342453, we would load the indirect block pointed to by pointer | |
243 | * 0 in the dinode. We would then load the indirect block pointed to by | |
244 | * pointer 48 in that indirect block. We would then load the data block | |
245 | * pointed to by pointer 165 in that indirect block. | |
246 | * | |
247 | * ---------------------------------------- | |
248 | * | Dinode | | | |
249 | * | | 4| | |
250 | * | |0 1 2 3 4 5 9| | |
251 | * | | 6| | |
252 | * ---------------------------------------- | |
253 | * | | |
254 | * | | |
255 | * V | |
256 | * ---------------------------------------- | |
257 | * | Indirect Block | | |
258 | * | 5| | |
259 | * | 4 4 4 4 4 5 5 1| | |
260 | * |0 5 6 7 8 9 0 1 2| | |
261 | * ---------------------------------------- | |
262 | * | | |
263 | * | | |
264 | * V | |
265 | * ---------------------------------------- | |
266 | * | Indirect Block | | |
267 | * | 1 1 1 1 1 5| | |
268 | * | 6 6 6 6 6 1| | |
269 | * |0 3 4 5 6 7 2| | |
270 | * ---------------------------------------- | |
271 | * | | |
272 | * | | |
273 | * V | |
274 | * ---------------------------------------- | |
275 | * | Data block containing offset | | |
276 | * | 101342453 | | |
277 | * | | | |
278 | * | | | |
279 | * ---------------------------------------- | |
280 | * | |
281 | */ | |
282 | ||
568f4c96 SW |
283 | static void find_metapath(struct gfs2_inode *ip, uint64_t block, |
284 | struct metapath *mp) | |
b3b94faa | 285 | { |
feaa7bba | 286 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
287 | uint64_t b = block; |
288 | unsigned int i; | |
289 | ||
290 | for (i = ip->i_di.di_height; i--;) | |
291 | mp->mp_list[i] = (__u16)do_div(b, sdp->sd_inptrs); | |
292 | ||
293 | } | |
294 | ||
295 | /** | |
296 | * metapointer - Return pointer to start of metadata in a buffer | |
297 | * @bh: The buffer | |
298 | * @height: The metadata height (0 = dinode) | |
299 | * @mp: The metapath | |
300 | * | |
301 | * Return a pointer to the block number of the next height of the metadata | |
302 | * tree given a buffer containing the pointer to the current height of the | |
303 | * metadata tree. | |
304 | */ | |
305 | ||
fd88de56 SW |
306 | static inline u64 *metapointer(struct buffer_head *bh, int *boundary, |
307 | unsigned int height, const struct metapath *mp) | |
b3b94faa DT |
308 | { |
309 | unsigned int head_size = (height > 0) ? | |
310 | sizeof(struct gfs2_meta_header) : sizeof(struct gfs2_dinode); | |
fd88de56 SW |
311 | u64 *ptr; |
312 | *boundary = 0; | |
313 | ptr = ((u64 *)(bh->b_data + head_size)) + mp->mp_list[height]; | |
314 | if (ptr + 1 == (u64*)(bh->b_data + bh->b_size)) | |
315 | *boundary = 1; | |
316 | return ptr; | |
b3b94faa DT |
317 | } |
318 | ||
319 | /** | |
320 | * lookup_block - Get the next metadata block in metadata tree | |
321 | * @ip: The GFS2 inode | |
322 | * @bh: Buffer containing the pointers to metadata blocks | |
323 | * @height: The height of the tree (0 = dinode) | |
324 | * @mp: The metapath | |
325 | * @create: Non-zero if we may create a new meatdata block | |
326 | * @new: Used to indicate if we did create a new metadata block | |
327 | * @block: the returned disk block number | |
328 | * | |
329 | * Given a metatree, complete to a particular height, checks to see if the next | |
330 | * height of the tree exists. If not the next height of the tree is created. | |
331 | * The block number of the next height of the metadata tree is returned. | |
332 | * | |
333 | */ | |
334 | ||
fd88de56 SW |
335 | static int lookup_block(struct gfs2_inode *ip, struct buffer_head *bh, |
336 | unsigned int height, struct metapath *mp, int create, | |
337 | int *new, uint64_t *block) | |
b3b94faa | 338 | { |
fd88de56 SW |
339 | int boundary; |
340 | uint64_t *ptr = metapointer(bh, &boundary, height, mp); | |
b3b94faa DT |
341 | |
342 | if (*ptr) { | |
343 | *block = be64_to_cpu(*ptr); | |
fd88de56 | 344 | return boundary; |
b3b94faa DT |
345 | } |
346 | ||
347 | *block = 0; | |
348 | ||
349 | if (!create) | |
fd88de56 | 350 | return 0; |
b3b94faa | 351 | |
fd88de56 | 352 | if (height == ip->i_di.di_height - 1 && !gfs2_is_dir(ip)) |
b3b94faa DT |
353 | *block = gfs2_alloc_data(ip); |
354 | else | |
355 | *block = gfs2_alloc_meta(ip); | |
356 | ||
d4e9c4c3 | 357 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
b3b94faa DT |
358 | |
359 | *ptr = cpu_to_be64(*block); | |
360 | ip->i_di.di_blocks++; | |
361 | ||
362 | *new = 1; | |
fd88de56 | 363 | return 0; |
b3b94faa DT |
364 | } |
365 | ||
366 | /** | |
fd88de56 SW |
367 | * gfs2_block_pointers - Map a block from an inode to a disk block |
368 | * @inode: The inode | |
b3b94faa DT |
369 | * @lblock: The logical block number |
370 | * @new: Value/Result argument (1 = may create/did create new blocks) | |
fd88de56 SW |
371 | * @boundary: gets set if we've hit a block boundary |
372 | * @mp: metapath to use | |
b3b94faa DT |
373 | * |
374 | * Find the block number on the current device which corresponds to an | |
375 | * inode's block. If the block had to be created, "new" will be set. | |
376 | * | |
377 | * Returns: errno | |
378 | */ | |
379 | ||
fd88de56 SW |
380 | static struct buffer_head *gfs2_block_pointers(struct inode *inode, u64 lblock, |
381 | int *new, u64 *dblock, | |
382 | int *boundary, | |
383 | struct metapath *mp) | |
b3b94faa | 384 | { |
feaa7bba SW |
385 | struct gfs2_inode *ip = GFS2_I(inode); |
386 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
b3b94faa | 387 | struct buffer_head *bh; |
b3b94faa DT |
388 | int create = *new; |
389 | unsigned int bsize; | |
390 | unsigned int height; | |
391 | unsigned int end_of_metadata; | |
392 | unsigned int x; | |
393 | int error = 0; | |
394 | ||
395 | *new = 0; | |
396 | *dblock = 0; | |
b3b94faa DT |
397 | |
398 | if (gfs2_assert_warn(sdp, !gfs2_is_stuffed(ip))) | |
399 | goto out; | |
400 | ||
18ec7d5c | 401 | bsize = (gfs2_is_dir(ip)) ? sdp->sd_jbsize : sdp->sd_sb.sb_bsize; |
b3b94faa DT |
402 | |
403 | height = calc_tree_height(ip, (lblock + 1) * bsize); | |
404 | if (ip->i_di.di_height < height) { | |
405 | if (!create) | |
406 | goto out; | |
407 | ||
e90c01e1 | 408 | error = build_height(inode, height); |
b3b94faa DT |
409 | if (error) |
410 | goto out; | |
411 | } | |
412 | ||
fd88de56 | 413 | find_metapath(ip, lblock, mp); |
b3b94faa DT |
414 | end_of_metadata = ip->i_di.di_height - 1; |
415 | ||
416 | error = gfs2_meta_inode_buffer(ip, &bh); | |
417 | if (error) | |
418 | goto out; | |
419 | ||
420 | for (x = 0; x < end_of_metadata; x++) { | |
fd88de56 | 421 | lookup_block(ip, bh, x, mp, create, new, dblock); |
b3b94faa DT |
422 | brelse(bh); |
423 | if (!*dblock) | |
424 | goto out; | |
425 | ||
426 | error = gfs2_meta_indirect_buffer(ip, x+1, *dblock, *new, &bh); | |
427 | if (error) | |
428 | goto out; | |
429 | } | |
430 | ||
fd88de56 | 431 | *boundary = lookup_block(ip, bh, end_of_metadata, mp, create, new, dblock); |
b3b94faa | 432 | if (*new) { |
fd88de56 SW |
433 | struct buffer_head *dibh; |
434 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
b3b94faa | 435 | if (!error) { |
fd88de56 SW |
436 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
437 | gfs2_dinode_out(&ip->i_di, dibh->b_data); | |
438 | brelse(dibh); | |
b3b94faa DT |
439 | } |
440 | } | |
fd88de56 SW |
441 | return bh; |
442 | out: | |
443 | return ERR_PTR(error); | |
444 | } | |
b3b94faa | 445 | |
fd88de56 SW |
446 | |
447 | static inline void bmap_lock(struct inode *inode, int create) | |
448 | { | |
feaa7bba | 449 | struct gfs2_inode *ip = GFS2_I(inode); |
fd88de56 SW |
450 | if (create) |
451 | down_write(&ip->i_rw_mutex); | |
452 | else | |
453 | down_read(&ip->i_rw_mutex); | |
454 | } | |
455 | ||
456 | static inline void bmap_unlock(struct inode *inode, int create) | |
457 | { | |
feaa7bba | 458 | struct gfs2_inode *ip = GFS2_I(inode); |
b3b94faa DT |
459 | if (create) |
460 | up_write(&ip->i_rw_mutex); | |
461 | else | |
462 | up_read(&ip->i_rw_mutex); | |
fd88de56 | 463 | } |
b3b94faa | 464 | |
fd88de56 SW |
465 | int gfs2_block_map(struct inode *inode, u64 lblock, int *new, u64 *dblock, int *boundary) |
466 | { | |
467 | struct metapath mp; | |
468 | struct buffer_head *bh; | |
469 | int create = *new; | |
470 | ||
471 | bmap_lock(inode, create); | |
472 | bh = gfs2_block_pointers(inode, lblock, new, dblock, boundary, &mp); | |
473 | bmap_unlock(inode, create); | |
474 | if (!bh) | |
475 | return 0; | |
476 | if (IS_ERR(bh)) | |
477 | return PTR_ERR(bh); | |
478 | brelse(bh); | |
479 | return 0; | |
480 | } | |
481 | ||
482 | int gfs2_extent_map(struct inode *inode, u64 lblock, int *new, u64 *dblock, unsigned *extlen) | |
483 | { | |
feaa7bba SW |
484 | struct gfs2_inode *ip = GFS2_I(inode); |
485 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
fd88de56 SW |
486 | struct metapath mp; |
487 | struct buffer_head *bh; | |
488 | int boundary; | |
489 | int create = *new; | |
490 | ||
491 | BUG_ON(!extlen); | |
492 | BUG_ON(!dblock); | |
493 | BUG_ON(!new); | |
494 | ||
495 | bmap_lock(inode, create); | |
496 | bh = gfs2_block_pointers(inode, lblock, new, dblock, &boundary, &mp); | |
497 | *extlen = 1; | |
498 | ||
499 | if (bh && !IS_ERR(bh) && *dblock && !*new) { | |
500 | u64 tmp_dblock; | |
501 | int tmp_new; | |
502 | unsigned int nptrs; | |
503 | unsigned end_of_metadata = ip->i_di.di_height - 1; | |
504 | ||
505 | nptrs = (end_of_metadata) ? sdp->sd_inptrs : sdp->sd_diptrs; | |
506 | while (++mp.mp_list[end_of_metadata] < nptrs) { | |
507 | lookup_block(ip, bh, end_of_metadata, &mp, 0, &tmp_new, &tmp_dblock); | |
508 | if (*dblock + *extlen != tmp_dblock) | |
509 | break; | |
510 | (*extlen)++; | |
511 | } | |
512 | } | |
513 | bmap_unlock(inode, create); | |
514 | if (!bh) | |
515 | return 0; | |
516 | if (IS_ERR(bh)) | |
517 | return PTR_ERR(bh); | |
518 | brelse(bh); | |
519 | return 0; | |
b3b94faa DT |
520 | } |
521 | ||
522 | /** | |
523 | * recursive_scan - recursively scan through the end of a file | |
524 | * @ip: the inode | |
525 | * @dibh: the dinode buffer | |
526 | * @mp: the path through the metadata to the point to start | |
527 | * @height: the height the recursion is at | |
528 | * @block: the indirect block to look at | |
529 | * @first: 1 if this is the first block | |
530 | * @bc: the call to make for each piece of metadata | |
531 | * @data: data opaque to this function to pass to @bc | |
532 | * | |
533 | * When this is first called @height and @block should be zero and | |
534 | * @first should be 1. | |
535 | * | |
536 | * Returns: errno | |
537 | */ | |
538 | ||
539 | static int recursive_scan(struct gfs2_inode *ip, struct buffer_head *dibh, | |
540 | struct metapath *mp, unsigned int height, | |
541 | uint64_t block, int first, block_call_t bc, | |
542 | void *data) | |
543 | { | |
feaa7bba | 544 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
545 | struct buffer_head *bh = NULL; |
546 | uint64_t *top, *bottom; | |
547 | uint64_t bn; | |
548 | int error; | |
549 | int mh_size = sizeof(struct gfs2_meta_header); | |
550 | ||
551 | if (!height) { | |
552 | error = gfs2_meta_inode_buffer(ip, &bh); | |
553 | if (error) | |
554 | return error; | |
555 | dibh = bh; | |
556 | ||
557 | top = (uint64_t *)(bh->b_data + sizeof(struct gfs2_dinode)) + | |
558 | mp->mp_list[0]; | |
559 | bottom = (uint64_t *)(bh->b_data + sizeof(struct gfs2_dinode)) + | |
560 | sdp->sd_diptrs; | |
561 | } else { | |
562 | error = gfs2_meta_indirect_buffer(ip, height, block, 0, &bh); | |
563 | if (error) | |
564 | return error; | |
565 | ||
566 | top = (uint64_t *)(bh->b_data + mh_size) + | |
567 | ((first) ? mp->mp_list[height] : 0); | |
568 | ||
569 | bottom = (uint64_t *)(bh->b_data + mh_size) + sdp->sd_inptrs; | |
570 | } | |
571 | ||
572 | error = bc(ip, dibh, bh, top, bottom, height, data); | |
573 | if (error) | |
574 | goto out; | |
575 | ||
576 | if (height < ip->i_di.di_height - 1) | |
577 | for (; top < bottom; top++, first = 0) { | |
578 | if (!*top) | |
579 | continue; | |
580 | ||
581 | bn = be64_to_cpu(*top); | |
582 | ||
583 | error = recursive_scan(ip, dibh, mp, height + 1, bn, | |
584 | first, bc, data); | |
585 | if (error) | |
586 | break; | |
587 | } | |
588 | ||
589 | out: | |
590 | brelse(bh); | |
591 | ||
592 | return error; | |
593 | } | |
594 | ||
595 | /** | |
596 | * do_strip - Look for a layer a particular layer of the file and strip it off | |
597 | * @ip: the inode | |
598 | * @dibh: the dinode buffer | |
599 | * @bh: A buffer of pointers | |
600 | * @top: The first pointer in the buffer | |
601 | * @bottom: One more than the last pointer | |
602 | * @height: the height this buffer is at | |
603 | * @data: a pointer to a struct strip_mine | |
604 | * | |
605 | * Returns: errno | |
606 | */ | |
607 | ||
608 | static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh, | |
609 | struct buffer_head *bh, uint64_t *top, uint64_t *bottom, | |
610 | unsigned int height, void *data) | |
611 | { | |
feaa7bba SW |
612 | struct strip_mine *sm = data; |
613 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
b3b94faa DT |
614 | struct gfs2_rgrp_list rlist; |
615 | uint64_t bn, bstart; | |
616 | uint32_t blen; | |
617 | uint64_t *p; | |
618 | unsigned int rg_blocks = 0; | |
619 | int metadata; | |
620 | unsigned int revokes = 0; | |
621 | int x; | |
622 | int error; | |
623 | ||
624 | if (!*top) | |
625 | sm->sm_first = 0; | |
626 | ||
627 | if (height != sm->sm_height) | |
628 | return 0; | |
629 | ||
630 | if (sm->sm_first) { | |
631 | top++; | |
632 | sm->sm_first = 0; | |
633 | } | |
634 | ||
18ec7d5c | 635 | metadata = (height != ip->i_di.di_height - 1); |
b3b94faa DT |
636 | if (metadata) |
637 | revokes = (height) ? sdp->sd_inptrs : sdp->sd_diptrs; | |
638 | ||
639 | error = gfs2_rindex_hold(sdp, &ip->i_alloc.al_ri_gh); | |
640 | if (error) | |
641 | return error; | |
642 | ||
643 | memset(&rlist, 0, sizeof(struct gfs2_rgrp_list)); | |
644 | bstart = 0; | |
645 | blen = 0; | |
646 | ||
647 | for (p = top; p < bottom; p++) { | |
648 | if (!*p) | |
649 | continue; | |
650 | ||
651 | bn = be64_to_cpu(*p); | |
652 | ||
653 | if (bstart + blen == bn) | |
654 | blen++; | |
655 | else { | |
656 | if (bstart) | |
657 | gfs2_rlist_add(sdp, &rlist, bstart); | |
658 | ||
659 | bstart = bn; | |
660 | blen = 1; | |
661 | } | |
662 | } | |
663 | ||
664 | if (bstart) | |
665 | gfs2_rlist_add(sdp, &rlist, bstart); | |
666 | else | |
667 | goto out; /* Nothing to do */ | |
668 | ||
669 | gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE, 0); | |
670 | ||
671 | for (x = 0; x < rlist.rl_rgrps; x++) { | |
672 | struct gfs2_rgrpd *rgd; | |
5c676f6d | 673 | rgd = rlist.rl_ghs[x].gh_gl->gl_object; |
b3b94faa DT |
674 | rg_blocks += rgd->rd_ri.ri_length; |
675 | } | |
676 | ||
677 | error = gfs2_glock_nq_m(rlist.rl_rgrps, rlist.rl_ghs); | |
678 | if (error) | |
679 | goto out_rlist; | |
680 | ||
681 | error = gfs2_trans_begin(sdp, rg_blocks + RES_DINODE + | |
682 | RES_INDIRECT + RES_STATFS + RES_QUOTA, | |
683 | revokes); | |
684 | if (error) | |
685 | goto out_rg_gunlock; | |
686 | ||
687 | down_write(&ip->i_rw_mutex); | |
688 | ||
d4e9c4c3 SW |
689 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
690 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
b3b94faa DT |
691 | |
692 | bstart = 0; | |
693 | blen = 0; | |
694 | ||
695 | for (p = top; p < bottom; p++) { | |
696 | if (!*p) | |
697 | continue; | |
698 | ||
699 | bn = be64_to_cpu(*p); | |
700 | ||
701 | if (bstart + blen == bn) | |
702 | blen++; | |
703 | else { | |
704 | if (bstart) { | |
705 | if (metadata) | |
706 | gfs2_free_meta(ip, bstart, blen); | |
707 | else | |
708 | gfs2_free_data(ip, bstart, blen); | |
709 | } | |
710 | ||
711 | bstart = bn; | |
712 | blen = 1; | |
713 | } | |
714 | ||
715 | *p = 0; | |
716 | if (!ip->i_di.di_blocks) | |
717 | gfs2_consist_inode(ip); | |
718 | ip->i_di.di_blocks--; | |
719 | } | |
720 | if (bstart) { | |
721 | if (metadata) | |
722 | gfs2_free_meta(ip, bstart, blen); | |
723 | else | |
724 | gfs2_free_data(ip, bstart, blen); | |
725 | } | |
726 | ||
727 | ip->i_di.di_mtime = ip->i_di.di_ctime = get_seconds(); | |
728 | ||
729 | gfs2_dinode_out(&ip->i_di, dibh->b_data); | |
730 | ||
731 | up_write(&ip->i_rw_mutex); | |
732 | ||
733 | gfs2_trans_end(sdp); | |
734 | ||
735 | out_rg_gunlock: | |
736 | gfs2_glock_dq_m(rlist.rl_rgrps, rlist.rl_ghs); | |
737 | ||
738 | out_rlist: | |
739 | gfs2_rlist_free(&rlist); | |
740 | ||
741 | out: | |
742 | gfs2_glock_dq_uninit(&ip->i_alloc.al_ri_gh); | |
743 | ||
744 | return error; | |
745 | } | |
746 | ||
747 | /** | |
748 | * do_grow - Make a file look bigger than it is | |
749 | * @ip: the inode | |
750 | * @size: the size to set the file to | |
751 | * | |
752 | * Called with an exclusive lock on @ip. | |
753 | * | |
754 | * Returns: errno | |
755 | */ | |
756 | ||
757 | static int do_grow(struct gfs2_inode *ip, uint64_t size) | |
758 | { | |
feaa7bba | 759 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
760 | struct gfs2_alloc *al; |
761 | struct buffer_head *dibh; | |
762 | unsigned int h; | |
763 | int error; | |
764 | ||
765 | al = gfs2_alloc_get(ip); | |
766 | ||
767 | error = gfs2_quota_lock(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
768 | if (error) | |
769 | goto out; | |
770 | ||
771 | error = gfs2_quota_check(ip, ip->i_di.di_uid, ip->i_di.di_gid); | |
772 | if (error) | |
773 | goto out_gunlock_q; | |
774 | ||
775 | al->al_requested = sdp->sd_max_height + RES_DATA; | |
776 | ||
777 | error = gfs2_inplace_reserve(ip); | |
778 | if (error) | |
779 | goto out_gunlock_q; | |
780 | ||
781 | error = gfs2_trans_begin(sdp, | |
782 | sdp->sd_max_height + al->al_rgd->rd_ri.ri_length + | |
783 | RES_JDATA + RES_DINODE + RES_STATFS + RES_QUOTA, 0); | |
784 | if (error) | |
785 | goto out_ipres; | |
786 | ||
787 | if (size > sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode)) { | |
788 | if (gfs2_is_stuffed(ip)) { | |
789 | error = gfs2_unstuff_dinode(ip, gfs2_unstuffer_page, | |
790 | NULL); | |
791 | if (error) | |
792 | goto out_end_trans; | |
793 | } | |
794 | ||
795 | h = calc_tree_height(ip, size); | |
796 | if (ip->i_di.di_height < h) { | |
797 | down_write(&ip->i_rw_mutex); | |
feaa7bba | 798 | error = build_height(&ip->i_inode, h); |
b3b94faa DT |
799 | up_write(&ip->i_rw_mutex); |
800 | if (error) | |
801 | goto out_end_trans; | |
802 | } | |
803 | } | |
804 | ||
805 | ip->i_di.di_size = size; | |
806 | ip->i_di.di_mtime = ip->i_di.di_ctime = get_seconds(); | |
807 | ||
808 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
809 | if (error) | |
810 | goto out_end_trans; | |
811 | ||
d4e9c4c3 | 812 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
b3b94faa DT |
813 | gfs2_dinode_out(&ip->i_di, dibh->b_data); |
814 | brelse(dibh); | |
815 | ||
816 | out_end_trans: | |
817 | gfs2_trans_end(sdp); | |
818 | ||
819 | out_ipres: | |
820 | gfs2_inplace_release(ip); | |
821 | ||
822 | out_gunlock_q: | |
823 | gfs2_quota_unlock(ip); | |
824 | ||
825 | out: | |
826 | gfs2_alloc_put(ip); | |
827 | ||
828 | return error; | |
829 | } | |
830 | ||
aa6a85a9 | 831 | static int trunc_start(struct gfs2_inode *ip, uint64_t size) |
b3b94faa | 832 | { |
feaa7bba | 833 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
834 | struct buffer_head *dibh; |
835 | int journaled = gfs2_is_jdata(ip); | |
836 | int error; | |
837 | ||
838 | error = gfs2_trans_begin(sdp, | |
839 | RES_DINODE + ((journaled) ? RES_JDATA : 0), 0); | |
840 | if (error) | |
841 | return error; | |
842 | ||
843 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
844 | if (error) | |
845 | goto out; | |
846 | ||
847 | if (gfs2_is_stuffed(ip)) { | |
848 | ip->i_di.di_size = size; | |
849 | ip->i_di.di_mtime = ip->i_di.di_ctime = get_seconds(); | |
d4e9c4c3 | 850 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
b3b94faa DT |
851 | gfs2_dinode_out(&ip->i_di, dibh->b_data); |
852 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode) + size); | |
853 | error = 1; | |
854 | ||
855 | } else { | |
18ec7d5c | 856 | if (size & (uint64_t)(sdp->sd_sb.sb_bsize - 1)) |
feaa7bba | 857 | error = gfs2_block_truncate_page(ip->i_inode.i_mapping); |
b3b94faa DT |
858 | |
859 | if (!error) { | |
860 | ip->i_di.di_size = size; | |
861 | ip->i_di.di_mtime = ip->i_di.di_ctime = get_seconds(); | |
862 | ip->i_di.di_flags |= GFS2_DIF_TRUNC_IN_PROG; | |
d4e9c4c3 | 863 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
b3b94faa DT |
864 | gfs2_dinode_out(&ip->i_di, dibh->b_data); |
865 | } | |
866 | } | |
867 | ||
868 | brelse(dibh); | |
869 | ||
870 | out: | |
871 | gfs2_trans_end(sdp); | |
872 | ||
873 | return error; | |
874 | } | |
875 | ||
876 | static int trunc_dealloc(struct gfs2_inode *ip, uint64_t size) | |
877 | { | |
878 | unsigned int height = ip->i_di.di_height; | |
879 | uint64_t lblock; | |
880 | struct metapath mp; | |
881 | int error; | |
882 | ||
883 | if (!size) | |
884 | lblock = 0; | |
18ec7d5c | 885 | else |
feaa7bba | 886 | lblock = (size - 1) >> GFS2_SB(&ip->i_inode)->sd_sb.sb_bsize_shift; |
b3b94faa DT |
887 | |
888 | find_metapath(ip, lblock, &mp); | |
889 | gfs2_alloc_get(ip); | |
890 | ||
891 | error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
892 | if (error) | |
893 | goto out; | |
894 | ||
895 | while (height--) { | |
896 | struct strip_mine sm; | |
897 | sm.sm_first = !!size; | |
898 | sm.sm_height = height; | |
899 | ||
900 | error = recursive_scan(ip, NULL, &mp, 0, 0, 1, do_strip, &sm); | |
901 | if (error) | |
902 | break; | |
903 | } | |
904 | ||
905 | gfs2_quota_unhold(ip); | |
906 | ||
907 | out: | |
908 | gfs2_alloc_put(ip); | |
909 | return error; | |
910 | } | |
911 | ||
912 | static int trunc_end(struct gfs2_inode *ip) | |
913 | { | |
feaa7bba | 914 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
915 | struct buffer_head *dibh; |
916 | int error; | |
917 | ||
918 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
919 | if (error) | |
920 | return error; | |
921 | ||
922 | down_write(&ip->i_rw_mutex); | |
923 | ||
924 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
925 | if (error) | |
926 | goto out; | |
927 | ||
928 | if (!ip->i_di.di_size) { | |
929 | ip->i_di.di_height = 0; | |
930 | ip->i_di.di_goal_meta = | |
931 | ip->i_di.di_goal_data = | |
932 | ip->i_num.no_addr; | |
933 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode)); | |
934 | } | |
935 | ip->i_di.di_mtime = ip->i_di.di_ctime = get_seconds(); | |
936 | ip->i_di.di_flags &= ~GFS2_DIF_TRUNC_IN_PROG; | |
937 | ||
d4e9c4c3 | 938 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
b3b94faa DT |
939 | gfs2_dinode_out(&ip->i_di, dibh->b_data); |
940 | brelse(dibh); | |
941 | ||
942 | out: | |
943 | up_write(&ip->i_rw_mutex); | |
944 | ||
945 | gfs2_trans_end(sdp); | |
946 | ||
947 | return error; | |
948 | } | |
949 | ||
950 | /** | |
951 | * do_shrink - make a file smaller | |
952 | * @ip: the inode | |
953 | * @size: the size to make the file | |
954 | * @truncator: function to truncate the last partial block | |
955 | * | |
956 | * Called with an exclusive lock on @ip. | |
957 | * | |
958 | * Returns: errno | |
959 | */ | |
960 | ||
aa6a85a9 | 961 | static int do_shrink(struct gfs2_inode *ip, uint64_t size) |
b3b94faa DT |
962 | { |
963 | int error; | |
964 | ||
aa6a85a9 | 965 | error = trunc_start(ip, size); |
b3b94faa DT |
966 | if (error < 0) |
967 | return error; | |
968 | if (error > 0) | |
969 | return 0; | |
970 | ||
971 | error = trunc_dealloc(ip, size); | |
972 | if (!error) | |
973 | error = trunc_end(ip); | |
974 | ||
975 | return error; | |
976 | } | |
977 | ||
978 | /** | |
666a2c53 | 979 | * gfs2_truncatei - make a file a given size |
b3b94faa DT |
980 | * @ip: the inode |
981 | * @size: the size to make the file | |
982 | * @truncator: function to truncate the last partial block | |
983 | * | |
984 | * The file size can grow, shrink, or stay the same size. | |
985 | * | |
986 | * Returns: errno | |
987 | */ | |
988 | ||
aa6a85a9 | 989 | int gfs2_truncatei(struct gfs2_inode *ip, uint64_t size) |
b3b94faa DT |
990 | { |
991 | int error; | |
992 | ||
feaa7bba | 993 | if (gfs2_assert_warn(GFS2_SB(&ip->i_inode), S_ISREG(ip->i_di.di_mode))) |
b3b94faa DT |
994 | return -EINVAL; |
995 | ||
996 | if (size > ip->i_di.di_size) | |
997 | error = do_grow(ip, size); | |
998 | else | |
aa6a85a9 | 999 | error = do_shrink(ip, size); |
b3b94faa DT |
1000 | |
1001 | return error; | |
1002 | } | |
1003 | ||
1004 | int gfs2_truncatei_resume(struct gfs2_inode *ip) | |
1005 | { | |
1006 | int error; | |
1007 | error = trunc_dealloc(ip, ip->i_di.di_size); | |
1008 | if (!error) | |
1009 | error = trunc_end(ip); | |
1010 | return error; | |
1011 | } | |
1012 | ||
1013 | int gfs2_file_dealloc(struct gfs2_inode *ip) | |
1014 | { | |
1015 | return trunc_dealloc(ip, 0); | |
1016 | } | |
1017 | ||
1018 | /** | |
1019 | * gfs2_write_calc_reserv - calculate number of blocks needed to write to a file | |
1020 | * @ip: the file | |
1021 | * @len: the number of bytes to be written to the file | |
1022 | * @data_blocks: returns the number of data blocks required | |
1023 | * @ind_blocks: returns the number of indirect blocks required | |
1024 | * | |
1025 | */ | |
1026 | ||
1027 | void gfs2_write_calc_reserv(struct gfs2_inode *ip, unsigned int len, | |
1028 | unsigned int *data_blocks, unsigned int *ind_blocks) | |
1029 | { | |
feaa7bba | 1030 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
1031 | unsigned int tmp; |
1032 | ||
18ec7d5c | 1033 | if (gfs2_is_dir(ip)) { |
5c676f6d | 1034 | *data_blocks = DIV_ROUND_UP(len, sdp->sd_jbsize) + 2; |
b3b94faa DT |
1035 | *ind_blocks = 3 * (sdp->sd_max_jheight - 1); |
1036 | } else { | |
1037 | *data_blocks = (len >> sdp->sd_sb.sb_bsize_shift) + 3; | |
1038 | *ind_blocks = 3 * (sdp->sd_max_height - 1); | |
1039 | } | |
1040 | ||
1041 | for (tmp = *data_blocks; tmp > sdp->sd_diptrs;) { | |
5c676f6d | 1042 | tmp = DIV_ROUND_UP(tmp, sdp->sd_inptrs); |
b3b94faa DT |
1043 | *ind_blocks += tmp; |
1044 | } | |
1045 | } | |
1046 | ||
1047 | /** | |
1048 | * gfs2_write_alloc_required - figure out if a write will require an allocation | |
1049 | * @ip: the file being written to | |
1050 | * @offset: the offset to write to | |
1051 | * @len: the number of bytes being written | |
1052 | * @alloc_required: set to 1 if an alloc is required, 0 otherwise | |
1053 | * | |
1054 | * Returns: errno | |
1055 | */ | |
1056 | ||
1057 | int gfs2_write_alloc_required(struct gfs2_inode *ip, uint64_t offset, | |
1058 | unsigned int len, int *alloc_required) | |
1059 | { | |
feaa7bba | 1060 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
1061 | uint64_t lblock, lblock_stop, dblock; |
1062 | uint32_t extlen; | |
1063 | int new = 0; | |
1064 | int error = 0; | |
1065 | ||
1066 | *alloc_required = 0; | |
1067 | ||
1068 | if (!len) | |
1069 | return 0; | |
1070 | ||
1071 | if (gfs2_is_stuffed(ip)) { | |
1072 | if (offset + len > | |
1073 | sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode)) | |
1074 | *alloc_required = 1; | |
1075 | return 0; | |
1076 | } | |
1077 | ||
18ec7d5c | 1078 | if (gfs2_is_dir(ip)) { |
b3b94faa DT |
1079 | unsigned int bsize = sdp->sd_jbsize; |
1080 | lblock = offset; | |
1081 | do_div(lblock, bsize); | |
1082 | lblock_stop = offset + len + bsize - 1; | |
1083 | do_div(lblock_stop, bsize); | |
1084 | } else { | |
1085 | unsigned int shift = sdp->sd_sb.sb_bsize_shift; | |
1086 | lblock = offset >> shift; | |
1087 | lblock_stop = (offset + len + sdp->sd_sb.sb_bsize - 1) >> shift; | |
1088 | } | |
1089 | ||
1090 | for (; lblock < lblock_stop; lblock += extlen) { | |
feaa7bba | 1091 | error = gfs2_extent_map(&ip->i_inode, lblock, &new, &dblock, &extlen); |
b3b94faa DT |
1092 | if (error) |
1093 | return error; | |
1094 | ||
1095 | if (!dblock) { | |
1096 | *alloc_required = 1; | |
1097 | return 0; | |
1098 | } | |
1099 | } | |
1100 | ||
1101 | return 0; | |
1102 | } | |
1103 |