Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
3 | * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved. | |
4 | * | |
5 | * This copyrighted material is made available to anyone wishing to use, | |
6 | * modify, copy, or redistribute it subject to the terms and conditions | |
7 | * of the GNU General Public License v.2. | |
8 | */ | |
9 | ||
10 | /* | |
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 block | |
46 | * pointer in the array that points to the same leaf. In fact, when a directory | |
47 | * is first converted from linear to exhash, all of the pointers point to the | |
48 | * 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 | */ | |
55 | ||
56 | #include <linux/sched.h> | |
57 | #include <linux/slab.h> | |
58 | #include <linux/spinlock.h> | |
59 | #include <linux/completion.h> | |
60 | #include <linux/buffer_head.h> | |
61 | #include <linux/sort.h> | |
62 | #include <asm/semaphore.h> | |
63 | ||
64 | #include "gfs2.h" | |
65 | #include "dir.h" | |
66 | #include "glock.h" | |
67 | #include "inode.h" | |
b3b94faa DT |
68 | #include "meta_io.h" |
69 | #include "quota.h" | |
70 | #include "rgrp.h" | |
71 | #include "trans.h" | |
e13940ba | 72 | #include "bmap.h" |
b3b94faa DT |
73 | |
74 | #define IS_LEAF 1 /* Hashed (leaf) directory */ | |
75 | #define IS_DINODE 2 /* Linear (stuffed dinode block) directory */ | |
76 | ||
77 | #if 1 | |
78 | #define gfs2_disk_hash2offset(h) (((uint64_t)(h)) >> 1) | |
79 | #define gfs2_dir_offset2hash(p) ((uint32_t)(((uint64_t)(p)) << 1)) | |
80 | #else | |
81 | #define gfs2_disk_hash2offset(h) (((uint64_t)(h))) | |
82 | #define gfs2_dir_offset2hash(p) ((uint32_t)(((uint64_t)(p)))) | |
83 | #endif | |
84 | ||
85 | typedef int (*leaf_call_t) (struct gfs2_inode *dip, | |
86 | uint32_t index, uint32_t len, uint64_t leaf_no, | |
87 | void *data); | |
88 | ||
18ec7d5c SW |
89 | int gfs2_dir_get_buffer(struct gfs2_inode *ip, uint64_t block, int new, |
90 | struct buffer_head **bhp) | |
e13940ba SW |
91 | { |
92 | struct buffer_head *bh; | |
93 | int error = 0; | |
94 | ||
95 | if (new) { | |
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)); | |
100 | } else { | |
101 | error = gfs2_meta_read(ip->i_gl, block, DIO_START | DIO_WAIT, &bh); | |
102 | if (error) | |
103 | return error; | |
104 | if (gfs2_metatype_check(ip->i_sbd, bh, GFS2_METATYPE_JD)) { | |
105 | brelse(bh); | |
106 | return -EIO; | |
107 | } | |
108 | } | |
109 | ||
110 | *bhp = bh; | |
111 | return 0; | |
112 | } | |
113 | ||
114 | ||
115 | ||
116 | static int gfs2_dir_write_stuffed(struct gfs2_inode *ip, const char *buf, | |
117 | unsigned int offset, unsigned int size) | |
118 | ||
119 | { | |
120 | struct buffer_head *dibh; | |
121 | int error; | |
122 | ||
123 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
124 | if (error) | |
125 | return error; | |
126 | ||
127 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
128 | memcpy(dibh->b_data + offset + sizeof(struct gfs2_inode), buf, size); | |
129 | if (ip->i_di.di_size < offset + size) | |
130 | ip->i_di.di_size = offset + size; | |
131 | ip->i_di.di_mtime = ip->i_di.di_ctime = get_seconds(); | |
132 | gfs2_dinode_out(&ip->i_di, dibh->b_data); | |
133 | ||
134 | brelse(dibh); | |
135 | ||
136 | return size; | |
137 | } | |
138 | ||
139 | ||
140 | ||
141 | /** | |
142 | * gfs2_dir_write_data - Write directory information to the inode | |
143 | * @ip: The GFS2 inode | |
144 | * @buf: The buffer containing information to be written | |
145 | * @offset: The file offset to start writing at | |
146 | * @size: The amount of data to write | |
147 | * | |
148 | * Returns: The number of bytes correctly written or error code | |
149 | */ | |
150 | static int gfs2_dir_write_data(struct gfs2_inode *ip, const char *buf, | |
151 | uint64_t offset, unsigned int size) | |
152 | { | |
153 | struct gfs2_sbd *sdp = ip->i_sbd; | |
154 | struct buffer_head *dibh; | |
155 | uint64_t lblock, dblock; | |
156 | uint32_t extlen = 0; | |
157 | unsigned int o; | |
158 | int copied = 0; | |
159 | int error = 0; | |
160 | ||
161 | if (!size) | |
162 | return 0; | |
163 | ||
164 | if (gfs2_is_stuffed(ip) && | |
165 | offset + size <= sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode)) | |
166 | return gfs2_dir_write_stuffed(ip, buf, (unsigned int)offset, size); | |
167 | ||
168 | if (gfs2_assert_warn(sdp, gfs2_is_jdata(ip))) | |
169 | return -EINVAL; | |
170 | ||
171 | if (gfs2_is_stuffed(ip)) { | |
172 | error = gfs2_unstuff_dinode(ip, NULL, NULL); | |
173 | if (error) | |
174 | return error; | |
175 | } | |
176 | ||
177 | lblock = offset; | |
178 | o = do_div(lblock, sdp->sd_jbsize) + sizeof(struct gfs2_meta_header); | |
179 | ||
180 | while (copied < size) { | |
181 | unsigned int amount; | |
182 | struct buffer_head *bh; | |
183 | int new; | |
184 | ||
185 | amount = size - copied; | |
186 | if (amount > sdp->sd_sb.sb_bsize - o) | |
187 | amount = sdp->sd_sb.sb_bsize - o; | |
188 | ||
189 | if (!extlen) { | |
190 | new = 1; | |
191 | error = gfs2_block_map(ip, lblock, &new, &dblock, &extlen); | |
192 | if (error) | |
193 | goto fail; | |
194 | error = -EIO; | |
195 | if (gfs2_assert_withdraw(sdp, dblock)) | |
196 | goto fail; | |
197 | } | |
198 | ||
199 | error = gfs2_dir_get_buffer(ip, dblock, (amount == sdp->sd_jbsize) ? 1 : new, &bh); | |
200 | if (error) | |
201 | goto fail; | |
202 | ||
203 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
204 | memcpy(bh->b_data + o, buf, amount); | |
205 | brelse(bh); | |
206 | if (error) | |
207 | goto fail; | |
208 | ||
209 | copied += amount; | |
210 | lblock++; | |
211 | dblock++; | |
212 | extlen--; | |
213 | ||
214 | o = sizeof(struct gfs2_meta_header); | |
215 | } | |
216 | ||
217 | out: | |
218 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
219 | if (error) | |
220 | return error; | |
221 | ||
222 | if (ip->i_di.di_size < offset + copied) | |
223 | ip->i_di.di_size = offset + copied; | |
224 | ip->i_di.di_mtime = ip->i_di.di_ctime = get_seconds(); | |
225 | ||
226 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
227 | gfs2_dinode_out(&ip->i_di, dibh->b_data); | |
228 | brelse(dibh); | |
229 | ||
230 | return copied; | |
231 | fail: | |
232 | if (copied) | |
233 | goto out; | |
234 | return error; | |
235 | } | |
236 | ||
237 | static int gfs2_dir_read_stuffed(struct gfs2_inode *ip, char *buf, | |
238 | unsigned int offset, unsigned int size) | |
239 | { | |
240 | struct buffer_head *dibh; | |
241 | int error; | |
242 | ||
243 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
244 | if (!error) { | |
245 | offset += sizeof(struct gfs2_dinode); | |
246 | memcpy(buf, dibh->b_data + offset, size); | |
247 | brelse(dibh); | |
248 | } | |
249 | ||
250 | return (error) ? error : size; | |
251 | } | |
252 | ||
253 | ||
254 | /** | |
255 | * gfs2_dir_read_data - Read a data from a directory inode | |
256 | * @ip: The GFS2 Inode | |
257 | * @buf: The buffer to place result into | |
258 | * @offset: File offset to begin jdata_readng from | |
259 | * @size: Amount of data to transfer | |
260 | * | |
261 | * Returns: The amount of data actually copied or the error | |
262 | */ | |
263 | static int gfs2_dir_read_data(struct gfs2_inode *ip, char *buf, | |
264 | uint64_t offset, unsigned int size) | |
265 | { | |
266 | struct gfs2_sbd *sdp = ip->i_sbd; | |
267 | uint64_t lblock, dblock; | |
268 | uint32_t extlen = 0; | |
269 | unsigned int o; | |
270 | int copied = 0; | |
271 | int error = 0; | |
272 | ||
273 | if (offset >= ip->i_di.di_size) | |
274 | return 0; | |
275 | ||
276 | if ((offset + size) > ip->i_di.di_size) | |
277 | size = ip->i_di.di_size - offset; | |
278 | ||
279 | if (!size) | |
280 | return 0; | |
281 | ||
282 | if (gfs2_is_stuffed(ip)) | |
283 | return gfs2_dir_read_stuffed(ip, buf, (unsigned int)offset, size); | |
284 | ||
285 | if (gfs2_assert_warn(sdp, gfs2_is_jdata(ip))) | |
286 | return -EINVAL; | |
287 | ||
288 | lblock = offset; | |
289 | o = do_div(lblock, sdp->sd_jbsize) + sizeof(struct gfs2_meta_header); | |
290 | ||
291 | while (copied < size) { | |
292 | unsigned int amount; | |
293 | struct buffer_head *bh; | |
294 | int new; | |
295 | ||
296 | amount = size - copied; | |
297 | if (amount > sdp->sd_sb.sb_bsize - o) | |
298 | amount = sdp->sd_sb.sb_bsize - o; | |
299 | ||
300 | if (!extlen) { | |
301 | new = 0; | |
302 | error = gfs2_block_map(ip, lblock, &new, &dblock, &extlen); | |
303 | if (error) | |
304 | goto fail; | |
305 | } | |
306 | ||
307 | if (extlen > 1) | |
308 | gfs2_meta_ra(ip->i_gl, dblock, extlen); | |
309 | ||
310 | if (dblock) { | |
311 | error = gfs2_dir_get_buffer(ip, dblock, new, &bh); | |
312 | if (error) | |
313 | goto fail; | |
314 | dblock++; | |
315 | extlen--; | |
316 | } else | |
317 | bh = NULL; | |
318 | ||
319 | memcpy(buf, bh->b_data + o, amount); | |
320 | brelse(bh); | |
321 | if (error) | |
322 | goto fail; | |
323 | ||
324 | copied += amount; | |
325 | lblock++; | |
326 | ||
327 | o = sizeof(struct gfs2_meta_header); | |
328 | } | |
329 | ||
330 | return copied; | |
331 | fail: | |
332 | return (copied) ? copied : error; | |
333 | } | |
334 | ||
b3b94faa DT |
335 | /** |
336 | * int gfs2_filecmp - Compare two filenames | |
337 | * @file1: The first filename | |
338 | * @file2: The second filename | |
339 | * @len_of_file2: The length of the second file | |
340 | * | |
341 | * This routine compares two filenames and returns 1 if they are equal. | |
342 | * | |
343 | * Returns: 1 if the files are the same, otherwise 0. | |
344 | */ | |
345 | ||
346 | int gfs2_filecmp(struct qstr *file1, char *file2, int len_of_file2) | |
347 | { | |
348 | if (file1->len != len_of_file2) | |
349 | return 0; | |
350 | if (memcmp(file1->name, file2, file1->len)) | |
351 | return 0; | |
352 | return 1; | |
353 | } | |
354 | ||
355 | /** | |
356 | * dirent_first - Return the first dirent | |
357 | * @dip: the directory | |
358 | * @bh: The buffer | |
359 | * @dent: Pointer to list of dirents | |
360 | * | |
361 | * return first dirent whether bh points to leaf or stuffed dinode | |
362 | * | |
363 | * Returns: IS_LEAF, IS_DINODE, or -errno | |
364 | */ | |
365 | ||
366 | static int dirent_first(struct gfs2_inode *dip, struct buffer_head *bh, | |
367 | struct gfs2_dirent **dent) | |
368 | { | |
369 | struct gfs2_meta_header *h = (struct gfs2_meta_header *)bh->b_data; | |
370 | ||
371 | if (be16_to_cpu(h->mh_type) == GFS2_METATYPE_LF) { | |
372 | if (gfs2_meta_check(dip->i_sbd, bh)) | |
373 | return -EIO; | |
374 | *dent = (struct gfs2_dirent *)(bh->b_data + | |
375 | sizeof(struct gfs2_leaf)); | |
376 | return IS_LEAF; | |
377 | } else { | |
378 | if (gfs2_metatype_check(dip->i_sbd, bh, GFS2_METATYPE_DI)) | |
379 | return -EIO; | |
380 | *dent = (struct gfs2_dirent *)(bh->b_data + | |
381 | sizeof(struct gfs2_dinode)); | |
382 | return IS_DINODE; | |
383 | } | |
384 | } | |
385 | ||
386 | /** | |
387 | * dirent_next - Next dirent | |
388 | * @dip: the directory | |
389 | * @bh: The buffer | |
390 | * @dent: Pointer to list of dirents | |
391 | * | |
392 | * Returns: 0 on success, error code otherwise | |
393 | */ | |
394 | ||
395 | static int dirent_next(struct gfs2_inode *dip, struct buffer_head *bh, | |
396 | struct gfs2_dirent **dent) | |
397 | { | |
398 | struct gfs2_dirent *tmp, *cur; | |
399 | char *bh_end; | |
400 | uint32_t cur_rec_len; | |
401 | ||
402 | cur = *dent; | |
403 | bh_end = bh->b_data + bh->b_size; | |
404 | cur_rec_len = be32_to_cpu(cur->de_rec_len); | |
405 | ||
406 | if ((char *)cur + cur_rec_len >= bh_end) { | |
407 | if ((char *)cur + cur_rec_len > bh_end) { | |
408 | gfs2_consist_inode(dip); | |
409 | return -EIO; | |
410 | } | |
411 | return -ENOENT; | |
412 | } | |
413 | ||
414 | tmp = (struct gfs2_dirent *)((char *)cur + cur_rec_len); | |
415 | ||
416 | if ((char *)tmp + be32_to_cpu(tmp->de_rec_len) > bh_end) { | |
417 | gfs2_consist_inode(dip); | |
418 | return -EIO; | |
419 | } | |
420 | /* Only the first dent could ever have de_inum.no_addr == 0 */ | |
421 | if (!tmp->de_inum.no_addr) { | |
422 | gfs2_consist_inode(dip); | |
423 | return -EIO; | |
424 | } | |
425 | ||
426 | *dent = tmp; | |
427 | ||
428 | return 0; | |
429 | } | |
430 | ||
431 | /** | |
432 | * dirent_del - Delete a dirent | |
433 | * @dip: The GFS2 inode | |
434 | * @bh: The buffer | |
435 | * @prev: The previous dirent | |
436 | * @cur: The current dirent | |
437 | * | |
438 | */ | |
439 | ||
440 | static void dirent_del(struct gfs2_inode *dip, struct buffer_head *bh, | |
441 | struct gfs2_dirent *prev, struct gfs2_dirent *cur) | |
442 | { | |
443 | uint32_t cur_rec_len, prev_rec_len; | |
444 | ||
445 | if (!cur->de_inum.no_addr) { | |
446 | gfs2_consist_inode(dip); | |
447 | return; | |
448 | } | |
449 | ||
d4e9c4c3 | 450 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
b3b94faa DT |
451 | |
452 | /* If there is no prev entry, this is the first entry in the block. | |
453 | The de_rec_len is already as big as it needs to be. Just zero | |
454 | out the inode number and return. */ | |
455 | ||
456 | if (!prev) { | |
457 | cur->de_inum.no_addr = 0; /* No endianess worries */ | |
458 | return; | |
459 | } | |
460 | ||
461 | /* Combine this dentry with the previous one. */ | |
462 | ||
463 | prev_rec_len = be32_to_cpu(prev->de_rec_len); | |
464 | cur_rec_len = be32_to_cpu(cur->de_rec_len); | |
465 | ||
466 | if ((char *)prev + prev_rec_len != (char *)cur) | |
467 | gfs2_consist_inode(dip); | |
468 | if ((char *)cur + cur_rec_len > bh->b_data + bh->b_size) | |
469 | gfs2_consist_inode(dip); | |
470 | ||
471 | prev_rec_len += cur_rec_len; | |
472 | prev->de_rec_len = cpu_to_be32(prev_rec_len); | |
473 | } | |
474 | ||
475 | /** | |
476 | * gfs2_dirent_alloc - Allocate a directory entry | |
477 | * @dip: The GFS2 inode | |
478 | * @bh: The buffer | |
479 | * @name_len: The length of the name | |
480 | * @dent_out: Pointer to list of dirents | |
481 | * | |
482 | * Returns: 0 on success, error code otherwise | |
483 | */ | |
484 | ||
485 | int gfs2_dirent_alloc(struct gfs2_inode *dip, struct buffer_head *bh, | |
486 | int name_len, struct gfs2_dirent **dent_out) | |
487 | { | |
488 | struct gfs2_dirent *dent, *new; | |
489 | unsigned int rec_len = GFS2_DIRENT_SIZE(name_len); | |
490 | unsigned int entries = 0, offset = 0; | |
491 | int type; | |
492 | ||
493 | type = dirent_first(dip, bh, &dent); | |
494 | if (type < 0) | |
495 | return type; | |
496 | ||
497 | if (type == IS_LEAF) { | |
498 | struct gfs2_leaf *leaf = (struct gfs2_leaf *)bh->b_data; | |
499 | entries = be16_to_cpu(leaf->lf_entries); | |
500 | offset = sizeof(struct gfs2_leaf); | |
501 | } else { | |
502 | struct gfs2_dinode *dinode = (struct gfs2_dinode *)bh->b_data; | |
503 | entries = be32_to_cpu(dinode->di_entries); | |
504 | offset = sizeof(struct gfs2_dinode); | |
505 | } | |
506 | ||
507 | if (!entries) { | |
508 | if (dent->de_inum.no_addr) { | |
509 | gfs2_consist_inode(dip); | |
510 | return -EIO; | |
511 | } | |
512 | ||
d4e9c4c3 | 513 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
b3b94faa DT |
514 | |
515 | dent->de_rec_len = bh->b_size - offset; | |
516 | dent->de_rec_len = cpu_to_be32(dent->de_rec_len); | |
517 | dent->de_name_len = name_len; | |
518 | ||
519 | *dent_out = dent; | |
520 | return 0; | |
521 | } | |
522 | ||
523 | do { | |
524 | uint32_t cur_rec_len, cur_name_len; | |
525 | ||
526 | cur_rec_len = be32_to_cpu(dent->de_rec_len); | |
527 | cur_name_len = dent->de_name_len; | |
528 | ||
529 | if ((!dent->de_inum.no_addr && cur_rec_len >= rec_len) || | |
530 | (cur_rec_len >= GFS2_DIRENT_SIZE(cur_name_len) + rec_len)) { | |
d4e9c4c3 | 531 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
b3b94faa DT |
532 | |
533 | if (dent->de_inum.no_addr) { | |
534 | new = (struct gfs2_dirent *)((char *)dent + | |
535 | GFS2_DIRENT_SIZE(cur_name_len)); | |
536 | memset(new, 0, sizeof(struct gfs2_dirent)); | |
537 | ||
538 | new->de_rec_len = cur_rec_len - GFS2_DIRENT_SIZE(cur_name_len); | |
539 | new->de_rec_len = cpu_to_be32(new->de_rec_len); | |
540 | new->de_name_len = name_len; | |
541 | ||
542 | dent->de_rec_len = cur_rec_len - be32_to_cpu(new->de_rec_len); | |
543 | dent->de_rec_len = cpu_to_be32(dent->de_rec_len); | |
544 | ||
545 | *dent_out = new; | |
546 | return 0; | |
547 | } | |
548 | ||
549 | dent->de_name_len = name_len; | |
550 | ||
551 | *dent_out = dent; | |
552 | return 0; | |
553 | } | |
554 | } while (dirent_next(dip, bh, &dent) == 0); | |
555 | ||
556 | return -ENOSPC; | |
557 | } | |
558 | ||
559 | /** | |
560 | * dirent_fits - See if we can fit a entry in this buffer | |
561 | * @dip: The GFS2 inode | |
562 | * @bh: The buffer | |
563 | * @name_len: The length of the name | |
564 | * | |
565 | * Returns: 1 if it can fit, 0 otherwise | |
566 | */ | |
567 | ||
568 | static int dirent_fits(struct gfs2_inode *dip, struct buffer_head *bh, | |
569 | int name_len) | |
570 | { | |
571 | struct gfs2_dirent *dent; | |
572 | unsigned int rec_len = GFS2_DIRENT_SIZE(name_len); | |
573 | unsigned int entries = 0; | |
574 | int type; | |
575 | ||
576 | type = dirent_first(dip, bh, &dent); | |
577 | if (type < 0) | |
578 | return type; | |
579 | ||
580 | if (type == IS_LEAF) { | |
581 | struct gfs2_leaf *leaf = (struct gfs2_leaf *)bh->b_data; | |
582 | entries = be16_to_cpu(leaf->lf_entries); | |
583 | } else { | |
584 | struct gfs2_dinode *dinode = (struct gfs2_dinode *)bh->b_data; | |
585 | entries = be32_to_cpu(dinode->di_entries); | |
586 | } | |
587 | ||
588 | if (!entries) | |
589 | return 1; | |
590 | ||
591 | do { | |
592 | uint32_t cur_rec_len, cur_name_len; | |
593 | ||
594 | cur_rec_len = be32_to_cpu(dent->de_rec_len); | |
595 | cur_name_len = dent->de_name_len; | |
596 | ||
597 | if ((!dent->de_inum.no_addr && cur_rec_len >= rec_len) || | |
598 | (cur_rec_len >= GFS2_DIRENT_SIZE(cur_name_len) + rec_len)) | |
599 | return 1; | |
600 | } while (dirent_next(dip, bh, &dent) == 0); | |
601 | ||
602 | return 0; | |
603 | } | |
604 | ||
605 | static int leaf_search(struct gfs2_inode *dip, struct buffer_head *bh, | |
606 | struct qstr *filename, struct gfs2_dirent **dent_out, | |
607 | struct gfs2_dirent **dent_prev) | |
608 | { | |
609 | uint32_t hash; | |
610 | struct gfs2_dirent *dent, *prev = NULL; | |
611 | unsigned int entries = 0; | |
612 | int type; | |
613 | ||
614 | type = dirent_first(dip, bh, &dent); | |
615 | if (type < 0) | |
616 | return type; | |
617 | ||
618 | if (type == IS_LEAF) { | |
619 | struct gfs2_leaf *leaf = (struct gfs2_leaf *)bh->b_data; | |
620 | entries = be16_to_cpu(leaf->lf_entries); | |
621 | } else if (type == IS_DINODE) { | |
622 | struct gfs2_dinode *dinode = (struct gfs2_dinode *)bh->b_data; | |
623 | entries = be32_to_cpu(dinode->di_entries); | |
624 | } | |
625 | ||
626 | hash = gfs2_disk_hash(filename->name, filename->len); | |
627 | ||
628 | do { | |
629 | if (!dent->de_inum.no_addr) { | |
630 | prev = dent; | |
631 | continue; | |
632 | } | |
633 | ||
634 | if (be32_to_cpu(dent->de_hash) == hash && | |
635 | gfs2_filecmp(filename, (char *)(dent + 1), | |
636 | dent->de_name_len)) { | |
637 | *dent_out = dent; | |
638 | if (dent_prev) | |
639 | *dent_prev = prev; | |
640 | ||
641 | return 0; | |
642 | } | |
643 | ||
644 | prev = dent; | |
645 | } while (dirent_next(dip, bh, &dent) == 0); | |
646 | ||
647 | return -ENOENT; | |
648 | } | |
649 | ||
650 | static int get_leaf(struct gfs2_inode *dip, uint64_t leaf_no, | |
651 | struct buffer_head **bhp) | |
652 | { | |
653 | int error; | |
654 | ||
655 | error = gfs2_meta_read(dip->i_gl, leaf_no, DIO_START | DIO_WAIT, bhp); | |
656 | if (!error && gfs2_metatype_check(dip->i_sbd, *bhp, GFS2_METATYPE_LF)) | |
657 | error = -EIO; | |
658 | ||
659 | return error; | |
660 | } | |
661 | ||
662 | /** | |
663 | * get_leaf_nr - Get a leaf number associated with the index | |
664 | * @dip: The GFS2 inode | |
665 | * @index: | |
666 | * @leaf_out: | |
667 | * | |
668 | * Returns: 0 on success, error code otherwise | |
669 | */ | |
670 | ||
671 | static int get_leaf_nr(struct gfs2_inode *dip, uint32_t index, | |
672 | uint64_t *leaf_out) | |
673 | { | |
674 | uint64_t leaf_no; | |
675 | int error; | |
676 | ||
e13940ba | 677 | error = gfs2_dir_read_data(dip, (char *)&leaf_no, |
b3b94faa DT |
678 | index * sizeof(uint64_t), |
679 | sizeof(uint64_t)); | |
680 | if (error != sizeof(uint64_t)) | |
681 | return (error < 0) ? error : -EIO; | |
682 | ||
683 | *leaf_out = be64_to_cpu(leaf_no); | |
684 | ||
685 | return 0; | |
686 | } | |
687 | ||
688 | static int get_first_leaf(struct gfs2_inode *dip, uint32_t index, | |
689 | struct buffer_head **bh_out) | |
690 | { | |
691 | uint64_t leaf_no; | |
692 | int error; | |
693 | ||
694 | error = get_leaf_nr(dip, index, &leaf_no); | |
695 | if (!error) | |
696 | error = get_leaf(dip, leaf_no, bh_out); | |
697 | ||
698 | return error; | |
699 | } | |
700 | ||
701 | static int get_next_leaf(struct gfs2_inode *dip, struct buffer_head *bh_in, | |
702 | struct buffer_head **bh_out) | |
703 | { | |
704 | struct gfs2_leaf *leaf; | |
705 | int error; | |
706 | ||
707 | leaf = (struct gfs2_leaf *)bh_in->b_data; | |
708 | ||
709 | if (!leaf->lf_next) | |
710 | error = -ENOENT; | |
711 | else | |
712 | error = get_leaf(dip, be64_to_cpu(leaf->lf_next), bh_out); | |
713 | ||
714 | return error; | |
715 | } | |
716 | ||
717 | static int linked_leaf_search(struct gfs2_inode *dip, struct qstr *filename, | |
718 | struct gfs2_dirent **dent_out, | |
719 | struct gfs2_dirent **dent_prev, | |
720 | struct buffer_head **bh_out) | |
721 | { | |
722 | struct buffer_head *bh = NULL, *bh_next; | |
723 | uint32_t hsize, index; | |
724 | uint32_t hash; | |
725 | int error; | |
726 | ||
727 | hsize = 1 << dip->i_di.di_depth; | |
728 | if (hsize * sizeof(uint64_t) != dip->i_di.di_size) { | |
729 | gfs2_consist_inode(dip); | |
730 | return -EIO; | |
731 | } | |
732 | ||
733 | /* Figure out the address of the leaf node. */ | |
734 | ||
735 | hash = gfs2_disk_hash(filename->name, filename->len); | |
736 | index = hash >> (32 - dip->i_di.di_depth); | |
737 | ||
738 | error = get_first_leaf(dip, index, &bh_next); | |
739 | if (error) | |
740 | return error; | |
741 | ||
742 | /* Find the entry */ | |
743 | ||
744 | do { | |
745 | brelse(bh); | |
746 | ||
747 | bh = bh_next; | |
748 | ||
749 | error = leaf_search(dip, bh, filename, dent_out, dent_prev); | |
750 | switch (error) { | |
751 | case 0: | |
752 | *bh_out = bh; | |
753 | return 0; | |
754 | ||
755 | case -ENOENT: | |
756 | break; | |
757 | ||
758 | default: | |
759 | brelse(bh); | |
760 | return error; | |
761 | } | |
762 | ||
763 | error = get_next_leaf(dip, bh, &bh_next); | |
764 | } | |
765 | while (!error); | |
766 | ||
767 | brelse(bh); | |
768 | ||
769 | return error; | |
770 | } | |
771 | ||
772 | /** | |
773 | * dir_make_exhash - Convert a stuffed directory into an ExHash directory | |
774 | * @dip: The GFS2 inode | |
775 | * | |
776 | * Returns: 0 on success, error code otherwise | |
777 | */ | |
778 | ||
779 | static int dir_make_exhash(struct gfs2_inode *dip) | |
780 | { | |
781 | struct gfs2_sbd *sdp = dip->i_sbd; | |
782 | struct gfs2_dirent *dent; | |
783 | struct buffer_head *bh, *dibh; | |
784 | struct gfs2_leaf *leaf; | |
785 | int y; | |
786 | uint32_t x; | |
787 | uint64_t *lp, bn; | |
788 | int error; | |
789 | ||
790 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
791 | if (error) | |
792 | return error; | |
793 | ||
794 | /* Allocate a new block for the first leaf node */ | |
795 | ||
796 | bn = gfs2_alloc_meta(dip); | |
797 | ||
798 | /* Turn over a new leaf */ | |
799 | ||
800 | bh = gfs2_meta_new(dip->i_gl, bn); | |
d4e9c4c3 | 801 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
b3b94faa DT |
802 | gfs2_metatype_set(bh, GFS2_METATYPE_LF, GFS2_FORMAT_LF); |
803 | gfs2_buffer_clear_tail(bh, sizeof(struct gfs2_meta_header)); | |
804 | ||
805 | /* Fill in the leaf structure */ | |
806 | ||
807 | leaf = (struct gfs2_leaf *)bh->b_data; | |
808 | ||
809 | gfs2_assert(sdp, dip->i_di.di_entries < (1 << 16)); | |
810 | ||
811 | leaf->lf_dirent_format = cpu_to_be32(GFS2_FORMAT_DE); | |
812 | leaf->lf_entries = cpu_to_be16(dip->i_di.di_entries); | |
813 | ||
814 | /* Copy dirents */ | |
815 | ||
816 | gfs2_buffer_copy_tail(bh, sizeof(struct gfs2_leaf), dibh, | |
817 | sizeof(struct gfs2_dinode)); | |
818 | ||
819 | /* Find last entry */ | |
820 | ||
821 | x = 0; | |
822 | dirent_first(dip, bh, &dent); | |
823 | ||
824 | do { | |
825 | if (!dent->de_inum.no_addr) | |
826 | continue; | |
827 | if (++x == dip->i_di.di_entries) | |
828 | break; | |
829 | } | |
830 | while (dirent_next(dip, bh, &dent) == 0); | |
831 | ||
832 | /* Adjust the last dirent's record length | |
833 | (Remember that dent still points to the last entry.) */ | |
834 | ||
835 | dent->de_rec_len = be32_to_cpu(dent->de_rec_len) + | |
836 | sizeof(struct gfs2_dinode) - | |
837 | sizeof(struct gfs2_leaf); | |
838 | dent->de_rec_len = cpu_to_be32(dent->de_rec_len); | |
839 | ||
840 | brelse(bh); | |
841 | ||
842 | /* We're done with the new leaf block, now setup the new | |
843 | hash table. */ | |
844 | ||
d4e9c4c3 | 845 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
b3b94faa DT |
846 | gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode)); |
847 | ||
848 | lp = (uint64_t *)(dibh->b_data + sizeof(struct gfs2_dinode)); | |
849 | ||
850 | for (x = sdp->sd_hash_ptrs; x--; lp++) | |
851 | *lp = cpu_to_be64(bn); | |
852 | ||
853 | dip->i_di.di_size = sdp->sd_sb.sb_bsize / 2; | |
854 | dip->i_di.di_blocks++; | |
855 | dip->i_di.di_flags |= GFS2_DIF_EXHASH; | |
856 | dip->i_di.di_payload_format = 0; | |
857 | ||
858 | for (x = sdp->sd_hash_ptrs, y = -1; x; x >>= 1, y++) ; | |
859 | dip->i_di.di_depth = y; | |
860 | ||
861 | gfs2_dinode_out(&dip->i_di, dibh->b_data); | |
862 | ||
863 | brelse(dibh); | |
864 | ||
865 | return 0; | |
866 | } | |
867 | ||
868 | /** | |
869 | * dir_split_leaf - Split a leaf block into two | |
870 | * @dip: The GFS2 inode | |
871 | * @index: | |
872 | * @leaf_no: | |
873 | * | |
874 | * Returns: 0 on success, error code on failure | |
875 | */ | |
876 | ||
877 | static int dir_split_leaf(struct gfs2_inode *dip, uint32_t index, | |
878 | uint64_t leaf_no) | |
879 | { | |
880 | struct buffer_head *nbh, *obh, *dibh; | |
881 | struct gfs2_leaf *nleaf, *oleaf; | |
882 | struct gfs2_dirent *dent, *prev = NULL, *next = NULL, *new; | |
883 | uint32_t start, len, half_len, divider; | |
884 | uint64_t bn, *lp; | |
885 | uint32_t name_len; | |
886 | int x, moved = 0; | |
887 | int error; | |
888 | ||
889 | /* Allocate the new leaf block */ | |
890 | ||
891 | bn = gfs2_alloc_meta(dip); | |
892 | ||
893 | /* Get the new leaf block */ | |
894 | ||
895 | nbh = gfs2_meta_new(dip->i_gl, bn); | |
d4e9c4c3 | 896 | gfs2_trans_add_bh(dip->i_gl, nbh, 1); |
b3b94faa DT |
897 | gfs2_metatype_set(nbh, GFS2_METATYPE_LF, GFS2_FORMAT_LF); |
898 | gfs2_buffer_clear_tail(nbh, sizeof(struct gfs2_meta_header)); | |
899 | ||
900 | nleaf = (struct gfs2_leaf *)nbh->b_data; | |
901 | ||
902 | nleaf->lf_dirent_format = cpu_to_be32(GFS2_FORMAT_DE); | |
903 | ||
904 | /* Get the old leaf block */ | |
905 | ||
906 | error = get_leaf(dip, leaf_no, &obh); | |
907 | if (error) | |
908 | goto fail; | |
909 | ||
d4e9c4c3 | 910 | gfs2_trans_add_bh(dip->i_gl, obh, 1); |
b3b94faa DT |
911 | |
912 | oleaf = (struct gfs2_leaf *)obh->b_data; | |
913 | ||
914 | /* Compute the start and len of leaf pointers in the hash table. */ | |
915 | ||
916 | len = 1 << (dip->i_di.di_depth - be16_to_cpu(oleaf->lf_depth)); | |
917 | half_len = len >> 1; | |
918 | if (!half_len) { | |
919 | gfs2_consist_inode(dip); | |
920 | error = -EIO; | |
921 | goto fail_brelse; | |
922 | } | |
923 | ||
924 | start = (index & ~(len - 1)); | |
925 | ||
926 | /* Change the pointers. | |
927 | Don't bother distinguishing stuffed from non-stuffed. | |
928 | This code is complicated enough already. */ | |
929 | ||
930 | lp = kcalloc(half_len, sizeof(uint64_t), GFP_KERNEL | __GFP_NOFAIL); | |
931 | ||
e13940ba | 932 | error = gfs2_dir_read_data(dip, (char *)lp, start * sizeof(uint64_t), |
b3b94faa DT |
933 | half_len * sizeof(uint64_t)); |
934 | if (error != half_len * sizeof(uint64_t)) { | |
935 | if (error >= 0) | |
936 | error = -EIO; | |
937 | goto fail_lpfree; | |
938 | } | |
939 | ||
940 | /* Change the pointers */ | |
941 | ||
942 | for (x = 0; x < half_len; x++) | |
943 | lp[x] = cpu_to_be64(bn); | |
944 | ||
e13940ba | 945 | error = gfs2_dir_write_data(dip, (char *)lp, start * sizeof(uint64_t), |
b3b94faa DT |
946 | half_len * sizeof(uint64_t)); |
947 | if (error != half_len * sizeof(uint64_t)) { | |
948 | if (error >= 0) | |
949 | error = -EIO; | |
950 | goto fail_lpfree; | |
951 | } | |
952 | ||
953 | kfree(lp); | |
954 | ||
955 | /* Compute the divider */ | |
956 | ||
957 | divider = (start + half_len) << (32 - dip->i_di.di_depth); | |
958 | ||
959 | /* Copy the entries */ | |
960 | ||
961 | dirent_first(dip, obh, &dent); | |
962 | ||
963 | do { | |
964 | next = dent; | |
965 | if (dirent_next(dip, obh, &next)) | |
966 | next = NULL; | |
967 | ||
968 | if (dent->de_inum.no_addr && | |
969 | be32_to_cpu(dent->de_hash) < divider) { | |
970 | name_len = dent->de_name_len; | |
971 | ||
972 | gfs2_dirent_alloc(dip, nbh, name_len, &new); | |
973 | ||
974 | new->de_inum = dent->de_inum; /* No endian worries */ | |
975 | new->de_hash = dent->de_hash; /* No endian worries */ | |
976 | new->de_type = dent->de_type; /* No endian worries */ | |
977 | memcpy((char *)(new + 1), (char *)(dent + 1), | |
978 | name_len); | |
979 | ||
980 | nleaf->lf_entries = be16_to_cpu(nleaf->lf_entries)+1; | |
981 | nleaf->lf_entries = cpu_to_be16(nleaf->lf_entries); | |
982 | ||
983 | dirent_del(dip, obh, prev, dent); | |
984 | ||
985 | if (!oleaf->lf_entries) | |
986 | gfs2_consist_inode(dip); | |
987 | oleaf->lf_entries = be16_to_cpu(oleaf->lf_entries)-1; | |
988 | oleaf->lf_entries = cpu_to_be16(oleaf->lf_entries); | |
989 | ||
990 | if (!prev) | |
991 | prev = dent; | |
992 | ||
993 | moved = 1; | |
994 | } else | |
995 | prev = dent; | |
996 | ||
997 | dent = next; | |
998 | } | |
999 | while (dent); | |
1000 | ||
1001 | /* If none of the entries got moved into the new leaf, | |
1002 | artificially fill in the first entry. */ | |
1003 | ||
1004 | if (!moved) { | |
1005 | gfs2_dirent_alloc(dip, nbh, 0, &new); | |
1006 | new->de_inum.no_addr = 0; | |
1007 | } | |
1008 | ||
1009 | oleaf->lf_depth = be16_to_cpu(oleaf->lf_depth) + 1; | |
1010 | oleaf->lf_depth = cpu_to_be16(oleaf->lf_depth); | |
1011 | nleaf->lf_depth = oleaf->lf_depth; | |
1012 | ||
1013 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1014 | if (!gfs2_assert_withdraw(dip->i_sbd, !error)) { | |
1015 | dip->i_di.di_blocks++; | |
1016 | gfs2_dinode_out(&dip->i_di, dibh->b_data); | |
1017 | brelse(dibh); | |
1018 | } | |
1019 | ||
1020 | brelse(obh); | |
1021 | brelse(nbh); | |
1022 | ||
1023 | return error; | |
1024 | ||
1025 | fail_lpfree: | |
1026 | kfree(lp); | |
1027 | ||
1028 | fail_brelse: | |
1029 | brelse(obh); | |
1030 | ||
1031 | fail: | |
1032 | brelse(nbh); | |
1033 | return error; | |
1034 | } | |
1035 | ||
1036 | /** | |
1037 | * dir_double_exhash - Double size of ExHash table | |
1038 | * @dip: The GFS2 dinode | |
1039 | * | |
1040 | * Returns: 0 on success, error code on failure | |
1041 | */ | |
1042 | ||
1043 | static int dir_double_exhash(struct gfs2_inode *dip) | |
1044 | { | |
1045 | struct gfs2_sbd *sdp = dip->i_sbd; | |
1046 | struct buffer_head *dibh; | |
1047 | uint32_t hsize; | |
1048 | uint64_t *buf; | |
1049 | uint64_t *from, *to; | |
1050 | uint64_t block; | |
1051 | int x; | |
1052 | int error = 0; | |
1053 | ||
1054 | hsize = 1 << dip->i_di.di_depth; | |
1055 | if (hsize * sizeof(uint64_t) != dip->i_di.di_size) { | |
1056 | gfs2_consist_inode(dip); | |
1057 | return -EIO; | |
1058 | } | |
1059 | ||
1060 | /* Allocate both the "from" and "to" buffers in one big chunk */ | |
1061 | ||
1062 | buf = kcalloc(3, sdp->sd_hash_bsize, GFP_KERNEL | __GFP_NOFAIL); | |
1063 | ||
1064 | for (block = dip->i_di.di_size >> sdp->sd_hash_bsize_shift; block--;) { | |
e13940ba | 1065 | error = gfs2_dir_read_data(dip, (char *)buf, |
b3b94faa DT |
1066 | block * sdp->sd_hash_bsize, |
1067 | sdp->sd_hash_bsize); | |
1068 | if (error != sdp->sd_hash_bsize) { | |
1069 | if (error >= 0) | |
1070 | error = -EIO; | |
1071 | goto fail; | |
1072 | } | |
1073 | ||
1074 | from = buf; | |
1075 | to = (uint64_t *)((char *)buf + sdp->sd_hash_bsize); | |
1076 | ||
1077 | for (x = sdp->sd_hash_ptrs; x--; from++) { | |
1078 | *to++ = *from; /* No endianess worries */ | |
1079 | *to++ = *from; | |
1080 | } | |
1081 | ||
e13940ba | 1082 | error = gfs2_dir_write_data(dip, |
b3b94faa DT |
1083 | (char *)buf + sdp->sd_hash_bsize, |
1084 | block * sdp->sd_sb.sb_bsize, | |
1085 | sdp->sd_sb.sb_bsize); | |
1086 | if (error != sdp->sd_sb.sb_bsize) { | |
1087 | if (error >= 0) | |
1088 | error = -EIO; | |
1089 | goto fail; | |
1090 | } | |
1091 | } | |
1092 | ||
1093 | kfree(buf); | |
1094 | ||
1095 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1096 | if (!gfs2_assert_withdraw(sdp, !error)) { | |
1097 | dip->i_di.di_depth++; | |
1098 | gfs2_dinode_out(&dip->i_di, dibh->b_data); | |
1099 | brelse(dibh); | |
1100 | } | |
1101 | ||
1102 | return error; | |
1103 | ||
1104 | fail: | |
1105 | kfree(buf); | |
1106 | ||
1107 | return error; | |
1108 | } | |
1109 | ||
1110 | /** | |
1111 | * compare_dents - compare directory entries by hash value | |
1112 | * @a: first dent | |
1113 | * @b: second dent | |
1114 | * | |
1115 | * When comparing the hash entries of @a to @b: | |
1116 | * gt: returns 1 | |
1117 | * lt: returns -1 | |
1118 | * eq: returns 0 | |
1119 | */ | |
1120 | ||
1121 | static int compare_dents(const void *a, const void *b) | |
1122 | { | |
1123 | struct gfs2_dirent *dent_a, *dent_b; | |
1124 | uint32_t hash_a, hash_b; | |
1125 | int ret = 0; | |
1126 | ||
1127 | dent_a = *(struct gfs2_dirent **)a; | |
1128 | hash_a = dent_a->de_hash; | |
1129 | hash_a = be32_to_cpu(hash_a); | |
1130 | ||
1131 | dent_b = *(struct gfs2_dirent **)b; | |
1132 | hash_b = dent_b->de_hash; | |
1133 | hash_b = be32_to_cpu(hash_b); | |
1134 | ||
1135 | if (hash_a > hash_b) | |
1136 | ret = 1; | |
1137 | else if (hash_a < hash_b) | |
1138 | ret = -1; | |
1139 | else { | |
1140 | unsigned int len_a = dent_a->de_name_len; | |
1141 | unsigned int len_b = dent_b->de_name_len; | |
1142 | ||
1143 | if (len_a > len_b) | |
1144 | ret = 1; | |
1145 | else if (len_a < len_b) | |
1146 | ret = -1; | |
1147 | else | |
1148 | ret = memcmp((char *)(dent_a + 1), | |
1149 | (char *)(dent_b + 1), | |
1150 | len_a); | |
1151 | } | |
1152 | ||
1153 | return ret; | |
1154 | } | |
1155 | ||
1156 | /** | |
1157 | * do_filldir_main - read out directory entries | |
1158 | * @dip: The GFS2 inode | |
1159 | * @offset: The offset in the file to read from | |
1160 | * @opaque: opaque data to pass to filldir | |
1161 | * @filldir: The function to pass entries to | |
1162 | * @darr: an array of struct gfs2_dirent pointers to read | |
1163 | * @entries: the number of entries in darr | |
1164 | * @copied: pointer to int that's non-zero if a entry has been copied out | |
1165 | * | |
1166 | * Jump through some hoops to make sure that if there are hash collsions, | |
1167 | * they are read out at the beginning of a buffer. We want to minimize | |
1168 | * the possibility that they will fall into different readdir buffers or | |
1169 | * that someone will want to seek to that location. | |
1170 | * | |
1171 | * Returns: errno, >0 on exception from filldir | |
1172 | */ | |
1173 | ||
1174 | static int do_filldir_main(struct gfs2_inode *dip, uint64_t *offset, | |
1175 | void *opaque, gfs2_filldir_t filldir, | |
1176 | struct gfs2_dirent **darr, uint32_t entries, | |
1177 | int *copied) | |
1178 | { | |
1179 | struct gfs2_dirent *dent, *dent_next; | |
1180 | struct gfs2_inum inum; | |
1181 | uint64_t off, off_next; | |
1182 | unsigned int x, y; | |
1183 | int run = 0; | |
1184 | int error = 0; | |
1185 | ||
1186 | sort(darr, entries, sizeof(struct gfs2_dirent *), compare_dents, NULL); | |
1187 | ||
1188 | dent_next = darr[0]; | |
1189 | off_next = be32_to_cpu(dent_next->de_hash); | |
1190 | off_next = gfs2_disk_hash2offset(off_next); | |
1191 | ||
1192 | for (x = 0, y = 1; x < entries; x++, y++) { | |
1193 | dent = dent_next; | |
1194 | off = off_next; | |
1195 | ||
1196 | if (y < entries) { | |
1197 | dent_next = darr[y]; | |
1198 | off_next = be32_to_cpu(dent_next->de_hash); | |
1199 | off_next = gfs2_disk_hash2offset(off_next); | |
1200 | ||
1201 | if (off < *offset) | |
1202 | continue; | |
1203 | *offset = off; | |
1204 | ||
1205 | if (off_next == off) { | |
1206 | if (*copied && !run) | |
1207 | return 1; | |
1208 | run = 1; | |
1209 | } else | |
1210 | run = 0; | |
1211 | } else { | |
1212 | if (off < *offset) | |
1213 | continue; | |
1214 | *offset = off; | |
1215 | } | |
1216 | ||
1217 | gfs2_inum_in(&inum, (char *)&dent->de_inum); | |
1218 | ||
1219 | error = filldir(opaque, (char *)(dent + 1), | |
1220 | dent->de_name_len, | |
1221 | off, &inum, | |
1222 | dent->de_type); | |
1223 | if (error) | |
1224 | return 1; | |
1225 | ||
1226 | *copied = 1; | |
1227 | } | |
1228 | ||
1229 | /* Increment the *offset by one, so the next time we come into the | |
1230 | do_filldir fxn, we get the next entry instead of the last one in the | |
1231 | current leaf */ | |
1232 | ||
1233 | (*offset)++; | |
1234 | ||
1235 | return 0; | |
1236 | } | |
1237 | ||
1238 | /** | |
1239 | * do_filldir_single - Read directory entries out of a single block | |
1240 | * @dip: The GFS2 inode | |
1241 | * @offset: The offset in the file to read from | |
1242 | * @opaque: opaque data to pass to filldir | |
1243 | * @filldir: The function to pass entries to | |
1244 | * @bh: the block | |
1245 | * @entries: the number of entries in the block | |
1246 | * @copied: pointer to int that's non-zero if a entry has been copied out | |
1247 | * | |
1248 | * Returns: errno, >0 on exception from filldir | |
1249 | */ | |
1250 | ||
1251 | static int do_filldir_single(struct gfs2_inode *dip, uint64_t *offset, | |
1252 | void *opaque, gfs2_filldir_t filldir, | |
1253 | struct buffer_head *bh, uint32_t entries, | |
1254 | int *copied) | |
1255 | { | |
1256 | struct gfs2_dirent **darr; | |
1257 | struct gfs2_dirent *de; | |
1258 | unsigned int e = 0; | |
1259 | int error; | |
1260 | ||
1261 | if (!entries) | |
1262 | return 0; | |
1263 | ||
1264 | darr = kcalloc(entries, sizeof(struct gfs2_dirent *), GFP_KERNEL); | |
1265 | if (!darr) | |
1266 | return -ENOMEM; | |
1267 | ||
1268 | dirent_first(dip, bh, &de); | |
1269 | do { | |
1270 | if (!de->de_inum.no_addr) | |
1271 | continue; | |
1272 | if (e >= entries) { | |
1273 | gfs2_consist_inode(dip); | |
1274 | error = -EIO; | |
1275 | goto out; | |
1276 | } | |
1277 | darr[e++] = de; | |
1278 | } | |
1279 | while (dirent_next(dip, bh, &de) == 0); | |
1280 | ||
1281 | if (e != entries) { | |
1282 | gfs2_consist_inode(dip); | |
1283 | error = -EIO; | |
1284 | goto out; | |
1285 | } | |
1286 | ||
1287 | error = do_filldir_main(dip, offset, opaque, filldir, darr, | |
1288 | entries, copied); | |
1289 | ||
1290 | out: | |
1291 | kfree(darr); | |
1292 | ||
1293 | return error; | |
1294 | } | |
1295 | ||
1296 | /** | |
1297 | * do_filldir_multi - Read directory entries out of a linked leaf list | |
1298 | * @dip: The GFS2 inode | |
1299 | * @offset: The offset in the file to read from | |
1300 | * @opaque: opaque data to pass to filldir | |
1301 | * @filldir: The function to pass entries to | |
1302 | * @bh: the first leaf in the list | |
1303 | * @copied: pointer to int that's non-zero if a entry has been copied out | |
1304 | * | |
1305 | * Returns: errno, >0 on exception from filldir | |
1306 | */ | |
1307 | ||
1308 | static int do_filldir_multi(struct gfs2_inode *dip, uint64_t *offset, | |
1309 | void *opaque, gfs2_filldir_t filldir, | |
1310 | struct buffer_head *bh, int *copied) | |
1311 | { | |
1312 | struct buffer_head **larr = NULL; | |
1313 | struct gfs2_dirent **darr; | |
1314 | struct gfs2_leaf *leaf; | |
1315 | struct buffer_head *tmp_bh; | |
1316 | struct gfs2_dirent *de; | |
1317 | unsigned int entries, e = 0; | |
1318 | unsigned int leaves = 0, l = 0; | |
1319 | unsigned int x; | |
1320 | uint64_t ln; | |
1321 | int error = 0; | |
1322 | ||
1323 | /* Count leaves and entries */ | |
1324 | ||
1325 | leaf = (struct gfs2_leaf *)bh->b_data; | |
1326 | entries = be16_to_cpu(leaf->lf_entries); | |
1327 | ln = leaf->lf_next; | |
1328 | ||
1329 | while (ln) { | |
1330 | ln = be64_to_cpu(ln); | |
1331 | ||
1332 | error = get_leaf(dip, ln, &tmp_bh); | |
1333 | if (error) | |
1334 | return error; | |
1335 | ||
1336 | leaf = (struct gfs2_leaf *)tmp_bh->b_data; | |
1337 | if (leaf->lf_entries) { | |
1338 | entries += be16_to_cpu(leaf->lf_entries); | |
1339 | leaves++; | |
1340 | } | |
1341 | ln = leaf->lf_next; | |
1342 | ||
1343 | brelse(tmp_bh); | |
1344 | } | |
1345 | ||
1346 | if (!entries) | |
1347 | return 0; | |
1348 | ||
1349 | if (leaves) { | |
1350 | larr = kcalloc(leaves, sizeof(struct buffer_head *),GFP_KERNEL); | |
1351 | if (!larr) | |
1352 | return -ENOMEM; | |
1353 | } | |
1354 | ||
1355 | darr = kcalloc(entries, sizeof(struct gfs2_dirent *), GFP_KERNEL); | |
1356 | if (!darr) { | |
1357 | kfree(larr); | |
1358 | return -ENOMEM; | |
1359 | } | |
1360 | ||
1361 | leaf = (struct gfs2_leaf *)bh->b_data; | |
1362 | if (leaf->lf_entries) { | |
1363 | dirent_first(dip, bh, &de); | |
1364 | do { | |
1365 | if (!de->de_inum.no_addr) | |
1366 | continue; | |
1367 | if (e >= entries) { | |
1368 | gfs2_consist_inode(dip); | |
1369 | error = -EIO; | |
1370 | goto out; | |
1371 | } | |
1372 | darr[e++] = de; | |
1373 | } | |
1374 | while (dirent_next(dip, bh, &de) == 0); | |
1375 | } | |
1376 | ln = leaf->lf_next; | |
1377 | ||
1378 | while (ln) { | |
1379 | ln = be64_to_cpu(ln); | |
1380 | ||
1381 | error = get_leaf(dip, ln, &tmp_bh); | |
1382 | if (error) | |
1383 | goto out; | |
1384 | ||
1385 | leaf = (struct gfs2_leaf *)tmp_bh->b_data; | |
1386 | if (leaf->lf_entries) { | |
1387 | dirent_first(dip, tmp_bh, &de); | |
1388 | do { | |
1389 | if (!de->de_inum.no_addr) | |
1390 | continue; | |
1391 | if (e >= entries) { | |
1392 | gfs2_consist_inode(dip); | |
1393 | error = -EIO; | |
1394 | goto out; | |
1395 | } | |
1396 | darr[e++] = de; | |
1397 | } | |
1398 | while (dirent_next(dip, tmp_bh, &de) == 0); | |
1399 | ||
1400 | larr[l++] = tmp_bh; | |
1401 | ||
1402 | ln = leaf->lf_next; | |
1403 | } else { | |
1404 | ln = leaf->lf_next; | |
1405 | brelse(tmp_bh); | |
1406 | } | |
1407 | } | |
1408 | ||
1409 | if (gfs2_assert_withdraw(dip->i_sbd, l == leaves)) { | |
1410 | error = -EIO; | |
1411 | goto out; | |
1412 | } | |
1413 | if (e != entries) { | |
1414 | gfs2_consist_inode(dip); | |
1415 | error = -EIO; | |
1416 | goto out; | |
1417 | } | |
1418 | ||
1419 | error = do_filldir_main(dip, offset, opaque, filldir, darr, | |
1420 | entries, copied); | |
1421 | ||
1422 | out: | |
1423 | kfree(darr); | |
1424 | for (x = 0; x < l; x++) | |
1425 | brelse(larr[x]); | |
1426 | kfree(larr); | |
1427 | ||
1428 | return error; | |
1429 | } | |
1430 | ||
1431 | /** | |
1432 | * dir_e_search - Search exhash (leaf) dir for inode matching name | |
1433 | * @dip: The GFS2 inode | |
1434 | * @filename: Filename string | |
1435 | * @inode: If non-NULL, function fills with formal inode # and block address | |
1436 | * @type: If non-NULL, function fills with DT_... dinode type | |
1437 | * | |
1438 | * Returns: | |
1439 | */ | |
1440 | ||
1441 | static int dir_e_search(struct gfs2_inode *dip, struct qstr *filename, | |
1442 | struct gfs2_inum *inum, unsigned int *type) | |
1443 | { | |
1444 | struct buffer_head *bh; | |
1445 | struct gfs2_dirent *dent; | |
1446 | int error; | |
1447 | ||
1448 | error = linked_leaf_search(dip, filename, &dent, NULL, &bh); | |
1449 | if (error) | |
1450 | return error; | |
1451 | ||
1452 | if (inum) | |
1453 | gfs2_inum_in(inum, (char *)&dent->de_inum); | |
1454 | if (type) | |
1455 | *type = dent->de_type; | |
1456 | ||
1457 | brelse(bh); | |
1458 | ||
1459 | return 0; | |
1460 | } | |
1461 | ||
1462 | static int dir_e_add(struct gfs2_inode *dip, struct qstr *filename, | |
1463 | struct gfs2_inum *inum, unsigned int type) | |
1464 | { | |
1465 | struct buffer_head *bh, *nbh, *dibh; | |
1466 | struct gfs2_leaf *leaf, *nleaf; | |
1467 | struct gfs2_dirent *dent; | |
1468 | uint32_t hsize, index; | |
1469 | uint32_t hash; | |
1470 | uint64_t leaf_no, bn; | |
1471 | int error; | |
1472 | ||
1473 | restart: | |
1474 | hsize = 1 << dip->i_di.di_depth; | |
1475 | if (hsize * sizeof(uint64_t) != dip->i_di.di_size) { | |
1476 | gfs2_consist_inode(dip); | |
1477 | return -EIO; | |
1478 | } | |
1479 | ||
1480 | /* Figure out the address of the leaf node. */ | |
1481 | ||
1482 | hash = gfs2_disk_hash(filename->name, filename->len); | |
1483 | index = hash >> (32 - dip->i_di.di_depth); | |
1484 | ||
1485 | error = get_leaf_nr(dip, index, &leaf_no); | |
1486 | if (error) | |
1487 | return error; | |
1488 | ||
1489 | /* Add entry to the leaf */ | |
1490 | ||
1491 | for (;;) { | |
1492 | error = get_leaf(dip, leaf_no, &bh); | |
1493 | if (error) | |
1494 | return error; | |
1495 | ||
1496 | leaf = (struct gfs2_leaf *)bh->b_data; | |
1497 | ||
1498 | if (gfs2_dirent_alloc(dip, bh, filename->len, &dent)) { | |
1499 | ||
1500 | if (be16_to_cpu(leaf->lf_depth) < dip->i_di.di_depth) { | |
1501 | /* Can we split the leaf? */ | |
1502 | ||
1503 | brelse(bh); | |
1504 | ||
1505 | error = dir_split_leaf(dip, index, leaf_no); | |
1506 | if (error) | |
1507 | return error; | |
1508 | ||
1509 | goto restart; | |
1510 | ||
1511 | } else if (dip->i_di.di_depth < GFS2_DIR_MAX_DEPTH) { | |
1512 | /* Can we double the hash table? */ | |
1513 | ||
1514 | brelse(bh); | |
1515 | ||
1516 | error = dir_double_exhash(dip); | |
1517 | if (error) | |
1518 | return error; | |
1519 | ||
1520 | goto restart; | |
1521 | ||
1522 | } else if (leaf->lf_next) { | |
1523 | /* Can we try the next leaf in the list? */ | |
1524 | leaf_no = be64_to_cpu(leaf->lf_next); | |
1525 | brelse(bh); | |
1526 | continue; | |
1527 | ||
1528 | } else { | |
1529 | /* Create a new leaf and add it to the list. */ | |
1530 | ||
1531 | bn = gfs2_alloc_meta(dip); | |
1532 | ||
1533 | nbh = gfs2_meta_new(dip->i_gl, bn); | |
d4e9c4c3 | 1534 | gfs2_trans_add_bh(dip->i_gl, nbh, 1); |
b3b94faa DT |
1535 | gfs2_metatype_set(nbh, |
1536 | GFS2_METATYPE_LF, | |
1537 | GFS2_FORMAT_LF); | |
1538 | gfs2_buffer_clear_tail(nbh, | |
1539 | sizeof(struct gfs2_meta_header)); | |
1540 | ||
d4e9c4c3 | 1541 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
b3b94faa DT |
1542 | leaf->lf_next = cpu_to_be64(bn); |
1543 | ||
1544 | nleaf = (struct gfs2_leaf *)nbh->b_data; | |
1545 | nleaf->lf_depth = leaf->lf_depth; | |
1546 | nleaf->lf_dirent_format = cpu_to_be32(GFS2_FORMAT_DE); | |
1547 | ||
1548 | gfs2_dirent_alloc(dip, nbh, filename->len, | |
1549 | &dent); | |
1550 | ||
1551 | dip->i_di.di_blocks++; | |
1552 | ||
1553 | brelse(bh); | |
1554 | ||
1555 | bh = nbh; | |
1556 | leaf = nleaf; | |
1557 | } | |
1558 | } | |
1559 | ||
1560 | /* If the gfs2_dirent_alloc() succeeded, it pinned the "bh" */ | |
1561 | ||
1562 | gfs2_inum_out(inum, (char *)&dent->de_inum); | |
1563 | dent->de_hash = cpu_to_be32(hash); | |
1564 | dent->de_type = type; | |
1565 | memcpy((char *)(dent + 1), filename->name, filename->len); | |
1566 | ||
1567 | leaf->lf_entries = be16_to_cpu(leaf->lf_entries) + 1; | |
1568 | leaf->lf_entries = cpu_to_be16(leaf->lf_entries); | |
1569 | ||
1570 | brelse(bh); | |
1571 | ||
1572 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1573 | if (error) | |
1574 | return error; | |
1575 | ||
1576 | dip->i_di.di_entries++; | |
1577 | dip->i_di.di_mtime = dip->i_di.di_ctime = get_seconds(); | |
1578 | ||
d4e9c4c3 | 1579 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
b3b94faa DT |
1580 | gfs2_dinode_out(&dip->i_di, dibh->b_data); |
1581 | brelse(dibh); | |
1582 | ||
1583 | return 0; | |
1584 | } | |
1585 | ||
1586 | return -ENOENT; | |
1587 | } | |
1588 | ||
1589 | static int dir_e_del(struct gfs2_inode *dip, struct qstr *filename) | |
1590 | { | |
1591 | struct buffer_head *bh, *dibh; | |
1592 | struct gfs2_dirent *dent, *prev; | |
1593 | struct gfs2_leaf *leaf; | |
1594 | unsigned int entries; | |
1595 | int error; | |
1596 | ||
1597 | error = linked_leaf_search(dip, filename, &dent, &prev, &bh); | |
1598 | if (error == -ENOENT) { | |
1599 | gfs2_consist_inode(dip); | |
1600 | return -EIO; | |
1601 | } | |
1602 | if (error) | |
1603 | return error; | |
1604 | ||
1605 | dirent_del(dip, bh, prev, dent); /* Pins bh */ | |
1606 | ||
1607 | leaf = (struct gfs2_leaf *)bh->b_data; | |
1608 | entries = be16_to_cpu(leaf->lf_entries); | |
1609 | if (!entries) | |
1610 | gfs2_consist_inode(dip); | |
1611 | entries--; | |
1612 | leaf->lf_entries = cpu_to_be16(entries); | |
1613 | ||
1614 | brelse(bh); | |
1615 | ||
1616 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1617 | if (error) | |
1618 | return error; | |
1619 | ||
1620 | if (!dip->i_di.di_entries) | |
1621 | gfs2_consist_inode(dip); | |
1622 | dip->i_di.di_entries--; | |
1623 | dip->i_di.di_mtime = dip->i_di.di_ctime = get_seconds(); | |
1624 | ||
d4e9c4c3 | 1625 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
b3b94faa DT |
1626 | gfs2_dinode_out(&dip->i_di, dibh->b_data); |
1627 | brelse(dibh); | |
1628 | ||
1629 | return 0; | |
1630 | } | |
1631 | ||
1632 | /** | |
1633 | * dir_e_read - Reads the entries from a directory into a filldir buffer | |
1634 | * @dip: dinode pointer | |
1635 | * @offset: the hash of the last entry read shifted to the right once | |
1636 | * @opaque: buffer for the filldir function to fill | |
1637 | * @filldir: points to the filldir function to use | |
1638 | * | |
1639 | * Returns: errno | |
1640 | */ | |
1641 | ||
1642 | static int dir_e_read(struct gfs2_inode *dip, uint64_t *offset, void *opaque, | |
1643 | gfs2_filldir_t filldir) | |
1644 | { | |
1645 | struct gfs2_sbd *sdp = dip->i_sbd; | |
1646 | struct buffer_head *bh; | |
1647 | struct gfs2_leaf leaf; | |
1648 | uint32_t hsize, len; | |
1649 | uint32_t ht_offset, lp_offset, ht_offset_cur = -1; | |
1650 | uint32_t hash, index; | |
1651 | uint64_t *lp; | |
1652 | int copied = 0; | |
1653 | int error = 0; | |
1654 | ||
1655 | hsize = 1 << dip->i_di.di_depth; | |
1656 | if (hsize * sizeof(uint64_t) != dip->i_di.di_size) { | |
1657 | gfs2_consist_inode(dip); | |
1658 | return -EIO; | |
1659 | } | |
1660 | ||
1661 | hash = gfs2_dir_offset2hash(*offset); | |
1662 | index = hash >> (32 - dip->i_di.di_depth); | |
1663 | ||
1664 | lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL); | |
1665 | if (!lp) | |
1666 | return -ENOMEM; | |
1667 | ||
1668 | while (index < hsize) { | |
1669 | lp_offset = index & (sdp->sd_hash_ptrs - 1); | |
1670 | ht_offset = index - lp_offset; | |
1671 | ||
1672 | if (ht_offset_cur != ht_offset) { | |
e13940ba | 1673 | error = gfs2_dir_read_data(dip, (char *)lp, |
b3b94faa DT |
1674 | ht_offset * sizeof(uint64_t), |
1675 | sdp->sd_hash_bsize); | |
1676 | if (error != sdp->sd_hash_bsize) { | |
1677 | if (error >= 0) | |
1678 | error = -EIO; | |
1679 | goto out; | |
1680 | } | |
1681 | ht_offset_cur = ht_offset; | |
1682 | } | |
1683 | ||
1684 | error = get_leaf(dip, be64_to_cpu(lp[lp_offset]), &bh); | |
1685 | if (error) | |
1686 | goto out; | |
1687 | ||
1688 | gfs2_leaf_in(&leaf, bh->b_data); | |
1689 | ||
1690 | if (leaf.lf_next) | |
1691 | error = do_filldir_multi(dip, offset, opaque, filldir, | |
1692 | bh, &copied); | |
1693 | else | |
1694 | error = do_filldir_single(dip, offset, opaque, filldir, | |
1695 | bh, leaf.lf_entries, &copied); | |
1696 | ||
1697 | brelse(bh); | |
1698 | ||
1699 | if (error) { | |
1700 | if (error > 0) | |
1701 | error = 0; | |
1702 | goto out; | |
1703 | } | |
1704 | ||
1705 | len = 1 << (dip->i_di.di_depth - leaf.lf_depth); | |
1706 | index = (index & ~(len - 1)) + len; | |
1707 | } | |
1708 | ||
1709 | out: | |
1710 | kfree(lp); | |
1711 | ||
1712 | return error; | |
1713 | } | |
1714 | ||
1715 | static int dir_e_mvino(struct gfs2_inode *dip, struct qstr *filename, | |
1716 | struct gfs2_inum *inum, unsigned int new_type) | |
1717 | { | |
1718 | struct buffer_head *bh, *dibh; | |
1719 | struct gfs2_dirent *dent; | |
1720 | int error; | |
1721 | ||
1722 | error = linked_leaf_search(dip, filename, &dent, NULL, &bh); | |
1723 | if (error == -ENOENT) { | |
1724 | gfs2_consist_inode(dip); | |
1725 | return -EIO; | |
1726 | } | |
1727 | if (error) | |
1728 | return error; | |
1729 | ||
d4e9c4c3 | 1730 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
b3b94faa DT |
1731 | |
1732 | gfs2_inum_out(inum, (char *)&dent->de_inum); | |
1733 | dent->de_type = new_type; | |
1734 | ||
1735 | brelse(bh); | |
1736 | ||
1737 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1738 | if (error) | |
1739 | return error; | |
1740 | ||
1741 | dip->i_di.di_mtime = dip->i_di.di_ctime = get_seconds(); | |
1742 | ||
d4e9c4c3 | 1743 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
b3b94faa DT |
1744 | gfs2_dinode_out(&dip->i_di, dibh->b_data); |
1745 | brelse(dibh); | |
1746 | ||
1747 | return 0; | |
1748 | } | |
1749 | ||
1750 | /** | |
1751 | * dir_l_search - Search linear (stuffed dinode) dir for inode matching name | |
1752 | * @dip: The GFS2 inode | |
1753 | * @filename: Filename string | |
1754 | * @inode: If non-NULL, function fills with formal inode # and block address | |
1755 | * @type: If non-NULL, function fills with DT_... dinode type | |
1756 | * | |
1757 | * Returns: | |
1758 | */ | |
1759 | ||
1760 | static int dir_l_search(struct gfs2_inode *dip, struct qstr *filename, | |
1761 | struct gfs2_inum *inum, unsigned int *type) | |
1762 | { | |
1763 | struct buffer_head *dibh; | |
1764 | struct gfs2_dirent *dent; | |
1765 | int error; | |
1766 | ||
1767 | if (!gfs2_is_stuffed(dip)) { | |
1768 | gfs2_consist_inode(dip); | |
1769 | return -EIO; | |
1770 | } | |
1771 | ||
1772 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1773 | if (error) | |
1774 | return error; | |
1775 | ||
1776 | error = leaf_search(dip, dibh, filename, &dent, NULL); | |
1777 | if (!error) { | |
1778 | if (inum) | |
1779 | gfs2_inum_in(inum, (char *)&dent->de_inum); | |
1780 | if (type) | |
1781 | *type = dent->de_type; | |
1782 | } | |
1783 | ||
1784 | brelse(dibh); | |
1785 | ||
1786 | return error; | |
1787 | } | |
1788 | ||
1789 | static int dir_l_add(struct gfs2_inode *dip, struct qstr *filename, | |
1790 | struct gfs2_inum *inum, unsigned int type) | |
1791 | { | |
1792 | struct buffer_head *dibh; | |
1793 | struct gfs2_dirent *dent; | |
1794 | int error; | |
1795 | ||
1796 | if (!gfs2_is_stuffed(dip)) { | |
1797 | gfs2_consist_inode(dip); | |
1798 | return -EIO; | |
1799 | } | |
1800 | ||
1801 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1802 | if (error) | |
1803 | return error; | |
1804 | ||
1805 | if (gfs2_dirent_alloc(dip, dibh, filename->len, &dent)) { | |
1806 | brelse(dibh); | |
1807 | ||
1808 | error = dir_make_exhash(dip); | |
1809 | if (!error) | |
1810 | error = dir_e_add(dip, filename, inum, type); | |
1811 | ||
1812 | return error; | |
1813 | } | |
1814 | ||
1815 | /* gfs2_dirent_alloc() pins */ | |
1816 | ||
1817 | gfs2_inum_out(inum, (char *)&dent->de_inum); | |
1818 | dent->de_hash = gfs2_disk_hash(filename->name, filename->len); | |
1819 | dent->de_hash = cpu_to_be32(dent->de_hash); | |
1820 | dent->de_type = type; | |
1821 | memcpy((char *)(dent + 1), filename->name, filename->len); | |
1822 | ||
1823 | dip->i_di.di_entries++; | |
1824 | dip->i_di.di_mtime = dip->i_di.di_ctime = get_seconds(); | |
1825 | ||
1826 | gfs2_dinode_out(&dip->i_di, dibh->b_data); | |
1827 | brelse(dibh); | |
1828 | ||
1829 | return 0; | |
1830 | } | |
1831 | ||
1832 | static int dir_l_del(struct gfs2_inode *dip, struct qstr *filename) | |
1833 | { | |
1834 | struct buffer_head *dibh; | |
1835 | struct gfs2_dirent *dent, *prev; | |
1836 | int error; | |
1837 | ||
1838 | if (!gfs2_is_stuffed(dip)) { | |
1839 | gfs2_consist_inode(dip); | |
1840 | return -EIO; | |
1841 | } | |
1842 | ||
1843 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1844 | if (error) | |
1845 | return error; | |
1846 | ||
1847 | error = leaf_search(dip, dibh, filename, &dent, &prev); | |
1848 | if (error == -ENOENT) { | |
1849 | gfs2_consist_inode(dip); | |
1850 | error = -EIO; | |
1851 | goto out; | |
1852 | } | |
1853 | if (error) | |
1854 | goto out; | |
1855 | ||
1856 | dirent_del(dip, dibh, prev, dent); | |
1857 | ||
1858 | /* dirent_del() pins */ | |
1859 | ||
1860 | if (!dip->i_di.di_entries) | |
1861 | gfs2_consist_inode(dip); | |
1862 | dip->i_di.di_entries--; | |
1863 | ||
1864 | dip->i_di.di_mtime = dip->i_di.di_ctime = get_seconds(); | |
1865 | ||
1866 | gfs2_dinode_out(&dip->i_di, dibh->b_data); | |
1867 | ||
1868 | out: | |
1869 | brelse(dibh); | |
1870 | ||
1871 | return error; | |
1872 | } | |
1873 | ||
1874 | static int dir_l_read(struct gfs2_inode *dip, uint64_t *offset, void *opaque, | |
1875 | gfs2_filldir_t filldir) | |
1876 | { | |
1877 | struct buffer_head *dibh; | |
1878 | int copied = 0; | |
1879 | int error; | |
1880 | ||
1881 | if (!gfs2_is_stuffed(dip)) { | |
1882 | gfs2_consist_inode(dip); | |
1883 | return -EIO; | |
1884 | } | |
1885 | ||
1886 | if (!dip->i_di.di_entries) | |
1887 | return 0; | |
1888 | ||
1889 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1890 | if (error) | |
1891 | return error; | |
1892 | ||
1893 | error = do_filldir_single(dip, offset, | |
1894 | opaque, filldir, | |
1895 | dibh, dip->i_di.di_entries, | |
1896 | &copied); | |
1897 | if (error > 0) | |
1898 | error = 0; | |
1899 | ||
1900 | brelse(dibh); | |
1901 | ||
1902 | return error; | |
1903 | } | |
1904 | ||
1905 | static int dir_l_mvino(struct gfs2_inode *dip, struct qstr *filename, | |
1906 | struct gfs2_inum *inum, unsigned int new_type) | |
1907 | { | |
1908 | struct buffer_head *dibh; | |
1909 | struct gfs2_dirent *dent; | |
1910 | int error; | |
1911 | ||
1912 | if (!gfs2_is_stuffed(dip)) { | |
1913 | gfs2_consist_inode(dip); | |
1914 | return -EIO; | |
1915 | } | |
1916 | ||
1917 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
1918 | if (error) | |
1919 | return error; | |
1920 | ||
1921 | error = leaf_search(dip, dibh, filename, &dent, NULL); | |
1922 | if (error == -ENOENT) { | |
1923 | gfs2_consist_inode(dip); | |
1924 | error = -EIO; | |
1925 | goto out; | |
1926 | } | |
1927 | if (error) | |
1928 | goto out; | |
1929 | ||
d4e9c4c3 | 1930 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
b3b94faa DT |
1931 | |
1932 | gfs2_inum_out(inum, (char *)&dent->de_inum); | |
1933 | dent->de_type = new_type; | |
1934 | ||
1935 | dip->i_di.di_mtime = dip->i_di.di_ctime = get_seconds(); | |
1936 | ||
1937 | gfs2_dinode_out(&dip->i_di, dibh->b_data); | |
1938 | ||
1939 | out: | |
1940 | brelse(dibh); | |
1941 | ||
1942 | return error; | |
1943 | } | |
1944 | ||
1945 | /** | |
1946 | * gfs2_dir_search - Search a directory | |
1947 | * @dip: The GFS2 inode | |
1948 | * @filename: | |
1949 | * @inode: | |
1950 | * | |
1951 | * This routine searches a directory for a file or another directory. | |
1952 | * Assumes a glock is held on dip. | |
1953 | * | |
1954 | * Returns: errno | |
1955 | */ | |
1956 | ||
1957 | int gfs2_dir_search(struct gfs2_inode *dip, struct qstr *filename, | |
1958 | struct gfs2_inum *inum, unsigned int *type) | |
1959 | { | |
1960 | int error; | |
1961 | ||
1962 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) | |
1963 | error = dir_e_search(dip, filename, inum, type); | |
1964 | else | |
1965 | error = dir_l_search(dip, filename, inum, type); | |
1966 | ||
1967 | return error; | |
1968 | } | |
1969 | ||
1970 | /** | |
1971 | * gfs2_dir_add - Add new filename into directory | |
1972 | * @dip: The GFS2 inode | |
1973 | * @filename: The new name | |
1974 | * @inode: The inode number of the entry | |
1975 | * @type: The type of the entry | |
1976 | * | |
1977 | * Returns: 0 on success, error code on failure | |
1978 | */ | |
1979 | ||
1980 | int gfs2_dir_add(struct gfs2_inode *dip, struct qstr *filename, | |
1981 | struct gfs2_inum *inum, unsigned int type) | |
1982 | { | |
1983 | int error; | |
1984 | ||
1985 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) | |
1986 | error = dir_e_add(dip, filename, inum, type); | |
1987 | else | |
1988 | error = dir_l_add(dip, filename, inum, type); | |
1989 | ||
1990 | return error; | |
1991 | } | |
1992 | ||
1993 | /** | |
1994 | * gfs2_dir_del - Delete a directory entry | |
1995 | * @dip: The GFS2 inode | |
1996 | * @filename: The filename | |
1997 | * | |
1998 | * Returns: 0 on success, error code on failure | |
1999 | */ | |
2000 | ||
2001 | int gfs2_dir_del(struct gfs2_inode *dip, struct qstr *filename) | |
2002 | { | |
2003 | int error; | |
2004 | ||
2005 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) | |
2006 | error = dir_e_del(dip, filename); | |
2007 | else | |
2008 | error = dir_l_del(dip, filename); | |
2009 | ||
2010 | return error; | |
2011 | } | |
2012 | ||
2013 | int gfs2_dir_read(struct gfs2_inode *dip, uint64_t *offset, void *opaque, | |
2014 | gfs2_filldir_t filldir) | |
2015 | { | |
2016 | int error; | |
2017 | ||
2018 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) | |
2019 | error = dir_e_read(dip, offset, opaque, filldir); | |
2020 | else | |
2021 | error = dir_l_read(dip, offset, opaque, filldir); | |
2022 | ||
2023 | return error; | |
2024 | } | |
2025 | ||
2026 | /** | |
2027 | * gfs2_dir_mvino - Change inode number of directory entry | |
2028 | * @dip: The GFS2 inode | |
2029 | * @filename: | |
2030 | * @new_inode: | |
2031 | * | |
2032 | * This routine changes the inode number of a directory entry. It's used | |
2033 | * by rename to change ".." when a directory is moved. | |
2034 | * Assumes a glock is held on dvp. | |
2035 | * | |
2036 | * Returns: errno | |
2037 | */ | |
2038 | ||
2039 | int gfs2_dir_mvino(struct gfs2_inode *dip, struct qstr *filename, | |
2040 | struct gfs2_inum *inum, unsigned int new_type) | |
2041 | { | |
2042 | int error; | |
2043 | ||
2044 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) | |
2045 | error = dir_e_mvino(dip, filename, inum, new_type); | |
2046 | else | |
2047 | error = dir_l_mvino(dip, filename, inum, new_type); | |
2048 | ||
2049 | return error; | |
2050 | } | |
2051 | ||
2052 | /** | |
2053 | * foreach_leaf - call a function for each leaf in a directory | |
2054 | * @dip: the directory | |
2055 | * @lc: the function to call for each each | |
2056 | * @data: private data to pass to it | |
2057 | * | |
2058 | * Returns: errno | |
2059 | */ | |
2060 | ||
2061 | static int foreach_leaf(struct gfs2_inode *dip, leaf_call_t lc, void *data) | |
2062 | { | |
2063 | struct gfs2_sbd *sdp = dip->i_sbd; | |
2064 | struct buffer_head *bh; | |
2065 | struct gfs2_leaf leaf; | |
2066 | uint32_t hsize, len; | |
2067 | uint32_t ht_offset, lp_offset, ht_offset_cur = -1; | |
2068 | uint32_t index = 0; | |
2069 | uint64_t *lp; | |
2070 | uint64_t leaf_no; | |
2071 | int error = 0; | |
2072 | ||
2073 | hsize = 1 << dip->i_di.di_depth; | |
2074 | if (hsize * sizeof(uint64_t) != dip->i_di.di_size) { | |
2075 | gfs2_consist_inode(dip); | |
2076 | return -EIO; | |
2077 | } | |
2078 | ||
2079 | lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL); | |
2080 | if (!lp) | |
2081 | return -ENOMEM; | |
2082 | ||
2083 | while (index < hsize) { | |
2084 | lp_offset = index & (sdp->sd_hash_ptrs - 1); | |
2085 | ht_offset = index - lp_offset; | |
2086 | ||
2087 | if (ht_offset_cur != ht_offset) { | |
e13940ba | 2088 | error = gfs2_dir_read_data(dip, (char *)lp, |
b3b94faa DT |
2089 | ht_offset * sizeof(uint64_t), |
2090 | sdp->sd_hash_bsize); | |
2091 | if (error != sdp->sd_hash_bsize) { | |
2092 | if (error >= 0) | |
2093 | error = -EIO; | |
2094 | goto out; | |
2095 | } | |
2096 | ht_offset_cur = ht_offset; | |
2097 | } | |
2098 | ||
2099 | leaf_no = be64_to_cpu(lp[lp_offset]); | |
2100 | if (leaf_no) { | |
2101 | error = get_leaf(dip, leaf_no, &bh); | |
2102 | if (error) | |
2103 | goto out; | |
2104 | gfs2_leaf_in(&leaf, bh->b_data); | |
2105 | brelse(bh); | |
2106 | ||
2107 | len = 1 << (dip->i_di.di_depth - leaf.lf_depth); | |
2108 | ||
2109 | error = lc(dip, index, len, leaf_no, data); | |
2110 | if (error) | |
2111 | goto out; | |
2112 | ||
2113 | index = (index & ~(len - 1)) + len; | |
2114 | } else | |
2115 | index++; | |
2116 | } | |
2117 | ||
2118 | if (index != hsize) { | |
2119 | gfs2_consist_inode(dip); | |
2120 | error = -EIO; | |
2121 | } | |
2122 | ||
2123 | out: | |
2124 | kfree(lp); | |
2125 | ||
2126 | return error; | |
2127 | } | |
2128 | ||
2129 | /** | |
2130 | * leaf_dealloc - Deallocate a directory leaf | |
2131 | * @dip: the directory | |
2132 | * @index: the hash table offset in the directory | |
2133 | * @len: the number of pointers to this leaf | |
2134 | * @leaf_no: the leaf number | |
2135 | * @data: not used | |
2136 | * | |
2137 | * Returns: errno | |
2138 | */ | |
2139 | ||
2140 | static int leaf_dealloc(struct gfs2_inode *dip, uint32_t index, uint32_t len, | |
2141 | uint64_t leaf_no, void *data) | |
2142 | { | |
2143 | struct gfs2_sbd *sdp = dip->i_sbd; | |
2144 | struct gfs2_leaf tmp_leaf; | |
2145 | struct gfs2_rgrp_list rlist; | |
2146 | struct buffer_head *bh, *dibh; | |
2147 | uint64_t blk; | |
2148 | unsigned int rg_blocks = 0, l_blocks = 0; | |
2149 | char *ht; | |
2150 | unsigned int x, size = len * sizeof(uint64_t); | |
2151 | int error; | |
2152 | ||
2153 | memset(&rlist, 0, sizeof(struct gfs2_rgrp_list)); | |
2154 | ||
2155 | ht = kzalloc(size, GFP_KERNEL); | |
2156 | if (!ht) | |
2157 | return -ENOMEM; | |
2158 | ||
2159 | gfs2_alloc_get(dip); | |
2160 | ||
2161 | error = gfs2_quota_hold(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE); | |
2162 | if (error) | |
2163 | goto out; | |
2164 | ||
2165 | error = gfs2_rindex_hold(sdp, &dip->i_alloc.al_ri_gh); | |
2166 | if (error) | |
2167 | goto out_qs; | |
2168 | ||
2169 | /* Count the number of leaves */ | |
2170 | ||
2171 | for (blk = leaf_no; blk; blk = tmp_leaf.lf_next) { | |
2172 | error = get_leaf(dip, blk, &bh); | |
2173 | if (error) | |
2174 | goto out_rlist; | |
2175 | gfs2_leaf_in(&tmp_leaf, (bh)->b_data); | |
2176 | brelse(bh); | |
2177 | ||
2178 | gfs2_rlist_add(sdp, &rlist, blk); | |
2179 | l_blocks++; | |
2180 | } | |
2181 | ||
2182 | gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE, 0); | |
2183 | ||
2184 | for (x = 0; x < rlist.rl_rgrps; x++) { | |
2185 | struct gfs2_rgrpd *rgd; | |
2186 | rgd = get_gl2rgd(rlist.rl_ghs[x].gh_gl); | |
2187 | rg_blocks += rgd->rd_ri.ri_length; | |
2188 | } | |
2189 | ||
2190 | error = gfs2_glock_nq_m(rlist.rl_rgrps, rlist.rl_ghs); | |
2191 | if (error) | |
2192 | goto out_rlist; | |
2193 | ||
2194 | error = gfs2_trans_begin(sdp, | |
2195 | rg_blocks + (DIV_RU(size, sdp->sd_jbsize) + 1) + | |
2196 | RES_DINODE + RES_STATFS + RES_QUOTA, l_blocks); | |
2197 | if (error) | |
2198 | goto out_rg_gunlock; | |
2199 | ||
2200 | for (blk = leaf_no; blk; blk = tmp_leaf.lf_next) { | |
2201 | error = get_leaf(dip, blk, &bh); | |
2202 | if (error) | |
2203 | goto out_end_trans; | |
2204 | gfs2_leaf_in(&tmp_leaf, bh->b_data); | |
2205 | brelse(bh); | |
2206 | ||
2207 | gfs2_free_meta(dip, blk, 1); | |
2208 | ||
2209 | if (!dip->i_di.di_blocks) | |
2210 | gfs2_consist_inode(dip); | |
2211 | dip->i_di.di_blocks--; | |
2212 | } | |
2213 | ||
e13940ba | 2214 | error = gfs2_dir_write_data(dip, ht, index * sizeof(uint64_t), size); |
b3b94faa DT |
2215 | if (error != size) { |
2216 | if (error >= 0) | |
2217 | error = -EIO; | |
2218 | goto out_end_trans; | |
2219 | } | |
2220 | ||
2221 | error = gfs2_meta_inode_buffer(dip, &dibh); | |
2222 | if (error) | |
2223 | goto out_end_trans; | |
2224 | ||
d4e9c4c3 | 2225 | gfs2_trans_add_bh(dip->i_gl, dibh, 1); |
b3b94faa DT |
2226 | gfs2_dinode_out(&dip->i_di, dibh->b_data); |
2227 | brelse(dibh); | |
2228 | ||
2229 | out_end_trans: | |
2230 | gfs2_trans_end(sdp); | |
2231 | ||
2232 | out_rg_gunlock: | |
2233 | gfs2_glock_dq_m(rlist.rl_rgrps, rlist.rl_ghs); | |
2234 | ||
2235 | out_rlist: | |
2236 | gfs2_rlist_free(&rlist); | |
2237 | gfs2_glock_dq_uninit(&dip->i_alloc.al_ri_gh); | |
2238 | ||
2239 | out_qs: | |
2240 | gfs2_quota_unhold(dip); | |
2241 | ||
2242 | out: | |
2243 | gfs2_alloc_put(dip); | |
2244 | kfree(ht); | |
2245 | ||
2246 | return error; | |
2247 | } | |
2248 | ||
2249 | /** | |
2250 | * gfs2_dir_exhash_dealloc - free all the leaf blocks in a directory | |
2251 | * @dip: the directory | |
2252 | * | |
2253 | * Dealloc all on-disk directory leaves to FREEMETA state | |
2254 | * Change on-disk inode type to "regular file" | |
2255 | * | |
2256 | * Returns: errno | |
2257 | */ | |
2258 | ||
2259 | int gfs2_dir_exhash_dealloc(struct gfs2_inode *dip) | |
2260 | { | |
2261 | struct gfs2_sbd *sdp = dip->i_sbd; | |
2262 | struct buffer_head *bh; | |
2263 | int error; | |
2264 | ||
2265 | /* Dealloc on-disk leaves to FREEMETA state */ | |
2266 | error = foreach_leaf(dip, leaf_dealloc, NULL); | |
2267 | if (error) | |
2268 | return error; | |
2269 | ||
2270 | /* Make this a regular file in case we crash. | |
2271 | (We don't want to free these blocks a second time.) */ | |
2272 | ||
2273 | error = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
2274 | if (error) | |
2275 | return error; | |
2276 | ||
2277 | error = gfs2_meta_inode_buffer(dip, &bh); | |
2278 | if (!error) { | |
d4e9c4c3 | 2279 | gfs2_trans_add_bh(dip->i_gl, bh, 1); |
b3b94faa DT |
2280 | ((struct gfs2_dinode *)bh->b_data)->di_mode = cpu_to_be32(S_IFREG); |
2281 | brelse(bh); | |
2282 | } | |
2283 | ||
2284 | gfs2_trans_end(sdp); | |
2285 | ||
2286 | return error; | |
2287 | } | |
2288 | ||
2289 | /** | |
2290 | * gfs2_diradd_alloc_required - find if adding entry will require an allocation | |
2291 | * @ip: the file being written to | |
2292 | * @filname: the filename that's going to be added | |
2293 | * @alloc_required: set to 1 if an alloc is required, 0 otherwise | |
2294 | * | |
2295 | * Returns: errno | |
2296 | */ | |
2297 | ||
2298 | int gfs2_diradd_alloc_required(struct gfs2_inode *dip, struct qstr *filename, | |
2299 | int *alloc_required) | |
2300 | { | |
2301 | struct buffer_head *bh = NULL, *bh_next; | |
2302 | uint32_t hsize, hash, index; | |
2303 | int error = 0; | |
2304 | ||
2305 | *alloc_required = 0; | |
2306 | ||
2307 | if (dip->i_di.di_flags & GFS2_DIF_EXHASH) { | |
2308 | hsize = 1 << dip->i_di.di_depth; | |
2309 | if (hsize * sizeof(uint64_t) != dip->i_di.di_size) { | |
2310 | gfs2_consist_inode(dip); | |
2311 | return -EIO; | |
2312 | } | |
2313 | ||
2314 | hash = gfs2_disk_hash(filename->name, filename->len); | |
2315 | index = hash >> (32 - dip->i_di.di_depth); | |
2316 | ||
2317 | error = get_first_leaf(dip, index, &bh_next); | |
2318 | if (error) | |
2319 | return error; | |
2320 | ||
2321 | do { | |
2322 | brelse(bh); | |
2323 | ||
2324 | bh = bh_next; | |
2325 | ||
2326 | if (dirent_fits(dip, bh, filename->len)) | |
2327 | break; | |
2328 | ||
2329 | error = get_next_leaf(dip, bh, &bh_next); | |
2330 | if (error == -ENOENT) { | |
2331 | *alloc_required = 1; | |
2332 | error = 0; | |
2333 | break; | |
2334 | } | |
2335 | } | |
2336 | while (!error); | |
2337 | ||
2338 | brelse(bh); | |
2339 | } else { | |
2340 | error = gfs2_meta_inode_buffer(dip, &bh); | |
2341 | if (error) | |
2342 | return error; | |
2343 | ||
2344 | if (!dirent_fits(dip, bh, filename->len)) | |
2345 | *alloc_required = 1; | |
2346 | ||
2347 | brelse(bh); | |
2348 | } | |
2349 | ||
2350 | return error; | |
2351 | } | |
2352 |