]>
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 | |
e9fc2aa0 | 7 | * of the GNU General Public License version 2. |
b3b94faa DT |
8 | */ |
9 | ||
10 | /* | |
61e085a8 SW |
11 | * Implements Extendible Hashing as described in: |
12 | * "Extendible Hashing" by Fagin, et al in | |
13 | * __ACM Trans. on Database Systems__, Sept 1979. | |
14 | * | |
15 | * | |
16 | * Here's the layout of dirents which is essentially the same as that of ext2 | |
17 | * within a single block. The field de_name_len is the number of bytes | |
18 | * actually required for the name (no null terminator). The field de_rec_len | |
19 | * is the number of bytes allocated to the dirent. The offset of the next | |
20 | * dirent in the block is (dirent + dirent->de_rec_len). When a dirent is | |
21 | * deleted, the preceding dirent inherits its allocated space, ie | |
22 | * prev->de_rec_len += deleted->de_rec_len. Since the next dirent is obtained | |
23 | * by adding de_rec_len to the current dirent, this essentially causes the | |
24 | * deleted dirent to get jumped over when iterating through all the dirents. | |
25 | * | |
26 | * When deleting the first dirent in a block, there is no previous dirent so | |
27 | * the field de_ino is set to zero to designate it as deleted. When allocating | |
28 | * a dirent, gfs2_dirent_alloc iterates through the dirents in a block. If the | |
29 | * first dirent has (de_ino == 0) and de_rec_len is large enough, this first | |
30 | * dirent is allocated. Otherwise it must go through all the 'used' dirents | |
31 | * searching for one in which the amount of total space minus the amount of | |
32 | * used space will provide enough space for the new dirent. | |
33 | * | |
34 | * There are two types of blocks in which dirents reside. In a stuffed dinode, | |
35 | * the dirents begin at offset sizeof(struct gfs2_dinode) from the beginning of | |
36 | * the block. In leaves, they begin at offset sizeof(struct gfs2_leaf) from the | |
37 | * beginning of the leaf block. The dirents reside in leaves when | |
38 | * | |
39 | * dip->i_di.di_flags & GFS2_DIF_EXHASH is true | |
40 | * | |
41 | * Otherwise, the dirents are "linear", within a single stuffed dinode block. | |
42 | * | |
43 | * When the dirents are in leaves, the actual contents of the directory file are | |
44 | * used as an array of 64-bit block pointers pointing to the leaf blocks. The | |
45 | * dirents are NOT in the directory file itself. There can be more than one | |
46 | * block pointer in the array that points to the same leaf. In fact, when a | |
47 | * directory is first converted from linear to exhash, all of the pointers | |
48 | * point to the same leaf. | |
49 | * | |
50 | * When a leaf is completely full, the size of the hash table can be | |
51 | * doubled unless it is already at the maximum size which is hard coded into | |
52 | * GFS2_DIR_MAX_DEPTH. After that, leaves are chained together in a linked list, | |
53 | * but never before the maximum hash table size has been reached. | |
54 | */ | |
b3b94faa DT |
55 | |
56 | #include <linux/sched.h> | |
57 | #include <linux/slab.h> | |
58 | #include <linux/spinlock.h> | |
b3b94faa DT |
59 | #include <linux/buffer_head.h> |
60 | #include <linux/sort.h> | |
5c676f6d | 61 | #include <linux/gfs2_ondisk.h> |
71b86f56 | 62 | #include <linux/crc32.h> |
fe1bdedc | 63 | #include <linux/vmalloc.h> |
7d308590 | 64 | #include <linux/lm_interface.h> |
b3b94faa DT |
65 | |
66 | #include "gfs2.h" | |
5c676f6d | 67 | #include "incore.h" |
b3b94faa DT |
68 | #include "dir.h" |
69 | #include "glock.h" | |
70 | #include "inode.h" | |
b3b94faa DT |
71 | #include "meta_io.h" |
72 | #include "quota.h" | |
73 | #include "rgrp.h" | |
74 | #include "trans.h" | |
e13940ba | 75 | #include "bmap.h" |
5c676f6d | 76 | #include "util.h" |
b3b94faa DT |
77 | |
78 | #define IS_LEAF 1 /* Hashed (leaf) directory */ | |
79 | #define IS_DINODE 2 /* Linear (stuffed dinode block) directory */ | |
80 | ||
cd915493 SW |
81 | #define gfs2_disk_hash2offset(h) (((u64)(h)) >> 1) |
82 | #define gfs2_dir_offset2hash(p) ((u32)(((u64)(p)) << 1)) | |
b3b94faa | 83 | |
907b9bce | 84 | typedef int (*leaf_call_t) (struct gfs2_inode *dip, u32 index, u32 len, |
2bdbc5d7 SW |
85 | u64 leaf_no, void *data); |
86 | typedef int (*gfs2_dscan_t)(const struct gfs2_dirent *dent, | |
87 | const struct qstr *name, void *opaque); | |
b3b94faa | 88 | |
61e085a8 | 89 | |
cd915493 | 90 | int gfs2_dir_get_new_buffer(struct gfs2_inode *ip, u64 block, |
61e085a8 | 91 | struct buffer_head **bhp) |
e13940ba SW |
92 | { |
93 | struct buffer_head *bh; | |
e13940ba | 94 | |
61e085a8 SW |
95 | bh = gfs2_meta_new(ip->i_gl, block); |
96 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
97 | gfs2_metatype_set(bh, GFS2_METATYPE_JD, GFS2_FORMAT_JD); | |
98 | gfs2_buffer_clear_tail(bh, sizeof(struct gfs2_meta_header)); | |
e13940ba SW |
99 | *bhp = bh; |
100 | return 0; | |
101 | } | |
102 | ||
cd915493 | 103 | static int gfs2_dir_get_existing_buffer(struct gfs2_inode *ip, u64 block, |
61e085a8 SW |
104 | struct buffer_head **bhp) |
105 | { | |
106 | struct buffer_head *bh; | |
107 | int error; | |
e13940ba | 108 | |
7276b3b0 | 109 | error = gfs2_meta_read(ip->i_gl, block, DIO_WAIT, &bh); |
61e085a8 SW |
110 | if (error) |
111 | return error; | |
feaa7bba | 112 | if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), bh, GFS2_METATYPE_JD)) { |
61e085a8 SW |
113 | brelse(bh); |
114 | return -EIO; | |
115 | } | |
116 | *bhp = bh; | |
117 | return 0; | |
118 | } | |
e13940ba SW |
119 | |
120 | static int gfs2_dir_write_stuffed(struct gfs2_inode *ip, const char *buf, | |
121 | unsigned int offset, unsigned int size) | |
e13940ba SW |
122 | { |
123 | struct buffer_head *dibh; | |
124 | int error; | |
125 | ||
126 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
127 | if (error) | |
128 | return error; | |
129 | ||
130 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
c752666c | 131 | memcpy(dibh->b_data + offset + sizeof(struct gfs2_dinode), buf, size); |
e13940ba SW |
132 | if (ip->i_di.di_size < offset + size) |
133 | ip->i_di.di_size = offset + size; | |
134 | ip->i_di.di_mtime = ip->i_di.di_ctime = get_seconds(); | |
135 | gfs2_dinode_out(&ip->i_di, dibh->b_data); | |
136 | ||
137 | brelse(dibh); | |
138 | ||
139 | return size; | |
140 | } | |
141 | ||
142 | ||
143 | ||
144 | /** | |
145 | * gfs2_dir_write_data - Write directory information to the inode | |
146 | * @ip: The GFS2 inode | |
147 | * @buf: The buffer containing information to be written | |
148 | * @offset: The file offset to start writing at | |
149 | * @size: The amount of data to write | |
150 | * | |
151 | * Returns: The number of bytes correctly written or error code | |
152 | */ | |
153 | static int gfs2_dir_write_data(struct gfs2_inode *ip, const char *buf, | |
cd915493 | 154 | u64 offset, unsigned int size) |
e13940ba | 155 | { |
feaa7bba | 156 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
e13940ba | 157 | struct buffer_head *dibh; |
cd915493 SW |
158 | u64 lblock, dblock; |
159 | u32 extlen = 0; | |
e13940ba SW |
160 | unsigned int o; |
161 | int copied = 0; | |
162 | int error = 0; | |
163 | ||
164 | if (!size) | |
165 | return 0; | |
166 | ||
167 | if (gfs2_is_stuffed(ip) && | |
168 | offset + size <= sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode)) | |
568f4c96 SW |
169 | return gfs2_dir_write_stuffed(ip, buf, (unsigned int)offset, |
170 | size); | |
e13940ba SW |
171 | |
172 | if (gfs2_assert_warn(sdp, gfs2_is_jdata(ip))) | |
173 | return -EINVAL; | |
174 | ||
175 | if (gfs2_is_stuffed(ip)) { | |
f25ef0c1 | 176 | error = gfs2_unstuff_dinode(ip, NULL); |
e13940ba | 177 | if (error) |
c752666c | 178 | return error; |
e13940ba SW |
179 | } |
180 | ||
181 | lblock = offset; | |
182 | o = do_div(lblock, sdp->sd_jbsize) + sizeof(struct gfs2_meta_header); | |
183 | ||
184 | while (copied < size) { | |
185 | unsigned int amount; | |
186 | struct buffer_head *bh; | |
187 | int new; | |
188 | ||
189 | amount = size - copied; | |
190 | if (amount > sdp->sd_sb.sb_bsize - o) | |
191 | amount = sdp->sd_sb.sb_bsize - o; | |
192 | ||
193 | if (!extlen) { | |
194 | new = 1; | |
feaa7bba | 195 | error = gfs2_extent_map(&ip->i_inode, lblock, &new, |
fd88de56 | 196 | &dblock, &extlen); |
e13940ba SW |
197 | if (error) |
198 | goto fail; | |
199 | error = -EIO; | |
200 | if (gfs2_assert_withdraw(sdp, dblock)) | |
201 | goto fail; | |
202 | } | |
203 | ||
61e085a8 SW |
204 | if (amount == sdp->sd_jbsize || new) |
205 | error = gfs2_dir_get_new_buffer(ip, dblock, &bh); | |
206 | else | |
207 | error = gfs2_dir_get_existing_buffer(ip, dblock, &bh); | |
208 | ||
e13940ba SW |
209 | if (error) |
210 | goto fail; | |
211 | ||
212 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
213 | memcpy(bh->b_data + o, buf, amount); | |
214 | brelse(bh); | |
215 | if (error) | |
216 | goto fail; | |
217 | ||
899bb264 | 218 | buf += amount; |
e13940ba SW |
219 | copied += amount; |
220 | lblock++; | |
221 | dblock++; | |
222 | extlen--; | |
223 | ||
224 | o = sizeof(struct gfs2_meta_header); | |
225 | } | |
226 | ||
227 | out: | |
228 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
229 | if (error) | |
230 | return error; | |
231 | ||
232 | if (ip->i_di.di_size < offset + copied) | |
233 | ip->i_di.di_size = offset + copied; | |
234 | ip->i_di.di_mtime = ip->i_di.di_ctime = get_seconds(); | |
235 | ||
236 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
237 | gfs2_dinode_out(&ip->i_di, dibh->b_data); | |
238 | brelse(dibh); | |
239 | ||
240 | return copied; | |
241 | fail: | |
242 | if (copied) | |
243 | goto out; | |
244 | return error; | |
245 | } | |
246 | ||
247 | static int gfs2_dir_read_stuffed(struct gfs2_inode *ip, char *buf, | |
7276b3b0 | 248 | u64 offset, unsigned int size) |
e13940ba SW |
249 | { |
250 | struct buffer_head *dibh; | |
251 | int error; | |
252 | ||
253 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
254 | if (!error) { | |
255 | offset += sizeof(struct gfs2_dinode); | |
256 | memcpy(buf, dibh->b_data + offset, size); | |
257 | brelse(dibh); | |
258 | } | |
259 | ||
260 | return (error) ? error : size; | |
261 | } | |
262 | ||
263 | ||
264 | /** | |
265 | * gfs2_dir_read_data - Read a data from a directory inode | |
266 | * @ip: The GFS2 Inode | |
267 | * @buf: The buffer to place result into | |
268 | * @offset: File offset to begin jdata_readng from | |
269 | * @size: Amount of data to transfer | |
270 | * | |
271 | * Returns: The amount of data actually copied or the error | |
272 | */ | |
7276b3b0 SW |
273 | static int gfs2_dir_read_data(struct gfs2_inode *ip, char *buf, u64 offset, |
274 | unsigned int size, unsigned ra) | |
e13940ba | 275 | { |
feaa7bba | 276 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
cd915493 SW |
277 | u64 lblock, dblock; |
278 | u32 extlen = 0; | |
e13940ba SW |
279 | unsigned int o; |
280 | int copied = 0; | |
281 | int error = 0; | |
282 | ||
283 | if (offset >= ip->i_di.di_size) | |
284 | return 0; | |
285 | ||
c5392124 | 286 | if (offset + size > ip->i_di.di_size) |
e13940ba SW |
287 | size = ip->i_di.di_size - offset; |
288 | ||
289 | if (!size) | |
290 | return 0; | |
291 | ||
292 | if (gfs2_is_stuffed(ip)) | |
7276b3b0 | 293 | return gfs2_dir_read_stuffed(ip, buf, offset, size); |
e13940ba SW |
294 | |
295 | if (gfs2_assert_warn(sdp, gfs2_is_jdata(ip))) | |
296 | return -EINVAL; | |
297 | ||
298 | lblock = offset; | |
299 | o = do_div(lblock, sdp->sd_jbsize) + sizeof(struct gfs2_meta_header); | |
300 | ||
301 | while (copied < size) { | |
302 | unsigned int amount; | |
303 | struct buffer_head *bh; | |
304 | int new; | |
305 | ||
306 | amount = size - copied; | |
307 | if (amount > sdp->sd_sb.sb_bsize - o) | |
308 | amount = sdp->sd_sb.sb_bsize - o; | |
309 | ||
310 | if (!extlen) { | |
311 | new = 0; | |
feaa7bba | 312 | error = gfs2_extent_map(&ip->i_inode, lblock, &new, |
fd88de56 | 313 | &dblock, &extlen); |
7276b3b0 | 314 | if (error || !dblock) |
e13940ba | 315 | goto fail; |
7276b3b0 SW |
316 | BUG_ON(extlen < 1); |
317 | if (!ra) | |
318 | extlen = 1; | |
319 | bh = gfs2_meta_ra(ip->i_gl, dblock, extlen); | |
e13940ba | 320 | } |
7276b3b0 SW |
321 | if (!bh) { |
322 | error = gfs2_meta_read(ip->i_gl, dblock, DIO_WAIT, &bh); | |
e13940ba SW |
323 | if (error) |
324 | goto fail; | |
7276b3b0 SW |
325 | } |
326 | error = gfs2_metatype_check(sdp, bh, GFS2_METATYPE_JD); | |
327 | if (error) { | |
328 | brelse(bh); | |
329 | goto fail; | |
330 | } | |
331 | dblock++; | |
332 | extlen--; | |
e13940ba SW |
333 | memcpy(buf, bh->b_data + o, amount); |
334 | brelse(bh); | |
7276b3b0 | 335 | bh = NULL; |
899bb264 | 336 | buf += amount; |
e13940ba SW |
337 | copied += amount; |
338 | lblock++; | |
e13940ba SW |
339 | o = sizeof(struct gfs2_meta_header); |
340 | } | |
341 | ||
342 | return copied; | |
343 | fail: | |
344 | return (copied) ? copied : error; | |
345 | } | |
346 | ||
c752666c SW |
347 | static inline int __gfs2_dirent_find(const struct gfs2_dirent *dent, |
348 | const struct qstr *name, int ret) | |
349 | { | |
350 | if (dent->de_inum.no_addr != 0 && | |
351 | be32_to_cpu(dent->de_hash) == name->hash && | |
352 | be16_to_cpu(dent->de_name_len) == name->len && | |
2bdbc5d7 | 353 | memcmp(dent+1, name->name, name->len) == 0) |
c752666c SW |
354 | return ret; |
355 | return 0; | |
356 | } | |
357 | ||
358 | static int gfs2_dirent_find(const struct gfs2_dirent *dent, | |
71b86f56 SW |
359 | const struct qstr *name, |
360 | void *opaque) | |
c752666c SW |
361 | { |
362 | return __gfs2_dirent_find(dent, name, 1); | |
363 | } | |
364 | ||
365 | static int gfs2_dirent_prev(const struct gfs2_dirent *dent, | |
71b86f56 SW |
366 | const struct qstr *name, |
367 | void *opaque) | |
c752666c SW |
368 | { |
369 | return __gfs2_dirent_find(dent, name, 2); | |
370 | } | |
371 | ||
372 | /* | |
373 | * name->name holds ptr to start of block. | |
374 | * name->len holds size of block. | |
b3b94faa | 375 | */ |
c752666c | 376 | static int gfs2_dirent_last(const struct gfs2_dirent *dent, |
71b86f56 SW |
377 | const struct qstr *name, |
378 | void *opaque) | |
c752666c SW |
379 | { |
380 | const char *start = name->name; | |
381 | const char *end = (const char *)dent + be16_to_cpu(dent->de_rec_len); | |
382 | if (name->len == (end - start)) | |
383 | return 1; | |
384 | return 0; | |
385 | } | |
b3b94faa | 386 | |
c752666c | 387 | static int gfs2_dirent_find_space(const struct gfs2_dirent *dent, |
71b86f56 SW |
388 | const struct qstr *name, |
389 | void *opaque) | |
b3b94faa | 390 | { |
c752666c SW |
391 | unsigned required = GFS2_DIRENT_SIZE(name->len); |
392 | unsigned actual = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len)); | |
393 | unsigned totlen = be16_to_cpu(dent->de_rec_len); | |
394 | ||
71b86f56 SW |
395 | if (!dent->de_inum.no_addr) |
396 | actual = GFS2_DIRENT_SIZE(0); | |
c5392124 | 397 | if (totlen - actual >= required) |
c752666c SW |
398 | return 1; |
399 | return 0; | |
400 | } | |
401 | ||
71b86f56 SW |
402 | struct dirent_gather { |
403 | const struct gfs2_dirent **pdent; | |
404 | unsigned offset; | |
405 | }; | |
406 | ||
407 | static int gfs2_dirent_gather(const struct gfs2_dirent *dent, | |
408 | const struct qstr *name, | |
409 | void *opaque) | |
410 | { | |
411 | struct dirent_gather *g = opaque; | |
412 | if (dent->de_inum.no_addr) { | |
413 | g->pdent[g->offset++] = dent; | |
414 | } | |
415 | return 0; | |
416 | } | |
417 | ||
c752666c SW |
418 | /* |
419 | * Other possible things to check: | |
420 | * - Inode located within filesystem size (and on valid block) | |
421 | * - Valid directory entry type | |
422 | * Not sure how heavy-weight we want to make this... could also check | |
423 | * hash is correct for example, but that would take a lot of extra time. | |
424 | * For now the most important thing is to check that the various sizes | |
425 | * are correct. | |
426 | */ | |
427 | static int gfs2_check_dirent(struct gfs2_dirent *dent, unsigned int offset, | |
428 | unsigned int size, unsigned int len, int first) | |
429 | { | |
430 | const char *msg = "gfs2_dirent too small"; | |
431 | if (unlikely(size < sizeof(struct gfs2_dirent))) | |
432 | goto error; | |
433 | msg = "gfs2_dirent misaligned"; | |
434 | if (unlikely(offset & 0x7)) | |
435 | goto error; | |
436 | msg = "gfs2_dirent points beyond end of block"; | |
437 | if (unlikely(offset + size > len)) | |
438 | goto error; | |
439 | msg = "zero inode number"; | |
440 | if (unlikely(!first && !dent->de_inum.no_addr)) | |
441 | goto error; | |
442 | msg = "name length is greater than space in dirent"; | |
443 | if (dent->de_inum.no_addr && | |
444 | unlikely(sizeof(struct gfs2_dirent)+be16_to_cpu(dent->de_name_len) > | |
445 | size)) | |
446 | goto error; | |
447 | return 0; | |
448 | error: | |
449 | printk(KERN_WARNING "gfs2_check_dirent: %s (%s)\n", msg, | |
450 | first ? "first in block" : "not first in block"); | |
451 | return -EIO; | |
b3b94faa DT |
452 | } |
453 | ||
71b86f56 | 454 | static int gfs2_dirent_offset(const void *buf) |
c752666c | 455 | { |
71b86f56 SW |
456 | const struct gfs2_meta_header *h = buf; |
457 | int offset; | |
c752666c SW |
458 | |
459 | BUG_ON(buf == NULL); | |
c752666c | 460 | |
e3167ded | 461 | switch(be32_to_cpu(h->mh_type)) { |
c752666c SW |
462 | case GFS2_METATYPE_LF: |
463 | offset = sizeof(struct gfs2_leaf); | |
464 | break; | |
465 | case GFS2_METATYPE_DI: | |
466 | offset = sizeof(struct gfs2_dinode); | |
467 | break; | |
468 | default: | |
469 | goto wrong_type; | |
470 | } | |
71b86f56 SW |
471 | return offset; |
472 | wrong_type: | |
473 | printk(KERN_WARNING "gfs2_scan_dirent: wrong block type %u\n", | |
e3167ded | 474 | be32_to_cpu(h->mh_type)); |
71b86f56 SW |
475 | return -1; |
476 | } | |
477 | ||
2bdbc5d7 | 478 | static struct gfs2_dirent *gfs2_dirent_scan(struct inode *inode, void *buf, |
71b86f56 SW |
479 | unsigned int len, gfs2_dscan_t scan, |
480 | const struct qstr *name, | |
481 | void *opaque) | |
482 | { | |
483 | struct gfs2_dirent *dent, *prev; | |
484 | unsigned offset; | |
485 | unsigned size; | |
486 | int ret = 0; | |
c752666c | 487 | |
71b86f56 SW |
488 | ret = gfs2_dirent_offset(buf); |
489 | if (ret < 0) | |
490 | goto consist_inode; | |
491 | ||
492 | offset = ret; | |
c752666c | 493 | prev = NULL; |
2bdbc5d7 | 494 | dent = buf + offset; |
c752666c SW |
495 | size = be16_to_cpu(dent->de_rec_len); |
496 | if (gfs2_check_dirent(dent, offset, size, len, 1)) | |
497 | goto consist_inode; | |
498 | do { | |
71b86f56 | 499 | ret = scan(dent, name, opaque); |
c752666c SW |
500 | if (ret) |
501 | break; | |
502 | offset += size; | |
503 | if (offset == len) | |
504 | break; | |
505 | prev = dent; | |
2bdbc5d7 | 506 | dent = buf + offset; |
c752666c SW |
507 | size = be16_to_cpu(dent->de_rec_len); |
508 | if (gfs2_check_dirent(dent, offset, size, len, 0)) | |
509 | goto consist_inode; | |
510 | } while(1); | |
511 | ||
512 | switch(ret) { | |
513 | case 0: | |
514 | return NULL; | |
515 | case 1: | |
516 | return dent; | |
517 | case 2: | |
518 | return prev ? prev : dent; | |
519 | default: | |
520 | BUG_ON(ret > 0); | |
521 | return ERR_PTR(ret); | |
522 | } | |
523 | ||
c752666c | 524 | consist_inode: |
feaa7bba | 525 | gfs2_consist_inode(GFS2_I(inode)); |
c752666c SW |
526 | return ERR_PTR(-EIO); |
527 | } | |
528 | ||
529 | ||
b3b94faa DT |
530 | /** |
531 | * dirent_first - Return the first dirent | |
532 | * @dip: the directory | |
533 | * @bh: The buffer | |
534 | * @dent: Pointer to list of dirents | |
535 | * | |
536 | * return first dirent whether bh points to leaf or stuffed dinode | |
537 | * | |
538 | * Returns: IS_LEAF, IS_DINODE, or -errno | |
539 | */ | |
540 | ||
541 | static int dirent_first(struct gfs2_inode *dip, struct buffer_head *bh, | |
542 | struct gfs2_dirent **dent) | |
543 | { | |
544 | struct gfs2_meta_header *h = (struct gfs2_meta_header *)bh->b_data; | |
545 | ||
e3167ded | 546 | if (be32_to_cpu(h->mh_type) == GFS2_METATYPE_LF) { |
feaa7bba | 547 | if (gfs2_meta_check(GFS2_SB(&dip->i_inode), bh)) |
b3b94faa DT |
548 | return -EIO; |
549 | *dent = (struct gfs2_dirent *)(bh->b_data + | |
550 | sizeof(struct gfs2_leaf)); | |
551 | return IS_LEAF; | |
552 | } else { | |
feaa7bba | 553 | if (gfs2_metatype_check(GFS2_SB(&dip->i_inode), bh, GFS2_METATYPE_DI)) |
b3b94faa DT |
554 | return -EIO; |
555 | *dent = (struct gfs2_dirent *)(bh->b_data + | |
556 | sizeof(struct gfs2_dinode)); | |
557 | return IS_DINODE; | |
558 | } | |
559 | } | |
560 | ||
2bdbc5d7 SW |
561 | static int dirent_check_reclen(struct gfs2_inode *dip, |
562 | const struct gfs2_dirent *d, const void *end_p) | |
563 | { | |
564 | const void *ptr = d; | |
565 | u16 rec_len = be16_to_cpu(d->de_rec_len); | |
566 | ||
567 | if (unlikely(rec_len < sizeof(struct gfs2_dirent))) | |
568 | goto broken; | |
569 | ptr += rec_len; | |
570 | if (ptr < end_p) | |
571 | return rec_len; | |
572 | if (ptr == end_p) | |
573 | return -ENOENT; | |
574 | broken: | |
575 | gfs2_consist_inode(dip); | |
576 | return -EIO; | |
577 | } | |
578 | ||
b3b94faa DT |
579 | /** |
580 | * dirent_next - Next dirent | |
581 | * @dip: the directory | |
582 | * @bh: The buffer | |
583 | * @dent: Pointer to list of dirents | |
584 | * | |
585 | * Returns: 0 on success, error code otherwise | |
586 | */ | |
587 | ||
588 | static int dirent_next(struct gfs2_inode *dip, struct buffer_head *bh, | |
589 | struct gfs2_dirent **dent) | |
590 | { | |
2bdbc5d7 SW |
591 | struct gfs2_dirent *cur = *dent, *tmp; |
592 | char *bh_end = bh->b_data + bh->b_size; | |
593 | int ret; | |
b3b94faa | 594 | |
2bdbc5d7 SW |
595 | ret = dirent_check_reclen(dip, cur, bh_end); |
596 | if (ret < 0) | |
597 | return ret; | |
4dd651ad | 598 | |
2bdbc5d7 SW |
599 | tmp = (void *)cur + ret; |
600 | ret = dirent_check_reclen(dip, tmp, bh_end); | |
601 | if (ret == -EIO) | |
602 | return ret; | |
4dd651ad | 603 | |
b3b94faa DT |
604 | /* Only the first dent could ever have de_inum.no_addr == 0 */ |
605 | if (!tmp->de_inum.no_addr) { | |
606 | gfs2_consist_inode(dip); | |
607 | return -EIO; | |
608 | } | |
609 | ||
610 | *dent = tmp; | |
b3b94faa DT |
611 | return 0; |
612 | } | |
613 | ||
614 | /** | |
615 | * dirent_del - Delete a dirent | |
616 | * @dip: The GFS2 inode | |
617 | * @bh: The buffer | |
618 | * @prev: The previous dirent | |
619 | * @cur: The current dirent | |
620 | * | |
621 | */ | |
622 | ||
623 | static void dirent_del(struct gfs2_inode *dip, struct buffer_head *bh, | |
624 | struct gfs2_dirent *prev, struct gfs2_dirent *cur) | |
625 | { | |
cd915493 | 626 | u16 cur_rec_len, prev_rec_len; |
b3b94faa DT |
627 | |
628 | if (!cur->de_inum.no_addr) { | |
629 | gfs2_consist_inode(dip); | |
630 | return; | |
631 | } | |
632 | ||
d4e9c4c3 | 633 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
b3b94faa DT |
634 | |
635 | /* If there is no prev entry, this is the first entry in the block. | |
636 | The de_rec_len is already as big as it needs to be. Just zero | |
637 | out the inode number and return. */ | |
638 | ||
639 | if (!prev) { | |
640 | cur->de_inum.no_addr = 0; /* No endianess worries */ | |
641 | return; | |
642 | } | |
643 | ||
644 | /* Combine this dentry with the previous one. */ | |
645 | ||
fc69d0d3 SW |
646 | prev_rec_len = be16_to_cpu(prev->de_rec_len); |
647 | cur_rec_len = be16_to_cpu(cur->de_rec_len); | |
b3b94faa DT |
648 | |
649 | if ((char *)prev + prev_rec_len != (char *)cur) | |
650 | gfs2_consist_inode(dip); | |
651 | if ((char *)cur + cur_rec_len > bh->b_data + bh->b_size) | |
652 | gfs2_consist_inode(dip); | |
653 | ||
654 | prev_rec_len += cur_rec_len; | |
fc69d0d3 | 655 | prev->de_rec_len = cpu_to_be16(prev_rec_len); |
b3b94faa DT |
656 | } |
657 | ||
c752666c SW |
658 | /* |
659 | * Takes a dent from which to grab space as an argument. Returns the | |
660 | * newly created dent. | |
b3b94faa | 661 | */ |
08bc2dbc AB |
662 | static struct gfs2_dirent *gfs2_init_dirent(struct inode *inode, |
663 | struct gfs2_dirent *dent, | |
664 | const struct qstr *name, | |
665 | struct buffer_head *bh) | |
b3b94faa | 666 | { |
feaa7bba | 667 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c SW |
668 | struct gfs2_dirent *ndent; |
669 | unsigned offset = 0, totlen; | |
670 | ||
671 | if (dent->de_inum.no_addr) | |
672 | offset = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len)); | |
673 | totlen = be16_to_cpu(dent->de_rec_len); | |
674 | BUG_ON(offset + name->len > totlen); | |
675 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
676 | ndent = (struct gfs2_dirent *)((char *)dent + offset); | |
677 | dent->de_rec_len = cpu_to_be16(offset); | |
678 | gfs2_qstr2dirent(name, totlen - offset, ndent); | |
679 | return ndent; | |
b3b94faa DT |
680 | } |
681 | ||
c752666c SW |
682 | static struct gfs2_dirent *gfs2_dirent_alloc(struct inode *inode, |
683 | struct buffer_head *bh, | |
684 | const struct qstr *name) | |
b3b94faa DT |
685 | { |
686 | struct gfs2_dirent *dent; | |
907b9bce | 687 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, |
71b86f56 | 688 | gfs2_dirent_find_space, name, NULL); |
c752666c SW |
689 | if (!dent || IS_ERR(dent)) |
690 | return dent; | |
691 | return gfs2_init_dirent(inode, dent, name, bh); | |
b3b94faa DT |
692 | } |
693 | ||
cd915493 | 694 | static int get_leaf(struct gfs2_inode *dip, u64 leaf_no, |
b3b94faa DT |
695 | struct buffer_head **bhp) |
696 | { | |
697 | int error; | |
698 | ||
7276b3b0 | 699 | error = gfs2_meta_read(dip->i_gl, leaf_no, DIO_WAIT, bhp); |
feaa7bba SW |
700 | if (!error && gfs2_metatype_check(GFS2_SB(&dip->i_inode), *bhp, GFS2_METATYPE_LF)) { |
701 | /* printk(KERN_INFO "block num=%llu\n", leaf_no); */ | |
b3b94faa | 702 | error = -EIO; |
feaa7bba | 703 | } |
b3b94faa DT |
704 | |
705 | return error; | |
706 | } | |
707 | ||
708 | /** | |
709 | * get_leaf_nr - Get a leaf number associated with the index | |
710 | * @dip: The GFS2 inode | |
711 | * @index: | |
712 | * @leaf_out: | |
713 | * | |
714 | * Returns: 0 on success, error code otherwise | |
715 | */ | |
716 | ||
cd915493 SW |
717 | static int get_leaf_nr(struct gfs2_inode *dip, u32 index, |
718 | u64 *leaf_out) | |
b3b94faa | 719 | { |
cd915493 | 720 | u64 leaf_no; |
b3b94faa DT |
721 | int error; |
722 | ||
e13940ba | 723 | error = gfs2_dir_read_data(dip, (char *)&leaf_no, |
cd915493 | 724 | index * sizeof(u64), |
7276b3b0 | 725 | sizeof(u64), 0); |
cd915493 | 726 | if (error != sizeof(u64)) |
b3b94faa DT |
727 | return (error < 0) ? error : -EIO; |
728 | ||
729 | *leaf_out = be64_to_cpu(leaf_no); | |
730 | ||
731 | return 0; | |
732 | } | |
733 | ||
cd915493 | 734 | static int get_first_leaf(struct gfs2_inode *dip, u32 index, |
b3b94faa DT |
735 | struct buffer_head **bh_out) |
736 | { | |
cd915493 | 737 | u64 leaf_no; |
b3b94faa DT |
738 | int error; |
739 | ||
740 | error = get_leaf_nr(dip, index, &leaf_no); | |
741 | if (!error) | |
742 | error = get_leaf(dip, leaf_no, bh_out); | |
743 | ||
744 | return error; | |
745 | } | |
746 | ||
c752666c SW |
747 | static struct gfs2_dirent *gfs2_dirent_search(struct inode *inode, |
748 | const struct qstr *name, | |
749 | gfs2_dscan_t scan, | |
750 | struct buffer_head **pbh) | |
b3b94faa | 751 | { |
c752666c SW |
752 | struct buffer_head *bh; |
753 | struct gfs2_dirent *dent; | |
feaa7bba | 754 | struct gfs2_inode *ip = GFS2_I(inode); |
b3b94faa DT |
755 | int error; |
756 | ||
c752666c SW |
757 | if (ip->i_di.di_flags & GFS2_DIF_EXHASH) { |
758 | struct gfs2_leaf *leaf; | |
759 | unsigned hsize = 1 << ip->i_di.di_depth; | |
760 | unsigned index; | |
761 | u64 ln; | |
762 | if (hsize * sizeof(u64) != ip->i_di.di_size) { | |
763 | gfs2_consist_inode(ip); | |
764 | return ERR_PTR(-EIO); | |
765 | } | |
907b9bce | 766 | |
c752666c SW |
767 | index = name->hash >> (32 - ip->i_di.di_depth); |
768 | error = get_first_leaf(ip, index, &bh); | |
769 | if (error) | |
770 | return ERR_PTR(error); | |
771 | do { | |
772 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, | |
71b86f56 | 773 | scan, name, NULL); |
c752666c SW |
774 | if (dent) |
775 | goto got_dent; | |
776 | leaf = (struct gfs2_leaf *)bh->b_data; | |
777 | ln = be64_to_cpu(leaf->lf_next); | |
f4154ea0 | 778 | brelse(bh); |
c752666c SW |
779 | if (!ln) |
780 | break; | |
907b9bce | 781 | |
c752666c SW |
782 | error = get_leaf(ip, ln, &bh); |
783 | } while(!error); | |
b3b94faa | 784 | |
c752666c | 785 | return error ? ERR_PTR(error) : NULL; |
b3b94faa | 786 | } |
b3b94faa | 787 | |
907b9bce | 788 | |
c752666c SW |
789 | error = gfs2_meta_inode_buffer(ip, &bh); |
790 | if (error) | |
791 | return ERR_PTR(error); | |
71b86f56 | 792 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, scan, name, NULL); |
c752666c | 793 | got_dent: |
f4154ea0 | 794 | if (unlikely(dent == NULL || IS_ERR(dent))) { |
ed386507 SW |
795 | brelse(bh); |
796 | bh = NULL; | |
797 | } | |
c752666c SW |
798 | *pbh = bh; |
799 | return dent; | |
800 | } | |
b3b94faa | 801 | |
c752666c SW |
802 | static struct gfs2_leaf *new_leaf(struct inode *inode, struct buffer_head **pbh, u16 depth) |
803 | { | |
feaa7bba | 804 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c SW |
805 | u64 bn = gfs2_alloc_meta(ip); |
806 | struct buffer_head *bh = gfs2_meta_new(ip->i_gl, bn); | |
807 | struct gfs2_leaf *leaf; | |
808 | struct gfs2_dirent *dent; | |
71b86f56 | 809 | struct qstr name = { .name = "", .len = 0, .hash = 0 }; |
c752666c SW |
810 | if (!bh) |
811 | return NULL; | |
907b9bce | 812 | |
c752666c SW |
813 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
814 | gfs2_metatype_set(bh, GFS2_METATYPE_LF, GFS2_FORMAT_LF); | |
815 | leaf = (struct gfs2_leaf *)bh->b_data; | |
816 | leaf->lf_depth = cpu_to_be16(depth); | |
2bdbc5d7 | 817 | leaf->lf_entries = 0; |
c752666c | 818 | leaf->lf_dirent_format = cpu_to_be16(GFS2_FORMAT_DE); |
2bdbc5d7 | 819 | leaf->lf_next = 0; |
c752666c SW |
820 | memset(leaf->lf_reserved, 0, sizeof(leaf->lf_reserved)); |
821 | dent = (struct gfs2_dirent *)(leaf+1); | |
71b86f56 | 822 | gfs2_qstr2dirent(&name, bh->b_size - sizeof(struct gfs2_leaf), dent); |
c752666c SW |
823 | *pbh = bh; |
824 | return leaf; | |
b3b94faa DT |
825 | } |
826 | ||
827 | /** | |
828 | * dir_make_exhash - Convert a stuffed directory into an ExHash directory | |
829 | * @dip: The GFS2 inode | |
830 | * | |
831 | * Returns: 0 on success, error code otherwise | |
832 | */ | |
833 | ||
c752666c | 834 | static int dir_make_exhash(struct inode *inode) |
b3b94faa | 835 | { |
feaa7bba SW |
836 | struct gfs2_inode *dip = GFS2_I(inode); |
837 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
b3b94faa | 838 | struct gfs2_dirent *dent; |
c752666c | 839 | struct qstr args; |
b3b94faa DT |
840 | struct buffer_head *bh, *dibh; |
841 | struct gfs2_leaf *leaf; | |
842 | int y; | |
cd915493 SW |
843 | u32 x; |
844 | u64 *lp, bn; | |
b3b94faa DT |
845 | int error; |
846 | ||
847 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
848 | if (error) | |
849 | return error; | |
850 | ||
b3b94faa DT |
851 | /* Turn over a new leaf */ |
852 | ||
c752666c SW |
853 | leaf = new_leaf(inode, &bh, 0); |
854 | if (!leaf) | |
855 | return -ENOSPC; | |
856 | bn = bh->b_blocknr; | |
b3b94faa DT |
857 | |
858 | gfs2_assert(sdp, dip->i_di.di_entries < (1 << 16)); | |
b3b94faa DT |
859 | leaf->lf_entries = cpu_to_be16(dip->i_di.di_entries); |
860 | ||
861 | /* Copy dirents */ | |
862 | ||
863 | gfs2_buffer_copy_tail(bh, sizeof(struct gfs2_leaf), dibh, | |
864 | sizeof(struct gfs2_dinode)); | |
865 | ||
866 | /* Find last entry */ | |
867 | ||
868 | x = 0; | |
c752666c SW |
869 | args.len = bh->b_size - sizeof(struct gfs2_dinode) + |
870 | sizeof(struct gfs2_leaf); | |
871 | args.name = bh->b_data; | |
feaa7bba | 872 | dent = gfs2_dirent_scan(&dip->i_inode, bh->b_data, bh->b_size, |
71b86f56 | 873 | gfs2_dirent_last, &args, NULL); |
c752666c SW |
874 | if (!dent) { |
875 | brelse(bh); | |
876 | brelse(dibh); | |
877 | return -EIO; | |
878 | } | |
879 | if (IS_ERR(dent)) { | |
880 | brelse(bh); | |
881 | brelse(dibh); | |
882 | return PTR_ERR(dent); | |
b3b94faa | 883 | } |
b3b94faa DT |
884 | |
885 | /* Adjust the last dirent's record length | |
886 | (Remember that dent still points to the last entry.) */ | |
887 | ||
4dd651ad | 888 | dent->de_rec_len = cpu_to_be16(be16_to_cpu(dent->de_rec_len) + |
b3b94faa | 889 | sizeof(struct gfs2_dinode) - |
4dd651ad | 890 | sizeof(struct gfs2_leaf)); |
b3b94faa DT |
891 | |
892 | brelse(bh); | |
893 | ||
894 | /* We're done with the new leaf block, now setup the new | |
895 | hash table. */ | |
896 | ||
d4e9c4c3 | 897 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
b3b94faa DT |
898 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode)); |
899 | ||
cd915493 | 900 | lp = (u64 *)(dibh->b_data + sizeof(struct gfs2_dinode)); |
b3b94faa DT |
901 | |
902 | for (x = sdp->sd_hash_ptrs; x--; lp++) | |
903 | *lp = cpu_to_be64(bn); | |
904 | ||
905 | dip->i_di.di_size = sdp->sd_sb.sb_bsize / 2; | |
906 | dip->i_di.di_blocks++; | |
907 | dip->i_di.di_flags |= GFS2_DIF_EXHASH; | |
908 | dip->i_di.di_payload_format = 0; | |
909 | ||
910 | for (x = sdp->sd_hash_ptrs, y = -1; x; x >>= 1, y++) ; | |
911 | dip->i_di.di_depth = y; | |
912 | ||
913 | gfs2_dinode_out(&dip->i_di, dibh->b_data); | |
914 | ||
915 | brelse(dibh); | |
916 | ||
917 | return 0; | |
918 | } | |
919 | ||
920 | /** | |
921 | * dir_split_leaf - Split a leaf block into two | |
922 | * @dip: The GFS2 inode | |
923 | * @index: | |
924 | * @leaf_no: | |
925 | * | |
926 | * Returns: 0 on success, error code on failure | |
927 | */ | |
928 | ||
c752666c | 929 | static int dir_split_leaf(struct inode *inode, const struct qstr *name) |
b3b94faa | 930 | { |
feaa7bba | 931 | struct gfs2_inode *dip = GFS2_I(inode); |
b3b94faa DT |
932 | struct buffer_head *nbh, *obh, *dibh; |
933 | struct gfs2_leaf *nleaf, *oleaf; | |
4da3c646 | 934 | struct gfs2_dirent *dent = NULL, *prev = NULL, *next = NULL, *new; |
cd915493 SW |
935 | u32 start, len, half_len, divider; |
936 | u64 bn, *lp, leaf_no; | |
937 | u32 index; | |
b3b94faa DT |
938 | int x, moved = 0; |
939 | int error; | |
940 | ||
c752666c SW |
941 | index = name->hash >> (32 - dip->i_di.di_depth); |
942 | error = get_leaf_nr(dip, index, &leaf_no); | |
943 | if (error) | |
944 | return error; | |
b3b94faa DT |
945 | |
946 | /* Get the old leaf block */ | |
b3b94faa DT |
947 | error = get_leaf(dip, leaf_no, &obh); |
948 | if (error) | |
e90deff5 | 949 | return error; |
b3b94faa | 950 | |
b3b94faa | 951 | oleaf = (struct gfs2_leaf *)obh->b_data; |
e90deff5 SW |
952 | if (dip->i_di.di_depth == be16_to_cpu(oleaf->lf_depth)) { |
953 | brelse(obh); | |
954 | return 1; /* can't split */ | |
955 | } | |
956 | ||
957 | gfs2_trans_add_bh(dip->i_gl, obh, 1); | |
b3b94faa | 958 | |
c752666c SW |
959 | nleaf = new_leaf(inode, &nbh, be16_to_cpu(oleaf->lf_depth) + 1); |
960 | if (!nleaf) { | |
961 | brelse(obh); | |
962 | return -ENOSPC; | |
963 | } | |
964 | bn = nbh->b_blocknr; | |
b3b94faa | 965 | |
c752666c | 966 | /* Compute the start and len of leaf pointers in the hash table. */ |
b3b94faa DT |
967 | len = 1 << (dip->i_di.di_depth - be16_to_cpu(oleaf->lf_depth)); |
968 | half_len = len >> 1; | |
969 | if (!half_len) { | |
e90deff5 | 970 | printk(KERN_WARNING "di_depth %u lf_depth %u index %u\n", dip->i_di.di_depth, be16_to_cpu(oleaf->lf_depth), index); |
b3b94faa DT |
971 | gfs2_consist_inode(dip); |
972 | error = -EIO; | |
973 | goto fail_brelse; | |
974 | } | |
975 | ||
976 | start = (index & ~(len - 1)); | |
977 | ||
978 | /* Change the pointers. | |
979 | Don't bother distinguishing stuffed from non-stuffed. | |
980 | This code is complicated enough already. */ | |
cd915493 | 981 | lp = kmalloc(half_len * sizeof(u64), GFP_NOFS | __GFP_NOFAIL); |
b3b94faa | 982 | /* Change the pointers */ |
b3b94faa DT |
983 | for (x = 0; x < half_len; x++) |
984 | lp[x] = cpu_to_be64(bn); | |
985 | ||
cd915493 SW |
986 | error = gfs2_dir_write_data(dip, (char *)lp, start * sizeof(u64), |
987 | half_len * sizeof(u64)); | |
988 | if (error != half_len * sizeof(u64)) { | |
b3b94faa DT |
989 | if (error >= 0) |
990 | error = -EIO; | |
991 | goto fail_lpfree; | |
992 | } | |
993 | ||
994 | kfree(lp); | |
995 | ||
996 | /* Compute the divider */ | |
b3b94faa DT |
997 | divider = (start + half_len) << (32 - dip->i_di.di_depth); |
998 | ||
999 | /* Copy the entries */ | |
b3b94faa DT |
1000 | dirent_first(dip, obh, &dent); |
1001 | ||
1002 | do { | |
1003 | next = dent; | |
1004 | if (dirent_next(dip, obh, &next)) | |
1005 | next = NULL; | |
1006 | ||
1007 | if (dent->de_inum.no_addr && | |
1008 | be32_to_cpu(dent->de_hash) < divider) { | |
c752666c SW |
1009 | struct qstr str; |
1010 | str.name = (char*)(dent+1); | |
1011 | str.len = be16_to_cpu(dent->de_name_len); | |
1012 | str.hash = be32_to_cpu(dent->de_hash); | |
71b86f56 | 1013 | new = gfs2_dirent_alloc(inode, nbh, &str); |
c752666c SW |
1014 | if (IS_ERR(new)) { |
1015 | error = PTR_ERR(new); | |
1016 | break; | |
1017 | } | |
b3b94faa DT |
1018 | |
1019 | new->de_inum = dent->de_inum; /* No endian worries */ | |
b3b94faa | 1020 | new->de_type = dent->de_type; /* No endian worries */ |
c752666c | 1021 | nleaf->lf_entries = cpu_to_be16(be16_to_cpu(nleaf->lf_entries)+1); |
b3b94faa DT |
1022 | |
1023 | dirent_del(dip, obh, prev, dent); | |
1024 | ||
1025 | if (!oleaf->lf_entries) | |
1026 | gfs2_consist_inode(dip); | |
c752666c | 1027 | oleaf->lf_entries = cpu_to_be16(be16_to_cpu(oleaf->lf_entries)-1); |
b3b94faa DT |
1028 | |
1029 | if (!prev) | |
1030 | prev = dent; | |
1031 | ||
1032 | moved = 1; | |
c752666c | 1033 | } else { |
b3b94faa | 1034 | prev = dent; |
c752666c | 1035 | } |
b3b94faa | 1036 | dent = next; |
c752666c | 1037 | } while (dent); |
b3b94faa | 1038 | |
c752666c | 1039 | oleaf->lf_depth = nleaf->lf_depth; |
b3b94faa DT |
1040 | |
1041 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
feaa7bba | 1042 | if (!gfs2_assert_withdraw(GFS2_SB(&dip->i_inode), !error)) { |
b3b94faa DT |
1043 | dip->i_di.di_blocks++; |
1044 | gfs2_dinode_out(&dip->i_di, dibh->b_data); | |
1045 | brelse(dibh); | |
1046 | } | |
1047 | ||
1048 | brelse(obh); | |
1049 | brelse(nbh); | |
1050 | ||
1051 | return error; | |
1052 | ||
e90deff5 | 1053 | fail_lpfree: |
b3b94faa DT |
1054 | kfree(lp); |
1055 | ||
e90deff5 | 1056 | fail_brelse: |
b3b94faa | 1057 | brelse(obh); |
b3b94faa DT |
1058 | brelse(nbh); |
1059 | return error; | |
1060 | } | |
1061 | ||
1062 | /** | |
1063 | * dir_double_exhash - Double size of ExHash table | |
1064 | * @dip: The GFS2 dinode | |
1065 | * | |
1066 | * Returns: 0 on success, error code on failure | |
1067 | */ | |
1068 | ||
1069 | static int dir_double_exhash(struct gfs2_inode *dip) | |
1070 | { | |
feaa7bba | 1071 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa | 1072 | struct buffer_head *dibh; |
cd915493 SW |
1073 | u32 hsize; |
1074 | u64 *buf; | |
1075 | u64 *from, *to; | |
1076 | u64 block; | |
b3b94faa DT |
1077 | int x; |
1078 | int error = 0; | |
1079 | ||
1080 | hsize = 1 << dip->i_di.di_depth; | |
cd915493 | 1081 | if (hsize * sizeof(u64) != dip->i_di.di_size) { |
b3b94faa DT |
1082 | gfs2_consist_inode(dip); |
1083 | return -EIO; | |
1084 | } | |
1085 | ||
1086 | /* Allocate both the "from" and "to" buffers in one big chunk */ | |
1087 | ||
1088 | buf = kcalloc(3, sdp->sd_hash_bsize, GFP_KERNEL | __GFP_NOFAIL); | |
1089 | ||
1090 | for (block = dip->i_di.di_size >> sdp->sd_hash_bsize_shift; block--;) { | |
e13940ba | 1091 | error = gfs2_dir_read_data(dip, (char *)buf, |
b3b94faa | 1092 | block * sdp->sd_hash_bsize, |
7276b3b0 | 1093 | sdp->sd_hash_bsize, 1); |
b3b94faa DT |
1094 | if (error != sdp->sd_hash_bsize) { |
1095 | if (error >= 0) | |
1096 | error = -EIO; | |
1097 | goto fail; | |
1098 | } | |
1099 | ||
1100 | from = buf; | |
cd915493 | 1101 | to = (u64 *)((char *)buf + sdp->sd_hash_bsize); |
b3b94faa DT |
1102 | |
1103 | for (x = sdp->sd_hash_ptrs; x--; from++) { | |
1104 | *to++ = *from; /* No endianess worries */ | |
1105 | *to++ = *from; | |
1106 | } | |
1107 | ||
e13940ba | 1108 | error = gfs2_dir_write_data(dip, |
b3b94faa DT |
1109 | (char *)buf + sdp->sd_hash_bsize, |
1110 | block * sdp->sd_sb.sb_bsize, | |
1111 | sdp->sd_sb.sb_bsize); | |
1112 | if (error != sdp->sd_sb.sb_bsize) { | |
1113 | if (error >= 0) | |
1114 | error = -EIO; | |
1115 | goto fail; | |
1116 | } | |
1117 | } | |
1118 | ||
1119 | kfree(buf); | |
1120 | ||
1121 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1122 | if (!gfs2_assert_withdraw(sdp, !error)) { | |
1123 | dip->i_di.di_depth++; | |
1124 | gfs2_dinode_out(&dip->i_di, dibh->b_data); | |
1125 | brelse(dibh); | |
1126 | } | |
1127 | ||
1128 | return error; | |
1129 | ||
a91ea69f | 1130 | fail: |
b3b94faa | 1131 | kfree(buf); |
b3b94faa DT |
1132 | return error; |
1133 | } | |
1134 | ||
1135 | /** | |
1136 | * compare_dents - compare directory entries by hash value | |
1137 | * @a: first dent | |
1138 | * @b: second dent | |
1139 | * | |
1140 | * When comparing the hash entries of @a to @b: | |
1141 | * gt: returns 1 | |
1142 | * lt: returns -1 | |
1143 | * eq: returns 0 | |
1144 | */ | |
1145 | ||
1146 | static int compare_dents(const void *a, const void *b) | |
1147 | { | |
2bdbc5d7 | 1148 | const struct gfs2_dirent *dent_a, *dent_b; |
cd915493 | 1149 | u32 hash_a, hash_b; |
b3b94faa DT |
1150 | int ret = 0; |
1151 | ||
2bdbc5d7 | 1152 | dent_a = *(const struct gfs2_dirent **)a; |
c752666c | 1153 | hash_a = be32_to_cpu(dent_a->de_hash); |
b3b94faa | 1154 | |
2bdbc5d7 | 1155 | dent_b = *(const struct gfs2_dirent **)b; |
c752666c | 1156 | hash_b = be32_to_cpu(dent_b->de_hash); |
b3b94faa DT |
1157 | |
1158 | if (hash_a > hash_b) | |
1159 | ret = 1; | |
1160 | else if (hash_a < hash_b) | |
1161 | ret = -1; | |
1162 | else { | |
4dd651ad SW |
1163 | unsigned int len_a = be16_to_cpu(dent_a->de_name_len); |
1164 | unsigned int len_b = be16_to_cpu(dent_b->de_name_len); | |
b3b94faa DT |
1165 | |
1166 | if (len_a > len_b) | |
1167 | ret = 1; | |
1168 | else if (len_a < len_b) | |
1169 | ret = -1; | |
1170 | else | |
2bdbc5d7 | 1171 | ret = memcmp(dent_a + 1, dent_b + 1, len_a); |
b3b94faa DT |
1172 | } |
1173 | ||
1174 | return ret; | |
1175 | } | |
1176 | ||
1177 | /** | |
1178 | * do_filldir_main - read out directory entries | |
1179 | * @dip: The GFS2 inode | |
1180 | * @offset: The offset in the file to read from | |
1181 | * @opaque: opaque data to pass to filldir | |
1182 | * @filldir: The function to pass entries to | |
1183 | * @darr: an array of struct gfs2_dirent pointers to read | |
1184 | * @entries: the number of entries in darr | |
1185 | * @copied: pointer to int that's non-zero if a entry has been copied out | |
1186 | * | |
1187 | * Jump through some hoops to make sure that if there are hash collsions, | |
1188 | * they are read out at the beginning of a buffer. We want to minimize | |
1189 | * the possibility that they will fall into different readdir buffers or | |
1190 | * that someone will want to seek to that location. | |
1191 | * | |
1192 | * Returns: errno, >0 on exception from filldir | |
1193 | */ | |
1194 | ||
cd915493 | 1195 | static int do_filldir_main(struct gfs2_inode *dip, u64 *offset, |
b3b94faa | 1196 | void *opaque, gfs2_filldir_t filldir, |
cd915493 | 1197 | const struct gfs2_dirent **darr, u32 entries, |
b3b94faa DT |
1198 | int *copied) |
1199 | { | |
71b86f56 | 1200 | const struct gfs2_dirent *dent, *dent_next; |
b3b94faa | 1201 | struct gfs2_inum inum; |
cd915493 | 1202 | u64 off, off_next; |
b3b94faa DT |
1203 | unsigned int x, y; |
1204 | int run = 0; | |
1205 | int error = 0; | |
1206 | ||
1207 | sort(darr, entries, sizeof(struct gfs2_dirent *), compare_dents, NULL); | |
1208 | ||
1209 | dent_next = darr[0]; | |
1210 | off_next = be32_to_cpu(dent_next->de_hash); | |
1211 | off_next = gfs2_disk_hash2offset(off_next); | |
1212 | ||
1213 | for (x = 0, y = 1; x < entries; x++, y++) { | |
1214 | dent = dent_next; | |
1215 | off = off_next; | |
1216 | ||
1217 | if (y < entries) { | |
1218 | dent_next = darr[y]; | |
1219 | off_next = be32_to_cpu(dent_next->de_hash); | |
1220 | off_next = gfs2_disk_hash2offset(off_next); | |
1221 | ||
1222 | if (off < *offset) | |
1223 | continue; | |
1224 | *offset = off; | |
1225 | ||
1226 | if (off_next == off) { | |
1227 | if (*copied && !run) | |
1228 | return 1; | |
1229 | run = 1; | |
1230 | } else | |
1231 | run = 0; | |
1232 | } else { | |
1233 | if (off < *offset) | |
1234 | continue; | |
1235 | *offset = off; | |
1236 | } | |
1237 | ||
1238 | gfs2_inum_in(&inum, (char *)&dent->de_inum); | |
1239 | ||
2bdbc5d7 | 1240 | error = filldir(opaque, (const char *)(dent + 1), |
4dd651ad | 1241 | be16_to_cpu(dent->de_name_len), |
b3b94faa | 1242 | off, &inum, |
4dd651ad | 1243 | be16_to_cpu(dent->de_type)); |
b3b94faa DT |
1244 | if (error) |
1245 | return 1; | |
1246 | ||
1247 | *copied = 1; | |
1248 | } | |
1249 | ||
1250 | /* Increment the *offset by one, so the next time we come into the | |
1251 | do_filldir fxn, we get the next entry instead of the last one in the | |
1252 | current leaf */ | |
1253 | ||
1254 | (*offset)++; | |
1255 | ||
1256 | return 0; | |
1257 | } | |
1258 | ||
71b86f56 SW |
1259 | static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque, |
1260 | gfs2_filldir_t filldir, int *copied, | |
1261 | unsigned *depth, u64 leaf_no) | |
b3b94faa | 1262 | { |
feaa7bba | 1263 | struct gfs2_inode *ip = GFS2_I(inode); |
71b86f56 SW |
1264 | struct buffer_head *bh; |
1265 | struct gfs2_leaf *lf; | |
1266 | unsigned entries = 0; | |
1267 | unsigned leaves = 0; | |
1268 | const struct gfs2_dirent **darr, *dent; | |
1269 | struct dirent_gather g; | |
1270 | struct buffer_head **larr; | |
1271 | int leaf = 0; | |
1272 | int error, i; | |
1273 | u64 lfn = leaf_no; | |
b3b94faa | 1274 | |
b3b94faa | 1275 | do { |
71b86f56 | 1276 | error = get_leaf(ip, lfn, &bh); |
b3b94faa | 1277 | if (error) |
71b86f56 SW |
1278 | goto out; |
1279 | lf = (struct gfs2_leaf *)bh->b_data; | |
1280 | if (leaves == 0) | |
1281 | *depth = be16_to_cpu(lf->lf_depth); | |
1282 | entries += be16_to_cpu(lf->lf_entries); | |
1283 | leaves++; | |
1284 | lfn = be64_to_cpu(lf->lf_next); | |
1285 | brelse(bh); | |
1286 | } while(lfn); | |
b3b94faa DT |
1287 | |
1288 | if (!entries) | |
1289 | return 0; | |
1290 | ||
71b86f56 | 1291 | error = -ENOMEM; |
2bdbc5d7 | 1292 | larr = vmalloc((leaves + entries) * sizeof(void *)); |
71b86f56 SW |
1293 | if (!larr) |
1294 | goto out; | |
1295 | darr = (const struct gfs2_dirent **)(larr + leaves); | |
1296 | g.pdent = darr; | |
1297 | g.offset = 0; | |
1298 | lfn = leaf_no; | |
b3b94faa | 1299 | |
71b86f56 SW |
1300 | do { |
1301 | error = get_leaf(ip, lfn, &bh); | |
b3b94faa | 1302 | if (error) |
71b86f56 SW |
1303 | goto out_kfree; |
1304 | lf = (struct gfs2_leaf *)bh->b_data; | |
1305 | lfn = be64_to_cpu(lf->lf_next); | |
1306 | if (lf->lf_entries) { | |
1307 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, | |
1308 | gfs2_dirent_gather, NULL, &g); | |
1309 | error = PTR_ERR(dent); | |
1310 | if (IS_ERR(dent)) { | |
1311 | goto out_kfree; | |
1312 | } | |
1313 | error = 0; | |
1314 | larr[leaf++] = bh; | |
b3b94faa | 1315 | } else { |
71b86f56 | 1316 | brelse(bh); |
b3b94faa | 1317 | } |
71b86f56 | 1318 | } while(lfn); |
b3b94faa | 1319 | |
71b86f56 | 1320 | error = do_filldir_main(ip, offset, opaque, filldir, darr, |
b3b94faa | 1321 | entries, copied); |
71b86f56 SW |
1322 | out_kfree: |
1323 | for(i = 0; i < leaf; i++) | |
1324 | brelse(larr[i]); | |
fe1bdedc | 1325 | vfree(larr); |
71b86f56 | 1326 | out: |
b3b94faa DT |
1327 | return error; |
1328 | } | |
1329 | ||
1330 | /** | |
c752666c SW |
1331 | * dir_e_read - Reads the entries from a directory into a filldir buffer |
1332 | * @dip: dinode pointer | |
1333 | * @offset: the hash of the last entry read shifted to the right once | |
1334 | * @opaque: buffer for the filldir function to fill | |
1335 | * @filldir: points to the filldir function to use | |
b3b94faa | 1336 | * |
c752666c | 1337 | * Returns: errno |
b3b94faa DT |
1338 | */ |
1339 | ||
cd915493 | 1340 | static int dir_e_read(struct inode *inode, u64 *offset, void *opaque, |
c752666c | 1341 | gfs2_filldir_t filldir) |
b3b94faa | 1342 | { |
feaa7bba SW |
1343 | struct gfs2_inode *dip = GFS2_I(inode); |
1344 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
cd915493 SW |
1345 | u32 hsize, len = 0; |
1346 | u32 ht_offset, lp_offset, ht_offset_cur = -1; | |
1347 | u32 hash, index; | |
1348 | u64 *lp; | |
c752666c SW |
1349 | int copied = 0; |
1350 | int error = 0; | |
4da3c646 | 1351 | unsigned depth = 0; |
b3b94faa DT |
1352 | |
1353 | hsize = 1 << dip->i_di.di_depth; | |
cd915493 | 1354 | if (hsize * sizeof(u64) != dip->i_di.di_size) { |
b3b94faa DT |
1355 | gfs2_consist_inode(dip); |
1356 | return -EIO; | |
1357 | } | |
1358 | ||
1359 | hash = gfs2_dir_offset2hash(*offset); | |
1360 | index = hash >> (32 - dip->i_di.di_depth); | |
1361 | ||
1362 | lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL); | |
1363 | if (!lp) | |
1364 | return -ENOMEM; | |
1365 | ||
1366 | while (index < hsize) { | |
1367 | lp_offset = index & (sdp->sd_hash_ptrs - 1); | |
1368 | ht_offset = index - lp_offset; | |
1369 | ||
1370 | if (ht_offset_cur != ht_offset) { | |
e13940ba | 1371 | error = gfs2_dir_read_data(dip, (char *)lp, |
cd915493 | 1372 | ht_offset * sizeof(u64), |
7276b3b0 | 1373 | sdp->sd_hash_bsize, 1); |
b3b94faa DT |
1374 | if (error != sdp->sd_hash_bsize) { |
1375 | if (error >= 0) | |
1376 | error = -EIO; | |
1377 | goto out; | |
1378 | } | |
1379 | ht_offset_cur = ht_offset; | |
1380 | } | |
1381 | ||
71b86f56 SW |
1382 | error = gfs2_dir_read_leaf(inode, offset, opaque, filldir, |
1383 | &copied, &depth, | |
1384 | be64_to_cpu(lp[lp_offset])); | |
b3b94faa | 1385 | if (error) |
71b86f56 | 1386 | break; |
b3b94faa | 1387 | |
71b86f56 | 1388 | len = 1 << (dip->i_di.di_depth - depth); |
b3b94faa DT |
1389 | index = (index & ~(len - 1)) + len; |
1390 | } | |
1391 | ||
71b86f56 | 1392 | out: |
b3b94faa | 1393 | kfree(lp); |
71b86f56 SW |
1394 | if (error > 0) |
1395 | error = 0; | |
b3b94faa DT |
1396 | return error; |
1397 | } | |
1398 | ||
cd915493 | 1399 | int gfs2_dir_read(struct inode *inode, u64 *offset, void *opaque, |
71b86f56 | 1400 | gfs2_filldir_t filldir) |
b3b94faa | 1401 | { |
feaa7bba | 1402 | struct gfs2_inode *dip = GFS2_I(inode); |
71b86f56 SW |
1403 | struct dirent_gather g; |
1404 | const struct gfs2_dirent **darr, *dent; | |
b3b94faa DT |
1405 | struct buffer_head *dibh; |
1406 | int copied = 0; | |
1407 | int error; | |
1408 | ||
71b86f56 SW |
1409 | if (!dip->i_di.di_entries) |
1410 | return 0; | |
1411 | ||
1412 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) | |
1413 | return dir_e_read(inode, offset, opaque, filldir); | |
1414 | ||
b3b94faa DT |
1415 | if (!gfs2_is_stuffed(dip)) { |
1416 | gfs2_consist_inode(dip); | |
1417 | return -EIO; | |
1418 | } | |
1419 | ||
b3b94faa DT |
1420 | error = gfs2_meta_inode_buffer(dip, &dibh); |
1421 | if (error) | |
1422 | return error; | |
1423 | ||
71b86f56 SW |
1424 | error = -ENOMEM; |
1425 | darr = kmalloc(dip->i_di.di_entries * sizeof(struct gfs2_dirent *), | |
1426 | GFP_KERNEL); | |
1427 | if (darr) { | |
1428 | g.pdent = darr; | |
1429 | g.offset = 0; | |
1430 | dent = gfs2_dirent_scan(inode, dibh->b_data, dibh->b_size, | |
1431 | gfs2_dirent_gather, NULL, &g); | |
1432 | if (IS_ERR(dent)) { | |
1433 | error = PTR_ERR(dent); | |
1434 | goto out; | |
1435 | } | |
1436 | error = do_filldir_main(dip, offset, opaque, filldir, darr, | |
1437 | dip->i_di.di_entries, &copied); | |
1438 | out: | |
1439 | kfree(darr); | |
1440 | } | |
1441 | ||
b3b94faa DT |
1442 | if (error > 0) |
1443 | error = 0; | |
1444 | ||
1445 | brelse(dibh); | |
1446 | ||
1447 | return error; | |
1448 | } | |
1449 | ||
b3b94faa DT |
1450 | /** |
1451 | * gfs2_dir_search - Search a directory | |
1452 | * @dip: The GFS2 inode | |
1453 | * @filename: | |
1454 | * @inode: | |
1455 | * | |
1456 | * This routine searches a directory for a file or another directory. | |
1457 | * Assumes a glock is held on dip. | |
1458 | * | |
1459 | * Returns: errno | |
1460 | */ | |
1461 | ||
c752666c | 1462 | int gfs2_dir_search(struct inode *dir, const struct qstr *name, |
b3b94faa DT |
1463 | struct gfs2_inum *inum, unsigned int *type) |
1464 | { | |
c752666c SW |
1465 | struct buffer_head *bh; |
1466 | struct gfs2_dirent *dent; | |
1467 | ||
1468 | dent = gfs2_dirent_search(dir, name, gfs2_dirent_find, &bh); | |
1469 | if (dent) { | |
1470 | if (IS_ERR(dent)) | |
1471 | return PTR_ERR(dent); | |
1472 | if (inum) | |
1473 | gfs2_inum_in(inum, (char *)&dent->de_inum); | |
1474 | if (type) | |
1475 | *type = be16_to_cpu(dent->de_type); | |
1476 | brelse(bh); | |
1477 | return 0; | |
1478 | } | |
1479 | return -ENOENT; | |
1480 | } | |
1481 | ||
1482 | static int dir_new_leaf(struct inode *inode, const struct qstr *name) | |
1483 | { | |
1484 | struct buffer_head *bh, *obh; | |
feaa7bba | 1485 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c | 1486 | struct gfs2_leaf *leaf, *oleaf; |
b3b94faa | 1487 | int error; |
c752666c SW |
1488 | u32 index; |
1489 | u64 bn; | |
b3b94faa | 1490 | |
c752666c SW |
1491 | index = name->hash >> (32 - ip->i_di.di_depth); |
1492 | error = get_first_leaf(ip, index, &obh); | |
1493 | if (error) | |
1494 | return error; | |
1495 | do { | |
1496 | oleaf = (struct gfs2_leaf *)obh->b_data; | |
1497 | bn = be64_to_cpu(oleaf->lf_next); | |
1498 | if (!bn) | |
1499 | break; | |
1500 | brelse(obh); | |
1501 | error = get_leaf(ip, bn, &obh); | |
1502 | if (error) | |
1503 | return error; | |
1504 | } while(1); | |
b3b94faa | 1505 | |
c752666c SW |
1506 | gfs2_trans_add_bh(ip->i_gl, obh, 1); |
1507 | ||
1508 | leaf = new_leaf(inode, &bh, be16_to_cpu(oleaf->lf_depth)); | |
1509 | if (!leaf) { | |
1510 | brelse(obh); | |
1511 | return -ENOSPC; | |
1512 | } | |
4d8012b6 | 1513 | oleaf->lf_next = cpu_to_be64(bh->b_blocknr); |
c752666c SW |
1514 | brelse(bh); |
1515 | brelse(obh); | |
1516 | ||
1517 | error = gfs2_meta_inode_buffer(ip, &bh); | |
1518 | if (error) | |
1519 | return error; | |
1520 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
1521 | ip->i_di.di_blocks++; | |
1522 | gfs2_dinode_out(&ip->i_di, bh->b_data); | |
1523 | brelse(bh); | |
1524 | return 0; | |
b3b94faa DT |
1525 | } |
1526 | ||
1527 | /** | |
1528 | * gfs2_dir_add - Add new filename into directory | |
1529 | * @dip: The GFS2 inode | |
1530 | * @filename: The new name | |
1531 | * @inode: The inode number of the entry | |
1532 | * @type: The type of the entry | |
1533 | * | |
1534 | * Returns: 0 on success, error code on failure | |
1535 | */ | |
1536 | ||
c752666c SW |
1537 | int gfs2_dir_add(struct inode *inode, const struct qstr *name, |
1538 | const struct gfs2_inum *inum, unsigned type) | |
b3b94faa | 1539 | { |
feaa7bba | 1540 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c SW |
1541 | struct buffer_head *bh; |
1542 | struct gfs2_dirent *dent; | |
1543 | struct gfs2_leaf *leaf; | |
b3b94faa DT |
1544 | int error; |
1545 | ||
c752666c SW |
1546 | while(1) { |
1547 | dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space, | |
1548 | &bh); | |
1549 | if (dent) { | |
1550 | if (IS_ERR(dent)) | |
1551 | return PTR_ERR(dent); | |
1552 | dent = gfs2_init_dirent(inode, dent, name, bh); | |
1553 | gfs2_inum_out(inum, (char *)&dent->de_inum); | |
1554 | dent->de_type = cpu_to_be16(type); | |
1555 | if (ip->i_di.di_flags & GFS2_DIF_EXHASH) { | |
1556 | leaf = (struct gfs2_leaf *)bh->b_data; | |
1557 | leaf->lf_entries = cpu_to_be16(be16_to_cpu(leaf->lf_entries) + 1); | |
1558 | } | |
1559 | brelse(bh); | |
1560 | error = gfs2_meta_inode_buffer(ip, &bh); | |
1561 | if (error) | |
1562 | break; | |
1563 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
1564 | ip->i_di.di_entries++; | |
1565 | ip->i_di.di_mtime = ip->i_di.di_ctime = get_seconds(); | |
1566 | gfs2_dinode_out(&ip->i_di, bh->b_data); | |
1567 | brelse(bh); | |
1568 | error = 0; | |
1569 | break; | |
1570 | } | |
1571 | if (!(ip->i_di.di_flags & GFS2_DIF_EXHASH)) { | |
1572 | error = dir_make_exhash(inode); | |
1573 | if (error) | |
1574 | break; | |
1575 | continue; | |
1576 | } | |
1577 | error = dir_split_leaf(inode, name); | |
1578 | if (error == 0) | |
1579 | continue; | |
e90deff5 | 1580 | if (error < 0) |
c752666c SW |
1581 | break; |
1582 | if (ip->i_di.di_depth < GFS2_DIR_MAX_DEPTH) { | |
1583 | error = dir_double_exhash(ip); | |
1584 | if (error) | |
1585 | break; | |
1586 | error = dir_split_leaf(inode, name); | |
e90deff5 | 1587 | if (error < 0) |
c752666c | 1588 | break; |
e90deff5 SW |
1589 | if (error == 0) |
1590 | continue; | |
c752666c SW |
1591 | } |
1592 | error = dir_new_leaf(inode, name); | |
1593 | if (!error) | |
1594 | continue; | |
1595 | error = -ENOSPC; | |
1596 | break; | |
1597 | } | |
b3b94faa DT |
1598 | return error; |
1599 | } | |
1600 | ||
c752666c | 1601 | |
b3b94faa DT |
1602 | /** |
1603 | * gfs2_dir_del - Delete a directory entry | |
1604 | * @dip: The GFS2 inode | |
1605 | * @filename: The filename | |
1606 | * | |
1607 | * Returns: 0 on success, error code on failure | |
1608 | */ | |
1609 | ||
c752666c | 1610 | int gfs2_dir_del(struct gfs2_inode *dip, const struct qstr *name) |
b3b94faa | 1611 | { |
c752666c SW |
1612 | struct gfs2_dirent *dent, *prev = NULL; |
1613 | struct buffer_head *bh; | |
b3b94faa DT |
1614 | int error; |
1615 | ||
c752666c SW |
1616 | /* Returns _either_ the entry (if its first in block) or the |
1617 | previous entry otherwise */ | |
feaa7bba | 1618 | dent = gfs2_dirent_search(&dip->i_inode, name, gfs2_dirent_prev, &bh); |
c752666c SW |
1619 | if (!dent) { |
1620 | gfs2_consist_inode(dip); | |
1621 | return -EIO; | |
1622 | } | |
1623 | if (IS_ERR(dent)) { | |
1624 | gfs2_consist_inode(dip); | |
1625 | return PTR_ERR(dent); | |
1626 | } | |
1627 | /* If not first in block, adjust pointers accordingly */ | |
71b86f56 | 1628 | if (gfs2_dirent_find(dent, name, NULL) == 0) { |
c752666c SW |
1629 | prev = dent; |
1630 | dent = (struct gfs2_dirent *)((char *)dent + be16_to_cpu(prev->de_rec_len)); | |
1631 | } | |
1632 | ||
1633 | dirent_del(dip, bh, prev, dent); | |
1634 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) { | |
1635 | struct gfs2_leaf *leaf = (struct gfs2_leaf *)bh->b_data; | |
1636 | u16 entries = be16_to_cpu(leaf->lf_entries); | |
1637 | if (!entries) | |
1638 | gfs2_consist_inode(dip); | |
1639 | leaf->lf_entries = cpu_to_be16(--entries); | |
c752666c | 1640 | } |
ed386507 | 1641 | brelse(bh); |
c752666c SW |
1642 | |
1643 | error = gfs2_meta_inode_buffer(dip, &bh); | |
1644 | if (error) | |
1645 | return error; | |
1646 | ||
1647 | if (!dip->i_di.di_entries) | |
1648 | gfs2_consist_inode(dip); | |
1649 | gfs2_trans_add_bh(dip->i_gl, bh, 1); | |
1650 | dip->i_di.di_entries--; | |
1651 | dip->i_di.di_mtime = dip->i_di.di_ctime = get_seconds(); | |
1652 | gfs2_dinode_out(&dip->i_di, bh->b_data); | |
1653 | brelse(bh); | |
feaa7bba | 1654 | mark_inode_dirty(&dip->i_inode); |
b3b94faa DT |
1655 | |
1656 | return error; | |
1657 | } | |
1658 | ||
b3b94faa DT |
1659 | /** |
1660 | * gfs2_dir_mvino - Change inode number of directory entry | |
1661 | * @dip: The GFS2 inode | |
1662 | * @filename: | |
1663 | * @new_inode: | |
1664 | * | |
1665 | * This routine changes the inode number of a directory entry. It's used | |
1666 | * by rename to change ".." when a directory is moved. | |
1667 | * Assumes a glock is held on dvp. | |
1668 | * | |
1669 | * Returns: errno | |
1670 | */ | |
1671 | ||
c752666c | 1672 | int gfs2_dir_mvino(struct gfs2_inode *dip, const struct qstr *filename, |
b3b94faa DT |
1673 | struct gfs2_inum *inum, unsigned int new_type) |
1674 | { | |
c752666c SW |
1675 | struct buffer_head *bh; |
1676 | struct gfs2_dirent *dent; | |
b3b94faa DT |
1677 | int error; |
1678 | ||
feaa7bba | 1679 | dent = gfs2_dirent_search(&dip->i_inode, filename, gfs2_dirent_find, &bh); |
c752666c SW |
1680 | if (!dent) { |
1681 | gfs2_consist_inode(dip); | |
1682 | return -EIO; | |
1683 | } | |
1684 | if (IS_ERR(dent)) | |
1685 | return PTR_ERR(dent); | |
b3b94faa | 1686 | |
c752666c SW |
1687 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
1688 | gfs2_inum_out(inum, (char *)&dent->de_inum); | |
1689 | dent->de_type = cpu_to_be16(new_type); | |
1690 | ||
1691 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) { | |
1692 | brelse(bh); | |
1693 | error = gfs2_meta_inode_buffer(dip, &bh); | |
1694 | if (error) | |
1695 | return error; | |
1696 | gfs2_trans_add_bh(dip->i_gl, bh, 1); | |
1697 | } | |
1698 | ||
1699 | dip->i_di.di_mtime = dip->i_di.di_ctime = get_seconds(); | |
1700 | gfs2_dinode_out(&dip->i_di, bh->b_data); | |
1701 | brelse(bh); | |
1702 | return 0; | |
b3b94faa DT |
1703 | } |
1704 | ||
1705 | /** | |
1706 | * foreach_leaf - call a function for each leaf in a directory | |
1707 | * @dip: the directory | |
1708 | * @lc: the function to call for each each | |
1709 | * @data: private data to pass to it | |
1710 | * | |
1711 | * Returns: errno | |
1712 | */ | |
1713 | ||
1714 | static int foreach_leaf(struct gfs2_inode *dip, leaf_call_t lc, void *data) | |
1715 | { | |
feaa7bba | 1716 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa | 1717 | struct buffer_head *bh; |
c752666c | 1718 | struct gfs2_leaf *leaf; |
cd915493 SW |
1719 | u32 hsize, len; |
1720 | u32 ht_offset, lp_offset, ht_offset_cur = -1; | |
1721 | u32 index = 0; | |
1722 | u64 *lp; | |
1723 | u64 leaf_no; | |
b3b94faa DT |
1724 | int error = 0; |
1725 | ||
1726 | hsize = 1 << dip->i_di.di_depth; | |
cd915493 | 1727 | if (hsize * sizeof(u64) != dip->i_di.di_size) { |
b3b94faa DT |
1728 | gfs2_consist_inode(dip); |
1729 | return -EIO; | |
1730 | } | |
1731 | ||
1732 | lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL); | |
1733 | if (!lp) | |
1734 | return -ENOMEM; | |
1735 | ||
1736 | while (index < hsize) { | |
1737 | lp_offset = index & (sdp->sd_hash_ptrs - 1); | |
1738 | ht_offset = index - lp_offset; | |
1739 | ||
1740 | if (ht_offset_cur != ht_offset) { | |
e13940ba | 1741 | error = gfs2_dir_read_data(dip, (char *)lp, |
cd915493 | 1742 | ht_offset * sizeof(u64), |
7276b3b0 | 1743 | sdp->sd_hash_bsize, 1); |
b3b94faa DT |
1744 | if (error != sdp->sd_hash_bsize) { |
1745 | if (error >= 0) | |
1746 | error = -EIO; | |
1747 | goto out; | |
1748 | } | |
1749 | ht_offset_cur = ht_offset; | |
1750 | } | |
1751 | ||
1752 | leaf_no = be64_to_cpu(lp[lp_offset]); | |
1753 | if (leaf_no) { | |
1754 | error = get_leaf(dip, leaf_no, &bh); | |
1755 | if (error) | |
1756 | goto out; | |
c752666c | 1757 | leaf = (struct gfs2_leaf *)bh->b_data; |
c752666c | 1758 | len = 1 << (dip->i_di.di_depth - be16_to_cpu(leaf->lf_depth)); |
634ee0b9 | 1759 | brelse(bh); |
b3b94faa DT |
1760 | |
1761 | error = lc(dip, index, len, leaf_no, data); | |
1762 | if (error) | |
1763 | goto out; | |
1764 | ||
1765 | index = (index & ~(len - 1)) + len; | |
1766 | } else | |
1767 | index++; | |
1768 | } | |
1769 | ||
1770 | if (index != hsize) { | |
1771 | gfs2_consist_inode(dip); | |
1772 | error = -EIO; | |
1773 | } | |
1774 | ||
634ee0b9 | 1775 | out: |
b3b94faa DT |
1776 | kfree(lp); |
1777 | ||
1778 | return error; | |
1779 | } | |
1780 | ||
1781 | /** | |
1782 | * leaf_dealloc - Deallocate a directory leaf | |
1783 | * @dip: the directory | |
1784 | * @index: the hash table offset in the directory | |
1785 | * @len: the number of pointers to this leaf | |
1786 | * @leaf_no: the leaf number | |
1787 | * @data: not used | |
1788 | * | |
1789 | * Returns: errno | |
1790 | */ | |
1791 | ||
cd915493 SW |
1792 | static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len, |
1793 | u64 leaf_no, void *data) | |
b3b94faa | 1794 | { |
feaa7bba | 1795 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
c752666c | 1796 | struct gfs2_leaf *tmp_leaf; |
b3b94faa DT |
1797 | struct gfs2_rgrp_list rlist; |
1798 | struct buffer_head *bh, *dibh; | |
cd915493 | 1799 | u64 blk, nblk; |
b3b94faa DT |
1800 | unsigned int rg_blocks = 0, l_blocks = 0; |
1801 | char *ht; | |
cd915493 | 1802 | unsigned int x, size = len * sizeof(u64); |
b3b94faa DT |
1803 | int error; |
1804 | ||
1805 | memset(&rlist, 0, sizeof(struct gfs2_rgrp_list)); | |
1806 | ||
1807 | ht = kzalloc(size, GFP_KERNEL); | |
1808 | if (!ht) | |
1809 | return -ENOMEM; | |
1810 | ||
1811 | gfs2_alloc_get(dip); | |
1812 | ||
1813 | error = gfs2_quota_hold(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
1814 | if (error) | |
1815 | goto out; | |
1816 | ||
1817 | error = gfs2_rindex_hold(sdp, &dip->i_alloc.al_ri_gh); | |
1818 | if (error) | |
1819 | goto out_qs; | |
1820 | ||
1821 | /* Count the number of leaves */ | |
1822 | ||
c752666c | 1823 | for (blk = leaf_no; blk; blk = nblk) { |
b3b94faa DT |
1824 | error = get_leaf(dip, blk, &bh); |
1825 | if (error) | |
1826 | goto out_rlist; | |
c752666c SW |
1827 | tmp_leaf = (struct gfs2_leaf *)bh->b_data; |
1828 | nblk = be64_to_cpu(tmp_leaf->lf_next); | |
b3b94faa DT |
1829 | brelse(bh); |
1830 | ||
1831 | gfs2_rlist_add(sdp, &rlist, blk); | |
1832 | l_blocks++; | |
1833 | } | |
1834 | ||
1835 | gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE, 0); | |
1836 | ||
1837 | for (x = 0; x < rlist.rl_rgrps; x++) { | |
1838 | struct gfs2_rgrpd *rgd; | |
5c676f6d | 1839 | rgd = rlist.rl_ghs[x].gh_gl->gl_object; |
b3b94faa DT |
1840 | rg_blocks += rgd->rd_ri.ri_length; |
1841 | } | |
1842 | ||
1843 | error = gfs2_glock_nq_m(rlist.rl_rgrps, rlist.rl_ghs); | |
1844 | if (error) | |
1845 | goto out_rlist; | |
1846 | ||
1847 | error = gfs2_trans_begin(sdp, | |
5c676f6d | 1848 | rg_blocks + (DIV_ROUND_UP(size, sdp->sd_jbsize) + 1) + |
b3b94faa DT |
1849 | RES_DINODE + RES_STATFS + RES_QUOTA, l_blocks); |
1850 | if (error) | |
1851 | goto out_rg_gunlock; | |
1852 | ||
c752666c | 1853 | for (blk = leaf_no; blk; blk = nblk) { |
b3b94faa DT |
1854 | error = get_leaf(dip, blk, &bh); |
1855 | if (error) | |
1856 | goto out_end_trans; | |
c752666c SW |
1857 | tmp_leaf = (struct gfs2_leaf *)bh->b_data; |
1858 | nblk = be64_to_cpu(tmp_leaf->lf_next); | |
b3b94faa DT |
1859 | brelse(bh); |
1860 | ||
1861 | gfs2_free_meta(dip, blk, 1); | |
1862 | ||
1863 | if (!dip->i_di.di_blocks) | |
1864 | gfs2_consist_inode(dip); | |
1865 | dip->i_di.di_blocks--; | |
1866 | } | |
1867 | ||
cd915493 | 1868 | error = gfs2_dir_write_data(dip, ht, index * sizeof(u64), size); |
b3b94faa DT |
1869 | if (error != size) { |
1870 | if (error >= 0) | |
1871 | error = -EIO; | |
1872 | goto out_end_trans; | |
1873 | } | |
1874 | ||
1875 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1876 | if (error) | |
1877 | goto out_end_trans; | |
1878 | ||
d4e9c4c3 | 1879 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
b3b94faa DT |
1880 | gfs2_dinode_out(&dip->i_di, dibh->b_data); |
1881 | brelse(dibh); | |
1882 | ||
a91ea69f | 1883 | out_end_trans: |
b3b94faa | 1884 | gfs2_trans_end(sdp); |
a91ea69f | 1885 | out_rg_gunlock: |
b3b94faa | 1886 | gfs2_glock_dq_m(rlist.rl_rgrps, rlist.rl_ghs); |
a91ea69f | 1887 | out_rlist: |
b3b94faa DT |
1888 | gfs2_rlist_free(&rlist); |
1889 | gfs2_glock_dq_uninit(&dip->i_alloc.al_ri_gh); | |
a91ea69f | 1890 | out_qs: |
b3b94faa | 1891 | gfs2_quota_unhold(dip); |
a91ea69f | 1892 | out: |
b3b94faa DT |
1893 | gfs2_alloc_put(dip); |
1894 | kfree(ht); | |
b3b94faa DT |
1895 | return error; |
1896 | } | |
1897 | ||
1898 | /** | |
1899 | * gfs2_dir_exhash_dealloc - free all the leaf blocks in a directory | |
1900 | * @dip: the directory | |
1901 | * | |
1902 | * Dealloc all on-disk directory leaves to FREEMETA state | |
1903 | * Change on-disk inode type to "regular file" | |
1904 | * | |
1905 | * Returns: errno | |
1906 | */ | |
1907 | ||
1908 | int gfs2_dir_exhash_dealloc(struct gfs2_inode *dip) | |
1909 | { | |
feaa7bba | 1910 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
1911 | struct buffer_head *bh; |
1912 | int error; | |
1913 | ||
1914 | /* Dealloc on-disk leaves to FREEMETA state */ | |
1915 | error = foreach_leaf(dip, leaf_dealloc, NULL); | |
1916 | if (error) | |
1917 | return error; | |
1918 | ||
1919 | /* Make this a regular file in case we crash. | |
1920 | (We don't want to free these blocks a second time.) */ | |
1921 | ||
1922 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
1923 | if (error) | |
1924 | return error; | |
1925 | ||
1926 | error = gfs2_meta_inode_buffer(dip, &bh); | |
1927 | if (!error) { | |
d4e9c4c3 | 1928 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
568f4c96 SW |
1929 | ((struct gfs2_dinode *)bh->b_data)->di_mode = |
1930 | cpu_to_be32(S_IFREG); | |
b3b94faa DT |
1931 | brelse(bh); |
1932 | } | |
1933 | ||
1934 | gfs2_trans_end(sdp); | |
1935 | ||
1936 | return error; | |
1937 | } | |
1938 | ||
1939 | /** | |
1940 | * gfs2_diradd_alloc_required - find if adding entry will require an allocation | |
1941 | * @ip: the file being written to | |
1942 | * @filname: the filename that's going to be added | |
b3b94faa | 1943 | * |
c752666c | 1944 | * Returns: 1 if alloc required, 0 if not, -ve on error |
b3b94faa DT |
1945 | */ |
1946 | ||
4d8012b6 | 1947 | int gfs2_diradd_alloc_required(struct inode *inode, const struct qstr *name) |
b3b94faa | 1948 | { |
c752666c SW |
1949 | struct gfs2_dirent *dent; |
1950 | struct buffer_head *bh; | |
b3b94faa | 1951 | |
c752666c | 1952 | dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space, &bh); |
ed386507 | 1953 | if (!dent) { |
c752666c | 1954 | return 1; |
ed386507 | 1955 | } |
c752666c SW |
1956 | if (IS_ERR(dent)) |
1957 | return PTR_ERR(dent); | |
1958 | brelse(bh); | |
1959 | return 0; | |
b3b94faa DT |
1960 | } |
1961 |