1 /*
2  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
3  * Copyright (C) 2004-2006 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 version 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_diskflags & GFS2_DIF_EXHASH is true
40  *
41  * Otherwise, the dirents are "linear", within a single stuffed dinode block.
42  *
43  * When the dirents are in leaves, the actual contents of the directory file are
44  * used as an array of 64-bit block pointers pointing to the leaf blocks. The
45  * dirents are NOT in the directory file itself. There can be more than one
46  * block pointer in the array that points to the same leaf. In fact, when a
47  * directory is first converted from linear to exhash, all of the pointers
48  * point to the same leaf.
49  *
50  * When a leaf is completely full, the size of the hash table can be
51  * doubled unless it is already at the maximum size which is hard coded into
52  * GFS2_DIR_MAX_DEPTH. After that, leaves are chained together in a linked list,
53  * but never before the maximum hash table size has been reached.
54  */
55 
56 #include <linux/slab.h>
57 #include <linux/spinlock.h>
58 #include <linux/buffer_head.h>
59 #include <linux/sort.h>
60 #include <linux/gfs2_ondisk.h>
61 #include <linux/crc32.h>
62 #include <linux/vmalloc.h>
63 
64 #include "gfs2.h"
65 #include "incore.h"
66 #include "dir.h"
67 #include "glock.h"
68 #include "inode.h"
69 #include "meta_io.h"
70 #include "quota.h"
71 #include "rgrp.h"
72 #include "trans.h"
73 #include "bmap.h"
74 #include "util.h"
75 
76 #define IS_LEAF     1 /* Hashed (leaf) directory */
77 #define IS_DINODE   2 /* Linear (stuffed dinode block) directory */
78 
79 #define MAX_RA_BLOCKS 32 /* max read-ahead blocks */
80 
81 #define gfs2_disk_hash2offset(h) (((u64)(h)) >> 1)
82 #define gfs2_dir_offset2hash(p) ((u32)(((u64)(p)) << 1))
83 
84 struct qstr gfs2_qdot __read_mostly;
85 struct qstr gfs2_qdotdot __read_mostly;
86 
87 typedef int (*gfs2_dscan_t)(const struct gfs2_dirent *dent,
88 			    const struct qstr *name, void *opaque);
89 
gfs2_dir_get_new_buffer(struct gfs2_inode * ip,u64 block,struct buffer_head ** bhp)90 int gfs2_dir_get_new_buffer(struct gfs2_inode *ip, u64 block,
91 			    struct buffer_head **bhp)
92 {
93 	struct buffer_head *bh;
94 
95 	bh = gfs2_meta_new(ip->i_gl, block);
96 	gfs2_trans_add_bh(ip->i_gl, bh, 1);
97 	gfs2_metatype_set(bh, GFS2_METATYPE_JD, GFS2_FORMAT_JD);
98 	gfs2_buffer_clear_tail(bh, sizeof(struct gfs2_meta_header));
99 	*bhp = bh;
100 	return 0;
101 }
102 
gfs2_dir_get_existing_buffer(struct gfs2_inode * ip,u64 block,struct buffer_head ** bhp)103 static int gfs2_dir_get_existing_buffer(struct gfs2_inode *ip, u64 block,
104 					struct buffer_head **bhp)
105 {
106 	struct buffer_head *bh;
107 	int error;
108 
109 	error = gfs2_meta_read(ip->i_gl, block, DIO_WAIT, &bh);
110 	if (error)
111 		return error;
112 	if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), bh, GFS2_METATYPE_JD)) {
113 		brelse(bh);
114 		return -EIO;
115 	}
116 	*bhp = bh;
117 	return 0;
118 }
119 
gfs2_dir_write_stuffed(struct gfs2_inode * ip,const char * buf,unsigned int offset,unsigned int size)120 static int gfs2_dir_write_stuffed(struct gfs2_inode *ip, const char *buf,
121 				  unsigned int offset, unsigned int size)
122 {
123 	struct buffer_head *dibh;
124 	int error;
125 
126 	error = gfs2_meta_inode_buffer(ip, &dibh);
127 	if (error)
128 		return error;
129 
130 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
131 	memcpy(dibh->b_data + offset + sizeof(struct gfs2_dinode), buf, size);
132 	if (ip->i_inode.i_size < offset + size)
133 		i_size_write(&ip->i_inode, offset + size);
134 	ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
135 	gfs2_dinode_out(ip, dibh->b_data);
136 
137 	brelse(dibh);
138 
139 	return size;
140 }
141 
142 
143 
144 /**
145  * gfs2_dir_write_data - Write directory information to the inode
146  * @ip: The GFS2 inode
147  * @buf: The buffer containing information to be written
148  * @offset: The file offset to start writing at
149  * @size: The amount of data to write
150  *
151  * Returns: The number of bytes correctly written or error code
152  */
gfs2_dir_write_data(struct gfs2_inode * ip,const char * buf,u64 offset,unsigned int size)153 static int gfs2_dir_write_data(struct gfs2_inode *ip, const char *buf,
154 			       u64 offset, unsigned int size)
155 {
156 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
157 	struct buffer_head *dibh;
158 	u64 lblock, dblock;
159 	u32 extlen = 0;
160 	unsigned int o;
161 	int copied = 0;
162 	int error = 0;
163 	int new = 0;
164 
165 	if (!size)
166 		return 0;
167 
168 	if (gfs2_is_stuffed(ip) &&
169 	    offset + size <= sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode))
170 		return gfs2_dir_write_stuffed(ip, buf, (unsigned int)offset,
171 					      size);
172 
173 	if (gfs2_assert_warn(sdp, gfs2_is_jdata(ip)))
174 		return -EINVAL;
175 
176 	if (gfs2_is_stuffed(ip)) {
177 		error = gfs2_unstuff_dinode(ip, NULL);
178 		if (error)
179 			return error;
180 	}
181 
182 	lblock = offset;
183 	o = do_div(lblock, sdp->sd_jbsize) + sizeof(struct gfs2_meta_header);
184 
185 	while (copied < size) {
186 		unsigned int amount;
187 		struct buffer_head *bh;
188 
189 		amount = size - copied;
190 		if (amount > sdp->sd_sb.sb_bsize - o)
191 			amount = sdp->sd_sb.sb_bsize - o;
192 
193 		if (!extlen) {
194 			new = 1;
195 			error = gfs2_extent_map(&ip->i_inode, lblock, &new,
196 						&dblock, &extlen);
197 			if (error)
198 				goto fail;
199 			error = -EIO;
200 			if (gfs2_assert_withdraw(sdp, dblock))
201 				goto fail;
202 		}
203 
204 		if (amount == sdp->sd_jbsize || new)
205 			error = gfs2_dir_get_new_buffer(ip, dblock, &bh);
206 		else
207 			error = gfs2_dir_get_existing_buffer(ip, dblock, &bh);
208 
209 		if (error)
210 			goto fail;
211 
212 		gfs2_trans_add_bh(ip->i_gl, bh, 1);
213 		memcpy(bh->b_data + o, buf, amount);
214 		brelse(bh);
215 
216 		buf += amount;
217 		copied += amount;
218 		lblock++;
219 		dblock++;
220 		extlen--;
221 
222 		o = sizeof(struct gfs2_meta_header);
223 	}
224 
225 out:
226 	error = gfs2_meta_inode_buffer(ip, &dibh);
227 	if (error)
228 		return error;
229 
230 	if (ip->i_inode.i_size < offset + copied)
231 		i_size_write(&ip->i_inode, offset + copied);
232 	ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
233 
234 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
235 	gfs2_dinode_out(ip, dibh->b_data);
236 	brelse(dibh);
237 
238 	return copied;
239 fail:
240 	if (copied)
241 		goto out;
242 	return error;
243 }
244 
gfs2_dir_read_stuffed(struct gfs2_inode * ip,__be64 * buf,unsigned int size)245 static int gfs2_dir_read_stuffed(struct gfs2_inode *ip, __be64 *buf,
246 				 unsigned int size)
247 {
248 	struct buffer_head *dibh;
249 	int error;
250 
251 	error = gfs2_meta_inode_buffer(ip, &dibh);
252 	if (!error) {
253 		memcpy(buf, dibh->b_data + sizeof(struct gfs2_dinode), size);
254 		brelse(dibh);
255 	}
256 
257 	return (error) ? error : size;
258 }
259 
260 
261 /**
262  * gfs2_dir_read_data - Read a data from a directory inode
263  * @ip: The GFS2 Inode
264  * @buf: The buffer to place result into
265  * @size: Amount of data to transfer
266  *
267  * Returns: The amount of data actually copied or the error
268  */
gfs2_dir_read_data(struct gfs2_inode * ip,__be64 * buf,unsigned int size)269 static int gfs2_dir_read_data(struct gfs2_inode *ip, __be64 *buf,
270 			      unsigned int size)
271 {
272 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
273 	u64 lblock, dblock;
274 	u32 extlen = 0;
275 	unsigned int o;
276 	int copied = 0;
277 	int error = 0;
278 
279 	if (gfs2_is_stuffed(ip))
280 		return gfs2_dir_read_stuffed(ip, buf, size);
281 
282 	if (gfs2_assert_warn(sdp, gfs2_is_jdata(ip)))
283 		return -EINVAL;
284 
285 	lblock = 0;
286 	o = do_div(lblock, sdp->sd_jbsize) + sizeof(struct gfs2_meta_header);
287 
288 	while (copied < size) {
289 		unsigned int amount;
290 		struct buffer_head *bh;
291 		int new;
292 
293 		amount = size - copied;
294 		if (amount > sdp->sd_sb.sb_bsize - o)
295 			amount = sdp->sd_sb.sb_bsize - o;
296 
297 		if (!extlen) {
298 			new = 0;
299 			error = gfs2_extent_map(&ip->i_inode, lblock, &new,
300 						&dblock, &extlen);
301 			if (error || !dblock)
302 				goto fail;
303 			BUG_ON(extlen < 1);
304 			bh = gfs2_meta_ra(ip->i_gl, dblock, extlen);
305 		} else {
306 			error = gfs2_meta_read(ip->i_gl, dblock, DIO_WAIT, &bh);
307 			if (error)
308 				goto fail;
309 		}
310 		error = gfs2_metatype_check(sdp, bh, GFS2_METATYPE_JD);
311 		if (error) {
312 			brelse(bh);
313 			goto fail;
314 		}
315 		dblock++;
316 		extlen--;
317 		memcpy(buf, bh->b_data + o, amount);
318 		brelse(bh);
319 		buf += (amount/sizeof(__be64));
320 		copied += amount;
321 		lblock++;
322 		o = sizeof(struct gfs2_meta_header);
323 	}
324 
325 	return copied;
326 fail:
327 	return (copied) ? copied : error;
328 }
329 
330 /**
331  * gfs2_dir_get_hash_table - Get pointer to the dir hash table
332  * @ip: The inode in question
333  *
334  * Returns: The hash table or an error
335  */
336 
gfs2_dir_get_hash_table(struct gfs2_inode * ip)337 static __be64 *gfs2_dir_get_hash_table(struct gfs2_inode *ip)
338 {
339 	struct inode *inode = &ip->i_inode;
340 	int ret;
341 	u32 hsize;
342 	__be64 *hc;
343 
344 	BUG_ON(!(ip->i_diskflags & GFS2_DIF_EXHASH));
345 
346 	hc = ip->i_hash_cache;
347 	if (hc)
348 		return hc;
349 
350 	hsize = 1 << ip->i_depth;
351 	hsize *= sizeof(__be64);
352 	if (hsize != i_size_read(&ip->i_inode)) {
353 		gfs2_consist_inode(ip);
354 		return ERR_PTR(-EIO);
355 	}
356 
357 	hc = kmalloc(hsize, GFP_NOFS);
358 	ret = -ENOMEM;
359 	if (hc == NULL)
360 		return ERR_PTR(-ENOMEM);
361 
362 	ret = gfs2_dir_read_data(ip, hc, hsize);
363 	if (ret < 0) {
364 		kfree(hc);
365 		return ERR_PTR(ret);
366 	}
367 
368 	spin_lock(&inode->i_lock);
369 	if (ip->i_hash_cache)
370 		kfree(hc);
371 	else
372 		ip->i_hash_cache = hc;
373 	spin_unlock(&inode->i_lock);
374 
375 	return ip->i_hash_cache;
376 }
377 
378 /**
379  * gfs2_dir_hash_inval - Invalidate dir hash
380  * @ip: The directory inode
381  *
382  * Must be called with an exclusive glock, or during glock invalidation.
383  */
gfs2_dir_hash_inval(struct gfs2_inode * ip)384 void gfs2_dir_hash_inval(struct gfs2_inode *ip)
385 {
386 	__be64 *hc = ip->i_hash_cache;
387 	ip->i_hash_cache = NULL;
388 	kfree(hc);
389 }
390 
gfs2_dirent_sentinel(const struct gfs2_dirent * dent)391 static inline int gfs2_dirent_sentinel(const struct gfs2_dirent *dent)
392 {
393 	return dent->de_inum.no_addr == 0 || dent->de_inum.no_formal_ino == 0;
394 }
395 
__gfs2_dirent_find(const struct gfs2_dirent * dent,const struct qstr * name,int ret)396 static inline int __gfs2_dirent_find(const struct gfs2_dirent *dent,
397 				     const struct qstr *name, int ret)
398 {
399 	if (!gfs2_dirent_sentinel(dent) &&
400 	    be32_to_cpu(dent->de_hash) == name->hash &&
401 	    be16_to_cpu(dent->de_name_len) == name->len &&
402 	    memcmp(dent+1, name->name, name->len) == 0)
403 		return ret;
404 	return 0;
405 }
406 
gfs2_dirent_find(const struct gfs2_dirent * dent,const struct qstr * name,void * opaque)407 static int gfs2_dirent_find(const struct gfs2_dirent *dent,
408 			    const struct qstr *name,
409 			    void *opaque)
410 {
411 	return __gfs2_dirent_find(dent, name, 1);
412 }
413 
gfs2_dirent_prev(const struct gfs2_dirent * dent,const struct qstr * name,void * opaque)414 static int gfs2_dirent_prev(const struct gfs2_dirent *dent,
415 			    const struct qstr *name,
416 			    void *opaque)
417 {
418 	return __gfs2_dirent_find(dent, name, 2);
419 }
420 
421 /*
422  * name->name holds ptr to start of block.
423  * name->len holds size of block.
424  */
gfs2_dirent_last(const struct gfs2_dirent * dent,const struct qstr * name,void * opaque)425 static int gfs2_dirent_last(const struct gfs2_dirent *dent,
426 			    const struct qstr *name,
427 			    void *opaque)
428 {
429 	const char *start = name->name;
430 	const char *end = (const char *)dent + be16_to_cpu(dent->de_rec_len);
431 	if (name->len == (end - start))
432 		return 1;
433 	return 0;
434 }
435 
gfs2_dirent_find_space(const struct gfs2_dirent * dent,const struct qstr * name,void * opaque)436 static int gfs2_dirent_find_space(const struct gfs2_dirent *dent,
437 				  const struct qstr *name,
438 				  void *opaque)
439 {
440 	unsigned required = GFS2_DIRENT_SIZE(name->len);
441 	unsigned actual = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len));
442 	unsigned totlen = be16_to_cpu(dent->de_rec_len);
443 
444 	if (gfs2_dirent_sentinel(dent))
445 		actual = 0;
446 	if (totlen - actual >= required)
447 		return 1;
448 	return 0;
449 }
450 
451 struct dirent_gather {
452 	const struct gfs2_dirent **pdent;
453 	unsigned offset;
454 };
455 
gfs2_dirent_gather(const struct gfs2_dirent * dent,const struct qstr * name,void * opaque)456 static int gfs2_dirent_gather(const struct gfs2_dirent *dent,
457 			      const struct qstr *name,
458 			      void *opaque)
459 {
460 	struct dirent_gather *g = opaque;
461 	if (!gfs2_dirent_sentinel(dent)) {
462 		g->pdent[g->offset++] = dent;
463 	}
464 	return 0;
465 }
466 
467 /*
468  * Other possible things to check:
469  * - Inode located within filesystem size (and on valid block)
470  * - Valid directory entry type
471  * Not sure how heavy-weight we want to make this... could also check
472  * hash is correct for example, but that would take a lot of extra time.
473  * For now the most important thing is to check that the various sizes
474  * are correct.
475  */
gfs2_check_dirent(struct gfs2_dirent * dent,unsigned int offset,unsigned int size,unsigned int len,int first)476 static int gfs2_check_dirent(struct gfs2_dirent *dent, unsigned int offset,
477 			     unsigned int size, unsigned int len, int first)
478 {
479 	const char *msg = "gfs2_dirent too small";
480 	if (unlikely(size < sizeof(struct gfs2_dirent)))
481 		goto error;
482 	msg = "gfs2_dirent misaligned";
483 	if (unlikely(offset & 0x7))
484 		goto error;
485 	msg = "gfs2_dirent points beyond end of block";
486 	if (unlikely(offset + size > len))
487 		goto error;
488 	msg = "zero inode number";
489 	if (unlikely(!first && gfs2_dirent_sentinel(dent)))
490 		goto error;
491 	msg = "name length is greater than space in dirent";
492 	if (!gfs2_dirent_sentinel(dent) &&
493 	    unlikely(sizeof(struct gfs2_dirent)+be16_to_cpu(dent->de_name_len) >
494 		     size))
495 		goto error;
496 	return 0;
497 error:
498 	printk(KERN_WARNING "gfs2_check_dirent: %s (%s)\n", msg,
499 	       first ? "first in block" : "not first in block");
500 	return -EIO;
501 }
502 
gfs2_dirent_offset(const void * buf)503 static int gfs2_dirent_offset(const void *buf)
504 {
505 	const struct gfs2_meta_header *h = buf;
506 	int offset;
507 
508 	BUG_ON(buf == NULL);
509 
510 	switch(be32_to_cpu(h->mh_type)) {
511 	case GFS2_METATYPE_LF:
512 		offset = sizeof(struct gfs2_leaf);
513 		break;
514 	case GFS2_METATYPE_DI:
515 		offset = sizeof(struct gfs2_dinode);
516 		break;
517 	default:
518 		goto wrong_type;
519 	}
520 	return offset;
521 wrong_type:
522 	printk(KERN_WARNING "gfs2_scan_dirent: wrong block type %u\n",
523 	       be32_to_cpu(h->mh_type));
524 	return -1;
525 }
526 
gfs2_dirent_scan(struct inode * inode,void * buf,unsigned int len,gfs2_dscan_t scan,const struct qstr * name,void * opaque)527 static struct gfs2_dirent *gfs2_dirent_scan(struct inode *inode, void *buf,
528 					    unsigned int len, gfs2_dscan_t scan,
529 					    const struct qstr *name,
530 					    void *opaque)
531 {
532 	struct gfs2_dirent *dent, *prev;
533 	unsigned offset;
534 	unsigned size;
535 	int ret = 0;
536 
537 	ret = gfs2_dirent_offset(buf);
538 	if (ret < 0)
539 		goto consist_inode;
540 
541 	offset = ret;
542 	prev = NULL;
543 	dent = buf + offset;
544 	size = be16_to_cpu(dent->de_rec_len);
545 	if (gfs2_check_dirent(dent, offset, size, len, 1))
546 		goto consist_inode;
547 	do {
548 		ret = scan(dent, name, opaque);
549 		if (ret)
550 			break;
551 		offset += size;
552 		if (offset == len)
553 			break;
554 		prev = dent;
555 		dent = buf + offset;
556 		size = be16_to_cpu(dent->de_rec_len);
557 		if (gfs2_check_dirent(dent, offset, size, len, 0))
558 			goto consist_inode;
559 	} while(1);
560 
561 	switch(ret) {
562 	case 0:
563 		return NULL;
564 	case 1:
565 		return dent;
566 	case 2:
567 		return prev ? prev : dent;
568 	default:
569 		BUG_ON(ret > 0);
570 		return ERR_PTR(ret);
571 	}
572 
573 consist_inode:
574 	gfs2_consist_inode(GFS2_I(inode));
575 	return ERR_PTR(-EIO);
576 }
577 
dirent_check_reclen(struct gfs2_inode * dip,const struct gfs2_dirent * d,const void * end_p)578 static int dirent_check_reclen(struct gfs2_inode *dip,
579 			       const struct gfs2_dirent *d, const void *end_p)
580 {
581 	const void *ptr = d;
582 	u16 rec_len = be16_to_cpu(d->de_rec_len);
583 
584 	if (unlikely(rec_len < sizeof(struct gfs2_dirent)))
585 		goto broken;
586 	ptr += rec_len;
587 	if (ptr < end_p)
588 		return rec_len;
589 	if (ptr == end_p)
590 		return -ENOENT;
591 broken:
592 	gfs2_consist_inode(dip);
593 	return -EIO;
594 }
595 
596 /**
597  * dirent_next - Next dirent
598  * @dip: the directory
599  * @bh: The buffer
600  * @dent: Pointer to list of dirents
601  *
602  * Returns: 0 on success, error code otherwise
603  */
604 
dirent_next(struct gfs2_inode * dip,struct buffer_head * bh,struct gfs2_dirent ** dent)605 static int dirent_next(struct gfs2_inode *dip, struct buffer_head *bh,
606 		       struct gfs2_dirent **dent)
607 {
608 	struct gfs2_dirent *cur = *dent, *tmp;
609 	char *bh_end = bh->b_data + bh->b_size;
610 	int ret;
611 
612 	ret = dirent_check_reclen(dip, cur, bh_end);
613 	if (ret < 0)
614 		return ret;
615 
616 	tmp = (void *)cur + ret;
617 	ret = dirent_check_reclen(dip, tmp, bh_end);
618 	if (ret == -EIO)
619 		return ret;
620 
621         /* Only the first dent could ever have de_inum.no_addr == 0 */
622 	if (gfs2_dirent_sentinel(tmp)) {
623 		gfs2_consist_inode(dip);
624 		return -EIO;
625 	}
626 
627 	*dent = tmp;
628 	return 0;
629 }
630 
631 /**
632  * dirent_del - Delete a dirent
633  * @dip: The GFS2 inode
634  * @bh: The buffer
635  * @prev: The previous dirent
636  * @cur: The current dirent
637  *
638  */
639 
dirent_del(struct gfs2_inode * dip,struct buffer_head * bh,struct gfs2_dirent * prev,struct gfs2_dirent * cur)640 static void dirent_del(struct gfs2_inode *dip, struct buffer_head *bh,
641 		       struct gfs2_dirent *prev, struct gfs2_dirent *cur)
642 {
643 	u16 cur_rec_len, prev_rec_len;
644 
645 	if (gfs2_dirent_sentinel(cur)) {
646 		gfs2_consist_inode(dip);
647 		return;
648 	}
649 
650 	gfs2_trans_add_bh(dip->i_gl, bh, 1);
651 
652 	/* If there is no prev entry, this is the first entry in the block.
653 	   The de_rec_len is already as big as it needs to be.  Just zero
654 	   out the inode number and return.  */
655 
656 	if (!prev) {
657 		cur->de_inum.no_addr = 0;
658 		cur->de_inum.no_formal_ino = 0;
659 		return;
660 	}
661 
662 	/*  Combine this dentry with the previous one.  */
663 
664 	prev_rec_len = be16_to_cpu(prev->de_rec_len);
665 	cur_rec_len = be16_to_cpu(cur->de_rec_len);
666 
667 	if ((char *)prev + prev_rec_len != (char *)cur)
668 		gfs2_consist_inode(dip);
669 	if ((char *)cur + cur_rec_len > bh->b_data + bh->b_size)
670 		gfs2_consist_inode(dip);
671 
672 	prev_rec_len += cur_rec_len;
673 	prev->de_rec_len = cpu_to_be16(prev_rec_len);
674 }
675 
676 /*
677  * Takes a dent from which to grab space as an argument. Returns the
678  * newly created dent.
679  */
gfs2_init_dirent(struct inode * inode,struct gfs2_dirent * dent,const struct qstr * name,struct buffer_head * bh)680 static struct gfs2_dirent *gfs2_init_dirent(struct inode *inode,
681 					    struct gfs2_dirent *dent,
682 					    const struct qstr *name,
683 					    struct buffer_head *bh)
684 {
685 	struct gfs2_inode *ip = GFS2_I(inode);
686 	struct gfs2_dirent *ndent;
687 	unsigned offset = 0, totlen;
688 
689 	if (!gfs2_dirent_sentinel(dent))
690 		offset = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len));
691 	totlen = be16_to_cpu(dent->de_rec_len);
692 	BUG_ON(offset + name->len > totlen);
693 	gfs2_trans_add_bh(ip->i_gl, bh, 1);
694 	ndent = (struct gfs2_dirent *)((char *)dent + offset);
695 	dent->de_rec_len = cpu_to_be16(offset);
696 	gfs2_qstr2dirent(name, totlen - offset, ndent);
697 	return ndent;
698 }
699 
gfs2_dirent_alloc(struct inode * inode,struct buffer_head * bh,const struct qstr * name)700 static struct gfs2_dirent *gfs2_dirent_alloc(struct inode *inode,
701 					     struct buffer_head *bh,
702 					     const struct qstr *name)
703 {
704 	struct gfs2_dirent *dent;
705 	dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size,
706 				gfs2_dirent_find_space, name, NULL);
707 	if (!dent || IS_ERR(dent))
708 		return dent;
709 	return gfs2_init_dirent(inode, dent, name, bh);
710 }
711 
get_leaf(struct gfs2_inode * dip,u64 leaf_no,struct buffer_head ** bhp)712 static int get_leaf(struct gfs2_inode *dip, u64 leaf_no,
713 		    struct buffer_head **bhp)
714 {
715 	int error;
716 
717 	error = gfs2_meta_read(dip->i_gl, leaf_no, DIO_WAIT, bhp);
718 	if (!error && gfs2_metatype_check(GFS2_SB(&dip->i_inode), *bhp, GFS2_METATYPE_LF)) {
719 		/* printk(KERN_INFO "block num=%llu\n", leaf_no); */
720 		error = -EIO;
721 	}
722 
723 	return error;
724 }
725 
726 /**
727  * get_leaf_nr - Get a leaf number associated with the index
728  * @dip: The GFS2 inode
729  * @index:
730  * @leaf_out:
731  *
732  * Returns: 0 on success, error code otherwise
733  */
734 
get_leaf_nr(struct gfs2_inode * dip,u32 index,u64 * leaf_out)735 static int get_leaf_nr(struct gfs2_inode *dip, u32 index,
736 		       u64 *leaf_out)
737 {
738 	__be64 *hash;
739 
740 	hash = gfs2_dir_get_hash_table(dip);
741 	if (IS_ERR(hash))
742 		return PTR_ERR(hash);
743 	*leaf_out = be64_to_cpu(*(hash + index));
744 	return 0;
745 }
746 
get_first_leaf(struct gfs2_inode * dip,u32 index,struct buffer_head ** bh_out)747 static int get_first_leaf(struct gfs2_inode *dip, u32 index,
748 			  struct buffer_head **bh_out)
749 {
750 	u64 leaf_no;
751 	int error;
752 
753 	error = get_leaf_nr(dip, index, &leaf_no);
754 	if (!error)
755 		error = get_leaf(dip, leaf_no, bh_out);
756 
757 	return error;
758 }
759 
gfs2_dirent_search(struct inode * inode,const struct qstr * name,gfs2_dscan_t scan,struct buffer_head ** pbh)760 static struct gfs2_dirent *gfs2_dirent_search(struct inode *inode,
761 					      const struct qstr *name,
762 					      gfs2_dscan_t scan,
763 					      struct buffer_head **pbh)
764 {
765 	struct buffer_head *bh;
766 	struct gfs2_dirent *dent;
767 	struct gfs2_inode *ip = GFS2_I(inode);
768 	int error;
769 
770 	if (ip->i_diskflags & GFS2_DIF_EXHASH) {
771 		struct gfs2_leaf *leaf;
772 		unsigned hsize = 1 << ip->i_depth;
773 		unsigned index;
774 		u64 ln;
775 		if (hsize * sizeof(u64) != i_size_read(inode)) {
776 			gfs2_consist_inode(ip);
777 			return ERR_PTR(-EIO);
778 		}
779 
780 		index = name->hash >> (32 - ip->i_depth);
781 		error = get_first_leaf(ip, index, &bh);
782 		if (error)
783 			return ERR_PTR(error);
784 		do {
785 			dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size,
786 						scan, name, NULL);
787 			if (dent)
788 				goto got_dent;
789 			leaf = (struct gfs2_leaf *)bh->b_data;
790 			ln = be64_to_cpu(leaf->lf_next);
791 			brelse(bh);
792 			if (!ln)
793 				break;
794 
795 			error = get_leaf(ip, ln, &bh);
796 		} while(!error);
797 
798 		return error ? ERR_PTR(error) : NULL;
799 	}
800 
801 
802 	error = gfs2_meta_inode_buffer(ip, &bh);
803 	if (error)
804 		return ERR_PTR(error);
805 	dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size, scan, name, NULL);
806 got_dent:
807 	if (unlikely(dent == NULL || IS_ERR(dent))) {
808 		brelse(bh);
809 		bh = NULL;
810 	}
811 	*pbh = bh;
812 	return dent;
813 }
814 
new_leaf(struct inode * inode,struct buffer_head ** pbh,u16 depth)815 static struct gfs2_leaf *new_leaf(struct inode *inode, struct buffer_head **pbh, u16 depth)
816 {
817 	struct gfs2_inode *ip = GFS2_I(inode);
818 	unsigned int n = 1;
819 	u64 bn;
820 	int error;
821 	struct buffer_head *bh;
822 	struct gfs2_leaf *leaf;
823 	struct gfs2_dirent *dent;
824 	struct qstr name = { .name = "", .len = 0, .hash = 0 };
825 
826 	error = gfs2_alloc_blocks(ip, &bn, &n, 0, NULL);
827 	if (error)
828 		return NULL;
829 	bh = gfs2_meta_new(ip->i_gl, bn);
830 	if (!bh)
831 		return NULL;
832 
833 	gfs2_trans_add_unrevoke(GFS2_SB(inode), bn, 1);
834 	gfs2_trans_add_bh(ip->i_gl, bh, 1);
835 	gfs2_metatype_set(bh, GFS2_METATYPE_LF, GFS2_FORMAT_LF);
836 	leaf = (struct gfs2_leaf *)bh->b_data;
837 	leaf->lf_depth = cpu_to_be16(depth);
838 	leaf->lf_entries = 0;
839 	leaf->lf_dirent_format = cpu_to_be32(GFS2_FORMAT_DE);
840 	leaf->lf_next = 0;
841 	memset(leaf->lf_reserved, 0, sizeof(leaf->lf_reserved));
842 	dent = (struct gfs2_dirent *)(leaf+1);
843 	gfs2_qstr2dirent(&name, bh->b_size - sizeof(struct gfs2_leaf), dent);
844 	*pbh = bh;
845 	return leaf;
846 }
847 
848 /**
849  * dir_make_exhash - Convert a stuffed directory into an ExHash directory
850  * @dip: The GFS2 inode
851  *
852  * Returns: 0 on success, error code otherwise
853  */
854 
dir_make_exhash(struct inode * inode)855 static int dir_make_exhash(struct inode *inode)
856 {
857 	struct gfs2_inode *dip = GFS2_I(inode);
858 	struct gfs2_sbd *sdp = GFS2_SB(inode);
859 	struct gfs2_dirent *dent;
860 	struct qstr args;
861 	struct buffer_head *bh, *dibh;
862 	struct gfs2_leaf *leaf;
863 	int y;
864 	u32 x;
865 	__be64 *lp;
866 	u64 bn;
867 	int error;
868 
869 	error = gfs2_meta_inode_buffer(dip, &dibh);
870 	if (error)
871 		return error;
872 
873 	/*  Turn over a new leaf  */
874 
875 	leaf = new_leaf(inode, &bh, 0);
876 	if (!leaf)
877 		return -ENOSPC;
878 	bn = bh->b_blocknr;
879 
880 	gfs2_assert(sdp, dip->i_entries < (1 << 16));
881 	leaf->lf_entries = cpu_to_be16(dip->i_entries);
882 
883 	/*  Copy dirents  */
884 
885 	gfs2_buffer_copy_tail(bh, sizeof(struct gfs2_leaf), dibh,
886 			     sizeof(struct gfs2_dinode));
887 
888 	/*  Find last entry  */
889 
890 	x = 0;
891 	args.len = bh->b_size - sizeof(struct gfs2_dinode) +
892 		   sizeof(struct gfs2_leaf);
893 	args.name = bh->b_data;
894 	dent = gfs2_dirent_scan(&dip->i_inode, bh->b_data, bh->b_size,
895 				gfs2_dirent_last, &args, NULL);
896 	if (!dent) {
897 		brelse(bh);
898 		brelse(dibh);
899 		return -EIO;
900 	}
901 	if (IS_ERR(dent)) {
902 		brelse(bh);
903 		brelse(dibh);
904 		return PTR_ERR(dent);
905 	}
906 
907 	/*  Adjust the last dirent's record length
908 	   (Remember that dent still points to the last entry.)  */
909 
910 	dent->de_rec_len = cpu_to_be16(be16_to_cpu(dent->de_rec_len) +
911 		sizeof(struct gfs2_dinode) -
912 		sizeof(struct gfs2_leaf));
913 
914 	brelse(bh);
915 
916 	/*  We're done with the new leaf block, now setup the new
917 	    hash table.  */
918 
919 	gfs2_trans_add_bh(dip->i_gl, dibh, 1);
920 	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
921 
922 	lp = (__be64 *)(dibh->b_data + sizeof(struct gfs2_dinode));
923 
924 	for (x = sdp->sd_hash_ptrs; x--; lp++)
925 		*lp = cpu_to_be64(bn);
926 
927 	i_size_write(inode, sdp->sd_sb.sb_bsize / 2);
928 	gfs2_add_inode_blocks(&dip->i_inode, 1);
929 	dip->i_diskflags |= GFS2_DIF_EXHASH;
930 
931 	for (x = sdp->sd_hash_ptrs, y = -1; x; x >>= 1, y++) ;
932 	dip->i_depth = y;
933 
934 	gfs2_dinode_out(dip, dibh->b_data);
935 
936 	brelse(dibh);
937 
938 	return 0;
939 }
940 
941 /**
942  * dir_split_leaf - Split a leaf block into two
943  * @dip: The GFS2 inode
944  * @index:
945  * @leaf_no:
946  *
947  * Returns: 0 on success, error code on failure
948  */
949 
dir_split_leaf(struct inode * inode,const struct qstr * name)950 static int dir_split_leaf(struct inode *inode, const struct qstr *name)
951 {
952 	struct gfs2_inode *dip = GFS2_I(inode);
953 	struct buffer_head *nbh, *obh, *dibh;
954 	struct gfs2_leaf *nleaf, *oleaf;
955 	struct gfs2_dirent *dent = NULL, *prev = NULL, *next = NULL, *new;
956 	u32 start, len, half_len, divider;
957 	u64 bn, leaf_no;
958 	__be64 *lp;
959 	u32 index;
960 	int x, moved = 0;
961 	int error;
962 
963 	index = name->hash >> (32 - dip->i_depth);
964 	error = get_leaf_nr(dip, index, &leaf_no);
965 	if (error)
966 		return error;
967 
968 	/*  Get the old leaf block  */
969 	error = get_leaf(dip, leaf_no, &obh);
970 	if (error)
971 		return error;
972 
973 	oleaf = (struct gfs2_leaf *)obh->b_data;
974 	if (dip->i_depth == be16_to_cpu(oleaf->lf_depth)) {
975 		brelse(obh);
976 		return 1; /* can't split */
977 	}
978 
979 	gfs2_trans_add_bh(dip->i_gl, obh, 1);
980 
981 	nleaf = new_leaf(inode, &nbh, be16_to_cpu(oleaf->lf_depth) + 1);
982 	if (!nleaf) {
983 		brelse(obh);
984 		return -ENOSPC;
985 	}
986 	bn = nbh->b_blocknr;
987 
988 	/*  Compute the start and len of leaf pointers in the hash table.  */
989 	len = 1 << (dip->i_depth - be16_to_cpu(oleaf->lf_depth));
990 	half_len = len >> 1;
991 	if (!half_len) {
992 		printk(KERN_WARNING "i_depth %u lf_depth %u index %u\n", dip->i_depth, be16_to_cpu(oleaf->lf_depth), index);
993 		gfs2_consist_inode(dip);
994 		error = -EIO;
995 		goto fail_brelse;
996 	}
997 
998 	start = (index & ~(len - 1));
999 
1000 	/* Change the pointers.
1001 	   Don't bother distinguishing stuffed from non-stuffed.
1002 	   This code is complicated enough already. */
1003 	lp = kmalloc(half_len * sizeof(__be64), GFP_NOFS);
1004 	if (!lp) {
1005 		error = -ENOMEM;
1006 		goto fail_brelse;
1007 	}
1008 
1009 	/*  Change the pointers  */
1010 	for (x = 0; x < half_len; x++)
1011 		lp[x] = cpu_to_be64(bn);
1012 
1013 	gfs2_dir_hash_inval(dip);
1014 
1015 	error = gfs2_dir_write_data(dip, (char *)lp, start * sizeof(u64),
1016 				    half_len * sizeof(u64));
1017 	if (error != half_len * sizeof(u64)) {
1018 		if (error >= 0)
1019 			error = -EIO;
1020 		goto fail_lpfree;
1021 	}
1022 
1023 	kfree(lp);
1024 
1025 	/*  Compute the divider  */
1026 	divider = (start + half_len) << (32 - dip->i_depth);
1027 
1028 	/*  Copy the entries  */
1029 	dent = (struct gfs2_dirent *)(obh->b_data + sizeof(struct gfs2_leaf));
1030 
1031 	do {
1032 		next = dent;
1033 		if (dirent_next(dip, obh, &next))
1034 			next = NULL;
1035 
1036 		if (!gfs2_dirent_sentinel(dent) &&
1037 		    be32_to_cpu(dent->de_hash) < divider) {
1038 			struct qstr str;
1039 			str.name = (char*)(dent+1);
1040 			str.len = be16_to_cpu(dent->de_name_len);
1041 			str.hash = be32_to_cpu(dent->de_hash);
1042 			new = gfs2_dirent_alloc(inode, nbh, &str);
1043 			if (IS_ERR(new)) {
1044 				error = PTR_ERR(new);
1045 				break;
1046 			}
1047 
1048 			new->de_inum = dent->de_inum; /* No endian worries */
1049 			new->de_type = dent->de_type; /* No endian worries */
1050 			be16_add_cpu(&nleaf->lf_entries, 1);
1051 
1052 			dirent_del(dip, obh, prev, dent);
1053 
1054 			if (!oleaf->lf_entries)
1055 				gfs2_consist_inode(dip);
1056 			be16_add_cpu(&oleaf->lf_entries, -1);
1057 
1058 			if (!prev)
1059 				prev = dent;
1060 
1061 			moved = 1;
1062 		} else {
1063 			prev = dent;
1064 		}
1065 		dent = next;
1066 	} while (dent);
1067 
1068 	oleaf->lf_depth = nleaf->lf_depth;
1069 
1070 	error = gfs2_meta_inode_buffer(dip, &dibh);
1071 	if (!gfs2_assert_withdraw(GFS2_SB(&dip->i_inode), !error)) {
1072 		gfs2_trans_add_bh(dip->i_gl, dibh, 1);
1073 		gfs2_add_inode_blocks(&dip->i_inode, 1);
1074 		gfs2_dinode_out(dip, dibh->b_data);
1075 		brelse(dibh);
1076 	}
1077 
1078 	brelse(obh);
1079 	brelse(nbh);
1080 
1081 	return error;
1082 
1083 fail_lpfree:
1084 	kfree(lp);
1085 
1086 fail_brelse:
1087 	brelse(obh);
1088 	brelse(nbh);
1089 	return error;
1090 }
1091 
1092 /**
1093  * dir_double_exhash - Double size of ExHash table
1094  * @dip: The GFS2 dinode
1095  *
1096  * Returns: 0 on success, error code on failure
1097  */
1098 
dir_double_exhash(struct gfs2_inode * dip)1099 static int dir_double_exhash(struct gfs2_inode *dip)
1100 {
1101 	struct buffer_head *dibh;
1102 	u32 hsize;
1103 	u32 hsize_bytes;
1104 	__be64 *hc;
1105 	__be64 *hc2, *h;
1106 	int x;
1107 	int error = 0;
1108 
1109 	hsize = 1 << dip->i_depth;
1110 	hsize_bytes = hsize * sizeof(__be64);
1111 
1112 	hc = gfs2_dir_get_hash_table(dip);
1113 	if (IS_ERR(hc))
1114 		return PTR_ERR(hc);
1115 
1116 	h = hc2 = kmalloc(hsize_bytes * 2, GFP_NOFS);
1117 	if (!hc2)
1118 		return -ENOMEM;
1119 
1120 	error = gfs2_meta_inode_buffer(dip, &dibh);
1121 	if (error)
1122 		goto out_kfree;
1123 
1124 	for (x = 0; x < hsize; x++) {
1125 		*h++ = *hc;
1126 		*h++ = *hc;
1127 		hc++;
1128 	}
1129 
1130 	error = gfs2_dir_write_data(dip, (char *)hc2, 0, hsize_bytes * 2);
1131 	if (error != (hsize_bytes * 2))
1132 		goto fail;
1133 
1134 	gfs2_dir_hash_inval(dip);
1135 	dip->i_hash_cache = hc2;
1136 	dip->i_depth++;
1137 	gfs2_dinode_out(dip, dibh->b_data);
1138 	brelse(dibh);
1139 	return 0;
1140 
1141 fail:
1142 	/* Replace original hash table & size */
1143 	gfs2_dir_write_data(dip, (char *)hc, 0, hsize_bytes);
1144 	i_size_write(&dip->i_inode, hsize_bytes);
1145 	gfs2_dinode_out(dip, dibh->b_data);
1146 	brelse(dibh);
1147 out_kfree:
1148 	kfree(hc2);
1149 	return error;
1150 }
1151 
1152 /**
1153  * compare_dents - compare directory entries by hash value
1154  * @a: first dent
1155  * @b: second dent
1156  *
1157  * When comparing the hash entries of @a to @b:
1158  *   gt: returns 1
1159  *   lt: returns -1
1160  *   eq: returns 0
1161  */
1162 
compare_dents(const void * a,const void * b)1163 static int compare_dents(const void *a, const void *b)
1164 {
1165 	const struct gfs2_dirent *dent_a, *dent_b;
1166 	u32 hash_a, hash_b;
1167 	int ret = 0;
1168 
1169 	dent_a = *(const struct gfs2_dirent **)a;
1170 	hash_a = be32_to_cpu(dent_a->de_hash);
1171 
1172 	dent_b = *(const struct gfs2_dirent **)b;
1173 	hash_b = be32_to_cpu(dent_b->de_hash);
1174 
1175 	if (hash_a > hash_b)
1176 		ret = 1;
1177 	else if (hash_a < hash_b)
1178 		ret = -1;
1179 	else {
1180 		unsigned int len_a = be16_to_cpu(dent_a->de_name_len);
1181 		unsigned int len_b = be16_to_cpu(dent_b->de_name_len);
1182 
1183 		if (len_a > len_b)
1184 			ret = 1;
1185 		else if (len_a < len_b)
1186 			ret = -1;
1187 		else
1188 			ret = memcmp(dent_a + 1, dent_b + 1, len_a);
1189 	}
1190 
1191 	return ret;
1192 }
1193 
1194 /**
1195  * do_filldir_main - read out directory entries
1196  * @dip: The GFS2 inode
1197  * @offset: The offset in the file to read from
1198  * @opaque: opaque data to pass to filldir
1199  * @filldir: The function to pass entries to
1200  * @darr: an array of struct gfs2_dirent pointers to read
1201  * @entries: the number of entries in darr
1202  * @copied: pointer to int that's non-zero if a entry has been copied out
1203  *
1204  * Jump through some hoops to make sure that if there are hash collsions,
1205  * they are read out at the beginning of a buffer.  We want to minimize
1206  * the possibility that they will fall into different readdir buffers or
1207  * that someone will want to seek to that location.
1208  *
1209  * Returns: errno, >0 on exception from filldir
1210  */
1211 
do_filldir_main(struct gfs2_inode * dip,u64 * offset,void * opaque,filldir_t filldir,const struct gfs2_dirent ** darr,u32 entries,int * copied)1212 static int do_filldir_main(struct gfs2_inode *dip, u64 *offset,
1213 			   void *opaque, filldir_t filldir,
1214 			   const struct gfs2_dirent **darr, u32 entries,
1215 			   int *copied)
1216 {
1217 	const struct gfs2_dirent *dent, *dent_next;
1218 	u64 off, off_next;
1219 	unsigned int x, y;
1220 	int run = 0;
1221 	int error = 0;
1222 
1223 	sort(darr, entries, sizeof(struct gfs2_dirent *), compare_dents, NULL);
1224 
1225 	dent_next = darr[0];
1226 	off_next = be32_to_cpu(dent_next->de_hash);
1227 	off_next = gfs2_disk_hash2offset(off_next);
1228 
1229 	for (x = 0, y = 1; x < entries; x++, y++) {
1230 		dent = dent_next;
1231 		off = off_next;
1232 
1233 		if (y < entries) {
1234 			dent_next = darr[y];
1235 			off_next = be32_to_cpu(dent_next->de_hash);
1236 			off_next = gfs2_disk_hash2offset(off_next);
1237 
1238 			if (off < *offset)
1239 				continue;
1240 			*offset = off;
1241 
1242 			if (off_next == off) {
1243 				if (*copied && !run)
1244 					return 1;
1245 				run = 1;
1246 			} else
1247 				run = 0;
1248 		} else {
1249 			if (off < *offset)
1250 				continue;
1251 			*offset = off;
1252 		}
1253 
1254 		error = filldir(opaque, (const char *)(dent + 1),
1255 				be16_to_cpu(dent->de_name_len),
1256 				off, be64_to_cpu(dent->de_inum.no_addr),
1257 				be16_to_cpu(dent->de_type));
1258 		if (error)
1259 			return 1;
1260 
1261 		*copied = 1;
1262 	}
1263 
1264 	/* Increment the *offset by one, so the next time we come into the
1265 	   do_filldir fxn, we get the next entry instead of the last one in the
1266 	   current leaf */
1267 
1268 	(*offset)++;
1269 
1270 	return 0;
1271 }
1272 
gfs2_alloc_sort_buffer(unsigned size)1273 static void *gfs2_alloc_sort_buffer(unsigned size)
1274 {
1275 	void *ptr = NULL;
1276 
1277 	if (size < KMALLOC_MAX_SIZE)
1278 		ptr = kmalloc(size, GFP_NOFS | __GFP_NOWARN);
1279 	if (!ptr)
1280 		ptr = __vmalloc(size, GFP_NOFS, PAGE_KERNEL);
1281 	return ptr;
1282 }
1283 
gfs2_free_sort_buffer(void * ptr)1284 static void gfs2_free_sort_buffer(void *ptr)
1285 {
1286 	if (is_vmalloc_addr(ptr))
1287 		vfree(ptr);
1288 	else
1289 		kfree(ptr);
1290 }
1291 
gfs2_dir_read_leaf(struct inode * inode,u64 * offset,void * opaque,filldir_t filldir,int * copied,unsigned * depth,u64 leaf_no)1292 static int gfs2_dir_read_leaf(struct inode *inode, u64 *offset, void *opaque,
1293 			      filldir_t filldir, int *copied, unsigned *depth,
1294 			      u64 leaf_no)
1295 {
1296 	struct gfs2_inode *ip = GFS2_I(inode);
1297 	struct gfs2_sbd *sdp = GFS2_SB(inode);
1298 	struct buffer_head *bh;
1299 	struct gfs2_leaf *lf;
1300 	unsigned entries = 0, entries2 = 0;
1301 	unsigned leaves = 0;
1302 	const struct gfs2_dirent **darr, *dent;
1303 	struct dirent_gather g;
1304 	struct buffer_head **larr;
1305 	int leaf = 0;
1306 	int error, i;
1307 	u64 lfn = leaf_no;
1308 
1309 	do {
1310 		error = get_leaf(ip, lfn, &bh);
1311 		if (error)
1312 			goto out;
1313 		lf = (struct gfs2_leaf *)bh->b_data;
1314 		if (leaves == 0)
1315 			*depth = be16_to_cpu(lf->lf_depth);
1316 		entries += be16_to_cpu(lf->lf_entries);
1317 		leaves++;
1318 		lfn = be64_to_cpu(lf->lf_next);
1319 		brelse(bh);
1320 	} while(lfn);
1321 
1322 	if (!entries)
1323 		return 0;
1324 
1325 	error = -ENOMEM;
1326 	/*
1327 	 * The extra 99 entries are not normally used, but are a buffer
1328 	 * zone in case the number of entries in the leaf is corrupt.
1329 	 * 99 is the maximum number of entries that can fit in a single
1330 	 * leaf block.
1331 	 */
1332 	larr = gfs2_alloc_sort_buffer((leaves + entries + 99) * sizeof(void *));
1333 	if (!larr)
1334 		goto out;
1335 	darr = (const struct gfs2_dirent **)(larr + leaves);
1336 	g.pdent = darr;
1337 	g.offset = 0;
1338 	lfn = leaf_no;
1339 
1340 	do {
1341 		error = get_leaf(ip, lfn, &bh);
1342 		if (error)
1343 			goto out_free;
1344 		lf = (struct gfs2_leaf *)bh->b_data;
1345 		lfn = be64_to_cpu(lf->lf_next);
1346 		if (lf->lf_entries) {
1347 			entries2 += be16_to_cpu(lf->lf_entries);
1348 			dent = gfs2_dirent_scan(inode, bh->b_data, bh->b_size,
1349 						gfs2_dirent_gather, NULL, &g);
1350 			error = PTR_ERR(dent);
1351 			if (IS_ERR(dent))
1352 				goto out_free;
1353 			if (entries2 != g.offset) {
1354 				fs_warn(sdp, "Number of entries corrupt in dir "
1355 						"leaf %llu, entries2 (%u) != "
1356 						"g.offset (%u)\n",
1357 					(unsigned long long)bh->b_blocknr,
1358 					entries2, g.offset);
1359 
1360 				error = -EIO;
1361 				goto out_free;
1362 			}
1363 			error = 0;
1364 			larr[leaf++] = bh;
1365 		} else {
1366 			brelse(bh);
1367 		}
1368 	} while(lfn);
1369 
1370 	BUG_ON(entries2 != entries);
1371 	error = do_filldir_main(ip, offset, opaque, filldir, darr,
1372 				entries, copied);
1373 out_free:
1374 	for(i = 0; i < leaf; i++)
1375 		brelse(larr[i]);
1376 	gfs2_free_sort_buffer(larr);
1377 out:
1378 	return error;
1379 }
1380 
1381 /**
1382  * gfs2_dir_readahead - Issue read-ahead requests for leaf blocks.
1383  *
1384  * Note: we can't calculate each index like dir_e_read can because we don't
1385  * have the leaf, and therefore we don't have the depth, and therefore we
1386  * don't have the length. So we have to just read enough ahead to make up
1387  * for the loss of information.
1388  */
gfs2_dir_readahead(struct inode * inode,unsigned hsize,u32 index,struct file_ra_state * f_ra)1389 static void gfs2_dir_readahead(struct inode *inode, unsigned hsize, u32 index,
1390 			       struct file_ra_state *f_ra)
1391 {
1392 	struct gfs2_inode *ip = GFS2_I(inode);
1393 	struct gfs2_glock *gl = ip->i_gl;
1394 	struct buffer_head *bh;
1395 	u64 blocknr = 0, last;
1396 	unsigned count;
1397 
1398 	/* First check if we've already read-ahead for the whole range. */
1399 	if (index + MAX_RA_BLOCKS < f_ra->start)
1400 		return;
1401 
1402 	f_ra->start = max((pgoff_t)index, f_ra->start);
1403 	for (count = 0; count < MAX_RA_BLOCKS; count++) {
1404 		if (f_ra->start >= hsize) /* if exceeded the hash table */
1405 			break;
1406 
1407 		last = blocknr;
1408 		blocknr = be64_to_cpu(ip->i_hash_cache[f_ra->start]);
1409 		f_ra->start++;
1410 		if (blocknr == last)
1411 			continue;
1412 
1413 		bh = gfs2_getbuf(gl, blocknr, 1);
1414 		if (trylock_buffer(bh)) {
1415 			if (buffer_uptodate(bh)) {
1416 				unlock_buffer(bh);
1417 				brelse(bh);
1418 				continue;
1419 			}
1420 			bh->b_end_io = end_buffer_read_sync;
1421 			submit_bh(READA | REQ_META, bh);
1422 			continue;
1423 		}
1424 		brelse(bh);
1425 	}
1426 }
1427 
1428 /**
1429  * dir_e_read - Reads the entries from a directory into a filldir buffer
1430  * @dip: dinode pointer
1431  * @offset: the hash of the last entry read shifted to the right once
1432  * @opaque: buffer for the filldir function to fill
1433  * @filldir: points to the filldir function to use
1434  *
1435  * Returns: errno
1436  */
1437 
dir_e_read(struct inode * inode,u64 * offset,void * opaque,filldir_t filldir,struct file_ra_state * f_ra)1438 static int dir_e_read(struct inode *inode, u64 *offset, void *opaque,
1439 		      filldir_t filldir, struct file_ra_state *f_ra)
1440 {
1441 	struct gfs2_inode *dip = GFS2_I(inode);
1442 	u32 hsize, len = 0;
1443 	u32 hash, index;
1444 	__be64 *lp;
1445 	int copied = 0;
1446 	int error = 0;
1447 	unsigned depth = 0;
1448 
1449 	hsize = 1 << dip->i_depth;
1450 	hash = gfs2_dir_offset2hash(*offset);
1451 	index = hash >> (32 - dip->i_depth);
1452 
1453 	if (dip->i_hash_cache == NULL)
1454 		f_ra->start = 0;
1455 	lp = gfs2_dir_get_hash_table(dip);
1456 	if (IS_ERR(lp))
1457 		return PTR_ERR(lp);
1458 
1459 	gfs2_dir_readahead(inode, hsize, index, f_ra);
1460 
1461 	while (index < hsize) {
1462 		error = gfs2_dir_read_leaf(inode, offset, opaque, filldir,
1463 					   &copied, &depth,
1464 					   be64_to_cpu(lp[index]));
1465 		if (error)
1466 			break;
1467 
1468 		len = 1 << (dip->i_depth - depth);
1469 		index = (index & ~(len - 1)) + len;
1470 	}
1471 
1472 	if (error > 0)
1473 		error = 0;
1474 	return error;
1475 }
1476 
gfs2_dir_read(struct inode * inode,u64 * offset,void * opaque,filldir_t filldir,struct file_ra_state * f_ra)1477 int gfs2_dir_read(struct inode *inode, u64 *offset, void *opaque,
1478 		  filldir_t filldir, struct file_ra_state *f_ra)
1479 {
1480 	struct gfs2_inode *dip = GFS2_I(inode);
1481 	struct gfs2_sbd *sdp = GFS2_SB(inode);
1482 	struct dirent_gather g;
1483 	const struct gfs2_dirent **darr, *dent;
1484 	struct buffer_head *dibh;
1485 	int copied = 0;
1486 	int error;
1487 
1488 	if (!dip->i_entries)
1489 		return 0;
1490 
1491 	if (dip->i_diskflags & GFS2_DIF_EXHASH)
1492 		return dir_e_read(inode, offset, opaque, filldir, f_ra);
1493 
1494 	if (!gfs2_is_stuffed(dip)) {
1495 		gfs2_consist_inode(dip);
1496 		return -EIO;
1497 	}
1498 
1499 	error = gfs2_meta_inode_buffer(dip, &dibh);
1500 	if (error)
1501 		return error;
1502 
1503 	error = -ENOMEM;
1504 	/* 96 is max number of dirents which can be stuffed into an inode */
1505 	darr = kmalloc(96 * sizeof(struct gfs2_dirent *), GFP_NOFS);
1506 	if (darr) {
1507 		g.pdent = darr;
1508 		g.offset = 0;
1509 		dent = gfs2_dirent_scan(inode, dibh->b_data, dibh->b_size,
1510 					gfs2_dirent_gather, NULL, &g);
1511 		if (IS_ERR(dent)) {
1512 			error = PTR_ERR(dent);
1513 			goto out;
1514 		}
1515 		if (dip->i_entries != g.offset) {
1516 			fs_warn(sdp, "Number of entries corrupt in dir %llu, "
1517 				"ip->i_entries (%u) != g.offset (%u)\n",
1518 				(unsigned long long)dip->i_no_addr,
1519 				dip->i_entries,
1520 				g.offset);
1521 			error = -EIO;
1522 			goto out;
1523 		}
1524 		error = do_filldir_main(dip, offset, opaque, filldir, darr,
1525 					dip->i_entries, &copied);
1526 out:
1527 		kfree(darr);
1528 	}
1529 
1530 	if (error > 0)
1531 		error = 0;
1532 
1533 	brelse(dibh);
1534 
1535 	return error;
1536 }
1537 
1538 /**
1539  * gfs2_dir_search - Search a directory
1540  * @dip: The GFS2 inode
1541  * @filename:
1542  * @inode:
1543  *
1544  * This routine searches a directory for a file or another directory.
1545  * Assumes a glock is held on dip.
1546  *
1547  * Returns: errno
1548  */
1549 
gfs2_dir_search(struct inode * dir,const struct qstr * name)1550 struct inode *gfs2_dir_search(struct inode *dir, const struct qstr *name)
1551 {
1552 	struct buffer_head *bh;
1553 	struct gfs2_dirent *dent;
1554 	struct inode *inode;
1555 
1556 	dent = gfs2_dirent_search(dir, name, gfs2_dirent_find, &bh);
1557 	if (dent) {
1558 		if (IS_ERR(dent))
1559 			return ERR_CAST(dent);
1560 		inode = gfs2_inode_lookup(dir->i_sb,
1561 				be16_to_cpu(dent->de_type),
1562 				be64_to_cpu(dent->de_inum.no_addr),
1563 				be64_to_cpu(dent->de_inum.no_formal_ino), 0);
1564 		brelse(bh);
1565 		return inode;
1566 	}
1567 	return ERR_PTR(-ENOENT);
1568 }
1569 
gfs2_dir_check(struct inode * dir,const struct qstr * name,const struct gfs2_inode * ip)1570 int gfs2_dir_check(struct inode *dir, const struct qstr *name,
1571 		   const struct gfs2_inode *ip)
1572 {
1573 	struct buffer_head *bh;
1574 	struct gfs2_dirent *dent;
1575 	int ret = -ENOENT;
1576 
1577 	dent = gfs2_dirent_search(dir, name, gfs2_dirent_find, &bh);
1578 	if (dent) {
1579 		if (IS_ERR(dent))
1580 			return PTR_ERR(dent);
1581 		if (ip) {
1582 			if (be64_to_cpu(dent->de_inum.no_addr) != ip->i_no_addr)
1583 				goto out;
1584 			if (be64_to_cpu(dent->de_inum.no_formal_ino) !=
1585 			    ip->i_no_formal_ino)
1586 				goto out;
1587 			if (unlikely(IF2DT(ip->i_inode.i_mode) !=
1588 			    be16_to_cpu(dent->de_type))) {
1589 				gfs2_consist_inode(GFS2_I(dir));
1590 				ret = -EIO;
1591 				goto out;
1592 			}
1593 		}
1594 		ret = 0;
1595 out:
1596 		brelse(bh);
1597 	}
1598 	return ret;
1599 }
1600 
dir_new_leaf(struct inode * inode,const struct qstr * name)1601 static int dir_new_leaf(struct inode *inode, const struct qstr *name)
1602 {
1603 	struct buffer_head *bh, *obh;
1604 	struct gfs2_inode *ip = GFS2_I(inode);
1605 	struct gfs2_leaf *leaf, *oleaf;
1606 	int error;
1607 	u32 index;
1608 	u64 bn;
1609 
1610 	index = name->hash >> (32 - ip->i_depth);
1611 	error = get_first_leaf(ip, index, &obh);
1612 	if (error)
1613 		return error;
1614 	do {
1615 		oleaf = (struct gfs2_leaf *)obh->b_data;
1616 		bn = be64_to_cpu(oleaf->lf_next);
1617 		if (!bn)
1618 			break;
1619 		brelse(obh);
1620 		error = get_leaf(ip, bn, &obh);
1621 		if (error)
1622 			return error;
1623 	} while(1);
1624 
1625 	gfs2_trans_add_bh(ip->i_gl, obh, 1);
1626 
1627 	leaf = new_leaf(inode, &bh, be16_to_cpu(oleaf->lf_depth));
1628 	if (!leaf) {
1629 		brelse(obh);
1630 		return -ENOSPC;
1631 	}
1632 	oleaf->lf_next = cpu_to_be64(bh->b_blocknr);
1633 	brelse(bh);
1634 	brelse(obh);
1635 
1636 	error = gfs2_meta_inode_buffer(ip, &bh);
1637 	if (error)
1638 		return error;
1639 	gfs2_trans_add_bh(ip->i_gl, bh, 1);
1640 	gfs2_add_inode_blocks(&ip->i_inode, 1);
1641 	gfs2_dinode_out(ip, bh->b_data);
1642 	brelse(bh);
1643 	return 0;
1644 }
1645 
1646 /**
1647  * gfs2_dir_add - Add new filename into directory
1648  * @dip: The GFS2 inode
1649  * @filename: The new name
1650  * @inode: The inode number of the entry
1651  * @type: The type of the entry
1652  *
1653  * Returns: 0 on success, error code on failure
1654  */
1655 
gfs2_dir_add(struct inode * inode,const struct qstr * name,const struct gfs2_inode * nip)1656 int gfs2_dir_add(struct inode *inode, const struct qstr *name,
1657 		 const struct gfs2_inode *nip)
1658 {
1659 	struct gfs2_inode *ip = GFS2_I(inode);
1660 	struct buffer_head *bh;
1661 	struct gfs2_dirent *dent;
1662 	struct gfs2_leaf *leaf;
1663 	int error;
1664 
1665 	while(1) {
1666 		dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space,
1667 					  &bh);
1668 		if (dent) {
1669 			if (IS_ERR(dent))
1670 				return PTR_ERR(dent);
1671 			dent = gfs2_init_dirent(inode, dent, name, bh);
1672 			gfs2_inum_out(nip, dent);
1673 			dent->de_type = cpu_to_be16(IF2DT(nip->i_inode.i_mode));
1674 			if (ip->i_diskflags & GFS2_DIF_EXHASH) {
1675 				leaf = (struct gfs2_leaf *)bh->b_data;
1676 				be16_add_cpu(&leaf->lf_entries, 1);
1677 			}
1678 			brelse(bh);
1679 			error = gfs2_meta_inode_buffer(ip, &bh);
1680 			if (error)
1681 				break;
1682 			gfs2_trans_add_bh(ip->i_gl, bh, 1);
1683 			ip->i_entries++;
1684 			ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
1685 			if (S_ISDIR(nip->i_inode.i_mode))
1686 				inc_nlink(&ip->i_inode);
1687 			gfs2_dinode_out(ip, bh->b_data);
1688 			brelse(bh);
1689 			error = 0;
1690 			break;
1691 		}
1692 		if (!(ip->i_diskflags & GFS2_DIF_EXHASH)) {
1693 			error = dir_make_exhash(inode);
1694 			if (error)
1695 				break;
1696 			continue;
1697 		}
1698 		error = dir_split_leaf(inode, name);
1699 		if (error == 0)
1700 			continue;
1701 		if (error < 0)
1702 			break;
1703 		if (ip->i_depth < GFS2_DIR_MAX_DEPTH) {
1704 			error = dir_double_exhash(ip);
1705 			if (error)
1706 				break;
1707 			error = dir_split_leaf(inode, name);
1708 			if (error < 0)
1709 				break;
1710 			if (error == 0)
1711 				continue;
1712 		}
1713 		error = dir_new_leaf(inode, name);
1714 		if (!error)
1715 			continue;
1716 		error = -ENOSPC;
1717 		break;
1718 	}
1719 	return error;
1720 }
1721 
1722 
1723 /**
1724  * gfs2_dir_del - Delete a directory entry
1725  * @dip: The GFS2 inode
1726  * @filename: The filename
1727  *
1728  * Returns: 0 on success, error code on failure
1729  */
1730 
gfs2_dir_del(struct gfs2_inode * dip,const struct dentry * dentry)1731 int gfs2_dir_del(struct gfs2_inode *dip, const struct dentry *dentry)
1732 {
1733 	const struct qstr *name = &dentry->d_name;
1734 	struct gfs2_dirent *dent, *prev = NULL;
1735 	struct buffer_head *bh;
1736 
1737 	/* Returns _either_ the entry (if its first in block) or the
1738 	   previous entry otherwise */
1739 	dent = gfs2_dirent_search(&dip->i_inode, name, gfs2_dirent_prev, &bh);
1740 	if (!dent) {
1741 		gfs2_consist_inode(dip);
1742 		return -EIO;
1743 	}
1744 	if (IS_ERR(dent)) {
1745 		gfs2_consist_inode(dip);
1746 		return PTR_ERR(dent);
1747 	}
1748 	/* If not first in block, adjust pointers accordingly */
1749 	if (gfs2_dirent_find(dent, name, NULL) == 0) {
1750 		prev = dent;
1751 		dent = (struct gfs2_dirent *)((char *)dent + be16_to_cpu(prev->de_rec_len));
1752 	}
1753 
1754 	dirent_del(dip, bh, prev, dent);
1755 	if (dip->i_diskflags & GFS2_DIF_EXHASH) {
1756 		struct gfs2_leaf *leaf = (struct gfs2_leaf *)bh->b_data;
1757 		u16 entries = be16_to_cpu(leaf->lf_entries);
1758 		if (!entries)
1759 			gfs2_consist_inode(dip);
1760 		leaf->lf_entries = cpu_to_be16(--entries);
1761 	}
1762 	brelse(bh);
1763 
1764 	if (!dip->i_entries)
1765 		gfs2_consist_inode(dip);
1766 	dip->i_entries--;
1767 	dip->i_inode.i_mtime = dip->i_inode.i_ctime = CURRENT_TIME;
1768 	if (S_ISDIR(dentry->d_inode->i_mode))
1769 		drop_nlink(&dip->i_inode);
1770 	mark_inode_dirty(&dip->i_inode);
1771 
1772 	return 0;
1773 }
1774 
1775 /**
1776  * gfs2_dir_mvino - Change inode number of directory entry
1777  * @dip: The GFS2 inode
1778  * @filename:
1779  * @new_inode:
1780  *
1781  * This routine changes the inode number of a directory entry.  It's used
1782  * by rename to change ".." when a directory is moved.
1783  * Assumes a glock is held on dvp.
1784  *
1785  * Returns: errno
1786  */
1787 
gfs2_dir_mvino(struct gfs2_inode * dip,const struct qstr * filename,const struct gfs2_inode * nip,unsigned int new_type)1788 int gfs2_dir_mvino(struct gfs2_inode *dip, const struct qstr *filename,
1789 		   const struct gfs2_inode *nip, unsigned int new_type)
1790 {
1791 	struct buffer_head *bh;
1792 	struct gfs2_dirent *dent;
1793 	int error;
1794 
1795 	dent = gfs2_dirent_search(&dip->i_inode, filename, gfs2_dirent_find, &bh);
1796 	if (!dent) {
1797 		gfs2_consist_inode(dip);
1798 		return -EIO;
1799 	}
1800 	if (IS_ERR(dent))
1801 		return PTR_ERR(dent);
1802 
1803 	gfs2_trans_add_bh(dip->i_gl, bh, 1);
1804 	gfs2_inum_out(nip, dent);
1805 	dent->de_type = cpu_to_be16(new_type);
1806 
1807 	if (dip->i_diskflags & GFS2_DIF_EXHASH) {
1808 		brelse(bh);
1809 		error = gfs2_meta_inode_buffer(dip, &bh);
1810 		if (error)
1811 			return error;
1812 		gfs2_trans_add_bh(dip->i_gl, bh, 1);
1813 	}
1814 
1815 	dip->i_inode.i_mtime = dip->i_inode.i_ctime = CURRENT_TIME;
1816 	gfs2_dinode_out(dip, bh->b_data);
1817 	brelse(bh);
1818 	return 0;
1819 }
1820 
1821 /**
1822  * leaf_dealloc - Deallocate a directory leaf
1823  * @dip: the directory
1824  * @index: the hash table offset in the directory
1825  * @len: the number of pointers to this leaf
1826  * @leaf_no: the leaf number
1827  * @leaf_bh: buffer_head for the starting leaf
1828  * last_dealloc: 1 if this is the final dealloc for the leaf, else 0
1829  *
1830  * Returns: errno
1831  */
1832 
leaf_dealloc(struct gfs2_inode * dip,u32 index,u32 len,u64 leaf_no,struct buffer_head * leaf_bh,int last_dealloc)1833 static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len,
1834 			u64 leaf_no, struct buffer_head *leaf_bh,
1835 			int last_dealloc)
1836 {
1837 	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
1838 	struct gfs2_leaf *tmp_leaf;
1839 	struct gfs2_rgrp_list rlist;
1840 	struct buffer_head *bh, *dibh;
1841 	u64 blk, nblk;
1842 	unsigned int rg_blocks = 0, l_blocks = 0;
1843 	char *ht;
1844 	unsigned int x, size = len * sizeof(u64);
1845 	int error;
1846 
1847 	error = gfs2_rindex_update(sdp);
1848 	if (error)
1849 		return error;
1850 
1851 	memset(&rlist, 0, sizeof(struct gfs2_rgrp_list));
1852 
1853 	ht = kzalloc(size, GFP_NOFS);
1854 	if (!ht)
1855 		return -ENOMEM;
1856 
1857 	if (!gfs2_qadata_get(dip)) {
1858 		error = -ENOMEM;
1859 		goto out;
1860 	}
1861 
1862 	error = gfs2_quota_hold(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
1863 	if (error)
1864 		goto out_put;
1865 
1866 	/*  Count the number of leaves  */
1867 	bh = leaf_bh;
1868 
1869 	for (blk = leaf_no; blk; blk = nblk) {
1870 		if (blk != leaf_no) {
1871 			error = get_leaf(dip, blk, &bh);
1872 			if (error)
1873 				goto out_rlist;
1874 		}
1875 		tmp_leaf = (struct gfs2_leaf *)bh->b_data;
1876 		nblk = be64_to_cpu(tmp_leaf->lf_next);
1877 		if (blk != leaf_no)
1878 			brelse(bh);
1879 
1880 		gfs2_rlist_add(dip, &rlist, blk);
1881 		l_blocks++;
1882 	}
1883 
1884 	gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE);
1885 
1886 	for (x = 0; x < rlist.rl_rgrps; x++) {
1887 		struct gfs2_rgrpd *rgd;
1888 		rgd = rlist.rl_ghs[x].gh_gl->gl_object;
1889 		rg_blocks += rgd->rd_length;
1890 	}
1891 
1892 	error = gfs2_glock_nq_m(rlist.rl_rgrps, rlist.rl_ghs);
1893 	if (error)
1894 		goto out_rlist;
1895 
1896 	error = gfs2_trans_begin(sdp,
1897 			rg_blocks + (DIV_ROUND_UP(size, sdp->sd_jbsize) + 1) +
1898 			RES_DINODE + RES_STATFS + RES_QUOTA, l_blocks);
1899 	if (error)
1900 		goto out_rg_gunlock;
1901 
1902 	bh = leaf_bh;
1903 
1904 	for (blk = leaf_no; blk; blk = nblk) {
1905 		if (blk != leaf_no) {
1906 			error = get_leaf(dip, blk, &bh);
1907 			if (error)
1908 				goto out_end_trans;
1909 		}
1910 		tmp_leaf = (struct gfs2_leaf *)bh->b_data;
1911 		nblk = be64_to_cpu(tmp_leaf->lf_next);
1912 		if (blk != leaf_no)
1913 			brelse(bh);
1914 
1915 		gfs2_free_meta(dip, blk, 1);
1916 		gfs2_add_inode_blocks(&dip->i_inode, -1);
1917 	}
1918 
1919 	error = gfs2_dir_write_data(dip, ht, index * sizeof(u64), size);
1920 	if (error != size) {
1921 		if (error >= 0)
1922 			error = -EIO;
1923 		goto out_end_trans;
1924 	}
1925 
1926 	error = gfs2_meta_inode_buffer(dip, &dibh);
1927 	if (error)
1928 		goto out_end_trans;
1929 
1930 	gfs2_trans_add_bh(dip->i_gl, dibh, 1);
1931 	/* On the last dealloc, make this a regular file in case we crash.
1932 	   (We don't want to free these blocks a second time.)  */
1933 	if (last_dealloc)
1934 		dip->i_inode.i_mode = S_IFREG;
1935 	gfs2_dinode_out(dip, dibh->b_data);
1936 	brelse(dibh);
1937 
1938 out_end_trans:
1939 	gfs2_trans_end(sdp);
1940 out_rg_gunlock:
1941 	gfs2_glock_dq_m(rlist.rl_rgrps, rlist.rl_ghs);
1942 out_rlist:
1943 	gfs2_rlist_free(&rlist);
1944 	gfs2_quota_unhold(dip);
1945 out_put:
1946 	gfs2_qadata_put(dip);
1947 out:
1948 	kfree(ht);
1949 	return error;
1950 }
1951 
1952 /**
1953  * gfs2_dir_exhash_dealloc - free all the leaf blocks in a directory
1954  * @dip: the directory
1955  *
1956  * Dealloc all on-disk directory leaves to FREEMETA state
1957  * Change on-disk inode type to "regular file"
1958  *
1959  * Returns: errno
1960  */
1961 
gfs2_dir_exhash_dealloc(struct gfs2_inode * dip)1962 int gfs2_dir_exhash_dealloc(struct gfs2_inode *dip)
1963 {
1964 	struct buffer_head *bh;
1965 	struct gfs2_leaf *leaf;
1966 	u32 hsize, len;
1967 	u32 index = 0, next_index;
1968 	__be64 *lp;
1969 	u64 leaf_no;
1970 	int error = 0, last;
1971 
1972 	hsize = 1 << dip->i_depth;
1973 
1974 	lp = gfs2_dir_get_hash_table(dip);
1975 	if (IS_ERR(lp))
1976 		return PTR_ERR(lp);
1977 
1978 	while (index < hsize) {
1979 		leaf_no = be64_to_cpu(lp[index]);
1980 		if (leaf_no) {
1981 			error = get_leaf(dip, leaf_no, &bh);
1982 			if (error)
1983 				goto out;
1984 			leaf = (struct gfs2_leaf *)bh->b_data;
1985 			len = 1 << (dip->i_depth - be16_to_cpu(leaf->lf_depth));
1986 
1987 			next_index = (index & ~(len - 1)) + len;
1988 			last = ((next_index >= hsize) ? 1 : 0);
1989 			error = leaf_dealloc(dip, index, len, leaf_no, bh,
1990 					     last);
1991 			brelse(bh);
1992 			if (error)
1993 				goto out;
1994 			index = next_index;
1995 		} else
1996 			index++;
1997 	}
1998 
1999 	if (index != hsize) {
2000 		gfs2_consist_inode(dip);
2001 		error = -EIO;
2002 	}
2003 
2004 out:
2005 
2006 	return error;
2007 }
2008 
2009 /**
2010  * gfs2_diradd_alloc_required - find if adding entry will require an allocation
2011  * @ip: the file being written to
2012  * @filname: the filename that's going to be added
2013  *
2014  * Returns: 1 if alloc required, 0 if not, -ve on error
2015  */
2016 
gfs2_diradd_alloc_required(struct inode * inode,const struct qstr * name)2017 int gfs2_diradd_alloc_required(struct inode *inode, const struct qstr *name)
2018 {
2019 	struct gfs2_dirent *dent;
2020 	struct buffer_head *bh;
2021 
2022 	dent = gfs2_dirent_search(inode, name, gfs2_dirent_find_space, &bh);
2023 	if (!dent) {
2024 		return 1;
2025 	}
2026 	if (IS_ERR(dent))
2027 		return PTR_ERR(dent);
2028 	brelse(bh);
2029 	return 0;
2030 }
2031 
2032