]>
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; | |
1a7b1eed | 134 | ip->i_inode.i_mtime.tv_sec = ip->i_inode.i_ctime.tv_sec = get_seconds(); |
539e5d6b | 135 | gfs2_dinode_out(ip, dibh->b_data); |
e13940ba SW |
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; | |
348acd48 | 187 | int new = 0; |
e13940ba SW |
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); | |
e13940ba | 215 | |
899bb264 | 216 | buf += amount; |
e13940ba SW |
217 | copied += amount; |
218 | lblock++; | |
219 | dblock++; | |
220 | extlen--; | |
221 | ||
222 | o = sizeof(struct gfs2_meta_header); | |
223 | } | |
224 | ||
225 | out: | |
226 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
227 | if (error) | |
228 | return error; | |
229 | ||
230 | if (ip->i_di.di_size < offset + copied) | |
231 | ip->i_di.di_size = offset + copied; | |
1a7b1eed | 232 | ip->i_inode.i_mtime.tv_sec = ip->i_inode.i_ctime.tv_sec = get_seconds(); |
e13940ba SW |
233 | |
234 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
539e5d6b | 235 | gfs2_dinode_out(ip, dibh->b_data); |
e13940ba SW |
236 | brelse(dibh); |
237 | ||
238 | return copied; | |
239 | fail: | |
240 | if (copied) | |
241 | goto out; | |
242 | return error; | |
243 | } | |
244 | ||
245 | static int gfs2_dir_read_stuffed(struct gfs2_inode *ip, char *buf, | |
7276b3b0 | 246 | u64 offset, unsigned int size) |
e13940ba SW |
247 | { |
248 | struct buffer_head *dibh; | |
249 | int error; | |
250 | ||
251 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
252 | if (!error) { | |
253 | offset += sizeof(struct gfs2_dinode); | |
254 | memcpy(buf, dibh->b_data + offset, size); | |
255 | brelse(dibh); | |
256 | } | |
257 | ||
258 | return (error) ? error : size; | |
259 | } | |
260 | ||
261 | ||
262 | /** | |
263 | * gfs2_dir_read_data - Read a data from a directory inode | |
264 | * @ip: The GFS2 Inode | |
265 | * @buf: The buffer to place result into | |
266 | * @offset: File offset to begin jdata_readng from | |
267 | * @size: Amount of data to transfer | |
268 | * | |
269 | * Returns: The amount of data actually copied or the error | |
270 | */ | |
7276b3b0 SW |
271 | static int gfs2_dir_read_data(struct gfs2_inode *ip, char *buf, u64 offset, |
272 | unsigned int size, unsigned ra) | |
e13940ba | 273 | { |
feaa7bba | 274 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
cd915493 SW |
275 | u64 lblock, dblock; |
276 | u32 extlen = 0; | |
e13940ba SW |
277 | unsigned int o; |
278 | int copied = 0; | |
279 | int error = 0; | |
280 | ||
281 | if (offset >= ip->i_di.di_size) | |
282 | return 0; | |
283 | ||
c5392124 | 284 | if (offset + size > ip->i_di.di_size) |
e13940ba SW |
285 | size = ip->i_di.di_size - offset; |
286 | ||
287 | if (!size) | |
288 | return 0; | |
289 | ||
290 | if (gfs2_is_stuffed(ip)) | |
7276b3b0 | 291 | return gfs2_dir_read_stuffed(ip, buf, offset, size); |
e13940ba SW |
292 | |
293 | if (gfs2_assert_warn(sdp, gfs2_is_jdata(ip))) | |
294 | return -EINVAL; | |
295 | ||
296 | lblock = offset; | |
297 | o = do_div(lblock, sdp->sd_jbsize) + sizeof(struct gfs2_meta_header); | |
298 | ||
299 | while (copied < size) { | |
300 | unsigned int amount; | |
301 | struct buffer_head *bh; | |
302 | int new; | |
303 | ||
304 | amount = size - copied; | |
305 | if (amount > sdp->sd_sb.sb_bsize - o) | |
306 | amount = sdp->sd_sb.sb_bsize - o; | |
307 | ||
308 | if (!extlen) { | |
309 | new = 0; | |
feaa7bba | 310 | error = gfs2_extent_map(&ip->i_inode, lblock, &new, |
fd88de56 | 311 | &dblock, &extlen); |
7276b3b0 | 312 | if (error || !dblock) |
e13940ba | 313 | goto fail; |
7276b3b0 SW |
314 | BUG_ON(extlen < 1); |
315 | if (!ra) | |
316 | extlen = 1; | |
317 | bh = gfs2_meta_ra(ip->i_gl, dblock, extlen); | |
b7d8ac3e | 318 | } else { |
7276b3b0 | 319 | error = gfs2_meta_read(ip->i_gl, dblock, DIO_WAIT, &bh); |
e13940ba SW |
320 | if (error) |
321 | goto fail; | |
7276b3b0 SW |
322 | } |
323 | error = gfs2_metatype_check(sdp, bh, GFS2_METATYPE_JD); | |
324 | if (error) { | |
325 | brelse(bh); | |
326 | goto fail; | |
327 | } | |
328 | dblock++; | |
329 | extlen--; | |
e13940ba SW |
330 | memcpy(buf, bh->b_data + o, amount); |
331 | brelse(bh); | |
899bb264 | 332 | buf += amount; |
e13940ba SW |
333 | copied += amount; |
334 | lblock++; | |
e13940ba SW |
335 | o = sizeof(struct gfs2_meta_header); |
336 | } | |
337 | ||
338 | return copied; | |
339 | fail: | |
340 | return (copied) ? copied : error; | |
341 | } | |
342 | ||
5e7d65cd SW |
343 | static inline int gfs2_dirent_sentinel(const struct gfs2_dirent *dent) |
344 | { | |
345 | return dent->de_inum.no_addr == 0 || dent->de_inum.no_formal_ino == 0; | |
346 | } | |
347 | ||
c752666c SW |
348 | static inline int __gfs2_dirent_find(const struct gfs2_dirent *dent, |
349 | const struct qstr *name, int ret) | |
350 | { | |
5e7d65cd | 351 | if (!gfs2_dirent_sentinel(dent) && |
c752666c SW |
352 | be32_to_cpu(dent->de_hash) == name->hash && |
353 | be16_to_cpu(dent->de_name_len) == name->len && | |
2bdbc5d7 | 354 | memcmp(dent+1, name->name, name->len) == 0) |
c752666c SW |
355 | return ret; |
356 | return 0; | |
357 | } | |
358 | ||
359 | static int gfs2_dirent_find(const struct gfs2_dirent *dent, | |
71b86f56 SW |
360 | const struct qstr *name, |
361 | void *opaque) | |
c752666c SW |
362 | { |
363 | return __gfs2_dirent_find(dent, name, 1); | |
364 | } | |
365 | ||
366 | static int gfs2_dirent_prev(const struct gfs2_dirent *dent, | |
71b86f56 SW |
367 | const struct qstr *name, |
368 | void *opaque) | |
c752666c SW |
369 | { |
370 | return __gfs2_dirent_find(dent, name, 2); | |
371 | } | |
372 | ||
373 | /* | |
374 | * name->name holds ptr to start of block. | |
375 | * name->len holds size of block. | |
b3b94faa | 376 | */ |
c752666c | 377 | static int gfs2_dirent_last(const struct gfs2_dirent *dent, |
71b86f56 SW |
378 | const struct qstr *name, |
379 | void *opaque) | |
c752666c SW |
380 | { |
381 | const char *start = name->name; | |
382 | const char *end = (const char *)dent + be16_to_cpu(dent->de_rec_len); | |
383 | if (name->len == (end - start)) | |
384 | return 1; | |
385 | return 0; | |
386 | } | |
b3b94faa | 387 | |
c752666c | 388 | static int gfs2_dirent_find_space(const struct gfs2_dirent *dent, |
71b86f56 SW |
389 | const struct qstr *name, |
390 | void *opaque) | |
b3b94faa | 391 | { |
c752666c SW |
392 | unsigned required = GFS2_DIRENT_SIZE(name->len); |
393 | unsigned actual = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len)); | |
394 | unsigned totlen = be16_to_cpu(dent->de_rec_len); | |
395 | ||
5e7d65cd | 396 | if (gfs2_dirent_sentinel(dent)) |
71b86f56 | 397 | actual = GFS2_DIRENT_SIZE(0); |
c5392124 | 398 | if (totlen - actual >= required) |
c752666c SW |
399 | return 1; |
400 | return 0; | |
401 | } | |
402 | ||
71b86f56 SW |
403 | struct dirent_gather { |
404 | const struct gfs2_dirent **pdent; | |
405 | unsigned offset; | |
406 | }; | |
407 | ||
408 | static int gfs2_dirent_gather(const struct gfs2_dirent *dent, | |
409 | const struct qstr *name, | |
410 | void *opaque) | |
411 | { | |
412 | struct dirent_gather *g = opaque; | |
5e7d65cd | 413 | if (!gfs2_dirent_sentinel(dent)) { |
71b86f56 SW |
414 | g->pdent[g->offset++] = dent; |
415 | } | |
416 | return 0; | |
417 | } | |
418 | ||
c752666c SW |
419 | /* |
420 | * Other possible things to check: | |
421 | * - Inode located within filesystem size (and on valid block) | |
422 | * - Valid directory entry type | |
423 | * Not sure how heavy-weight we want to make this... could also check | |
424 | * hash is correct for example, but that would take a lot of extra time. | |
425 | * For now the most important thing is to check that the various sizes | |
426 | * are correct. | |
427 | */ | |
428 | static int gfs2_check_dirent(struct gfs2_dirent *dent, unsigned int offset, | |
429 | unsigned int size, unsigned int len, int first) | |
430 | { | |
431 | const char *msg = "gfs2_dirent too small"; | |
432 | if (unlikely(size < sizeof(struct gfs2_dirent))) | |
433 | goto error; | |
434 | msg = "gfs2_dirent misaligned"; | |
435 | if (unlikely(offset & 0x7)) | |
436 | goto error; | |
437 | msg = "gfs2_dirent points beyond end of block"; | |
438 | if (unlikely(offset + size > len)) | |
439 | goto error; | |
440 | msg = "zero inode number"; | |
5e7d65cd | 441 | if (unlikely(!first && gfs2_dirent_sentinel(dent))) |
c752666c SW |
442 | goto error; |
443 | msg = "name length is greater than space in dirent"; | |
5e7d65cd | 444 | if (!gfs2_dirent_sentinel(dent) && |
c752666c SW |
445 | unlikely(sizeof(struct gfs2_dirent)+be16_to_cpu(dent->de_name_len) > |
446 | size)) | |
447 | goto error; | |
448 | return 0; | |
449 | error: | |
450 | printk(KERN_WARNING "gfs2_check_dirent: %s (%s)\n", msg, | |
451 | first ? "first in block" : "not first in block"); | |
452 | return -EIO; | |
b3b94faa DT |
453 | } |
454 | ||
71b86f56 | 455 | static int gfs2_dirent_offset(const void *buf) |
c752666c | 456 | { |
71b86f56 SW |
457 | const struct gfs2_meta_header *h = buf; |
458 | int offset; | |
c752666c SW |
459 | |
460 | BUG_ON(buf == NULL); | |
c752666c | 461 | |
e3167ded | 462 | switch(be32_to_cpu(h->mh_type)) { |
c752666c SW |
463 | case GFS2_METATYPE_LF: |
464 | offset = sizeof(struct gfs2_leaf); | |
465 | break; | |
466 | case GFS2_METATYPE_DI: | |
467 | offset = sizeof(struct gfs2_dinode); | |
468 | break; | |
469 | default: | |
470 | goto wrong_type; | |
471 | } | |
71b86f56 SW |
472 | return offset; |
473 | wrong_type: | |
474 | printk(KERN_WARNING "gfs2_scan_dirent: wrong block type %u\n", | |
e3167ded | 475 | be32_to_cpu(h->mh_type)); |
71b86f56 SW |
476 | return -1; |
477 | } | |
478 | ||
2bdbc5d7 | 479 | static struct gfs2_dirent *gfs2_dirent_scan(struct inode *inode, void *buf, |
71b86f56 SW |
480 | unsigned int len, gfs2_dscan_t scan, |
481 | const struct qstr *name, | |
482 | void *opaque) | |
483 | { | |
484 | struct gfs2_dirent *dent, *prev; | |
485 | unsigned offset; | |
486 | unsigned size; | |
487 | int ret = 0; | |
c752666c | 488 | |
71b86f56 SW |
489 | ret = gfs2_dirent_offset(buf); |
490 | if (ret < 0) | |
491 | goto consist_inode; | |
492 | ||
493 | offset = ret; | |
c752666c | 494 | prev = NULL; |
2bdbc5d7 | 495 | dent = buf + offset; |
c752666c SW |
496 | size = be16_to_cpu(dent->de_rec_len); |
497 | if (gfs2_check_dirent(dent, offset, size, len, 1)) | |
498 | goto consist_inode; | |
499 | do { | |
71b86f56 | 500 | ret = scan(dent, name, opaque); |
c752666c SW |
501 | if (ret) |
502 | break; | |
503 | offset += size; | |
504 | if (offset == len) | |
505 | break; | |
506 | prev = dent; | |
2bdbc5d7 | 507 | dent = buf + offset; |
c752666c SW |
508 | size = be16_to_cpu(dent->de_rec_len); |
509 | if (gfs2_check_dirent(dent, offset, size, len, 0)) | |
510 | goto consist_inode; | |
511 | } while(1); | |
512 | ||
513 | switch(ret) { | |
514 | case 0: | |
515 | return NULL; | |
516 | case 1: | |
517 | return dent; | |
518 | case 2: | |
519 | return prev ? prev : dent; | |
520 | default: | |
521 | BUG_ON(ret > 0); | |
522 | return ERR_PTR(ret); | |
523 | } | |
524 | ||
c752666c | 525 | consist_inode: |
feaa7bba | 526 | gfs2_consist_inode(GFS2_I(inode)); |
c752666c SW |
527 | return ERR_PTR(-EIO); |
528 | } | |
529 | ||
530 | ||
b3b94faa DT |
531 | /** |
532 | * dirent_first - Return the first dirent | |
533 | * @dip: the directory | |
534 | * @bh: The buffer | |
535 | * @dent: Pointer to list of dirents | |
536 | * | |
537 | * return first dirent whether bh points to leaf or stuffed dinode | |
538 | * | |
539 | * Returns: IS_LEAF, IS_DINODE, or -errno | |
540 | */ | |
541 | ||
542 | static int dirent_first(struct gfs2_inode *dip, struct buffer_head *bh, | |
543 | struct gfs2_dirent **dent) | |
544 | { | |
545 | struct gfs2_meta_header *h = (struct gfs2_meta_header *)bh->b_data; | |
546 | ||
e3167ded | 547 | if (be32_to_cpu(h->mh_type) == GFS2_METATYPE_LF) { |
feaa7bba | 548 | if (gfs2_meta_check(GFS2_SB(&dip->i_inode), bh)) |
b3b94faa DT |
549 | return -EIO; |
550 | *dent = (struct gfs2_dirent *)(bh->b_data + | |
551 | sizeof(struct gfs2_leaf)); | |
552 | return IS_LEAF; | |
553 | } else { | |
feaa7bba | 554 | if (gfs2_metatype_check(GFS2_SB(&dip->i_inode), bh, GFS2_METATYPE_DI)) |
b3b94faa DT |
555 | return -EIO; |
556 | *dent = (struct gfs2_dirent *)(bh->b_data + | |
557 | sizeof(struct gfs2_dinode)); | |
558 | return IS_DINODE; | |
559 | } | |
560 | } | |
561 | ||
2bdbc5d7 SW |
562 | static int dirent_check_reclen(struct gfs2_inode *dip, |
563 | const struct gfs2_dirent *d, const void *end_p) | |
564 | { | |
565 | const void *ptr = d; | |
566 | u16 rec_len = be16_to_cpu(d->de_rec_len); | |
567 | ||
568 | if (unlikely(rec_len < sizeof(struct gfs2_dirent))) | |
569 | goto broken; | |
570 | ptr += rec_len; | |
571 | if (ptr < end_p) | |
572 | return rec_len; | |
573 | if (ptr == end_p) | |
574 | return -ENOENT; | |
575 | broken: | |
576 | gfs2_consist_inode(dip); | |
577 | return -EIO; | |
578 | } | |
579 | ||
b3b94faa DT |
580 | /** |
581 | * dirent_next - Next dirent | |
582 | * @dip: the directory | |
583 | * @bh: The buffer | |
584 | * @dent: Pointer to list of dirents | |
585 | * | |
586 | * Returns: 0 on success, error code otherwise | |
587 | */ | |
588 | ||
589 | static int dirent_next(struct gfs2_inode *dip, struct buffer_head *bh, | |
590 | struct gfs2_dirent **dent) | |
591 | { | |
2bdbc5d7 SW |
592 | struct gfs2_dirent *cur = *dent, *tmp; |
593 | char *bh_end = bh->b_data + bh->b_size; | |
594 | int ret; | |
b3b94faa | 595 | |
2bdbc5d7 SW |
596 | ret = dirent_check_reclen(dip, cur, bh_end); |
597 | if (ret < 0) | |
598 | return ret; | |
4dd651ad | 599 | |
2bdbc5d7 SW |
600 | tmp = (void *)cur + ret; |
601 | ret = dirent_check_reclen(dip, tmp, bh_end); | |
602 | if (ret == -EIO) | |
603 | return ret; | |
4dd651ad | 604 | |
b3b94faa | 605 | /* Only the first dent could ever have de_inum.no_addr == 0 */ |
5e7d65cd | 606 | if (gfs2_dirent_sentinel(tmp)) { |
b3b94faa DT |
607 | gfs2_consist_inode(dip); |
608 | return -EIO; | |
609 | } | |
610 | ||
611 | *dent = tmp; | |
b3b94faa DT |
612 | return 0; |
613 | } | |
614 | ||
615 | /** | |
616 | * dirent_del - Delete a dirent | |
617 | * @dip: The GFS2 inode | |
618 | * @bh: The buffer | |
619 | * @prev: The previous dirent | |
620 | * @cur: The current dirent | |
621 | * | |
622 | */ | |
623 | ||
624 | static void dirent_del(struct gfs2_inode *dip, struct buffer_head *bh, | |
625 | struct gfs2_dirent *prev, struct gfs2_dirent *cur) | |
626 | { | |
cd915493 | 627 | u16 cur_rec_len, prev_rec_len; |
b3b94faa | 628 | |
5e7d65cd | 629 | if (gfs2_dirent_sentinel(cur)) { |
b3b94faa DT |
630 | gfs2_consist_inode(dip); |
631 | return; | |
632 | } | |
633 | ||
d4e9c4c3 | 634 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
b3b94faa DT |
635 | |
636 | /* If there is no prev entry, this is the first entry in the block. | |
637 | The de_rec_len is already as big as it needs to be. Just zero | |
638 | out the inode number and return. */ | |
639 | ||
640 | if (!prev) { | |
5e7d65cd SW |
641 | cur->de_inum.no_addr = 0; |
642 | cur->de_inum.no_formal_ino = 0; | |
b3b94faa DT |
643 | return; |
644 | } | |
645 | ||
646 | /* Combine this dentry with the previous one. */ | |
647 | ||
fc69d0d3 SW |
648 | prev_rec_len = be16_to_cpu(prev->de_rec_len); |
649 | cur_rec_len = be16_to_cpu(cur->de_rec_len); | |
b3b94faa DT |
650 | |
651 | if ((char *)prev + prev_rec_len != (char *)cur) | |
652 | gfs2_consist_inode(dip); | |
653 | if ((char *)cur + cur_rec_len > bh->b_data + bh->b_size) | |
654 | gfs2_consist_inode(dip); | |
655 | ||
656 | prev_rec_len += cur_rec_len; | |
fc69d0d3 | 657 | prev->de_rec_len = cpu_to_be16(prev_rec_len); |
b3b94faa DT |
658 | } |
659 | ||
c752666c SW |
660 | /* |
661 | * Takes a dent from which to grab space as an argument. Returns the | |
662 | * newly created dent. | |
b3b94faa | 663 | */ |
08bc2dbc AB |
664 | static struct gfs2_dirent *gfs2_init_dirent(struct inode *inode, |
665 | struct gfs2_dirent *dent, | |
666 | const struct qstr *name, | |
667 | struct buffer_head *bh) | |
b3b94faa | 668 | { |
feaa7bba | 669 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c SW |
670 | struct gfs2_dirent *ndent; |
671 | unsigned offset = 0, totlen; | |
672 | ||
5e7d65cd | 673 | if (!gfs2_dirent_sentinel(dent)) |
c752666c SW |
674 | offset = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len)); |
675 | totlen = be16_to_cpu(dent->de_rec_len); | |
676 | BUG_ON(offset + name->len > totlen); | |
677 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
678 | ndent = (struct gfs2_dirent *)((char *)dent + offset); | |
679 | dent->de_rec_len = cpu_to_be16(offset); | |
680 | gfs2_qstr2dirent(name, totlen - offset, ndent); | |
681 | return ndent; | |
b3b94faa DT |
682 | } |
683 | ||
c752666c SW |
684 | static struct gfs2_dirent *gfs2_dirent_alloc(struct inode *inode, |
685 | struct buffer_head *bh, | |
686 | const struct qstr *name) | |
b3b94faa DT |
687 | { |
688 | struct gfs2_dirent *dent; | |
907b9bce | 689 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, |
71b86f56 | 690 | gfs2_dirent_find_space, name, NULL); |
c752666c SW |
691 | if (!dent || IS_ERR(dent)) |
692 | return dent; | |
693 | return gfs2_init_dirent(inode, dent, name, bh); | |
b3b94faa DT |
694 | } |
695 | ||
cd915493 | 696 | static int get_leaf(struct gfs2_inode *dip, u64 leaf_no, |
b3b94faa DT |
697 | struct buffer_head **bhp) |
698 | { | |
699 | int error; | |
700 | ||
7276b3b0 | 701 | error = gfs2_meta_read(dip->i_gl, leaf_no, DIO_WAIT, bhp); |
feaa7bba SW |
702 | if (!error && gfs2_metatype_check(GFS2_SB(&dip->i_inode), *bhp, GFS2_METATYPE_LF)) { |
703 | /* printk(KERN_INFO "block num=%llu\n", leaf_no); */ | |
b3b94faa | 704 | error = -EIO; |
feaa7bba | 705 | } |
b3b94faa DT |
706 | |
707 | return error; | |
708 | } | |
709 | ||
710 | /** | |
711 | * get_leaf_nr - Get a leaf number associated with the index | |
712 | * @dip: The GFS2 inode | |
713 | * @index: | |
714 | * @leaf_out: | |
715 | * | |
716 | * Returns: 0 on success, error code otherwise | |
717 | */ | |
718 | ||
cd915493 SW |
719 | static int get_leaf_nr(struct gfs2_inode *dip, u32 index, |
720 | u64 *leaf_out) | |
b3b94faa | 721 | { |
b44b84d7 | 722 | __be64 leaf_no; |
b3b94faa DT |
723 | int error; |
724 | ||
e13940ba | 725 | error = gfs2_dir_read_data(dip, (char *)&leaf_no, |
b44b84d7 AV |
726 | index * sizeof(__be64), |
727 | sizeof(__be64), 0); | |
cd915493 | 728 | if (error != sizeof(u64)) |
b3b94faa DT |
729 | return (error < 0) ? error : -EIO; |
730 | ||
731 | *leaf_out = be64_to_cpu(leaf_no); | |
732 | ||
733 | return 0; | |
734 | } | |
735 | ||
cd915493 | 736 | static int get_first_leaf(struct gfs2_inode *dip, u32 index, |
b3b94faa DT |
737 | struct buffer_head **bh_out) |
738 | { | |
cd915493 | 739 | u64 leaf_no; |
b3b94faa DT |
740 | int error; |
741 | ||
742 | error = get_leaf_nr(dip, index, &leaf_no); | |
743 | if (!error) | |
744 | error = get_leaf(dip, leaf_no, bh_out); | |
745 | ||
746 | return error; | |
747 | } | |
748 | ||
c752666c SW |
749 | static struct gfs2_dirent *gfs2_dirent_search(struct inode *inode, |
750 | const struct qstr *name, | |
751 | gfs2_dscan_t scan, | |
752 | struct buffer_head **pbh) | |
b3b94faa | 753 | { |
c752666c SW |
754 | struct buffer_head *bh; |
755 | struct gfs2_dirent *dent; | |
feaa7bba | 756 | struct gfs2_inode *ip = GFS2_I(inode); |
b3b94faa DT |
757 | int error; |
758 | ||
c752666c SW |
759 | if (ip->i_di.di_flags & GFS2_DIF_EXHASH) { |
760 | struct gfs2_leaf *leaf; | |
761 | unsigned hsize = 1 << ip->i_di.di_depth; | |
762 | unsigned index; | |
763 | u64 ln; | |
764 | if (hsize * sizeof(u64) != ip->i_di.di_size) { | |
765 | gfs2_consist_inode(ip); | |
766 | return ERR_PTR(-EIO); | |
767 | } | |
907b9bce | 768 | |
c752666c SW |
769 | index = name->hash >> (32 - ip->i_di.di_depth); |
770 | error = get_first_leaf(ip, index, &bh); | |
771 | if (error) | |
772 | return ERR_PTR(error); | |
773 | do { | |
774 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, | |
71b86f56 | 775 | scan, name, NULL); |
c752666c SW |
776 | if (dent) |
777 | goto got_dent; | |
778 | leaf = (struct gfs2_leaf *)bh->b_data; | |
779 | ln = be64_to_cpu(leaf->lf_next); | |
f4154ea0 | 780 | brelse(bh); |
c752666c SW |
781 | if (!ln) |
782 | break; | |
907b9bce | 783 | |
c752666c SW |
784 | error = get_leaf(ip, ln, &bh); |
785 | } while(!error); | |
b3b94faa | 786 | |
c752666c | 787 | return error ? ERR_PTR(error) : NULL; |
b3b94faa | 788 | } |
b3b94faa | 789 | |
907b9bce | 790 | |
c752666c SW |
791 | error = gfs2_meta_inode_buffer(ip, &bh); |
792 | if (error) | |
793 | return ERR_PTR(error); | |
71b86f56 | 794 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, scan, name, NULL); |
c752666c | 795 | got_dent: |
f4154ea0 | 796 | if (unlikely(dent == NULL || IS_ERR(dent))) { |
ed386507 SW |
797 | brelse(bh); |
798 | bh = NULL; | |
799 | } | |
c752666c SW |
800 | *pbh = bh; |
801 | return dent; | |
802 | } | |
b3b94faa | 803 | |
c752666c SW |
804 | static struct gfs2_leaf *new_leaf(struct inode *inode, struct buffer_head **pbh, u16 depth) |
805 | { | |
feaa7bba | 806 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c SW |
807 | u64 bn = gfs2_alloc_meta(ip); |
808 | struct buffer_head *bh = gfs2_meta_new(ip->i_gl, bn); | |
809 | struct gfs2_leaf *leaf; | |
810 | struct gfs2_dirent *dent; | |
71b86f56 | 811 | struct qstr name = { .name = "", .len = 0, .hash = 0 }; |
c752666c SW |
812 | if (!bh) |
813 | return NULL; | |
907b9bce | 814 | |
c752666c SW |
815 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
816 | gfs2_metatype_set(bh, GFS2_METATYPE_LF, GFS2_FORMAT_LF); | |
817 | leaf = (struct gfs2_leaf *)bh->b_data; | |
818 | leaf->lf_depth = cpu_to_be16(depth); | |
2bdbc5d7 | 819 | leaf->lf_entries = 0; |
a2d7d021 | 820 | leaf->lf_dirent_format = cpu_to_be32(GFS2_FORMAT_DE); |
2bdbc5d7 | 821 | leaf->lf_next = 0; |
c752666c SW |
822 | memset(leaf->lf_reserved, 0, sizeof(leaf->lf_reserved)); |
823 | dent = (struct gfs2_dirent *)(leaf+1); | |
71b86f56 | 824 | gfs2_qstr2dirent(&name, bh->b_size - sizeof(struct gfs2_leaf), dent); |
c752666c SW |
825 | *pbh = bh; |
826 | return leaf; | |
b3b94faa DT |
827 | } |
828 | ||
829 | /** | |
830 | * dir_make_exhash - Convert a stuffed directory into an ExHash directory | |
831 | * @dip: The GFS2 inode | |
832 | * | |
833 | * Returns: 0 on success, error code otherwise | |
834 | */ | |
835 | ||
c752666c | 836 | static int dir_make_exhash(struct inode *inode) |
b3b94faa | 837 | { |
feaa7bba SW |
838 | struct gfs2_inode *dip = GFS2_I(inode); |
839 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
b3b94faa | 840 | struct gfs2_dirent *dent; |
c752666c | 841 | struct qstr args; |
b3b94faa DT |
842 | struct buffer_head *bh, *dibh; |
843 | struct gfs2_leaf *leaf; | |
844 | int y; | |
cd915493 | 845 | u32 x; |
b44b84d7 AV |
846 | __be64 *lp; |
847 | u64 bn; | |
b3b94faa DT |
848 | int error; |
849 | ||
850 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
851 | if (error) | |
852 | return error; | |
853 | ||
b3b94faa DT |
854 | /* Turn over a new leaf */ |
855 | ||
c752666c SW |
856 | leaf = new_leaf(inode, &bh, 0); |
857 | if (!leaf) | |
858 | return -ENOSPC; | |
859 | bn = bh->b_blocknr; | |
b3b94faa DT |
860 | |
861 | gfs2_assert(sdp, dip->i_di.di_entries < (1 << 16)); | |
b3b94faa DT |
862 | leaf->lf_entries = cpu_to_be16(dip->i_di.di_entries); |
863 | ||
864 | /* Copy dirents */ | |
865 | ||
866 | gfs2_buffer_copy_tail(bh, sizeof(struct gfs2_leaf), dibh, | |
867 | sizeof(struct gfs2_dinode)); | |
868 | ||
869 | /* Find last entry */ | |
870 | ||
871 | x = 0; | |
c752666c SW |
872 | args.len = bh->b_size - sizeof(struct gfs2_dinode) + |
873 | sizeof(struct gfs2_leaf); | |
874 | args.name = bh->b_data; | |
feaa7bba | 875 | dent = gfs2_dirent_scan(&dip->i_inode, bh->b_data, bh->b_size, |
71b86f56 | 876 | gfs2_dirent_last, &args, NULL); |
c752666c SW |
877 | if (!dent) { |
878 | brelse(bh); | |
879 | brelse(dibh); | |
880 | return -EIO; | |
881 | } | |
882 | if (IS_ERR(dent)) { | |
883 | brelse(bh); | |
884 | brelse(dibh); | |
885 | return PTR_ERR(dent); | |
b3b94faa | 886 | } |
b3b94faa DT |
887 | |
888 | /* Adjust the last dirent's record length | |
889 | (Remember that dent still points to the last entry.) */ | |
890 | ||
4dd651ad | 891 | dent->de_rec_len = cpu_to_be16(be16_to_cpu(dent->de_rec_len) + |
b3b94faa | 892 | sizeof(struct gfs2_dinode) - |
4dd651ad | 893 | sizeof(struct gfs2_leaf)); |
b3b94faa DT |
894 | |
895 | brelse(bh); | |
896 | ||
897 | /* We're done with the new leaf block, now setup the new | |
898 | hash table. */ | |
899 | ||
d4e9c4c3 | 900 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
b3b94faa DT |
901 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode)); |
902 | ||
b44b84d7 | 903 | lp = (__be64 *)(dibh->b_data + sizeof(struct gfs2_dinode)); |
b3b94faa DT |
904 | |
905 | for (x = sdp->sd_hash_ptrs; x--; lp++) | |
906 | *lp = cpu_to_be64(bn); | |
907 | ||
908 | dip->i_di.di_size = sdp->sd_sb.sb_bsize / 2; | |
909 | dip->i_di.di_blocks++; | |
9e2dbdac | 910 | gfs2_set_inode_blocks(&dip->i_inode); |
b3b94faa | 911 | dip->i_di.di_flags |= GFS2_DIF_EXHASH; |
b3b94faa DT |
912 | |
913 | for (x = sdp->sd_hash_ptrs, y = -1; x; x >>= 1, y++) ; | |
914 | dip->i_di.di_depth = y; | |
915 | ||
539e5d6b | 916 | gfs2_dinode_out(dip, dibh->b_data); |
b3b94faa DT |
917 | |
918 | brelse(dibh); | |
919 | ||
920 | return 0; | |
921 | } | |
922 | ||
923 | /** | |
924 | * dir_split_leaf - Split a leaf block into two | |
925 | * @dip: The GFS2 inode | |
926 | * @index: | |
927 | * @leaf_no: | |
928 | * | |
929 | * Returns: 0 on success, error code on failure | |
930 | */ | |
931 | ||
c752666c | 932 | static int dir_split_leaf(struct inode *inode, const struct qstr *name) |
b3b94faa | 933 | { |
feaa7bba | 934 | struct gfs2_inode *dip = GFS2_I(inode); |
b3b94faa DT |
935 | struct buffer_head *nbh, *obh, *dibh; |
936 | struct gfs2_leaf *nleaf, *oleaf; | |
4da3c646 | 937 | struct gfs2_dirent *dent = NULL, *prev = NULL, *next = NULL, *new; |
cd915493 | 938 | u32 start, len, half_len, divider; |
b44b84d7 AV |
939 | u64 bn, leaf_no; |
940 | __be64 *lp; | |
cd915493 | 941 | u32 index; |
b3b94faa DT |
942 | int x, moved = 0; |
943 | int error; | |
944 | ||
c752666c SW |
945 | index = name->hash >> (32 - dip->i_di.di_depth); |
946 | error = get_leaf_nr(dip, index, &leaf_no); | |
947 | if (error) | |
948 | return error; | |
b3b94faa DT |
949 | |
950 | /* Get the old leaf block */ | |
b3b94faa DT |
951 | error = get_leaf(dip, leaf_no, &obh); |
952 | if (error) | |
e90deff5 | 953 | return error; |
b3b94faa | 954 | |
b3b94faa | 955 | oleaf = (struct gfs2_leaf *)obh->b_data; |
e90deff5 SW |
956 | if (dip->i_di.di_depth == be16_to_cpu(oleaf->lf_depth)) { |
957 | brelse(obh); | |
958 | return 1; /* can't split */ | |
959 | } | |
960 | ||
961 | gfs2_trans_add_bh(dip->i_gl, obh, 1); | |
b3b94faa | 962 | |
c752666c SW |
963 | nleaf = new_leaf(inode, &nbh, be16_to_cpu(oleaf->lf_depth) + 1); |
964 | if (!nleaf) { | |
965 | brelse(obh); | |
966 | return -ENOSPC; | |
967 | } | |
968 | bn = nbh->b_blocknr; | |
b3b94faa | 969 | |
c752666c | 970 | /* Compute the start and len of leaf pointers in the hash table. */ |
b3b94faa DT |
971 | len = 1 << (dip->i_di.di_depth - be16_to_cpu(oleaf->lf_depth)); |
972 | half_len = len >> 1; | |
973 | if (!half_len) { | |
e90deff5 | 974 | 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 |
975 | gfs2_consist_inode(dip); |
976 | error = -EIO; | |
977 | goto fail_brelse; | |
978 | } | |
979 | ||
980 | start = (index & ~(len - 1)); | |
981 | ||
982 | /* Change the pointers. | |
983 | Don't bother distinguishing stuffed from non-stuffed. | |
984 | This code is complicated enough already. */ | |
b44b84d7 | 985 | lp = kmalloc(half_len * sizeof(__be64), GFP_NOFS | __GFP_NOFAIL); |
b3b94faa | 986 | /* Change the pointers */ |
b3b94faa DT |
987 | for (x = 0; x < half_len; x++) |
988 | lp[x] = cpu_to_be64(bn); | |
989 | ||
cd915493 SW |
990 | error = gfs2_dir_write_data(dip, (char *)lp, start * sizeof(u64), |
991 | half_len * sizeof(u64)); | |
992 | if (error != half_len * sizeof(u64)) { | |
b3b94faa DT |
993 | if (error >= 0) |
994 | error = -EIO; | |
995 | goto fail_lpfree; | |
996 | } | |
997 | ||
998 | kfree(lp); | |
999 | ||
1000 | /* Compute the divider */ | |
b3b94faa DT |
1001 | divider = (start + half_len) << (32 - dip->i_di.di_depth); |
1002 | ||
1003 | /* Copy the entries */ | |
b3b94faa DT |
1004 | dirent_first(dip, obh, &dent); |
1005 | ||
1006 | do { | |
1007 | next = dent; | |
1008 | if (dirent_next(dip, obh, &next)) | |
1009 | next = NULL; | |
1010 | ||
5e7d65cd | 1011 | if (!gfs2_dirent_sentinel(dent) && |
b3b94faa | 1012 | be32_to_cpu(dent->de_hash) < divider) { |
c752666c SW |
1013 | struct qstr str; |
1014 | str.name = (char*)(dent+1); | |
1015 | str.len = be16_to_cpu(dent->de_name_len); | |
1016 | str.hash = be32_to_cpu(dent->de_hash); | |
71b86f56 | 1017 | new = gfs2_dirent_alloc(inode, nbh, &str); |
c752666c SW |
1018 | if (IS_ERR(new)) { |
1019 | error = PTR_ERR(new); | |
1020 | break; | |
1021 | } | |
b3b94faa DT |
1022 | |
1023 | new->de_inum = dent->de_inum; /* No endian worries */ | |
b3b94faa | 1024 | new->de_type = dent->de_type; /* No endian worries */ |
c752666c | 1025 | nleaf->lf_entries = cpu_to_be16(be16_to_cpu(nleaf->lf_entries)+1); |
b3b94faa DT |
1026 | |
1027 | dirent_del(dip, obh, prev, dent); | |
1028 | ||
1029 | if (!oleaf->lf_entries) | |
1030 | gfs2_consist_inode(dip); | |
c752666c | 1031 | oleaf->lf_entries = cpu_to_be16(be16_to_cpu(oleaf->lf_entries)-1); |
b3b94faa DT |
1032 | |
1033 | if (!prev) | |
1034 | prev = dent; | |
1035 | ||
1036 | moved = 1; | |
c752666c | 1037 | } else { |
b3b94faa | 1038 | prev = dent; |
c752666c | 1039 | } |
b3b94faa | 1040 | dent = next; |
c752666c | 1041 | } while (dent); |
b3b94faa | 1042 | |
c752666c | 1043 | oleaf->lf_depth = nleaf->lf_depth; |
b3b94faa DT |
1044 | |
1045 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
feaa7bba | 1046 | if (!gfs2_assert_withdraw(GFS2_SB(&dip->i_inode), !error)) { |
b3b94faa | 1047 | dip->i_di.di_blocks++; |
9e2dbdac | 1048 | gfs2_set_inode_blocks(&dip->i_inode); |
539e5d6b | 1049 | gfs2_dinode_out(dip, dibh->b_data); |
b3b94faa DT |
1050 | brelse(dibh); |
1051 | } | |
1052 | ||
1053 | brelse(obh); | |
1054 | brelse(nbh); | |
1055 | ||
1056 | return error; | |
1057 | ||
e90deff5 | 1058 | fail_lpfree: |
b3b94faa DT |
1059 | kfree(lp); |
1060 | ||
e90deff5 | 1061 | fail_brelse: |
b3b94faa | 1062 | brelse(obh); |
b3b94faa DT |
1063 | brelse(nbh); |
1064 | return error; | |
1065 | } | |
1066 | ||
1067 | /** | |
1068 | * dir_double_exhash - Double size of ExHash table | |
1069 | * @dip: The GFS2 dinode | |
1070 | * | |
1071 | * Returns: 0 on success, error code on failure | |
1072 | */ | |
1073 | ||
1074 | static int dir_double_exhash(struct gfs2_inode *dip) | |
1075 | { | |
feaa7bba | 1076 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa | 1077 | struct buffer_head *dibh; |
cd915493 SW |
1078 | u32 hsize; |
1079 | u64 *buf; | |
1080 | u64 *from, *to; | |
1081 | u64 block; | |
b3b94faa DT |
1082 | int x; |
1083 | int error = 0; | |
1084 | ||
1085 | hsize = 1 << dip->i_di.di_depth; | |
cd915493 | 1086 | if (hsize * sizeof(u64) != dip->i_di.di_size) { |
b3b94faa DT |
1087 | gfs2_consist_inode(dip); |
1088 | return -EIO; | |
1089 | } | |
1090 | ||
1091 | /* Allocate both the "from" and "to" buffers in one big chunk */ | |
1092 | ||
1093 | buf = kcalloc(3, sdp->sd_hash_bsize, GFP_KERNEL | __GFP_NOFAIL); | |
1094 | ||
1095 | for (block = dip->i_di.di_size >> sdp->sd_hash_bsize_shift; block--;) { | |
e13940ba | 1096 | error = gfs2_dir_read_data(dip, (char *)buf, |
b3b94faa | 1097 | block * sdp->sd_hash_bsize, |
7276b3b0 | 1098 | sdp->sd_hash_bsize, 1); |
b3b94faa DT |
1099 | if (error != sdp->sd_hash_bsize) { |
1100 | if (error >= 0) | |
1101 | error = -EIO; | |
1102 | goto fail; | |
1103 | } | |
1104 | ||
1105 | from = buf; | |
cd915493 | 1106 | to = (u64 *)((char *)buf + sdp->sd_hash_bsize); |
b3b94faa DT |
1107 | |
1108 | for (x = sdp->sd_hash_ptrs; x--; from++) { | |
1109 | *to++ = *from; /* No endianess worries */ | |
1110 | *to++ = *from; | |
1111 | } | |
1112 | ||
e13940ba | 1113 | error = gfs2_dir_write_data(dip, |
b3b94faa DT |
1114 | (char *)buf + sdp->sd_hash_bsize, |
1115 | block * sdp->sd_sb.sb_bsize, | |
1116 | sdp->sd_sb.sb_bsize); | |
1117 | if (error != sdp->sd_sb.sb_bsize) { | |
1118 | if (error >= 0) | |
1119 | error = -EIO; | |
1120 | goto fail; | |
1121 | } | |
1122 | } | |
1123 | ||
1124 | kfree(buf); | |
1125 | ||
1126 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1127 | if (!gfs2_assert_withdraw(sdp, !error)) { | |
1128 | dip->i_di.di_depth++; | |
539e5d6b | 1129 | gfs2_dinode_out(dip, dibh->b_data); |
b3b94faa DT |
1130 | brelse(dibh); |
1131 | } | |
1132 | ||
1133 | return error; | |
1134 | ||
a91ea69f | 1135 | fail: |
b3b94faa | 1136 | kfree(buf); |
b3b94faa DT |
1137 | return error; |
1138 | } | |
1139 | ||
1140 | /** | |
1141 | * compare_dents - compare directory entries by hash value | |
1142 | * @a: first dent | |
1143 | * @b: second dent | |
1144 | * | |
1145 | * When comparing the hash entries of @a to @b: | |
1146 | * gt: returns 1 | |
1147 | * lt: returns -1 | |
1148 | * eq: returns 0 | |
1149 | */ | |
1150 | ||
1151 | static int compare_dents(const void *a, const void *b) | |
1152 | { | |
2bdbc5d7 | 1153 | const struct gfs2_dirent *dent_a, *dent_b; |
cd915493 | 1154 | u32 hash_a, hash_b; |
b3b94faa DT |
1155 | int ret = 0; |
1156 | ||
2bdbc5d7 | 1157 | dent_a = *(const struct gfs2_dirent **)a; |
c752666c | 1158 | hash_a = be32_to_cpu(dent_a->de_hash); |
b3b94faa | 1159 | |
2bdbc5d7 | 1160 | dent_b = *(const struct gfs2_dirent **)b; |
c752666c | 1161 | hash_b = be32_to_cpu(dent_b->de_hash); |
b3b94faa DT |
1162 | |
1163 | if (hash_a > hash_b) | |
1164 | ret = 1; | |
1165 | else if (hash_a < hash_b) | |
1166 | ret = -1; | |
1167 | else { | |
4dd651ad SW |
1168 | unsigned int len_a = be16_to_cpu(dent_a->de_name_len); |
1169 | unsigned int len_b = be16_to_cpu(dent_b->de_name_len); | |
b3b94faa DT |
1170 | |
1171 | if (len_a > len_b) | |
1172 | ret = 1; | |
1173 | else if (len_a < len_b) | |
1174 | ret = -1; | |
1175 | else | |
2bdbc5d7 | 1176 | ret = memcmp(dent_a + 1, dent_b + 1, len_a); |
b3b94faa DT |
1177 | } |
1178 | ||
1179 | return ret; | |
1180 | } | |
1181 | ||
1182 | /** | |
1183 | * do_filldir_main - read out directory entries | |
1184 | * @dip: The GFS2 inode | |
1185 | * @offset: The offset in the file to read from | |
1186 | * @opaque: opaque data to pass to filldir | |
1187 | * @filldir: The function to pass entries to | |
1188 | * @darr: an array of struct gfs2_dirent pointers to read | |
1189 | * @entries: the number of entries in darr | |
1190 | * @copied: pointer to int that's non-zero if a entry has been copied out | |
1191 | * | |
1192 | * Jump through some hoops to make sure that if there are hash collsions, | |
1193 | * they are read out at the beginning of a buffer. We want to minimize | |
1194 | * the possibility that they will fall into different readdir buffers or | |
1195 | * that someone will want to seek to that location. | |
1196 | * | |
1197 | * Returns: errno, >0 on exception from filldir | |
1198 | */ | |
1199 | ||
cd915493 | 1200 | static int do_filldir_main(struct gfs2_inode *dip, u64 *offset, |
b3b94faa | 1201 | void *opaque, gfs2_filldir_t filldir, |
cd915493 | 1202 | const struct gfs2_dirent **darr, u32 entries, |
b3b94faa DT |
1203 | int *copied) |
1204 | { | |
71b86f56 | 1205 | const struct gfs2_dirent *dent, *dent_next; |
629a21e7 | 1206 | struct gfs2_inum_host inum; |
cd915493 | 1207 | u64 off, off_next; |
b3b94faa DT |
1208 | unsigned int x, y; |
1209 | int run = 0; | |
1210 | int error = 0; | |
1211 | ||
1212 | sort(darr, entries, sizeof(struct gfs2_dirent *), compare_dents, NULL); | |
1213 | ||
1214 | dent_next = darr[0]; | |
1215 | off_next = be32_to_cpu(dent_next->de_hash); | |
1216 | off_next = gfs2_disk_hash2offset(off_next); | |
1217 | ||
1218 | for (x = 0, y = 1; x < entries; x++, y++) { | |
1219 | dent = dent_next; | |
1220 | off = off_next; | |
1221 | ||
1222 | if (y < entries) { | |
1223 | dent_next = darr[y]; | |
1224 | off_next = be32_to_cpu(dent_next->de_hash); | |
1225 | off_next = gfs2_disk_hash2offset(off_next); | |
1226 | ||
1227 | if (off < *offset) | |
1228 | continue; | |
1229 | *offset = off; | |
1230 | ||
1231 | if (off_next == off) { | |
1232 | if (*copied && !run) | |
1233 | return 1; | |
1234 | run = 1; | |
1235 | } else | |
1236 | run = 0; | |
1237 | } else { | |
1238 | if (off < *offset) | |
1239 | continue; | |
1240 | *offset = off; | |
1241 | } | |
1242 | ||
1243 | gfs2_inum_in(&inum, (char *)&dent->de_inum); | |
1244 | ||
2bdbc5d7 | 1245 | error = filldir(opaque, (const char *)(dent + 1), |
4dd651ad | 1246 | be16_to_cpu(dent->de_name_len), |
b3b94faa | 1247 | off, &inum, |
4dd651ad | 1248 | be16_to_cpu(dent->de_type)); |
b3b94faa DT |
1249 | if (error) |
1250 | return 1; | |
1251 | ||
1252 | *copied = 1; | |
1253 | } | |
1254 | ||
1255 | /* Increment the *offset by one, so the next time we come into the | |
1256 | do_filldir fxn, we get the next entry instead of the last one in the | |
1257 | current leaf */ | |
1258 | ||
1259 | (*offset)++; | |
1260 | ||
1261 | return 0; | |
1262 | } | |
1263 | ||
71b86f56 SW |
1264 | static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque, |
1265 | gfs2_filldir_t filldir, int *copied, | |
1266 | unsigned *depth, u64 leaf_no) | |
b3b94faa | 1267 | { |
feaa7bba | 1268 | struct gfs2_inode *ip = GFS2_I(inode); |
71b86f56 SW |
1269 | struct buffer_head *bh; |
1270 | struct gfs2_leaf *lf; | |
1271 | unsigned entries = 0; | |
1272 | unsigned leaves = 0; | |
1273 | const struct gfs2_dirent **darr, *dent; | |
1274 | struct dirent_gather g; | |
1275 | struct buffer_head **larr; | |
1276 | int leaf = 0; | |
1277 | int error, i; | |
1278 | u64 lfn = leaf_no; | |
b3b94faa | 1279 | |
b3b94faa | 1280 | do { |
71b86f56 | 1281 | error = get_leaf(ip, lfn, &bh); |
b3b94faa | 1282 | if (error) |
71b86f56 SW |
1283 | goto out; |
1284 | lf = (struct gfs2_leaf *)bh->b_data; | |
1285 | if (leaves == 0) | |
1286 | *depth = be16_to_cpu(lf->lf_depth); | |
1287 | entries += be16_to_cpu(lf->lf_entries); | |
1288 | leaves++; | |
1289 | lfn = be64_to_cpu(lf->lf_next); | |
1290 | brelse(bh); | |
1291 | } while(lfn); | |
b3b94faa DT |
1292 | |
1293 | if (!entries) | |
1294 | return 0; | |
1295 | ||
71b86f56 | 1296 | error = -ENOMEM; |
2bdbc5d7 | 1297 | larr = vmalloc((leaves + entries) * sizeof(void *)); |
71b86f56 SW |
1298 | if (!larr) |
1299 | goto out; | |
1300 | darr = (const struct gfs2_dirent **)(larr + leaves); | |
1301 | g.pdent = darr; | |
1302 | g.offset = 0; | |
1303 | lfn = leaf_no; | |
b3b94faa | 1304 | |
71b86f56 SW |
1305 | do { |
1306 | error = get_leaf(ip, lfn, &bh); | |
b3b94faa | 1307 | if (error) |
71b86f56 SW |
1308 | goto out_kfree; |
1309 | lf = (struct gfs2_leaf *)bh->b_data; | |
1310 | lfn = be64_to_cpu(lf->lf_next); | |
1311 | if (lf->lf_entries) { | |
1312 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, | |
1313 | gfs2_dirent_gather, NULL, &g); | |
1314 | error = PTR_ERR(dent); | |
1315 | if (IS_ERR(dent)) { | |
1316 | goto out_kfree; | |
1317 | } | |
1318 | error = 0; | |
1319 | larr[leaf++] = bh; | |
b3b94faa | 1320 | } else { |
71b86f56 | 1321 | brelse(bh); |
b3b94faa | 1322 | } |
71b86f56 | 1323 | } while(lfn); |
b3b94faa | 1324 | |
71b86f56 | 1325 | error = do_filldir_main(ip, offset, opaque, filldir, darr, |
b3b94faa | 1326 | entries, copied); |
71b86f56 SW |
1327 | out_kfree: |
1328 | for(i = 0; i < leaf; i++) | |
1329 | brelse(larr[i]); | |
fe1bdedc | 1330 | vfree(larr); |
71b86f56 | 1331 | out: |
b3b94faa DT |
1332 | return error; |
1333 | } | |
1334 | ||
1335 | /** | |
c752666c SW |
1336 | * dir_e_read - Reads the entries from a directory into a filldir buffer |
1337 | * @dip: dinode pointer | |
1338 | * @offset: the hash of the last entry read shifted to the right once | |
1339 | * @opaque: buffer for the filldir function to fill | |
1340 | * @filldir: points to the filldir function to use | |
b3b94faa | 1341 | * |
c752666c | 1342 | * Returns: errno |
b3b94faa DT |
1343 | */ |
1344 | ||
cd915493 | 1345 | static int dir_e_read(struct inode *inode, u64 *offset, void *opaque, |
c752666c | 1346 | gfs2_filldir_t filldir) |
b3b94faa | 1347 | { |
feaa7bba SW |
1348 | struct gfs2_inode *dip = GFS2_I(inode); |
1349 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
cd915493 SW |
1350 | u32 hsize, len = 0; |
1351 | u32 ht_offset, lp_offset, ht_offset_cur = -1; | |
1352 | u32 hash, index; | |
b44b84d7 | 1353 | __be64 *lp; |
c752666c SW |
1354 | int copied = 0; |
1355 | int error = 0; | |
4da3c646 | 1356 | unsigned depth = 0; |
b3b94faa DT |
1357 | |
1358 | hsize = 1 << dip->i_di.di_depth; | |
cd915493 | 1359 | if (hsize * sizeof(u64) != dip->i_di.di_size) { |
b3b94faa DT |
1360 | gfs2_consist_inode(dip); |
1361 | return -EIO; | |
1362 | } | |
1363 | ||
1364 | hash = gfs2_dir_offset2hash(*offset); | |
1365 | index = hash >> (32 - dip->i_di.di_depth); | |
1366 | ||
1367 | lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL); | |
1368 | if (!lp) | |
1369 | return -ENOMEM; | |
1370 | ||
1371 | while (index < hsize) { | |
1372 | lp_offset = index & (sdp->sd_hash_ptrs - 1); | |
1373 | ht_offset = index - lp_offset; | |
1374 | ||
1375 | if (ht_offset_cur != ht_offset) { | |
e13940ba | 1376 | error = gfs2_dir_read_data(dip, (char *)lp, |
b44b84d7 | 1377 | ht_offset * sizeof(__be64), |
7276b3b0 | 1378 | sdp->sd_hash_bsize, 1); |
b3b94faa DT |
1379 | if (error != sdp->sd_hash_bsize) { |
1380 | if (error >= 0) | |
1381 | error = -EIO; | |
1382 | goto out; | |
1383 | } | |
1384 | ht_offset_cur = ht_offset; | |
1385 | } | |
1386 | ||
71b86f56 SW |
1387 | error = gfs2_dir_read_leaf(inode, offset, opaque, filldir, |
1388 | &copied, &depth, | |
1389 | be64_to_cpu(lp[lp_offset])); | |
b3b94faa | 1390 | if (error) |
71b86f56 | 1391 | break; |
b3b94faa | 1392 | |
71b86f56 | 1393 | len = 1 << (dip->i_di.di_depth - depth); |
b3b94faa DT |
1394 | index = (index & ~(len - 1)) + len; |
1395 | } | |
1396 | ||
71b86f56 | 1397 | out: |
b3b94faa | 1398 | kfree(lp); |
71b86f56 SW |
1399 | if (error > 0) |
1400 | error = 0; | |
b3b94faa DT |
1401 | return error; |
1402 | } | |
1403 | ||
cd915493 | 1404 | int gfs2_dir_read(struct inode *inode, u64 *offset, void *opaque, |
71b86f56 | 1405 | gfs2_filldir_t filldir) |
b3b94faa | 1406 | { |
feaa7bba | 1407 | struct gfs2_inode *dip = GFS2_I(inode); |
71b86f56 SW |
1408 | struct dirent_gather g; |
1409 | const struct gfs2_dirent **darr, *dent; | |
b3b94faa DT |
1410 | struct buffer_head *dibh; |
1411 | int copied = 0; | |
1412 | int error; | |
1413 | ||
71b86f56 SW |
1414 | if (!dip->i_di.di_entries) |
1415 | return 0; | |
1416 | ||
1417 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) | |
1418 | return dir_e_read(inode, offset, opaque, filldir); | |
1419 | ||
b3b94faa DT |
1420 | if (!gfs2_is_stuffed(dip)) { |
1421 | gfs2_consist_inode(dip); | |
1422 | return -EIO; | |
1423 | } | |
1424 | ||
b3b94faa DT |
1425 | error = gfs2_meta_inode_buffer(dip, &dibh); |
1426 | if (error) | |
1427 | return error; | |
1428 | ||
71b86f56 SW |
1429 | error = -ENOMEM; |
1430 | darr = kmalloc(dip->i_di.di_entries * sizeof(struct gfs2_dirent *), | |
1431 | GFP_KERNEL); | |
1432 | if (darr) { | |
1433 | g.pdent = darr; | |
1434 | g.offset = 0; | |
1435 | dent = gfs2_dirent_scan(inode, dibh->b_data, dibh->b_size, | |
1436 | gfs2_dirent_gather, NULL, &g); | |
1437 | if (IS_ERR(dent)) { | |
1438 | error = PTR_ERR(dent); | |
1439 | goto out; | |
1440 | } | |
1441 | error = do_filldir_main(dip, offset, opaque, filldir, darr, | |
1442 | dip->i_di.di_entries, &copied); | |
1443 | out: | |
1444 | kfree(darr); | |
1445 | } | |
1446 | ||
b3b94faa DT |
1447 | if (error > 0) |
1448 | error = 0; | |
1449 | ||
1450 | brelse(dibh); | |
1451 | ||
1452 | return error; | |
1453 | } | |
1454 | ||
b3b94faa DT |
1455 | /** |
1456 | * gfs2_dir_search - Search a directory | |
1457 | * @dip: The GFS2 inode | |
1458 | * @filename: | |
1459 | * @inode: | |
1460 | * | |
1461 | * This routine searches a directory for a file or another directory. | |
1462 | * Assumes a glock is held on dip. | |
1463 | * | |
1464 | * Returns: errno | |
1465 | */ | |
1466 | ||
c752666c | 1467 | int gfs2_dir_search(struct inode *dir, const struct qstr *name, |
629a21e7 | 1468 | struct gfs2_inum_host *inum, unsigned int *type) |
b3b94faa | 1469 | { |
c752666c SW |
1470 | struct buffer_head *bh; |
1471 | struct gfs2_dirent *dent; | |
1472 | ||
1473 | dent = gfs2_dirent_search(dir, name, gfs2_dirent_find, &bh); | |
1474 | if (dent) { | |
1475 | if (IS_ERR(dent)) | |
1476 | return PTR_ERR(dent); | |
1477 | if (inum) | |
1478 | gfs2_inum_in(inum, (char *)&dent->de_inum); | |
1479 | if (type) | |
1480 | *type = be16_to_cpu(dent->de_type); | |
1481 | brelse(bh); | |
1482 | return 0; | |
1483 | } | |
1484 | return -ENOENT; | |
1485 | } | |
1486 | ||
1487 | static int dir_new_leaf(struct inode *inode, const struct qstr *name) | |
1488 | { | |
1489 | struct buffer_head *bh, *obh; | |
feaa7bba | 1490 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c | 1491 | struct gfs2_leaf *leaf, *oleaf; |
b3b94faa | 1492 | int error; |
c752666c SW |
1493 | u32 index; |
1494 | u64 bn; | |
b3b94faa | 1495 | |
c752666c SW |
1496 | index = name->hash >> (32 - ip->i_di.di_depth); |
1497 | error = get_first_leaf(ip, index, &obh); | |
1498 | if (error) | |
1499 | return error; | |
1500 | do { | |
1501 | oleaf = (struct gfs2_leaf *)obh->b_data; | |
1502 | bn = be64_to_cpu(oleaf->lf_next); | |
1503 | if (!bn) | |
1504 | break; | |
1505 | brelse(obh); | |
1506 | error = get_leaf(ip, bn, &obh); | |
1507 | if (error) | |
1508 | return error; | |
1509 | } while(1); | |
b3b94faa | 1510 | |
c752666c SW |
1511 | gfs2_trans_add_bh(ip->i_gl, obh, 1); |
1512 | ||
1513 | leaf = new_leaf(inode, &bh, be16_to_cpu(oleaf->lf_depth)); | |
1514 | if (!leaf) { | |
1515 | brelse(obh); | |
1516 | return -ENOSPC; | |
1517 | } | |
4d8012b6 | 1518 | oleaf->lf_next = cpu_to_be64(bh->b_blocknr); |
c752666c SW |
1519 | brelse(bh); |
1520 | brelse(obh); | |
1521 | ||
1522 | error = gfs2_meta_inode_buffer(ip, &bh); | |
1523 | if (error) | |
1524 | return error; | |
1525 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
1526 | ip->i_di.di_blocks++; | |
9e2dbdac | 1527 | gfs2_set_inode_blocks(&ip->i_inode); |
539e5d6b | 1528 | gfs2_dinode_out(ip, bh->b_data); |
c752666c SW |
1529 | brelse(bh); |
1530 | return 0; | |
b3b94faa DT |
1531 | } |
1532 | ||
1533 | /** | |
1534 | * gfs2_dir_add - Add new filename into directory | |
1535 | * @dip: The GFS2 inode | |
1536 | * @filename: The new name | |
1537 | * @inode: The inode number of the entry | |
1538 | * @type: The type of the entry | |
1539 | * | |
1540 | * Returns: 0 on success, error code on failure | |
1541 | */ | |
1542 | ||
c752666c | 1543 | int gfs2_dir_add(struct inode *inode, const struct qstr *name, |
629a21e7 | 1544 | const struct gfs2_inum_host *inum, unsigned type) |
b3b94faa | 1545 | { |
feaa7bba | 1546 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c SW |
1547 | struct buffer_head *bh; |
1548 | struct gfs2_dirent *dent; | |
1549 | struct gfs2_leaf *leaf; | |
b3b94faa DT |
1550 | int error; |
1551 | ||
c752666c SW |
1552 | while(1) { |
1553 | dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space, | |
1554 | &bh); | |
1555 | if (dent) { | |
1556 | if (IS_ERR(dent)) | |
1557 | return PTR_ERR(dent); | |
1558 | dent = gfs2_init_dirent(inode, dent, name, bh); | |
1559 | gfs2_inum_out(inum, (char *)&dent->de_inum); | |
1560 | dent->de_type = cpu_to_be16(type); | |
1561 | if (ip->i_di.di_flags & GFS2_DIF_EXHASH) { | |
1562 | leaf = (struct gfs2_leaf *)bh->b_data; | |
1563 | leaf->lf_entries = cpu_to_be16(be16_to_cpu(leaf->lf_entries) + 1); | |
1564 | } | |
1565 | brelse(bh); | |
1566 | error = gfs2_meta_inode_buffer(ip, &bh); | |
1567 | if (error) | |
1568 | break; | |
1569 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
1570 | ip->i_di.di_entries++; | |
1a7b1eed | 1571 | ip->i_inode.i_mtime.tv_sec = ip->i_inode.i_ctime.tv_sec = get_seconds(); |
539e5d6b | 1572 | gfs2_dinode_out(ip, bh->b_data); |
c752666c SW |
1573 | brelse(bh); |
1574 | error = 0; | |
1575 | break; | |
1576 | } | |
1577 | if (!(ip->i_di.di_flags & GFS2_DIF_EXHASH)) { | |
1578 | error = dir_make_exhash(inode); | |
1579 | if (error) | |
1580 | break; | |
1581 | continue; | |
1582 | } | |
1583 | error = dir_split_leaf(inode, name); | |
1584 | if (error == 0) | |
1585 | continue; | |
e90deff5 | 1586 | if (error < 0) |
c752666c SW |
1587 | break; |
1588 | if (ip->i_di.di_depth < GFS2_DIR_MAX_DEPTH) { | |
1589 | error = dir_double_exhash(ip); | |
1590 | if (error) | |
1591 | break; | |
1592 | error = dir_split_leaf(inode, name); | |
e90deff5 | 1593 | if (error < 0) |
c752666c | 1594 | break; |
e90deff5 SW |
1595 | if (error == 0) |
1596 | continue; | |
c752666c SW |
1597 | } |
1598 | error = dir_new_leaf(inode, name); | |
1599 | if (!error) | |
1600 | continue; | |
1601 | error = -ENOSPC; | |
1602 | break; | |
1603 | } | |
b3b94faa DT |
1604 | return error; |
1605 | } | |
1606 | ||
c752666c | 1607 | |
b3b94faa DT |
1608 | /** |
1609 | * gfs2_dir_del - Delete a directory entry | |
1610 | * @dip: The GFS2 inode | |
1611 | * @filename: The filename | |
1612 | * | |
1613 | * Returns: 0 on success, error code on failure | |
1614 | */ | |
1615 | ||
c752666c | 1616 | int gfs2_dir_del(struct gfs2_inode *dip, const struct qstr *name) |
b3b94faa | 1617 | { |
c752666c SW |
1618 | struct gfs2_dirent *dent, *prev = NULL; |
1619 | struct buffer_head *bh; | |
b3b94faa DT |
1620 | int error; |
1621 | ||
c752666c SW |
1622 | /* Returns _either_ the entry (if its first in block) or the |
1623 | previous entry otherwise */ | |
feaa7bba | 1624 | dent = gfs2_dirent_search(&dip->i_inode, name, gfs2_dirent_prev, &bh); |
c752666c SW |
1625 | if (!dent) { |
1626 | gfs2_consist_inode(dip); | |
1627 | return -EIO; | |
1628 | } | |
1629 | if (IS_ERR(dent)) { | |
1630 | gfs2_consist_inode(dip); | |
1631 | return PTR_ERR(dent); | |
1632 | } | |
1633 | /* If not first in block, adjust pointers accordingly */ | |
71b86f56 | 1634 | if (gfs2_dirent_find(dent, name, NULL) == 0) { |
c752666c SW |
1635 | prev = dent; |
1636 | dent = (struct gfs2_dirent *)((char *)dent + be16_to_cpu(prev->de_rec_len)); | |
1637 | } | |
1638 | ||
1639 | dirent_del(dip, bh, prev, dent); | |
1640 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) { | |
1641 | struct gfs2_leaf *leaf = (struct gfs2_leaf *)bh->b_data; | |
1642 | u16 entries = be16_to_cpu(leaf->lf_entries); | |
1643 | if (!entries) | |
1644 | gfs2_consist_inode(dip); | |
1645 | leaf->lf_entries = cpu_to_be16(--entries); | |
c752666c | 1646 | } |
ed386507 | 1647 | brelse(bh); |
c752666c SW |
1648 | |
1649 | error = gfs2_meta_inode_buffer(dip, &bh); | |
1650 | if (error) | |
1651 | return error; | |
1652 | ||
1653 | if (!dip->i_di.di_entries) | |
1654 | gfs2_consist_inode(dip); | |
1655 | gfs2_trans_add_bh(dip->i_gl, bh, 1); | |
1656 | dip->i_di.di_entries--; | |
1a7b1eed | 1657 | dip->i_inode.i_mtime.tv_sec = dip->i_inode.i_ctime.tv_sec = get_seconds(); |
539e5d6b | 1658 | gfs2_dinode_out(dip, bh->b_data); |
c752666c | 1659 | brelse(bh); |
feaa7bba | 1660 | mark_inode_dirty(&dip->i_inode); |
b3b94faa DT |
1661 | |
1662 | return error; | |
1663 | } | |
1664 | ||
b3b94faa DT |
1665 | /** |
1666 | * gfs2_dir_mvino - Change inode number of directory entry | |
1667 | * @dip: The GFS2 inode | |
1668 | * @filename: | |
1669 | * @new_inode: | |
1670 | * | |
1671 | * This routine changes the inode number of a directory entry. It's used | |
1672 | * by rename to change ".." when a directory is moved. | |
1673 | * Assumes a glock is held on dvp. | |
1674 | * | |
1675 | * Returns: errno | |
1676 | */ | |
1677 | ||
c752666c | 1678 | int gfs2_dir_mvino(struct gfs2_inode *dip, const struct qstr *filename, |
629a21e7 | 1679 | struct gfs2_inum_host *inum, unsigned int new_type) |
b3b94faa | 1680 | { |
c752666c SW |
1681 | struct buffer_head *bh; |
1682 | struct gfs2_dirent *dent; | |
b3b94faa DT |
1683 | int error; |
1684 | ||
feaa7bba | 1685 | dent = gfs2_dirent_search(&dip->i_inode, filename, gfs2_dirent_find, &bh); |
c752666c SW |
1686 | if (!dent) { |
1687 | gfs2_consist_inode(dip); | |
1688 | return -EIO; | |
1689 | } | |
1690 | if (IS_ERR(dent)) | |
1691 | return PTR_ERR(dent); | |
b3b94faa | 1692 | |
c752666c SW |
1693 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
1694 | gfs2_inum_out(inum, (char *)&dent->de_inum); | |
1695 | dent->de_type = cpu_to_be16(new_type); | |
1696 | ||
1697 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) { | |
1698 | brelse(bh); | |
1699 | error = gfs2_meta_inode_buffer(dip, &bh); | |
1700 | if (error) | |
1701 | return error; | |
1702 | gfs2_trans_add_bh(dip->i_gl, bh, 1); | |
1703 | } | |
1704 | ||
1a7b1eed | 1705 | dip->i_inode.i_mtime.tv_sec = dip->i_inode.i_ctime.tv_sec = get_seconds(); |
539e5d6b | 1706 | gfs2_dinode_out(dip, bh->b_data); |
c752666c SW |
1707 | brelse(bh); |
1708 | return 0; | |
b3b94faa DT |
1709 | } |
1710 | ||
1711 | /** | |
1712 | * foreach_leaf - call a function for each leaf in a directory | |
1713 | * @dip: the directory | |
1714 | * @lc: the function to call for each each | |
1715 | * @data: private data to pass to it | |
1716 | * | |
1717 | * Returns: errno | |
1718 | */ | |
1719 | ||
1720 | static int foreach_leaf(struct gfs2_inode *dip, leaf_call_t lc, void *data) | |
1721 | { | |
feaa7bba | 1722 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa | 1723 | struct buffer_head *bh; |
c752666c | 1724 | struct gfs2_leaf *leaf; |
cd915493 SW |
1725 | u32 hsize, len; |
1726 | u32 ht_offset, lp_offset, ht_offset_cur = -1; | |
1727 | u32 index = 0; | |
b44b84d7 | 1728 | __be64 *lp; |
cd915493 | 1729 | u64 leaf_no; |
b3b94faa DT |
1730 | int error = 0; |
1731 | ||
1732 | hsize = 1 << dip->i_di.di_depth; | |
cd915493 | 1733 | if (hsize * sizeof(u64) != dip->i_di.di_size) { |
b3b94faa DT |
1734 | gfs2_consist_inode(dip); |
1735 | return -EIO; | |
1736 | } | |
1737 | ||
1738 | lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL); | |
1739 | if (!lp) | |
1740 | return -ENOMEM; | |
1741 | ||
1742 | while (index < hsize) { | |
1743 | lp_offset = index & (sdp->sd_hash_ptrs - 1); | |
1744 | ht_offset = index - lp_offset; | |
1745 | ||
1746 | if (ht_offset_cur != ht_offset) { | |
e13940ba | 1747 | error = gfs2_dir_read_data(dip, (char *)lp, |
b44b84d7 | 1748 | ht_offset * sizeof(__be64), |
7276b3b0 | 1749 | sdp->sd_hash_bsize, 1); |
b3b94faa DT |
1750 | if (error != sdp->sd_hash_bsize) { |
1751 | if (error >= 0) | |
1752 | error = -EIO; | |
1753 | goto out; | |
1754 | } | |
1755 | ht_offset_cur = ht_offset; | |
1756 | } | |
1757 | ||
1758 | leaf_no = be64_to_cpu(lp[lp_offset]); | |
1759 | if (leaf_no) { | |
1760 | error = get_leaf(dip, leaf_no, &bh); | |
1761 | if (error) | |
1762 | goto out; | |
c752666c | 1763 | leaf = (struct gfs2_leaf *)bh->b_data; |
c752666c | 1764 | len = 1 << (dip->i_di.di_depth - be16_to_cpu(leaf->lf_depth)); |
634ee0b9 | 1765 | brelse(bh); |
b3b94faa DT |
1766 | |
1767 | error = lc(dip, index, len, leaf_no, data); | |
1768 | if (error) | |
1769 | goto out; | |
1770 | ||
1771 | index = (index & ~(len - 1)) + len; | |
1772 | } else | |
1773 | index++; | |
1774 | } | |
1775 | ||
1776 | if (index != hsize) { | |
1777 | gfs2_consist_inode(dip); | |
1778 | error = -EIO; | |
1779 | } | |
1780 | ||
634ee0b9 | 1781 | out: |
b3b94faa DT |
1782 | kfree(lp); |
1783 | ||
1784 | return error; | |
1785 | } | |
1786 | ||
1787 | /** | |
1788 | * leaf_dealloc - Deallocate a directory leaf | |
1789 | * @dip: the directory | |
1790 | * @index: the hash table offset in the directory | |
1791 | * @len: the number of pointers to this leaf | |
1792 | * @leaf_no: the leaf number | |
1793 | * @data: not used | |
1794 | * | |
1795 | * Returns: errno | |
1796 | */ | |
1797 | ||
cd915493 SW |
1798 | static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len, |
1799 | u64 leaf_no, void *data) | |
b3b94faa | 1800 | { |
feaa7bba | 1801 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
c752666c | 1802 | struct gfs2_leaf *tmp_leaf; |
b3b94faa DT |
1803 | struct gfs2_rgrp_list rlist; |
1804 | struct buffer_head *bh, *dibh; | |
cd915493 | 1805 | u64 blk, nblk; |
b3b94faa DT |
1806 | unsigned int rg_blocks = 0, l_blocks = 0; |
1807 | char *ht; | |
cd915493 | 1808 | unsigned int x, size = len * sizeof(u64); |
b3b94faa DT |
1809 | int error; |
1810 | ||
1811 | memset(&rlist, 0, sizeof(struct gfs2_rgrp_list)); | |
1812 | ||
1813 | ht = kzalloc(size, GFP_KERNEL); | |
1814 | if (!ht) | |
1815 | return -ENOMEM; | |
1816 | ||
1817 | gfs2_alloc_get(dip); | |
1818 | ||
1819 | error = gfs2_quota_hold(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
1820 | if (error) | |
1821 | goto out; | |
1822 | ||
1823 | error = gfs2_rindex_hold(sdp, &dip->i_alloc.al_ri_gh); | |
1824 | if (error) | |
1825 | goto out_qs; | |
1826 | ||
1827 | /* Count the number of leaves */ | |
1828 | ||
c752666c | 1829 | for (blk = leaf_no; blk; blk = nblk) { |
b3b94faa DT |
1830 | error = get_leaf(dip, blk, &bh); |
1831 | if (error) | |
1832 | goto out_rlist; | |
c752666c SW |
1833 | tmp_leaf = (struct gfs2_leaf *)bh->b_data; |
1834 | nblk = be64_to_cpu(tmp_leaf->lf_next); | |
b3b94faa DT |
1835 | brelse(bh); |
1836 | ||
1837 | gfs2_rlist_add(sdp, &rlist, blk); | |
1838 | l_blocks++; | |
1839 | } | |
1840 | ||
1841 | gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE, 0); | |
1842 | ||
1843 | for (x = 0; x < rlist.rl_rgrps; x++) { | |
1844 | struct gfs2_rgrpd *rgd; | |
5c676f6d | 1845 | rgd = rlist.rl_ghs[x].gh_gl->gl_object; |
b3b94faa DT |
1846 | rg_blocks += rgd->rd_ri.ri_length; |
1847 | } | |
1848 | ||
1849 | error = gfs2_glock_nq_m(rlist.rl_rgrps, rlist.rl_ghs); | |
1850 | if (error) | |
1851 | goto out_rlist; | |
1852 | ||
1853 | error = gfs2_trans_begin(sdp, | |
5c676f6d | 1854 | rg_blocks + (DIV_ROUND_UP(size, sdp->sd_jbsize) + 1) + |
b3b94faa DT |
1855 | RES_DINODE + RES_STATFS + RES_QUOTA, l_blocks); |
1856 | if (error) | |
1857 | goto out_rg_gunlock; | |
1858 | ||
c752666c | 1859 | for (blk = leaf_no; blk; blk = nblk) { |
b3b94faa DT |
1860 | error = get_leaf(dip, blk, &bh); |
1861 | if (error) | |
1862 | goto out_end_trans; | |
c752666c SW |
1863 | tmp_leaf = (struct gfs2_leaf *)bh->b_data; |
1864 | nblk = be64_to_cpu(tmp_leaf->lf_next); | |
b3b94faa DT |
1865 | brelse(bh); |
1866 | ||
1867 | gfs2_free_meta(dip, blk, 1); | |
1868 | ||
1869 | if (!dip->i_di.di_blocks) | |
1870 | gfs2_consist_inode(dip); | |
1871 | dip->i_di.di_blocks--; | |
9e2dbdac | 1872 | gfs2_set_inode_blocks(&dip->i_inode); |
b3b94faa DT |
1873 | } |
1874 | ||
cd915493 | 1875 | error = gfs2_dir_write_data(dip, ht, index * sizeof(u64), size); |
b3b94faa DT |
1876 | if (error != size) { |
1877 | if (error >= 0) | |
1878 | error = -EIO; | |
1879 | goto out_end_trans; | |
1880 | } | |
1881 | ||
1882 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1883 | if (error) | |
1884 | goto out_end_trans; | |
1885 | ||
d4e9c4c3 | 1886 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
539e5d6b | 1887 | gfs2_dinode_out(dip, dibh->b_data); |
b3b94faa DT |
1888 | brelse(dibh); |
1889 | ||
a91ea69f | 1890 | out_end_trans: |
b3b94faa | 1891 | gfs2_trans_end(sdp); |
a91ea69f | 1892 | out_rg_gunlock: |
b3b94faa | 1893 | gfs2_glock_dq_m(rlist.rl_rgrps, rlist.rl_ghs); |
a91ea69f | 1894 | out_rlist: |
b3b94faa DT |
1895 | gfs2_rlist_free(&rlist); |
1896 | gfs2_glock_dq_uninit(&dip->i_alloc.al_ri_gh); | |
a91ea69f | 1897 | out_qs: |
b3b94faa | 1898 | gfs2_quota_unhold(dip); |
a91ea69f | 1899 | out: |
b3b94faa DT |
1900 | gfs2_alloc_put(dip); |
1901 | kfree(ht); | |
b3b94faa DT |
1902 | return error; |
1903 | } | |
1904 | ||
1905 | /** | |
1906 | * gfs2_dir_exhash_dealloc - free all the leaf blocks in a directory | |
1907 | * @dip: the directory | |
1908 | * | |
1909 | * Dealloc all on-disk directory leaves to FREEMETA state | |
1910 | * Change on-disk inode type to "regular file" | |
1911 | * | |
1912 | * Returns: errno | |
1913 | */ | |
1914 | ||
1915 | int gfs2_dir_exhash_dealloc(struct gfs2_inode *dip) | |
1916 | { | |
feaa7bba | 1917 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa DT |
1918 | struct buffer_head *bh; |
1919 | int error; | |
1920 | ||
1921 | /* Dealloc on-disk leaves to FREEMETA state */ | |
1922 | error = foreach_leaf(dip, leaf_dealloc, NULL); | |
1923 | if (error) | |
1924 | return error; | |
1925 | ||
1926 | /* Make this a regular file in case we crash. | |
1927 | (We don't want to free these blocks a second time.) */ | |
1928 | ||
1929 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
1930 | if (error) | |
1931 | return error; | |
1932 | ||
1933 | error = gfs2_meta_inode_buffer(dip, &bh); | |
1934 | if (!error) { | |
d4e9c4c3 | 1935 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
568f4c96 SW |
1936 | ((struct gfs2_dinode *)bh->b_data)->di_mode = |
1937 | cpu_to_be32(S_IFREG); | |
b3b94faa DT |
1938 | brelse(bh); |
1939 | } | |
1940 | ||
1941 | gfs2_trans_end(sdp); | |
1942 | ||
1943 | return error; | |
1944 | } | |
1945 | ||
1946 | /** | |
1947 | * gfs2_diradd_alloc_required - find if adding entry will require an allocation | |
1948 | * @ip: the file being written to | |
1949 | * @filname: the filename that's going to be added | |
b3b94faa | 1950 | * |
c752666c | 1951 | * Returns: 1 if alloc required, 0 if not, -ve on error |
b3b94faa DT |
1952 | */ |
1953 | ||
4d8012b6 | 1954 | int gfs2_diradd_alloc_required(struct inode *inode, const struct qstr *name) |
b3b94faa | 1955 | { |
c752666c SW |
1956 | struct gfs2_dirent *dent; |
1957 | struct buffer_head *bh; | |
b3b94faa | 1958 | |
c752666c | 1959 | dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space, &bh); |
ed386507 | 1960 | if (!dent) { |
c752666c | 1961 | return 1; |
ed386507 | 1962 | } |
c752666c SW |
1963 | if (IS_ERR(dent)) |
1964 | return PTR_ERR(dent); | |
1965 | brelse(bh); | |
1966 | return 0; | |
b3b94faa DT |
1967 | } |
1968 |