]>
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 | * | |
383f01fb | 39 | * dip->i_diskflags & GFS2_DIF_EXHASH is true |
61e085a8 SW |
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 | 55 | |
b3b94faa DT |
56 | #include <linux/slab.h> |
57 | #include <linux/spinlock.h> | |
b3b94faa DT |
58 | #include <linux/buffer_head.h> |
59 | #include <linux/sort.h> | |
5c676f6d | 60 | #include <linux/gfs2_ondisk.h> |
71b86f56 | 61 | #include <linux/crc32.h> |
fe1bdedc | 62 | #include <linux/vmalloc.h> |
b3b94faa DT |
63 | |
64 | #include "gfs2.h" | |
5c676f6d | 65 | #include "incore.h" |
b3b94faa DT |
66 | #include "dir.h" |
67 | #include "glock.h" | |
68 | #include "inode.h" | |
b3b94faa DT |
69 | #include "meta_io.h" |
70 | #include "quota.h" | |
71 | #include "rgrp.h" | |
72 | #include "trans.h" | |
e13940ba | 73 | #include "bmap.h" |
5c676f6d | 74 | #include "util.h" |
b3b94faa DT |
75 | |
76 | #define IS_LEAF 1 /* Hashed (leaf) directory */ | |
77 | #define IS_DINODE 2 /* Linear (stuffed dinode block) directory */ | |
78 | ||
cd915493 SW |
79 | #define gfs2_disk_hash2offset(h) (((u64)(h)) >> 1) |
80 | #define gfs2_dir_offset2hash(p) ((u32)(((u64)(p)) << 1)) | |
b3b94faa | 81 | |
8d123585 SW |
82 | struct qstr gfs2_qdot __read_mostly; |
83 | struct qstr gfs2_qdotdot __read_mostly; | |
84 | ||
2bdbc5d7 SW |
85 | typedef int (*gfs2_dscan_t)(const struct gfs2_dirent *dent, |
86 | const struct qstr *name, void *opaque); | |
b3b94faa | 87 | |
0d95326d | 88 | static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len, |
d24a7a43 | 89 | u64 leaf_no, int last_dealloc); |
61e085a8 | 90 | |
cd915493 | 91 | int gfs2_dir_get_new_buffer(struct gfs2_inode *ip, u64 block, |
61e085a8 | 92 | struct buffer_head **bhp) |
e13940ba SW |
93 | { |
94 | struct buffer_head *bh; | |
e13940ba | 95 | |
61e085a8 SW |
96 | bh = gfs2_meta_new(ip->i_gl, block); |
97 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
98 | gfs2_metatype_set(bh, GFS2_METATYPE_JD, GFS2_FORMAT_JD); | |
99 | gfs2_buffer_clear_tail(bh, sizeof(struct gfs2_meta_header)); | |
e13940ba SW |
100 | *bhp = bh; |
101 | return 0; | |
102 | } | |
103 | ||
cd915493 | 104 | static int gfs2_dir_get_existing_buffer(struct gfs2_inode *ip, u64 block, |
61e085a8 SW |
105 | struct buffer_head **bhp) |
106 | { | |
107 | struct buffer_head *bh; | |
108 | int error; | |
e13940ba | 109 | |
7276b3b0 | 110 | error = gfs2_meta_read(ip->i_gl, block, DIO_WAIT, &bh); |
61e085a8 SW |
111 | if (error) |
112 | return error; | |
feaa7bba | 113 | if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), bh, GFS2_METATYPE_JD)) { |
61e085a8 SW |
114 | brelse(bh); |
115 | return -EIO; | |
116 | } | |
117 | *bhp = bh; | |
118 | return 0; | |
119 | } | |
e13940ba SW |
120 | |
121 | static int gfs2_dir_write_stuffed(struct gfs2_inode *ip, const char *buf, | |
122 | unsigned int offset, unsigned int size) | |
e13940ba SW |
123 | { |
124 | struct buffer_head *dibh; | |
125 | int error; | |
126 | ||
127 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
128 | if (error) | |
129 | return error; | |
130 | ||
131 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
c752666c | 132 | memcpy(dibh->b_data + offset + sizeof(struct gfs2_dinode), buf, size); |
a2e0f799 SW |
133 | if (ip->i_inode.i_size < offset + size) |
134 | i_size_write(&ip->i_inode, offset + size); | |
4bd91ba1 | 135 | ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME; |
539e5d6b | 136 | gfs2_dinode_out(ip, dibh->b_data); |
e13940ba SW |
137 | |
138 | brelse(dibh); | |
139 | ||
140 | return size; | |
141 | } | |
142 | ||
143 | ||
144 | ||
145 | /** | |
146 | * gfs2_dir_write_data - Write directory information to the inode | |
147 | * @ip: The GFS2 inode | |
148 | * @buf: The buffer containing information to be written | |
149 | * @offset: The file offset to start writing at | |
150 | * @size: The amount of data to write | |
151 | * | |
152 | * Returns: The number of bytes correctly written or error code | |
153 | */ | |
154 | static int gfs2_dir_write_data(struct gfs2_inode *ip, const char *buf, | |
cd915493 | 155 | u64 offset, unsigned int size) |
e13940ba | 156 | { |
feaa7bba | 157 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
e13940ba | 158 | struct buffer_head *dibh; |
cd915493 SW |
159 | u64 lblock, dblock; |
160 | u32 extlen = 0; | |
e13940ba SW |
161 | unsigned int o; |
162 | int copied = 0; | |
163 | int error = 0; | |
9b8c81d1 | 164 | int new = 0; |
e13940ba SW |
165 | |
166 | if (!size) | |
167 | return 0; | |
168 | ||
169 | if (gfs2_is_stuffed(ip) && | |
170 | offset + size <= sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode)) | |
568f4c96 SW |
171 | return gfs2_dir_write_stuffed(ip, buf, (unsigned int)offset, |
172 | size); | |
e13940ba SW |
173 | |
174 | if (gfs2_assert_warn(sdp, gfs2_is_jdata(ip))) | |
175 | return -EINVAL; | |
176 | ||
177 | if (gfs2_is_stuffed(ip)) { | |
f25ef0c1 | 178 | error = gfs2_unstuff_dinode(ip, NULL); |
e13940ba | 179 | if (error) |
c752666c | 180 | return error; |
e13940ba SW |
181 | } |
182 | ||
183 | lblock = offset; | |
184 | o = do_div(lblock, sdp->sd_jbsize) + sizeof(struct gfs2_meta_header); | |
185 | ||
186 | while (copied < size) { | |
187 | unsigned int amount; | |
188 | struct buffer_head *bh; | |
e13940ba SW |
189 | |
190 | amount = size - copied; | |
191 | if (amount > sdp->sd_sb.sb_bsize - o) | |
192 | amount = sdp->sd_sb.sb_bsize - o; | |
193 | ||
194 | if (!extlen) { | |
195 | new = 1; | |
feaa7bba | 196 | error = gfs2_extent_map(&ip->i_inode, lblock, &new, |
fd88de56 | 197 | &dblock, &extlen); |
e13940ba SW |
198 | if (error) |
199 | goto fail; | |
200 | error = -EIO; | |
201 | if (gfs2_assert_withdraw(sdp, dblock)) | |
202 | goto fail; | |
203 | } | |
204 | ||
61e085a8 SW |
205 | if (amount == sdp->sd_jbsize || new) |
206 | error = gfs2_dir_get_new_buffer(ip, dblock, &bh); | |
207 | else | |
208 | error = gfs2_dir_get_existing_buffer(ip, dblock, &bh); | |
209 | ||
e13940ba SW |
210 | if (error) |
211 | goto fail; | |
212 | ||
213 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
214 | memcpy(bh->b_data + o, buf, amount); | |
215 | brelse(bh); | |
e13940ba | 216 | |
899bb264 | 217 | buf += amount; |
e13940ba SW |
218 | copied += amount; |
219 | lblock++; | |
220 | dblock++; | |
221 | extlen--; | |
222 | ||
223 | o = sizeof(struct gfs2_meta_header); | |
224 | } | |
225 | ||
226 | out: | |
227 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
228 | if (error) | |
229 | return error; | |
230 | ||
a2e0f799 SW |
231 | if (ip->i_inode.i_size < offset + copied) |
232 | i_size_write(&ip->i_inode, offset + copied); | |
4bd91ba1 | 233 | ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME; |
e13940ba SW |
234 | |
235 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
539e5d6b | 236 | gfs2_dinode_out(ip, dibh->b_data); |
e13940ba SW |
237 | brelse(dibh); |
238 | ||
239 | return copied; | |
240 | fail: | |
241 | if (copied) | |
242 | goto out; | |
243 | return error; | |
244 | } | |
245 | ||
246 | static int gfs2_dir_read_stuffed(struct gfs2_inode *ip, char *buf, | |
7276b3b0 | 247 | u64 offset, unsigned int size) |
e13940ba SW |
248 | { |
249 | struct buffer_head *dibh; | |
250 | int error; | |
251 | ||
252 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
253 | if (!error) { | |
254 | offset += sizeof(struct gfs2_dinode); | |
255 | memcpy(buf, dibh->b_data + offset, size); | |
256 | brelse(dibh); | |
257 | } | |
258 | ||
259 | return (error) ? error : size; | |
260 | } | |
261 | ||
262 | ||
263 | /** | |
264 | * gfs2_dir_read_data - Read a data from a directory inode | |
265 | * @ip: The GFS2 Inode | |
266 | * @buf: The buffer to place result into | |
267 | * @offset: File offset to begin jdata_readng from | |
268 | * @size: Amount of data to transfer | |
269 | * | |
270 | * Returns: The amount of data actually copied or the error | |
271 | */ | |
7276b3b0 SW |
272 | static int gfs2_dir_read_data(struct gfs2_inode *ip, char *buf, u64 offset, |
273 | unsigned int size, unsigned ra) | |
e13940ba | 274 | { |
feaa7bba | 275 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
cd915493 SW |
276 | u64 lblock, dblock; |
277 | u32 extlen = 0; | |
e13940ba SW |
278 | unsigned int o; |
279 | int copied = 0; | |
280 | int error = 0; | |
a2e0f799 | 281 | u64 disksize = i_size_read(&ip->i_inode); |
e13940ba | 282 | |
a2e0f799 | 283 | if (offset >= disksize) |
e13940ba SW |
284 | return 0; |
285 | ||
a2e0f799 SW |
286 | if (offset + size > disksize) |
287 | size = disksize - offset; | |
e13940ba SW |
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); | |
b7d8ac3e | 320 | } else { |
7276b3b0 | 321 | error = gfs2_meta_read(ip->i_gl, dblock, DIO_WAIT, &bh); |
e13940ba SW |
322 | if (error) |
323 | goto fail; | |
7276b3b0 SW |
324 | } |
325 | error = gfs2_metatype_check(sdp, bh, GFS2_METATYPE_JD); | |
326 | if (error) { | |
327 | brelse(bh); | |
328 | goto fail; | |
329 | } | |
330 | dblock++; | |
331 | extlen--; | |
e13940ba SW |
332 | memcpy(buf, bh->b_data + o, amount); |
333 | brelse(bh); | |
899bb264 | 334 | buf += amount; |
e13940ba SW |
335 | copied += amount; |
336 | lblock++; | |
e13940ba SW |
337 | o = sizeof(struct gfs2_meta_header); |
338 | } | |
339 | ||
340 | return copied; | |
341 | fail: | |
342 | return (copied) ? copied : error; | |
343 | } | |
344 | ||
5e7d65cd SW |
345 | static inline int gfs2_dirent_sentinel(const struct gfs2_dirent *dent) |
346 | { | |
347 | return dent->de_inum.no_addr == 0 || dent->de_inum.no_formal_ino == 0; | |
348 | } | |
349 | ||
c752666c SW |
350 | static inline int __gfs2_dirent_find(const struct gfs2_dirent *dent, |
351 | const struct qstr *name, int ret) | |
352 | { | |
5e7d65cd | 353 | if (!gfs2_dirent_sentinel(dent) && |
c752666c SW |
354 | be32_to_cpu(dent->de_hash) == name->hash && |
355 | be16_to_cpu(dent->de_name_len) == name->len && | |
2bdbc5d7 | 356 | memcmp(dent+1, name->name, name->len) == 0) |
c752666c SW |
357 | return ret; |
358 | return 0; | |
359 | } | |
360 | ||
361 | static int gfs2_dirent_find(const struct gfs2_dirent *dent, | |
71b86f56 SW |
362 | const struct qstr *name, |
363 | void *opaque) | |
c752666c SW |
364 | { |
365 | return __gfs2_dirent_find(dent, name, 1); | |
366 | } | |
367 | ||
368 | static int gfs2_dirent_prev(const struct gfs2_dirent *dent, | |
71b86f56 SW |
369 | const struct qstr *name, |
370 | void *opaque) | |
c752666c SW |
371 | { |
372 | return __gfs2_dirent_find(dent, name, 2); | |
373 | } | |
374 | ||
375 | /* | |
376 | * name->name holds ptr to start of block. | |
377 | * name->len holds size of block. | |
b3b94faa | 378 | */ |
c752666c | 379 | static int gfs2_dirent_last(const struct gfs2_dirent *dent, |
71b86f56 SW |
380 | const struct qstr *name, |
381 | void *opaque) | |
c752666c SW |
382 | { |
383 | const char *start = name->name; | |
384 | const char *end = (const char *)dent + be16_to_cpu(dent->de_rec_len); | |
385 | if (name->len == (end - start)) | |
386 | return 1; | |
387 | return 0; | |
388 | } | |
b3b94faa | 389 | |
c752666c | 390 | static int gfs2_dirent_find_space(const struct gfs2_dirent *dent, |
71b86f56 SW |
391 | const struct qstr *name, |
392 | void *opaque) | |
b3b94faa | 393 | { |
c752666c SW |
394 | unsigned required = GFS2_DIRENT_SIZE(name->len); |
395 | unsigned actual = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len)); | |
396 | unsigned totlen = be16_to_cpu(dent->de_rec_len); | |
397 | ||
5e7d65cd | 398 | if (gfs2_dirent_sentinel(dent)) |
728a756b | 399 | actual = 0; |
c5392124 | 400 | if (totlen - actual >= required) |
c752666c SW |
401 | return 1; |
402 | return 0; | |
403 | } | |
404 | ||
71b86f56 SW |
405 | struct dirent_gather { |
406 | const struct gfs2_dirent **pdent; | |
407 | unsigned offset; | |
408 | }; | |
409 | ||
410 | static int gfs2_dirent_gather(const struct gfs2_dirent *dent, | |
411 | const struct qstr *name, | |
412 | void *opaque) | |
413 | { | |
414 | struct dirent_gather *g = opaque; | |
5e7d65cd | 415 | if (!gfs2_dirent_sentinel(dent)) { |
71b86f56 SW |
416 | g->pdent[g->offset++] = dent; |
417 | } | |
418 | return 0; | |
419 | } | |
420 | ||
c752666c SW |
421 | /* |
422 | * Other possible things to check: | |
423 | * - Inode located within filesystem size (and on valid block) | |
424 | * - Valid directory entry type | |
425 | * Not sure how heavy-weight we want to make this... could also check | |
426 | * hash is correct for example, but that would take a lot of extra time. | |
427 | * For now the most important thing is to check that the various sizes | |
428 | * are correct. | |
429 | */ | |
430 | static int gfs2_check_dirent(struct gfs2_dirent *dent, unsigned int offset, | |
431 | unsigned int size, unsigned int len, int first) | |
432 | { | |
433 | const char *msg = "gfs2_dirent too small"; | |
434 | if (unlikely(size < sizeof(struct gfs2_dirent))) | |
435 | goto error; | |
436 | msg = "gfs2_dirent misaligned"; | |
437 | if (unlikely(offset & 0x7)) | |
438 | goto error; | |
439 | msg = "gfs2_dirent points beyond end of block"; | |
440 | if (unlikely(offset + size > len)) | |
441 | goto error; | |
442 | msg = "zero inode number"; | |
5e7d65cd | 443 | if (unlikely(!first && gfs2_dirent_sentinel(dent))) |
c752666c SW |
444 | goto error; |
445 | msg = "name length is greater than space in dirent"; | |
5e7d65cd | 446 | if (!gfs2_dirent_sentinel(dent) && |
c752666c SW |
447 | unlikely(sizeof(struct gfs2_dirent)+be16_to_cpu(dent->de_name_len) > |
448 | size)) | |
449 | goto error; | |
450 | return 0; | |
451 | error: | |
452 | printk(KERN_WARNING "gfs2_check_dirent: %s (%s)\n", msg, | |
453 | first ? "first in block" : "not first in block"); | |
454 | return -EIO; | |
b3b94faa DT |
455 | } |
456 | ||
71b86f56 | 457 | static int gfs2_dirent_offset(const void *buf) |
c752666c | 458 | { |
71b86f56 SW |
459 | const struct gfs2_meta_header *h = buf; |
460 | int offset; | |
c752666c SW |
461 | |
462 | BUG_ON(buf == NULL); | |
c752666c | 463 | |
e3167ded | 464 | switch(be32_to_cpu(h->mh_type)) { |
c752666c SW |
465 | case GFS2_METATYPE_LF: |
466 | offset = sizeof(struct gfs2_leaf); | |
467 | break; | |
468 | case GFS2_METATYPE_DI: | |
469 | offset = sizeof(struct gfs2_dinode); | |
470 | break; | |
471 | default: | |
472 | goto wrong_type; | |
473 | } | |
71b86f56 SW |
474 | return offset; |
475 | wrong_type: | |
476 | printk(KERN_WARNING "gfs2_scan_dirent: wrong block type %u\n", | |
e3167ded | 477 | be32_to_cpu(h->mh_type)); |
71b86f56 SW |
478 | return -1; |
479 | } | |
480 | ||
2bdbc5d7 | 481 | static struct gfs2_dirent *gfs2_dirent_scan(struct inode *inode, void *buf, |
71b86f56 SW |
482 | unsigned int len, gfs2_dscan_t scan, |
483 | const struct qstr *name, | |
484 | void *opaque) | |
485 | { | |
486 | struct gfs2_dirent *dent, *prev; | |
487 | unsigned offset; | |
488 | unsigned size; | |
489 | int ret = 0; | |
c752666c | 490 | |
71b86f56 SW |
491 | ret = gfs2_dirent_offset(buf); |
492 | if (ret < 0) | |
493 | goto consist_inode; | |
494 | ||
495 | offset = ret; | |
c752666c | 496 | prev = NULL; |
2bdbc5d7 | 497 | dent = buf + offset; |
c752666c SW |
498 | size = be16_to_cpu(dent->de_rec_len); |
499 | if (gfs2_check_dirent(dent, offset, size, len, 1)) | |
500 | goto consist_inode; | |
501 | do { | |
71b86f56 | 502 | ret = scan(dent, name, opaque); |
c752666c SW |
503 | if (ret) |
504 | break; | |
505 | offset += size; | |
506 | if (offset == len) | |
507 | break; | |
508 | prev = dent; | |
2bdbc5d7 | 509 | dent = buf + offset; |
c752666c SW |
510 | size = be16_to_cpu(dent->de_rec_len); |
511 | if (gfs2_check_dirent(dent, offset, size, len, 0)) | |
512 | goto consist_inode; | |
513 | } while(1); | |
514 | ||
515 | switch(ret) { | |
516 | case 0: | |
517 | return NULL; | |
518 | case 1: | |
519 | return dent; | |
520 | case 2: | |
521 | return prev ? prev : dent; | |
522 | default: | |
523 | BUG_ON(ret > 0); | |
524 | return ERR_PTR(ret); | |
525 | } | |
526 | ||
c752666c | 527 | consist_inode: |
feaa7bba | 528 | gfs2_consist_inode(GFS2_I(inode)); |
c752666c SW |
529 | return ERR_PTR(-EIO); |
530 | } | |
531 | ||
2bdbc5d7 SW |
532 | static int dirent_check_reclen(struct gfs2_inode *dip, |
533 | const struct gfs2_dirent *d, const void *end_p) | |
534 | { | |
535 | const void *ptr = d; | |
536 | u16 rec_len = be16_to_cpu(d->de_rec_len); | |
537 | ||
538 | if (unlikely(rec_len < sizeof(struct gfs2_dirent))) | |
539 | goto broken; | |
540 | ptr += rec_len; | |
541 | if (ptr < end_p) | |
542 | return rec_len; | |
543 | if (ptr == end_p) | |
544 | return -ENOENT; | |
545 | broken: | |
546 | gfs2_consist_inode(dip); | |
547 | return -EIO; | |
548 | } | |
549 | ||
b3b94faa DT |
550 | /** |
551 | * dirent_next - Next dirent | |
552 | * @dip: the directory | |
553 | * @bh: The buffer | |
554 | * @dent: Pointer to list of dirents | |
555 | * | |
556 | * Returns: 0 on success, error code otherwise | |
557 | */ | |
558 | ||
559 | static int dirent_next(struct gfs2_inode *dip, struct buffer_head *bh, | |
560 | struct gfs2_dirent **dent) | |
561 | { | |
2bdbc5d7 SW |
562 | struct gfs2_dirent *cur = *dent, *tmp; |
563 | char *bh_end = bh->b_data + bh->b_size; | |
564 | int ret; | |
b3b94faa | 565 | |
2bdbc5d7 SW |
566 | ret = dirent_check_reclen(dip, cur, bh_end); |
567 | if (ret < 0) | |
568 | return ret; | |
4dd651ad | 569 | |
2bdbc5d7 SW |
570 | tmp = (void *)cur + ret; |
571 | ret = dirent_check_reclen(dip, tmp, bh_end); | |
572 | if (ret == -EIO) | |
573 | return ret; | |
4dd651ad | 574 | |
b3b94faa | 575 | /* Only the first dent could ever have de_inum.no_addr == 0 */ |
5e7d65cd | 576 | if (gfs2_dirent_sentinel(tmp)) { |
b3b94faa DT |
577 | gfs2_consist_inode(dip); |
578 | return -EIO; | |
579 | } | |
580 | ||
581 | *dent = tmp; | |
b3b94faa DT |
582 | return 0; |
583 | } | |
584 | ||
585 | /** | |
586 | * dirent_del - Delete a dirent | |
587 | * @dip: The GFS2 inode | |
588 | * @bh: The buffer | |
589 | * @prev: The previous dirent | |
590 | * @cur: The current dirent | |
591 | * | |
592 | */ | |
593 | ||
594 | static void dirent_del(struct gfs2_inode *dip, struct buffer_head *bh, | |
595 | struct gfs2_dirent *prev, struct gfs2_dirent *cur) | |
596 | { | |
cd915493 | 597 | u16 cur_rec_len, prev_rec_len; |
b3b94faa | 598 | |
5e7d65cd | 599 | if (gfs2_dirent_sentinel(cur)) { |
b3b94faa DT |
600 | gfs2_consist_inode(dip); |
601 | return; | |
602 | } | |
603 | ||
d4e9c4c3 | 604 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
b3b94faa DT |
605 | |
606 | /* If there is no prev entry, this is the first entry in the block. | |
607 | The de_rec_len is already as big as it needs to be. Just zero | |
608 | out the inode number and return. */ | |
609 | ||
610 | if (!prev) { | |
5e7d65cd SW |
611 | cur->de_inum.no_addr = 0; |
612 | cur->de_inum.no_formal_ino = 0; | |
b3b94faa DT |
613 | return; |
614 | } | |
615 | ||
616 | /* Combine this dentry with the previous one. */ | |
617 | ||
fc69d0d3 SW |
618 | prev_rec_len = be16_to_cpu(prev->de_rec_len); |
619 | cur_rec_len = be16_to_cpu(cur->de_rec_len); | |
b3b94faa DT |
620 | |
621 | if ((char *)prev + prev_rec_len != (char *)cur) | |
622 | gfs2_consist_inode(dip); | |
623 | if ((char *)cur + cur_rec_len > bh->b_data + bh->b_size) | |
624 | gfs2_consist_inode(dip); | |
625 | ||
626 | prev_rec_len += cur_rec_len; | |
fc69d0d3 | 627 | prev->de_rec_len = cpu_to_be16(prev_rec_len); |
b3b94faa DT |
628 | } |
629 | ||
c752666c SW |
630 | /* |
631 | * Takes a dent from which to grab space as an argument. Returns the | |
632 | * newly created dent. | |
b3b94faa | 633 | */ |
08bc2dbc AB |
634 | static struct gfs2_dirent *gfs2_init_dirent(struct inode *inode, |
635 | struct gfs2_dirent *dent, | |
636 | const struct qstr *name, | |
637 | struct buffer_head *bh) | |
b3b94faa | 638 | { |
feaa7bba | 639 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c SW |
640 | struct gfs2_dirent *ndent; |
641 | unsigned offset = 0, totlen; | |
642 | ||
5e7d65cd | 643 | if (!gfs2_dirent_sentinel(dent)) |
c752666c SW |
644 | offset = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len)); |
645 | totlen = be16_to_cpu(dent->de_rec_len); | |
646 | BUG_ON(offset + name->len > totlen); | |
647 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
648 | ndent = (struct gfs2_dirent *)((char *)dent + offset); | |
649 | dent->de_rec_len = cpu_to_be16(offset); | |
650 | gfs2_qstr2dirent(name, totlen - offset, ndent); | |
651 | return ndent; | |
b3b94faa DT |
652 | } |
653 | ||
c752666c SW |
654 | static struct gfs2_dirent *gfs2_dirent_alloc(struct inode *inode, |
655 | struct buffer_head *bh, | |
656 | const struct qstr *name) | |
b3b94faa DT |
657 | { |
658 | struct gfs2_dirent *dent; | |
907b9bce | 659 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, |
71b86f56 | 660 | gfs2_dirent_find_space, name, NULL); |
c752666c SW |
661 | if (!dent || IS_ERR(dent)) |
662 | return dent; | |
663 | return gfs2_init_dirent(inode, dent, name, bh); | |
b3b94faa DT |
664 | } |
665 | ||
cd915493 | 666 | static int get_leaf(struct gfs2_inode *dip, u64 leaf_no, |
b3b94faa DT |
667 | struct buffer_head **bhp) |
668 | { | |
669 | int error; | |
670 | ||
7276b3b0 | 671 | error = gfs2_meta_read(dip->i_gl, leaf_no, DIO_WAIT, bhp); |
feaa7bba SW |
672 | if (!error && gfs2_metatype_check(GFS2_SB(&dip->i_inode), *bhp, GFS2_METATYPE_LF)) { |
673 | /* printk(KERN_INFO "block num=%llu\n", leaf_no); */ | |
b3b94faa | 674 | error = -EIO; |
feaa7bba | 675 | } |
b3b94faa DT |
676 | |
677 | return error; | |
678 | } | |
679 | ||
680 | /** | |
681 | * get_leaf_nr - Get a leaf number associated with the index | |
682 | * @dip: The GFS2 inode | |
683 | * @index: | |
684 | * @leaf_out: | |
685 | * | |
686 | * Returns: 0 on success, error code otherwise | |
687 | */ | |
688 | ||
cd915493 SW |
689 | static int get_leaf_nr(struct gfs2_inode *dip, u32 index, |
690 | u64 *leaf_out) | |
b3b94faa | 691 | { |
b44b84d7 | 692 | __be64 leaf_no; |
b3b94faa DT |
693 | int error; |
694 | ||
e13940ba | 695 | error = gfs2_dir_read_data(dip, (char *)&leaf_no, |
b44b84d7 AV |
696 | index * sizeof(__be64), |
697 | sizeof(__be64), 0); | |
cd915493 | 698 | if (error != sizeof(u64)) |
b3b94faa DT |
699 | return (error < 0) ? error : -EIO; |
700 | ||
701 | *leaf_out = be64_to_cpu(leaf_no); | |
702 | ||
703 | return 0; | |
704 | } | |
705 | ||
cd915493 | 706 | static int get_first_leaf(struct gfs2_inode *dip, u32 index, |
b3b94faa DT |
707 | struct buffer_head **bh_out) |
708 | { | |
cd915493 | 709 | u64 leaf_no; |
b3b94faa DT |
710 | int error; |
711 | ||
712 | error = get_leaf_nr(dip, index, &leaf_no); | |
713 | if (!error) | |
714 | error = get_leaf(dip, leaf_no, bh_out); | |
715 | ||
716 | return error; | |
717 | } | |
718 | ||
c752666c SW |
719 | static struct gfs2_dirent *gfs2_dirent_search(struct inode *inode, |
720 | const struct qstr *name, | |
721 | gfs2_dscan_t scan, | |
722 | struct buffer_head **pbh) | |
b3b94faa | 723 | { |
c752666c SW |
724 | struct buffer_head *bh; |
725 | struct gfs2_dirent *dent; | |
feaa7bba | 726 | struct gfs2_inode *ip = GFS2_I(inode); |
b3b94faa DT |
727 | int error; |
728 | ||
383f01fb | 729 | if (ip->i_diskflags & GFS2_DIF_EXHASH) { |
c752666c | 730 | struct gfs2_leaf *leaf; |
9a004508 | 731 | unsigned hsize = 1 << ip->i_depth; |
c752666c SW |
732 | unsigned index; |
733 | u64 ln; | |
a2e0f799 | 734 | if (hsize * sizeof(u64) != i_size_read(inode)) { |
c752666c SW |
735 | gfs2_consist_inode(ip); |
736 | return ERR_PTR(-EIO); | |
737 | } | |
907b9bce | 738 | |
9a004508 | 739 | index = name->hash >> (32 - ip->i_depth); |
c752666c SW |
740 | error = get_first_leaf(ip, index, &bh); |
741 | if (error) | |
742 | return ERR_PTR(error); | |
743 | do { | |
744 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, | |
71b86f56 | 745 | scan, name, NULL); |
c752666c SW |
746 | if (dent) |
747 | goto got_dent; | |
748 | leaf = (struct gfs2_leaf *)bh->b_data; | |
749 | ln = be64_to_cpu(leaf->lf_next); | |
f4154ea0 | 750 | brelse(bh); |
c752666c SW |
751 | if (!ln) |
752 | break; | |
907b9bce | 753 | |
c752666c SW |
754 | error = get_leaf(ip, ln, &bh); |
755 | } while(!error); | |
b3b94faa | 756 | |
c752666c | 757 | return error ? ERR_PTR(error) : NULL; |
b3b94faa | 758 | } |
b3b94faa | 759 | |
907b9bce | 760 | |
c752666c SW |
761 | error = gfs2_meta_inode_buffer(ip, &bh); |
762 | if (error) | |
763 | return ERR_PTR(error); | |
71b86f56 | 764 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, scan, name, NULL); |
c752666c | 765 | got_dent: |
f4154ea0 | 766 | if (unlikely(dent == NULL || IS_ERR(dent))) { |
ed386507 SW |
767 | brelse(bh); |
768 | bh = NULL; | |
769 | } | |
c752666c SW |
770 | *pbh = bh; |
771 | return dent; | |
772 | } | |
b3b94faa | 773 | |
c752666c SW |
774 | static struct gfs2_leaf *new_leaf(struct inode *inode, struct buffer_head **pbh, u16 depth) |
775 | { | |
feaa7bba | 776 | struct gfs2_inode *ip = GFS2_I(inode); |
b45e41d7 | 777 | unsigned int n = 1; |
09010978 SW |
778 | u64 bn; |
779 | int error; | |
780 | struct buffer_head *bh; | |
c752666c SW |
781 | struct gfs2_leaf *leaf; |
782 | struct gfs2_dirent *dent; | |
71b86f56 | 783 | struct qstr name = { .name = "", .len = 0, .hash = 0 }; |
09010978 SW |
784 | |
785 | error = gfs2_alloc_block(ip, &bn, &n); | |
786 | if (error) | |
787 | return NULL; | |
788 | bh = gfs2_meta_new(ip->i_gl, bn); | |
c752666c SW |
789 | if (!bh) |
790 | return NULL; | |
09010978 | 791 | |
5731be53 | 792 | gfs2_trans_add_unrevoke(GFS2_SB(inode), bn, 1); |
c752666c SW |
793 | gfs2_trans_add_bh(ip->i_gl, bh, 1); |
794 | gfs2_metatype_set(bh, GFS2_METATYPE_LF, GFS2_FORMAT_LF); | |
795 | leaf = (struct gfs2_leaf *)bh->b_data; | |
796 | leaf->lf_depth = cpu_to_be16(depth); | |
2bdbc5d7 | 797 | leaf->lf_entries = 0; |
a2d7d021 | 798 | leaf->lf_dirent_format = cpu_to_be32(GFS2_FORMAT_DE); |
2bdbc5d7 | 799 | leaf->lf_next = 0; |
c752666c SW |
800 | memset(leaf->lf_reserved, 0, sizeof(leaf->lf_reserved)); |
801 | dent = (struct gfs2_dirent *)(leaf+1); | |
71b86f56 | 802 | gfs2_qstr2dirent(&name, bh->b_size - sizeof(struct gfs2_leaf), dent); |
c752666c SW |
803 | *pbh = bh; |
804 | return leaf; | |
b3b94faa DT |
805 | } |
806 | ||
807 | /** | |
808 | * dir_make_exhash - Convert a stuffed directory into an ExHash directory | |
809 | * @dip: The GFS2 inode | |
810 | * | |
811 | * Returns: 0 on success, error code otherwise | |
812 | */ | |
813 | ||
c752666c | 814 | static int dir_make_exhash(struct inode *inode) |
b3b94faa | 815 | { |
feaa7bba SW |
816 | struct gfs2_inode *dip = GFS2_I(inode); |
817 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
b3b94faa | 818 | struct gfs2_dirent *dent; |
c752666c | 819 | struct qstr args; |
b3b94faa DT |
820 | struct buffer_head *bh, *dibh; |
821 | struct gfs2_leaf *leaf; | |
822 | int y; | |
cd915493 | 823 | u32 x; |
b44b84d7 AV |
824 | __be64 *lp; |
825 | u64 bn; | |
b3b94faa DT |
826 | int error; |
827 | ||
828 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
829 | if (error) | |
830 | return error; | |
831 | ||
b3b94faa DT |
832 | /* Turn over a new leaf */ |
833 | ||
c752666c SW |
834 | leaf = new_leaf(inode, &bh, 0); |
835 | if (!leaf) | |
836 | return -ENOSPC; | |
837 | bn = bh->b_blocknr; | |
b3b94faa | 838 | |
ad6203f2 SW |
839 | gfs2_assert(sdp, dip->i_entries < (1 << 16)); |
840 | leaf->lf_entries = cpu_to_be16(dip->i_entries); | |
b3b94faa DT |
841 | |
842 | /* Copy dirents */ | |
843 | ||
844 | gfs2_buffer_copy_tail(bh, sizeof(struct gfs2_leaf), dibh, | |
845 | sizeof(struct gfs2_dinode)); | |
846 | ||
847 | /* Find last entry */ | |
848 | ||
849 | x = 0; | |
c752666c SW |
850 | args.len = bh->b_size - sizeof(struct gfs2_dinode) + |
851 | sizeof(struct gfs2_leaf); | |
852 | args.name = bh->b_data; | |
feaa7bba | 853 | dent = gfs2_dirent_scan(&dip->i_inode, bh->b_data, bh->b_size, |
71b86f56 | 854 | gfs2_dirent_last, &args, NULL); |
c752666c SW |
855 | if (!dent) { |
856 | brelse(bh); | |
857 | brelse(dibh); | |
858 | return -EIO; | |
859 | } | |
860 | if (IS_ERR(dent)) { | |
861 | brelse(bh); | |
862 | brelse(dibh); | |
863 | return PTR_ERR(dent); | |
b3b94faa | 864 | } |
b3b94faa DT |
865 | |
866 | /* Adjust the last dirent's record length | |
867 | (Remember that dent still points to the last entry.) */ | |
868 | ||
4dd651ad | 869 | dent->de_rec_len = cpu_to_be16(be16_to_cpu(dent->de_rec_len) + |
b3b94faa | 870 | sizeof(struct gfs2_dinode) - |
4dd651ad | 871 | sizeof(struct gfs2_leaf)); |
b3b94faa DT |
872 | |
873 | brelse(bh); | |
874 | ||
875 | /* We're done with the new leaf block, now setup the new | |
876 | hash table. */ | |
877 | ||
d4e9c4c3 | 878 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
b3b94faa DT |
879 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode)); |
880 | ||
b44b84d7 | 881 | lp = (__be64 *)(dibh->b_data + sizeof(struct gfs2_dinode)); |
b3b94faa DT |
882 | |
883 | for (x = sdp->sd_hash_ptrs; x--; lp++) | |
884 | *lp = cpu_to_be64(bn); | |
885 | ||
a2e0f799 | 886 | i_size_write(inode, sdp->sd_sb.sb_bsize / 2); |
77658aad | 887 | gfs2_add_inode_blocks(&dip->i_inode, 1); |
383f01fb | 888 | dip->i_diskflags |= GFS2_DIF_EXHASH; |
b3b94faa DT |
889 | |
890 | for (x = sdp->sd_hash_ptrs, y = -1; x; x >>= 1, y++) ; | |
9a004508 | 891 | dip->i_depth = y; |
b3b94faa | 892 | |
539e5d6b | 893 | gfs2_dinode_out(dip, dibh->b_data); |
b3b94faa DT |
894 | |
895 | brelse(dibh); | |
896 | ||
897 | return 0; | |
898 | } | |
899 | ||
900 | /** | |
901 | * dir_split_leaf - Split a leaf block into two | |
902 | * @dip: The GFS2 inode | |
903 | * @index: | |
904 | * @leaf_no: | |
905 | * | |
906 | * Returns: 0 on success, error code on failure | |
907 | */ | |
908 | ||
c752666c | 909 | static int dir_split_leaf(struct inode *inode, const struct qstr *name) |
b3b94faa | 910 | { |
feaa7bba | 911 | struct gfs2_inode *dip = GFS2_I(inode); |
b3b94faa DT |
912 | struct buffer_head *nbh, *obh, *dibh; |
913 | struct gfs2_leaf *nleaf, *oleaf; | |
4da3c646 | 914 | struct gfs2_dirent *dent = NULL, *prev = NULL, *next = NULL, *new; |
cd915493 | 915 | u32 start, len, half_len, divider; |
b44b84d7 AV |
916 | u64 bn, leaf_no; |
917 | __be64 *lp; | |
cd915493 | 918 | u32 index; |
b3b94faa DT |
919 | int x, moved = 0; |
920 | int error; | |
921 | ||
9a004508 | 922 | index = name->hash >> (32 - dip->i_depth); |
c752666c SW |
923 | error = get_leaf_nr(dip, index, &leaf_no); |
924 | if (error) | |
925 | return error; | |
b3b94faa DT |
926 | |
927 | /* Get the old leaf block */ | |
b3b94faa DT |
928 | error = get_leaf(dip, leaf_no, &obh); |
929 | if (error) | |
e90deff5 | 930 | return error; |
b3b94faa | 931 | |
b3b94faa | 932 | oleaf = (struct gfs2_leaf *)obh->b_data; |
9a004508 | 933 | if (dip->i_depth == be16_to_cpu(oleaf->lf_depth)) { |
e90deff5 SW |
934 | brelse(obh); |
935 | return 1; /* can't split */ | |
936 | } | |
937 | ||
938 | gfs2_trans_add_bh(dip->i_gl, obh, 1); | |
b3b94faa | 939 | |
c752666c SW |
940 | nleaf = new_leaf(inode, &nbh, be16_to_cpu(oleaf->lf_depth) + 1); |
941 | if (!nleaf) { | |
942 | brelse(obh); | |
943 | return -ENOSPC; | |
944 | } | |
945 | bn = nbh->b_blocknr; | |
b3b94faa | 946 | |
c752666c | 947 | /* Compute the start and len of leaf pointers in the hash table. */ |
9a004508 | 948 | len = 1 << (dip->i_depth - be16_to_cpu(oleaf->lf_depth)); |
b3b94faa DT |
949 | half_len = len >> 1; |
950 | if (!half_len) { | |
9a004508 | 951 | printk(KERN_WARNING "i_depth %u lf_depth %u index %u\n", dip->i_depth, be16_to_cpu(oleaf->lf_depth), index); |
b3b94faa DT |
952 | gfs2_consist_inode(dip); |
953 | error = -EIO; | |
954 | goto fail_brelse; | |
955 | } | |
956 | ||
957 | start = (index & ~(len - 1)); | |
958 | ||
959 | /* Change the pointers. | |
960 | Don't bother distinguishing stuffed from non-stuffed. | |
961 | This code is complicated enough already. */ | |
4244b52e DR |
962 | lp = kmalloc(half_len * sizeof(__be64), GFP_NOFS); |
963 | if (!lp) { | |
964 | error = -ENOMEM; | |
965 | goto fail_brelse; | |
966 | } | |
967 | ||
b3b94faa | 968 | /* Change the pointers */ |
b3b94faa DT |
969 | for (x = 0; x < half_len; x++) |
970 | lp[x] = cpu_to_be64(bn); | |
971 | ||
cd915493 SW |
972 | error = gfs2_dir_write_data(dip, (char *)lp, start * sizeof(u64), |
973 | half_len * sizeof(u64)); | |
974 | if (error != half_len * sizeof(u64)) { | |
b3b94faa DT |
975 | if (error >= 0) |
976 | error = -EIO; | |
977 | goto fail_lpfree; | |
978 | } | |
979 | ||
980 | kfree(lp); | |
981 | ||
982 | /* Compute the divider */ | |
9a004508 | 983 | divider = (start + half_len) << (32 - dip->i_depth); |
b3b94faa DT |
984 | |
985 | /* Copy the entries */ | |
1579343a | 986 | dent = (struct gfs2_dirent *)(obh->b_data + sizeof(struct gfs2_leaf)); |
b3b94faa DT |
987 | |
988 | do { | |
989 | next = dent; | |
990 | if (dirent_next(dip, obh, &next)) | |
991 | next = NULL; | |
992 | ||
5e7d65cd | 993 | if (!gfs2_dirent_sentinel(dent) && |
b3b94faa | 994 | be32_to_cpu(dent->de_hash) < divider) { |
c752666c SW |
995 | struct qstr str; |
996 | str.name = (char*)(dent+1); | |
997 | str.len = be16_to_cpu(dent->de_name_len); | |
998 | str.hash = be32_to_cpu(dent->de_hash); | |
71b86f56 | 999 | new = gfs2_dirent_alloc(inode, nbh, &str); |
c752666c SW |
1000 | if (IS_ERR(new)) { |
1001 | error = PTR_ERR(new); | |
1002 | break; | |
1003 | } | |
b3b94faa DT |
1004 | |
1005 | new->de_inum = dent->de_inum; /* No endian worries */ | |
b3b94faa | 1006 | new->de_type = dent->de_type; /* No endian worries */ |
bb16b342 | 1007 | be16_add_cpu(&nleaf->lf_entries, 1); |
b3b94faa DT |
1008 | |
1009 | dirent_del(dip, obh, prev, dent); | |
1010 | ||
1011 | if (!oleaf->lf_entries) | |
1012 | gfs2_consist_inode(dip); | |
bb16b342 | 1013 | be16_add_cpu(&oleaf->lf_entries, -1); |
b3b94faa DT |
1014 | |
1015 | if (!prev) | |
1016 | prev = dent; | |
1017 | ||
1018 | moved = 1; | |
c752666c | 1019 | } else { |
b3b94faa | 1020 | prev = dent; |
c752666c | 1021 | } |
b3b94faa | 1022 | dent = next; |
c752666c | 1023 | } while (dent); |
b3b94faa | 1024 | |
c752666c | 1025 | oleaf->lf_depth = nleaf->lf_depth; |
b3b94faa DT |
1026 | |
1027 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
feaa7bba | 1028 | if (!gfs2_assert_withdraw(GFS2_SB(&dip->i_inode), !error)) { |
382e6e25 | 1029 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
77658aad | 1030 | gfs2_add_inode_blocks(&dip->i_inode, 1); |
539e5d6b | 1031 | gfs2_dinode_out(dip, dibh->b_data); |
b3b94faa DT |
1032 | brelse(dibh); |
1033 | } | |
1034 | ||
1035 | brelse(obh); | |
1036 | brelse(nbh); | |
1037 | ||
1038 | return error; | |
1039 | ||
e90deff5 | 1040 | fail_lpfree: |
b3b94faa DT |
1041 | kfree(lp); |
1042 | ||
e90deff5 | 1043 | fail_brelse: |
b3b94faa | 1044 | brelse(obh); |
b3b94faa DT |
1045 | brelse(nbh); |
1046 | return error; | |
1047 | } | |
1048 | ||
1049 | /** | |
1050 | * dir_double_exhash - Double size of ExHash table | |
1051 | * @dip: The GFS2 dinode | |
1052 | * | |
1053 | * Returns: 0 on success, error code on failure | |
1054 | */ | |
1055 | ||
1056 | static int dir_double_exhash(struct gfs2_inode *dip) | |
1057 | { | |
feaa7bba | 1058 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa | 1059 | struct buffer_head *dibh; |
cd915493 SW |
1060 | u32 hsize; |
1061 | u64 *buf; | |
1062 | u64 *from, *to; | |
1063 | u64 block; | |
a2e0f799 | 1064 | u64 disksize = i_size_read(&dip->i_inode); |
b3b94faa DT |
1065 | int x; |
1066 | int error = 0; | |
1067 | ||
9a004508 | 1068 | hsize = 1 << dip->i_depth; |
a2e0f799 | 1069 | if (hsize * sizeof(u64) != disksize) { |
b3b94faa DT |
1070 | gfs2_consist_inode(dip); |
1071 | return -EIO; | |
1072 | } | |
1073 | ||
1074 | /* Allocate both the "from" and "to" buffers in one big chunk */ | |
1075 | ||
4244b52e DR |
1076 | buf = kcalloc(3, sdp->sd_hash_bsize, GFP_NOFS); |
1077 | if (!buf) | |
1078 | return -ENOMEM; | |
b3b94faa | 1079 | |
a2e0f799 | 1080 | for (block = disksize >> sdp->sd_hash_bsize_shift; block--;) { |
e13940ba | 1081 | error = gfs2_dir_read_data(dip, (char *)buf, |
b3b94faa | 1082 | block * sdp->sd_hash_bsize, |
7276b3b0 | 1083 | sdp->sd_hash_bsize, 1); |
b3b94faa DT |
1084 | if (error != sdp->sd_hash_bsize) { |
1085 | if (error >= 0) | |
1086 | error = -EIO; | |
1087 | goto fail; | |
1088 | } | |
1089 | ||
1090 | from = buf; | |
cd915493 | 1091 | to = (u64 *)((char *)buf + sdp->sd_hash_bsize); |
b3b94faa DT |
1092 | |
1093 | for (x = sdp->sd_hash_ptrs; x--; from++) { | |
1094 | *to++ = *from; /* No endianess worries */ | |
1095 | *to++ = *from; | |
1096 | } | |
1097 | ||
e13940ba | 1098 | error = gfs2_dir_write_data(dip, |
b3b94faa DT |
1099 | (char *)buf + sdp->sd_hash_bsize, |
1100 | block * sdp->sd_sb.sb_bsize, | |
1101 | sdp->sd_sb.sb_bsize); | |
1102 | if (error != sdp->sd_sb.sb_bsize) { | |
1103 | if (error >= 0) | |
1104 | error = -EIO; | |
1105 | goto fail; | |
1106 | } | |
1107 | } | |
1108 | ||
1109 | kfree(buf); | |
1110 | ||
1111 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1112 | if (!gfs2_assert_withdraw(sdp, !error)) { | |
9a004508 | 1113 | dip->i_depth++; |
539e5d6b | 1114 | gfs2_dinode_out(dip, dibh->b_data); |
b3b94faa DT |
1115 | brelse(dibh); |
1116 | } | |
1117 | ||
1118 | return error; | |
1119 | ||
a91ea69f | 1120 | fail: |
b3b94faa | 1121 | kfree(buf); |
b3b94faa DT |
1122 | return error; |
1123 | } | |
1124 | ||
1125 | /** | |
1126 | * compare_dents - compare directory entries by hash value | |
1127 | * @a: first dent | |
1128 | * @b: second dent | |
1129 | * | |
1130 | * When comparing the hash entries of @a to @b: | |
1131 | * gt: returns 1 | |
1132 | * lt: returns -1 | |
1133 | * eq: returns 0 | |
1134 | */ | |
1135 | ||
1136 | static int compare_dents(const void *a, const void *b) | |
1137 | { | |
2bdbc5d7 | 1138 | const struct gfs2_dirent *dent_a, *dent_b; |
cd915493 | 1139 | u32 hash_a, hash_b; |
b3b94faa DT |
1140 | int ret = 0; |
1141 | ||
2bdbc5d7 | 1142 | dent_a = *(const struct gfs2_dirent **)a; |
c752666c | 1143 | hash_a = be32_to_cpu(dent_a->de_hash); |
b3b94faa | 1144 | |
2bdbc5d7 | 1145 | dent_b = *(const struct gfs2_dirent **)b; |
c752666c | 1146 | hash_b = be32_to_cpu(dent_b->de_hash); |
b3b94faa DT |
1147 | |
1148 | if (hash_a > hash_b) | |
1149 | ret = 1; | |
1150 | else if (hash_a < hash_b) | |
1151 | ret = -1; | |
1152 | else { | |
4dd651ad SW |
1153 | unsigned int len_a = be16_to_cpu(dent_a->de_name_len); |
1154 | unsigned int len_b = be16_to_cpu(dent_b->de_name_len); | |
b3b94faa DT |
1155 | |
1156 | if (len_a > len_b) | |
1157 | ret = 1; | |
1158 | else if (len_a < len_b) | |
1159 | ret = -1; | |
1160 | else | |
2bdbc5d7 | 1161 | ret = memcmp(dent_a + 1, dent_b + 1, len_a); |
b3b94faa DT |
1162 | } |
1163 | ||
1164 | return ret; | |
1165 | } | |
1166 | ||
1167 | /** | |
1168 | * do_filldir_main - read out directory entries | |
1169 | * @dip: The GFS2 inode | |
1170 | * @offset: The offset in the file to read from | |
1171 | * @opaque: opaque data to pass to filldir | |
1172 | * @filldir: The function to pass entries to | |
1173 | * @darr: an array of struct gfs2_dirent pointers to read | |
1174 | * @entries: the number of entries in darr | |
1175 | * @copied: pointer to int that's non-zero if a entry has been copied out | |
1176 | * | |
1177 | * Jump through some hoops to make sure that if there are hash collsions, | |
1178 | * they are read out at the beginning of a buffer. We want to minimize | |
1179 | * the possibility that they will fall into different readdir buffers or | |
1180 | * that someone will want to seek to that location. | |
1181 | * | |
1182 | * Returns: errno, >0 on exception from filldir | |
1183 | */ | |
1184 | ||
cd915493 | 1185 | static int do_filldir_main(struct gfs2_inode *dip, u64 *offset, |
3699e3a4 | 1186 | void *opaque, filldir_t filldir, |
cd915493 | 1187 | const struct gfs2_dirent **darr, u32 entries, |
b3b94faa DT |
1188 | int *copied) |
1189 | { | |
71b86f56 | 1190 | const struct gfs2_dirent *dent, *dent_next; |
cd915493 | 1191 | u64 off, off_next; |
b3b94faa DT |
1192 | unsigned int x, y; |
1193 | int run = 0; | |
1194 | int error = 0; | |
1195 | ||
1196 | sort(darr, entries, sizeof(struct gfs2_dirent *), compare_dents, NULL); | |
1197 | ||
1198 | dent_next = darr[0]; | |
1199 | off_next = be32_to_cpu(dent_next->de_hash); | |
1200 | off_next = gfs2_disk_hash2offset(off_next); | |
1201 | ||
1202 | for (x = 0, y = 1; x < entries; x++, y++) { | |
1203 | dent = dent_next; | |
1204 | off = off_next; | |
1205 | ||
1206 | if (y < entries) { | |
1207 | dent_next = darr[y]; | |
1208 | off_next = be32_to_cpu(dent_next->de_hash); | |
1209 | off_next = gfs2_disk_hash2offset(off_next); | |
1210 | ||
1211 | if (off < *offset) | |
1212 | continue; | |
1213 | *offset = off; | |
1214 | ||
1215 | if (off_next == off) { | |
1216 | if (*copied && !run) | |
1217 | return 1; | |
1218 | run = 1; | |
1219 | } else | |
1220 | run = 0; | |
1221 | } else { | |
1222 | if (off < *offset) | |
1223 | continue; | |
1224 | *offset = off; | |
1225 | } | |
1226 | ||
2bdbc5d7 | 1227 | error = filldir(opaque, (const char *)(dent + 1), |
4dd651ad | 1228 | be16_to_cpu(dent->de_name_len), |
3699e3a4 | 1229 | off, be64_to_cpu(dent->de_inum.no_addr), |
4dd651ad | 1230 | be16_to_cpu(dent->de_type)); |
b3b94faa DT |
1231 | if (error) |
1232 | return 1; | |
1233 | ||
1234 | *copied = 1; | |
1235 | } | |
1236 | ||
1237 | /* Increment the *offset by one, so the next time we come into the | |
1238 | do_filldir fxn, we get the next entry instead of the last one in the | |
1239 | current leaf */ | |
1240 | ||
1241 | (*offset)++; | |
1242 | ||
1243 | return 0; | |
1244 | } | |
1245 | ||
d2a97a4e SW |
1246 | static void *gfs2_alloc_sort_buffer(unsigned size) |
1247 | { | |
1248 | void *ptr = NULL; | |
1249 | ||
1250 | if (size < KMALLOC_MAX_SIZE) | |
1251 | ptr = kmalloc(size, GFP_NOFS | __GFP_NOWARN); | |
1252 | if (!ptr) | |
1253 | ptr = __vmalloc(size, GFP_NOFS, PAGE_KERNEL); | |
1254 | return ptr; | |
1255 | } | |
1256 | ||
1257 | static void gfs2_free_sort_buffer(void *ptr) | |
1258 | { | |
1259 | if (is_vmalloc_addr(ptr)) | |
1260 | vfree(ptr); | |
1261 | else | |
1262 | kfree(ptr); | |
1263 | } | |
1264 | ||
71b86f56 | 1265 | static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque, |
3699e3a4 SW |
1266 | filldir_t filldir, int *copied, unsigned *depth, |
1267 | u64 leaf_no) | |
b3b94faa | 1268 | { |
feaa7bba | 1269 | struct gfs2_inode *ip = GFS2_I(inode); |
bdd19a22 | 1270 | struct gfs2_sbd *sdp = GFS2_SB(inode); |
71b86f56 SW |
1271 | struct buffer_head *bh; |
1272 | struct gfs2_leaf *lf; | |
bdd19a22 | 1273 | unsigned entries = 0, entries2 = 0; |
71b86f56 SW |
1274 | unsigned leaves = 0; |
1275 | const struct gfs2_dirent **darr, *dent; | |
1276 | struct dirent_gather g; | |
1277 | struct buffer_head **larr; | |
1278 | int leaf = 0; | |
1279 | int error, i; | |
1280 | u64 lfn = leaf_no; | |
b3b94faa | 1281 | |
b3b94faa | 1282 | do { |
71b86f56 | 1283 | error = get_leaf(ip, lfn, &bh); |
b3b94faa | 1284 | if (error) |
71b86f56 SW |
1285 | goto out; |
1286 | lf = (struct gfs2_leaf *)bh->b_data; | |
1287 | if (leaves == 0) | |
1288 | *depth = be16_to_cpu(lf->lf_depth); | |
1289 | entries += be16_to_cpu(lf->lf_entries); | |
1290 | leaves++; | |
1291 | lfn = be64_to_cpu(lf->lf_next); | |
1292 | brelse(bh); | |
1293 | } while(lfn); | |
b3b94faa DT |
1294 | |
1295 | if (!entries) | |
1296 | return 0; | |
1297 | ||
71b86f56 | 1298 | error = -ENOMEM; |
bdd19a22 SW |
1299 | /* |
1300 | * The extra 99 entries are not normally used, but are a buffer | |
1301 | * zone in case the number of entries in the leaf is corrupt. | |
1302 | * 99 is the maximum number of entries that can fit in a single | |
1303 | * leaf block. | |
1304 | */ | |
d2a97a4e | 1305 | larr = gfs2_alloc_sort_buffer((leaves + entries + 99) * sizeof(void *)); |
71b86f56 SW |
1306 | if (!larr) |
1307 | goto out; | |
1308 | darr = (const struct gfs2_dirent **)(larr + leaves); | |
1309 | g.pdent = darr; | |
1310 | g.offset = 0; | |
1311 | lfn = leaf_no; | |
b3b94faa | 1312 | |
71b86f56 SW |
1313 | do { |
1314 | error = get_leaf(ip, lfn, &bh); | |
b3b94faa | 1315 | if (error) |
d2a97a4e | 1316 | goto out_free; |
71b86f56 SW |
1317 | lf = (struct gfs2_leaf *)bh->b_data; |
1318 | lfn = be64_to_cpu(lf->lf_next); | |
1319 | if (lf->lf_entries) { | |
bdd19a22 | 1320 | entries2 += be16_to_cpu(lf->lf_entries); |
71b86f56 SW |
1321 | dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, |
1322 | gfs2_dirent_gather, NULL, &g); | |
1323 | error = PTR_ERR(dent); | |
bdd19a22 | 1324 | if (IS_ERR(dent)) |
d2a97a4e | 1325 | goto out_free; |
bdd19a22 | 1326 | if (entries2 != g.offset) { |
f391a4ea AM |
1327 | fs_warn(sdp, "Number of entries corrupt in dir " |
1328 | "leaf %llu, entries2 (%u) != " | |
1329 | "g.offset (%u)\n", | |
1330 | (unsigned long long)bh->b_blocknr, | |
1331 | entries2, g.offset); | |
bdd19a22 SW |
1332 | |
1333 | error = -EIO; | |
d2a97a4e | 1334 | goto out_free; |
71b86f56 SW |
1335 | } |
1336 | error = 0; | |
1337 | larr[leaf++] = bh; | |
b3b94faa | 1338 | } else { |
71b86f56 | 1339 | brelse(bh); |
b3b94faa | 1340 | } |
71b86f56 | 1341 | } while(lfn); |
b3b94faa | 1342 | |
bdd19a22 | 1343 | BUG_ON(entries2 != entries); |
71b86f56 | 1344 | error = do_filldir_main(ip, offset, opaque, filldir, darr, |
b3b94faa | 1345 | entries, copied); |
d2a97a4e | 1346 | out_free: |
71b86f56 SW |
1347 | for(i = 0; i < leaf; i++) |
1348 | brelse(larr[i]); | |
d2a97a4e | 1349 | gfs2_free_sort_buffer(larr); |
71b86f56 | 1350 | out: |
b3b94faa DT |
1351 | return error; |
1352 | } | |
1353 | ||
1354 | /** | |
c752666c SW |
1355 | * dir_e_read - Reads the entries from a directory into a filldir buffer |
1356 | * @dip: dinode pointer | |
1357 | * @offset: the hash of the last entry read shifted to the right once | |
1358 | * @opaque: buffer for the filldir function to fill | |
1359 | * @filldir: points to the filldir function to use | |
b3b94faa | 1360 | * |
c752666c | 1361 | * Returns: errno |
b3b94faa DT |
1362 | */ |
1363 | ||
cd915493 | 1364 | static int dir_e_read(struct inode *inode, u64 *offset, void *opaque, |
3699e3a4 | 1365 | filldir_t filldir) |
b3b94faa | 1366 | { |
feaa7bba SW |
1367 | struct gfs2_inode *dip = GFS2_I(inode); |
1368 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
cd915493 SW |
1369 | u32 hsize, len = 0; |
1370 | u32 ht_offset, lp_offset, ht_offset_cur = -1; | |
1371 | u32 hash, index; | |
b44b84d7 | 1372 | __be64 *lp; |
c752666c SW |
1373 | int copied = 0; |
1374 | int error = 0; | |
4da3c646 | 1375 | unsigned depth = 0; |
b3b94faa | 1376 | |
9a004508 | 1377 | hsize = 1 << dip->i_depth; |
a2e0f799 | 1378 | if (hsize * sizeof(u64) != i_size_read(inode)) { |
b3b94faa DT |
1379 | gfs2_consist_inode(dip); |
1380 | return -EIO; | |
1381 | } | |
1382 | ||
1383 | hash = gfs2_dir_offset2hash(*offset); | |
9a004508 | 1384 | index = hash >> (32 - dip->i_depth); |
b3b94faa | 1385 | |
16c5f06f | 1386 | lp = kmalloc(sdp->sd_hash_bsize, GFP_NOFS); |
b3b94faa DT |
1387 | if (!lp) |
1388 | return -ENOMEM; | |
1389 | ||
1390 | while (index < hsize) { | |
1391 | lp_offset = index & (sdp->sd_hash_ptrs - 1); | |
1392 | ht_offset = index - lp_offset; | |
1393 | ||
1394 | if (ht_offset_cur != ht_offset) { | |
e13940ba | 1395 | error = gfs2_dir_read_data(dip, (char *)lp, |
b44b84d7 | 1396 | ht_offset * sizeof(__be64), |
7276b3b0 | 1397 | sdp->sd_hash_bsize, 1); |
b3b94faa DT |
1398 | if (error != sdp->sd_hash_bsize) { |
1399 | if (error >= 0) | |
1400 | error = -EIO; | |
1401 | goto out; | |
1402 | } | |
1403 | ht_offset_cur = ht_offset; | |
1404 | } | |
1405 | ||
71b86f56 SW |
1406 | error = gfs2_dir_read_leaf(inode, offset, opaque, filldir, |
1407 | &copied, &depth, | |
1408 | be64_to_cpu(lp[lp_offset])); | |
b3b94faa | 1409 | if (error) |
71b86f56 | 1410 | break; |
b3b94faa | 1411 | |
9a004508 | 1412 | len = 1 << (dip->i_depth - depth); |
b3b94faa DT |
1413 | index = (index & ~(len - 1)) + len; |
1414 | } | |
1415 | ||
71b86f56 | 1416 | out: |
b3b94faa | 1417 | kfree(lp); |
71b86f56 SW |
1418 | if (error > 0) |
1419 | error = 0; | |
b3b94faa DT |
1420 | return error; |
1421 | } | |
1422 | ||
cd915493 | 1423 | int gfs2_dir_read(struct inode *inode, u64 *offset, void *opaque, |
3699e3a4 | 1424 | filldir_t filldir) |
b3b94faa | 1425 | { |
feaa7bba | 1426 | struct gfs2_inode *dip = GFS2_I(inode); |
bdd19a22 | 1427 | struct gfs2_sbd *sdp = GFS2_SB(inode); |
71b86f56 SW |
1428 | struct dirent_gather g; |
1429 | const struct gfs2_dirent **darr, *dent; | |
b3b94faa DT |
1430 | struct buffer_head *dibh; |
1431 | int copied = 0; | |
1432 | int error; | |
1433 | ||
ad6203f2 | 1434 | if (!dip->i_entries) |
71b86f56 SW |
1435 | return 0; |
1436 | ||
383f01fb | 1437 | if (dip->i_diskflags & GFS2_DIF_EXHASH) |
71b86f56 SW |
1438 | return dir_e_read(inode, offset, opaque, filldir); |
1439 | ||
b3b94faa DT |
1440 | if (!gfs2_is_stuffed(dip)) { |
1441 | gfs2_consist_inode(dip); | |
1442 | return -EIO; | |
1443 | } | |
1444 | ||
b3b94faa DT |
1445 | error = gfs2_meta_inode_buffer(dip, &dibh); |
1446 | if (error) | |
1447 | return error; | |
1448 | ||
71b86f56 | 1449 | error = -ENOMEM; |
bdd19a22 | 1450 | /* 96 is max number of dirents which can be stuffed into an inode */ |
16c5f06f | 1451 | darr = kmalloc(96 * sizeof(struct gfs2_dirent *), GFP_NOFS); |
71b86f56 SW |
1452 | if (darr) { |
1453 | g.pdent = darr; | |
1454 | g.offset = 0; | |
1455 | dent = gfs2_dirent_scan(inode, dibh->b_data, dibh->b_size, | |
1456 | gfs2_dirent_gather, NULL, &g); | |
1457 | if (IS_ERR(dent)) { | |
1458 | error = PTR_ERR(dent); | |
1459 | goto out; | |
1460 | } | |
ad6203f2 | 1461 | if (dip->i_entries != g.offset) { |
bdd19a22 | 1462 | fs_warn(sdp, "Number of entries corrupt in dir %llu, " |
ad6203f2 | 1463 | "ip->i_entries (%u) != g.offset (%u)\n", |
dbb7cae2 | 1464 | (unsigned long long)dip->i_no_addr, |
ad6203f2 | 1465 | dip->i_entries, |
bdd19a22 SW |
1466 | g.offset); |
1467 | error = -EIO; | |
1468 | goto out; | |
1469 | } | |
71b86f56 | 1470 | error = do_filldir_main(dip, offset, opaque, filldir, darr, |
ad6203f2 | 1471 | dip->i_entries, &copied); |
71b86f56 SW |
1472 | out: |
1473 | kfree(darr); | |
1474 | } | |
1475 | ||
b3b94faa DT |
1476 | if (error > 0) |
1477 | error = 0; | |
1478 | ||
1479 | brelse(dibh); | |
1480 | ||
1481 | return error; | |
1482 | } | |
1483 | ||
b3b94faa DT |
1484 | /** |
1485 | * gfs2_dir_search - Search a directory | |
1486 | * @dip: The GFS2 inode | |
1487 | * @filename: | |
1488 | * @inode: | |
1489 | * | |
1490 | * This routine searches a directory for a file or another directory. | |
1491 | * Assumes a glock is held on dip. | |
1492 | * | |
1493 | * Returns: errno | |
1494 | */ | |
1495 | ||
dbb7cae2 | 1496 | struct inode *gfs2_dir_search(struct inode *dir, const struct qstr *name) |
b3b94faa | 1497 | { |
c752666c SW |
1498 | struct buffer_head *bh; |
1499 | struct gfs2_dirent *dent; | |
dbb7cae2 SW |
1500 | struct inode *inode; |
1501 | ||
1502 | dent = gfs2_dirent_search(dir, name, gfs2_dirent_find, &bh); | |
1503 | if (dent) { | |
1504 | if (IS_ERR(dent)) | |
e231c2ee | 1505 | return ERR_CAST(dent); |
bb9bcf06 WC |
1506 | inode = gfs2_inode_lookup(dir->i_sb, |
1507 | be16_to_cpu(dent->de_type), | |
1508 | be64_to_cpu(dent->de_inum.no_addr), | |
44ad37d6 | 1509 | be64_to_cpu(dent->de_inum.no_formal_ino), 0); |
dbb7cae2 SW |
1510 | brelse(bh); |
1511 | return inode; | |
1512 | } | |
1513 | return ERR_PTR(-ENOENT); | |
1514 | } | |
1515 | ||
1516 | int gfs2_dir_check(struct inode *dir, const struct qstr *name, | |
1517 | const struct gfs2_inode *ip) | |
1518 | { | |
1519 | struct buffer_head *bh; | |
1520 | struct gfs2_dirent *dent; | |
1521 | int ret = -ENOENT; | |
c752666c SW |
1522 | |
1523 | dent = gfs2_dirent_search(dir, name, gfs2_dirent_find, &bh); | |
1524 | if (dent) { | |
1525 | if (IS_ERR(dent)) | |
1526 | return PTR_ERR(dent); | |
dbb7cae2 SW |
1527 | if (ip) { |
1528 | if (be64_to_cpu(dent->de_inum.no_addr) != ip->i_no_addr) | |
1529 | goto out; | |
1530 | if (be64_to_cpu(dent->de_inum.no_formal_ino) != | |
1531 | ip->i_no_formal_ino) | |
1532 | goto out; | |
1533 | if (unlikely(IF2DT(ip->i_inode.i_mode) != | |
1534 | be16_to_cpu(dent->de_type))) { | |
1535 | gfs2_consist_inode(GFS2_I(dir)); | |
1536 | ret = -EIO; | |
1537 | goto out; | |
1538 | } | |
1539 | } | |
1540 | ret = 0; | |
1541 | out: | |
c752666c | 1542 | brelse(bh); |
c752666c | 1543 | } |
dbb7cae2 | 1544 | return ret; |
c752666c SW |
1545 | } |
1546 | ||
1547 | static int dir_new_leaf(struct inode *inode, const struct qstr *name) | |
1548 | { | |
1549 | struct buffer_head *bh, *obh; | |
feaa7bba | 1550 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c | 1551 | struct gfs2_leaf *leaf, *oleaf; |
b3b94faa | 1552 | int error; |
c752666c SW |
1553 | u32 index; |
1554 | u64 bn; | |
b3b94faa | 1555 | |
9a004508 | 1556 | index = name->hash >> (32 - ip->i_depth); |
c752666c SW |
1557 | error = get_first_leaf(ip, index, &obh); |
1558 | if (error) | |
1559 | return error; | |
1560 | do { | |
1561 | oleaf = (struct gfs2_leaf *)obh->b_data; | |
1562 | bn = be64_to_cpu(oleaf->lf_next); | |
1563 | if (!bn) | |
1564 | break; | |
1565 | brelse(obh); | |
1566 | error = get_leaf(ip, bn, &obh); | |
1567 | if (error) | |
1568 | return error; | |
1569 | } while(1); | |
b3b94faa | 1570 | |
c752666c SW |
1571 | gfs2_trans_add_bh(ip->i_gl, obh, 1); |
1572 | ||
1573 | leaf = new_leaf(inode, &bh, be16_to_cpu(oleaf->lf_depth)); | |
1574 | if (!leaf) { | |
1575 | brelse(obh); | |
1576 | return -ENOSPC; | |
1577 | } | |
4d8012b6 | 1578 | oleaf->lf_next = cpu_to_be64(bh->b_blocknr); |
c752666c SW |
1579 | brelse(bh); |
1580 | brelse(obh); | |
1581 | ||
1582 | error = gfs2_meta_inode_buffer(ip, &bh); | |
1583 | if (error) | |
1584 | return error; | |
1585 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
77658aad | 1586 | gfs2_add_inode_blocks(&ip->i_inode, 1); |
539e5d6b | 1587 | gfs2_dinode_out(ip, bh->b_data); |
c752666c SW |
1588 | brelse(bh); |
1589 | return 0; | |
b3b94faa DT |
1590 | } |
1591 | ||
1592 | /** | |
1593 | * gfs2_dir_add - Add new filename into directory | |
1594 | * @dip: The GFS2 inode | |
1595 | * @filename: The new name | |
1596 | * @inode: The inode number of the entry | |
1597 | * @type: The type of the entry | |
1598 | * | |
1599 | * Returns: 0 on success, error code on failure | |
1600 | */ | |
1601 | ||
c752666c | 1602 | int gfs2_dir_add(struct inode *inode, const struct qstr *name, |
dbb7cae2 | 1603 | const struct gfs2_inode *nip, unsigned type) |
b3b94faa | 1604 | { |
feaa7bba | 1605 | struct gfs2_inode *ip = GFS2_I(inode); |
c752666c SW |
1606 | struct buffer_head *bh; |
1607 | struct gfs2_dirent *dent; | |
1608 | struct gfs2_leaf *leaf; | |
b3b94faa DT |
1609 | int error; |
1610 | ||
c752666c SW |
1611 | while(1) { |
1612 | dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space, | |
1613 | &bh); | |
1614 | if (dent) { | |
1615 | if (IS_ERR(dent)) | |
1616 | return PTR_ERR(dent); | |
1617 | dent = gfs2_init_dirent(inode, dent, name, bh); | |
dbb7cae2 | 1618 | gfs2_inum_out(nip, dent); |
c752666c | 1619 | dent->de_type = cpu_to_be16(type); |
383f01fb | 1620 | if (ip->i_diskflags & GFS2_DIF_EXHASH) { |
c752666c | 1621 | leaf = (struct gfs2_leaf *)bh->b_data; |
bb16b342 | 1622 | be16_add_cpu(&leaf->lf_entries, 1); |
c752666c SW |
1623 | } |
1624 | brelse(bh); | |
1625 | error = gfs2_meta_inode_buffer(ip, &bh); | |
1626 | if (error) | |
1627 | break; | |
1628 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
ad6203f2 | 1629 | ip->i_entries++; |
4bd91ba1 | 1630 | ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME; |
539e5d6b | 1631 | gfs2_dinode_out(ip, bh->b_data); |
c752666c SW |
1632 | brelse(bh); |
1633 | error = 0; | |
1634 | break; | |
1635 | } | |
383f01fb | 1636 | if (!(ip->i_diskflags & GFS2_DIF_EXHASH)) { |
c752666c SW |
1637 | error = dir_make_exhash(inode); |
1638 | if (error) | |
1639 | break; | |
1640 | continue; | |
1641 | } | |
1642 | error = dir_split_leaf(inode, name); | |
1643 | if (error == 0) | |
1644 | continue; | |
e90deff5 | 1645 | if (error < 0) |
c752666c | 1646 | break; |
9a004508 | 1647 | if (ip->i_depth < GFS2_DIR_MAX_DEPTH) { |
c752666c SW |
1648 | error = dir_double_exhash(ip); |
1649 | if (error) | |
1650 | break; | |
1651 | error = dir_split_leaf(inode, name); | |
e90deff5 | 1652 | if (error < 0) |
c752666c | 1653 | break; |
e90deff5 SW |
1654 | if (error == 0) |
1655 | continue; | |
c752666c SW |
1656 | } |
1657 | error = dir_new_leaf(inode, name); | |
1658 | if (!error) | |
1659 | continue; | |
1660 | error = -ENOSPC; | |
1661 | break; | |
1662 | } | |
b3b94faa DT |
1663 | return error; |
1664 | } | |
1665 | ||
c752666c | 1666 | |
b3b94faa DT |
1667 | /** |
1668 | * gfs2_dir_del - Delete a directory entry | |
1669 | * @dip: The GFS2 inode | |
1670 | * @filename: The filename | |
1671 | * | |
1672 | * Returns: 0 on success, error code on failure | |
1673 | */ | |
1674 | ||
c752666c | 1675 | int gfs2_dir_del(struct gfs2_inode *dip, const struct qstr *name) |
b3b94faa | 1676 | { |
c752666c SW |
1677 | struct gfs2_dirent *dent, *prev = NULL; |
1678 | struct buffer_head *bh; | |
b3b94faa DT |
1679 | int error; |
1680 | ||
c752666c SW |
1681 | /* Returns _either_ the entry (if its first in block) or the |
1682 | previous entry otherwise */ | |
feaa7bba | 1683 | dent = gfs2_dirent_search(&dip->i_inode, name, gfs2_dirent_prev, &bh); |
c752666c SW |
1684 | if (!dent) { |
1685 | gfs2_consist_inode(dip); | |
1686 | return -EIO; | |
1687 | } | |
1688 | if (IS_ERR(dent)) { | |
1689 | gfs2_consist_inode(dip); | |
1690 | return PTR_ERR(dent); | |
1691 | } | |
1692 | /* If not first in block, adjust pointers accordingly */ | |
71b86f56 | 1693 | if (gfs2_dirent_find(dent, name, NULL) == 0) { |
c752666c SW |
1694 | prev = dent; |
1695 | dent = (struct gfs2_dirent *)((char *)dent + be16_to_cpu(prev->de_rec_len)); | |
1696 | } | |
1697 | ||
1698 | dirent_del(dip, bh, prev, dent); | |
383f01fb | 1699 | if (dip->i_diskflags & GFS2_DIF_EXHASH) { |
c752666c SW |
1700 | struct gfs2_leaf *leaf = (struct gfs2_leaf *)bh->b_data; |
1701 | u16 entries = be16_to_cpu(leaf->lf_entries); | |
1702 | if (!entries) | |
1703 | gfs2_consist_inode(dip); | |
1704 | leaf->lf_entries = cpu_to_be16(--entries); | |
c752666c | 1705 | } |
ed386507 | 1706 | brelse(bh); |
c752666c SW |
1707 | |
1708 | error = gfs2_meta_inode_buffer(dip, &bh); | |
1709 | if (error) | |
1710 | return error; | |
1711 | ||
ad6203f2 | 1712 | if (!dip->i_entries) |
c752666c SW |
1713 | gfs2_consist_inode(dip); |
1714 | gfs2_trans_add_bh(dip->i_gl, bh, 1); | |
ad6203f2 | 1715 | dip->i_entries--; |
4bd91ba1 | 1716 | dip->i_inode.i_mtime = dip->i_inode.i_ctime = CURRENT_TIME; |
539e5d6b | 1717 | gfs2_dinode_out(dip, bh->b_data); |
c752666c | 1718 | brelse(bh); |
feaa7bba | 1719 | mark_inode_dirty(&dip->i_inode); |
b3b94faa DT |
1720 | |
1721 | return error; | |
1722 | } | |
1723 | ||
b3b94faa DT |
1724 | /** |
1725 | * gfs2_dir_mvino - Change inode number of directory entry | |
1726 | * @dip: The GFS2 inode | |
1727 | * @filename: | |
1728 | * @new_inode: | |
1729 | * | |
1730 | * This routine changes the inode number of a directory entry. It's used | |
1731 | * by rename to change ".." when a directory is moved. | |
1732 | * Assumes a glock is held on dvp. | |
1733 | * | |
1734 | * Returns: errno | |
1735 | */ | |
1736 | ||
c752666c | 1737 | int gfs2_dir_mvino(struct gfs2_inode *dip, const struct qstr *filename, |
dbb7cae2 | 1738 | const struct gfs2_inode *nip, unsigned int new_type) |
b3b94faa | 1739 | { |
c752666c SW |
1740 | struct buffer_head *bh; |
1741 | struct gfs2_dirent *dent; | |
b3b94faa DT |
1742 | int error; |
1743 | ||
feaa7bba | 1744 | dent = gfs2_dirent_search(&dip->i_inode, filename, gfs2_dirent_find, &bh); |
c752666c SW |
1745 | if (!dent) { |
1746 | gfs2_consist_inode(dip); | |
1747 | return -EIO; | |
1748 | } | |
1749 | if (IS_ERR(dent)) | |
1750 | return PTR_ERR(dent); | |
b3b94faa | 1751 | |
c752666c | 1752 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
dbb7cae2 | 1753 | gfs2_inum_out(nip, dent); |
c752666c SW |
1754 | dent->de_type = cpu_to_be16(new_type); |
1755 | ||
383f01fb | 1756 | if (dip->i_diskflags & GFS2_DIF_EXHASH) { |
c752666c SW |
1757 | brelse(bh); |
1758 | error = gfs2_meta_inode_buffer(dip, &bh); | |
1759 | if (error) | |
1760 | return error; | |
1761 | gfs2_trans_add_bh(dip->i_gl, bh, 1); | |
1762 | } | |
1763 | ||
4bd91ba1 | 1764 | dip->i_inode.i_mtime = dip->i_inode.i_ctime = CURRENT_TIME; |
539e5d6b | 1765 | gfs2_dinode_out(dip, bh->b_data); |
c752666c SW |
1766 | brelse(bh); |
1767 | return 0; | |
b3b94faa DT |
1768 | } |
1769 | ||
1770 | /** | |
1771 | * foreach_leaf - call a function for each leaf in a directory | |
1772 | * @dip: the directory | |
b3b94faa DT |
1773 | * |
1774 | * Returns: errno | |
1775 | */ | |
1776 | ||
0d95326d | 1777 | static int foreach_leaf(struct gfs2_inode *dip) |
b3b94faa | 1778 | { |
feaa7bba | 1779 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
b3b94faa | 1780 | struct buffer_head *bh; |
c752666c | 1781 | struct gfs2_leaf *leaf; |
cd915493 SW |
1782 | u32 hsize, len; |
1783 | u32 ht_offset, lp_offset, ht_offset_cur = -1; | |
d24a7a43 | 1784 | u32 index = 0, next_index; |
b44b84d7 | 1785 | __be64 *lp; |
cd915493 | 1786 | u64 leaf_no; |
d24a7a43 | 1787 | int error = 0, last; |
b3b94faa | 1788 | |
9a004508 | 1789 | hsize = 1 << dip->i_depth; |
a2e0f799 | 1790 | if (hsize * sizeof(u64) != i_size_read(&dip->i_inode)) { |
b3b94faa DT |
1791 | gfs2_consist_inode(dip); |
1792 | return -EIO; | |
1793 | } | |
1794 | ||
16c5f06f | 1795 | lp = kmalloc(sdp->sd_hash_bsize, GFP_NOFS); |
b3b94faa DT |
1796 | if (!lp) |
1797 | return -ENOMEM; | |
1798 | ||
1799 | while (index < hsize) { | |
1800 | lp_offset = index & (sdp->sd_hash_ptrs - 1); | |
1801 | ht_offset = index - lp_offset; | |
1802 | ||
1803 | if (ht_offset_cur != ht_offset) { | |
e13940ba | 1804 | error = gfs2_dir_read_data(dip, (char *)lp, |
b44b84d7 | 1805 | ht_offset * sizeof(__be64), |
7276b3b0 | 1806 | sdp->sd_hash_bsize, 1); |
b3b94faa DT |
1807 | if (error != sdp->sd_hash_bsize) { |
1808 | if (error >= 0) | |
1809 | error = -EIO; | |
1810 | goto out; | |
1811 | } | |
1812 | ht_offset_cur = ht_offset; | |
1813 | } | |
1814 | ||
1815 | leaf_no = be64_to_cpu(lp[lp_offset]); | |
1816 | if (leaf_no) { | |
1817 | error = get_leaf(dip, leaf_no, &bh); | |
1818 | if (error) | |
1819 | goto out; | |
c752666c | 1820 | leaf = (struct gfs2_leaf *)bh->b_data; |
9a004508 | 1821 | len = 1 << (dip->i_depth - be16_to_cpu(leaf->lf_depth)); |
d24a7a43 BP |
1822 | next_index = (index & ~(len - 1)) + len; |
1823 | last = ((next_index >= hsize) ? 1 : 0); | |
634ee0b9 | 1824 | brelse(bh); |
d24a7a43 | 1825 | error = leaf_dealloc(dip, index, len, leaf_no, last); |
b3b94faa DT |
1826 | if (error) |
1827 | goto out; | |
d24a7a43 | 1828 | index = next_index; |
b3b94faa DT |
1829 | } else |
1830 | index++; | |
1831 | } | |
1832 | ||
1833 | if (index != hsize) { | |
1834 | gfs2_consist_inode(dip); | |
1835 | error = -EIO; | |
1836 | } | |
1837 | ||
634ee0b9 | 1838 | out: |
b3b94faa DT |
1839 | kfree(lp); |
1840 | ||
1841 | return error; | |
1842 | } | |
1843 | ||
1844 | /** | |
1845 | * leaf_dealloc - Deallocate a directory leaf | |
1846 | * @dip: the directory | |
1847 | * @index: the hash table offset in the directory | |
1848 | * @len: the number of pointers to this leaf | |
1849 | * @leaf_no: the leaf number | |
d24a7a43 | 1850 | * last_dealloc: 1 if this is the final dealloc for the leaf, else 0 |
b3b94faa DT |
1851 | * |
1852 | * Returns: errno | |
1853 | */ | |
1854 | ||
cd915493 | 1855 | static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len, |
d24a7a43 | 1856 | u64 leaf_no, int last_dealloc) |
b3b94faa | 1857 | { |
feaa7bba | 1858 | struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); |
c752666c | 1859 | struct gfs2_leaf *tmp_leaf; |
b3b94faa DT |
1860 | struct gfs2_rgrp_list rlist; |
1861 | struct buffer_head *bh, *dibh; | |
cd915493 | 1862 | u64 blk, nblk; |
b3b94faa DT |
1863 | unsigned int rg_blocks = 0, l_blocks = 0; |
1864 | char *ht; | |
cd915493 | 1865 | unsigned int x, size = len * sizeof(u64); |
b3b94faa DT |
1866 | int error; |
1867 | ||
1868 | memset(&rlist, 0, sizeof(struct gfs2_rgrp_list)); | |
1869 | ||
16c5f06f | 1870 | ht = kzalloc(size, GFP_NOFS); |
b3b94faa DT |
1871 | if (!ht) |
1872 | return -ENOMEM; | |
1873 | ||
182fe5ab CG |
1874 | if (!gfs2_alloc_get(dip)) { |
1875 | error = -ENOMEM; | |
1876 | goto out; | |
1877 | } | |
b3b94faa DT |
1878 | |
1879 | error = gfs2_quota_hold(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
1880 | if (error) | |
182fe5ab | 1881 | goto out_put; |
b3b94faa | 1882 | |
6dbd8224 | 1883 | error = gfs2_rindex_hold(sdp, &dip->i_alloc->al_ri_gh); |
b3b94faa DT |
1884 | if (error) |
1885 | goto out_qs; | |
1886 | ||
1887 | /* Count the number of leaves */ | |
1888 | ||
c752666c | 1889 | for (blk = leaf_no; blk; blk = nblk) { |
b3b94faa DT |
1890 | error = get_leaf(dip, blk, &bh); |
1891 | if (error) | |
1892 | goto out_rlist; | |
c752666c SW |
1893 | tmp_leaf = (struct gfs2_leaf *)bh->b_data; |
1894 | nblk = be64_to_cpu(tmp_leaf->lf_next); | |
b3b94faa DT |
1895 | brelse(bh); |
1896 | ||
1897 | gfs2_rlist_add(sdp, &rlist, blk); | |
1898 | l_blocks++; | |
1899 | } | |
1900 | ||
fe6c991c | 1901 | gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE); |
b3b94faa DT |
1902 | |
1903 | for (x = 0; x < rlist.rl_rgrps; x++) { | |
1904 | struct gfs2_rgrpd *rgd; | |
5c676f6d | 1905 | rgd = rlist.rl_ghs[x].gh_gl->gl_object; |
bb8d8a6f | 1906 | rg_blocks += rgd->rd_length; |
b3b94faa DT |
1907 | } |
1908 | ||
1909 | error = gfs2_glock_nq_m(rlist.rl_rgrps, rlist.rl_ghs); | |
1910 | if (error) | |
1911 | goto out_rlist; | |
1912 | ||
1913 | error = gfs2_trans_begin(sdp, | |
5c676f6d | 1914 | rg_blocks + (DIV_ROUND_UP(size, sdp->sd_jbsize) + 1) + |
b3b94faa DT |
1915 | RES_DINODE + RES_STATFS + RES_QUOTA, l_blocks); |
1916 | if (error) | |
1917 | goto out_rg_gunlock; | |
1918 | ||
c752666c | 1919 | for (blk = leaf_no; blk; blk = nblk) { |
b3b94faa DT |
1920 | error = get_leaf(dip, blk, &bh); |
1921 | if (error) | |
1922 | goto out_end_trans; | |
c752666c SW |
1923 | tmp_leaf = (struct gfs2_leaf *)bh->b_data; |
1924 | nblk = be64_to_cpu(tmp_leaf->lf_next); | |
b3b94faa DT |
1925 | brelse(bh); |
1926 | ||
1927 | gfs2_free_meta(dip, blk, 1); | |
77658aad | 1928 | gfs2_add_inode_blocks(&dip->i_inode, -1); |
b3b94faa DT |
1929 | } |
1930 | ||
cd915493 | 1931 | error = gfs2_dir_write_data(dip, ht, index * sizeof(u64), size); |
b3b94faa DT |
1932 | if (error != size) { |
1933 | if (error >= 0) | |
1934 | error = -EIO; | |
1935 | goto out_end_trans; | |
1936 | } | |
1937 | ||
1938 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1939 | if (error) | |
1940 | goto out_end_trans; | |
1941 | ||
d4e9c4c3 | 1942 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
d24a7a43 BP |
1943 | /* On the last dealloc, make this a regular file in case we crash. |
1944 | (We don't want to free these blocks a second time.) */ | |
1945 | if (last_dealloc) | |
1946 | dip->i_inode.i_mode = S_IFREG; | |
539e5d6b | 1947 | gfs2_dinode_out(dip, dibh->b_data); |
b3b94faa DT |
1948 | brelse(dibh); |
1949 | ||
a91ea69f | 1950 | out_end_trans: |
b3b94faa | 1951 | gfs2_trans_end(sdp); |
a91ea69f | 1952 | out_rg_gunlock: |
b3b94faa | 1953 | gfs2_glock_dq_m(rlist.rl_rgrps, rlist.rl_ghs); |
a91ea69f | 1954 | out_rlist: |
b3b94faa | 1955 | gfs2_rlist_free(&rlist); |
6dbd8224 | 1956 | gfs2_glock_dq_uninit(&dip->i_alloc->al_ri_gh); |
a91ea69f | 1957 | out_qs: |
b3b94faa | 1958 | gfs2_quota_unhold(dip); |
182fe5ab | 1959 | out_put: |
b3b94faa | 1960 | gfs2_alloc_put(dip); |
182fe5ab | 1961 | out: |
b3b94faa | 1962 | kfree(ht); |
b3b94faa DT |
1963 | return error; |
1964 | } | |
1965 | ||
1966 | /** | |
1967 | * gfs2_dir_exhash_dealloc - free all the leaf blocks in a directory | |
1968 | * @dip: the directory | |
1969 | * | |
1970 | * Dealloc all on-disk directory leaves to FREEMETA state | |
1971 | * Change on-disk inode type to "regular file" | |
1972 | * | |
1973 | * Returns: errno | |
1974 | */ | |
1975 | ||
1976 | int gfs2_dir_exhash_dealloc(struct gfs2_inode *dip) | |
1977 | { | |
b3b94faa | 1978 | /* Dealloc on-disk leaves to FREEMETA state */ |
d24a7a43 | 1979 | return foreach_leaf(dip); |
b3b94faa DT |
1980 | } |
1981 | ||
1982 | /** | |
1983 | * gfs2_diradd_alloc_required - find if adding entry will require an allocation | |
1984 | * @ip: the file being written to | |
1985 | * @filname: the filename that's going to be added | |
b3b94faa | 1986 | * |
c752666c | 1987 | * Returns: 1 if alloc required, 0 if not, -ve on error |
b3b94faa DT |
1988 | */ |
1989 | ||
4d8012b6 | 1990 | int gfs2_diradd_alloc_required(struct inode *inode, const struct qstr *name) |
b3b94faa | 1991 | { |
c752666c SW |
1992 | struct gfs2_dirent *dent; |
1993 | struct buffer_head *bh; | |
b3b94faa | 1994 | |
c752666c | 1995 | dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space, &bh); |
ed386507 | 1996 | if (!dent) { |
c752666c | 1997 | return 1; |
ed386507 | 1998 | } |
c752666c SW |
1999 | if (IS_ERR(dent)) |
2000 | return PTR_ERR(dent); | |
2001 | brelse(bh); | |
2002 | return 0; | |
b3b94faa DT |
2003 | } |
2004 |