xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_htree.c (revision 9bd4ba417d045dba399c191a33f736de4600c34b)
1 /*	$NetBSD: ext2fs_htree.c,v 1.11 2023/08/26 05:22:50 riastradh Exp $	*/
2 
3 /*-
4  * Copyright (c) 2010, 2012 Zheng Liu <lz@freebsd.org>
5  * Copyright (c) 2012, Vyacheslav Matyushin
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  * $FreeBSD: head/sys/fs/ext2fs/ext2fs_htree.c 294653 2016-01-24 02:41:49Z pfg $
30  */
31 #include <sys/cdefs.h>
32 __KERNEL_RCSID(0, "$NetBSD: ext2fs_htree.c,v 1.11 2023/08/26 05:22:50 riastradh Exp $");
33 
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/resourcevar.h>
37 #include <sys/kernel.h>
38 #include <sys/file.h>
39 #include <sys/stat.h>
40 #include <sys/buf.h>
41 #include <sys/proc.h>
42 #include <sys/mount.h>
43 #include <sys/namei.h>
44 #include <sys/vnode.h>
45 #include <sys/lockf.h>
46 #include <sys/pool.h>
47 #include <sys/signalvar.h>
48 #include <sys/kauth.h>
49 #include <sys/malloc.h>
50 #include <ufs/ufs/dir.h>
51 
52 #include <ufs/ufs/inode.h>
53 #include <ufs/ufs/ufsmount.h>
54 #include <ufs/ext2fs/ext2fs.h>
55 
56 #include <ufs/ext2fs/ext2fs_extern.h>
57 #include <ufs/ext2fs/ext2fs_dinode.h>
58 #include <ufs/ext2fs/ext2fs_dir.h>
59 #include <ufs/ext2fs/ext2fs_htree.h>
60 #include <ufs/ext2fs/ext2fs_hash.h>
61 
62 static int ext2fs_htree_find_leaf(struct inode *, const char *, int ,
63     uint32_t *, uint8_t *, struct ext2fs_htree_lookup_info *);
64 
65 int
ext2fs_htree_has_idx(struct inode * ip)66 ext2fs_htree_has_idx(struct inode *ip)
67 {
68 	return EXT2F_HAS_COMPAT_FEATURE(ip->i_e2fs, EXT2F_COMPAT_DIRHASHINDEX)
69 	    && (ip->i_din.e2fs_din->e2di_flags & EXT2_INDEX);
70 }
71 
72 static off_t
ext2fs_htree_get_block(struct ext2fs_htree_entry * ep)73 ext2fs_htree_get_block(struct ext2fs_htree_entry *ep)
74 {
75 	return ep->h_blk & 0x00FFFFFF;
76 }
77 
78 static void
ext2fs_htree_release(struct ext2fs_htree_lookup_info * info)79 ext2fs_htree_release(struct ext2fs_htree_lookup_info *info)
80 {
81 	for (u_int i = 0; i < info->h_levels_num; i++) {
82 		struct buf *bp = info->h_levels[i].h_bp;
83 		if (bp != NULL)
84 			brelse(bp, 0);
85 	}
86 }
87 
88 static uint16_t
ext2fs_htree_get_limit(struct ext2fs_htree_entry * ep)89 ext2fs_htree_get_limit(struct ext2fs_htree_entry *ep)
90 {
91 	return ((struct ext2fs_htree_count *)(ep))->h_entries_max;
92 }
93 
94 static uint32_t
ext2fs_htree_root_limit(struct inode * ip,int len)95 ext2fs_htree_root_limit(struct inode *ip, int len)
96 {
97 	uint32_t space = ip->i_e2fs->e2fs_bsize - EXT2_DIR_REC_LEN(1) -
98 	    EXT2_DIR_REC_LEN(2) - len;
99 	return space / sizeof(struct ext2fs_htree_entry);
100 }
101 
102 static uint16_t
ext2fs_htree_get_count(struct ext2fs_htree_entry * ep)103 ext2fs_htree_get_count(struct ext2fs_htree_entry *ep)
104 {
105 	return ((struct ext2fs_htree_count *)(ep))->h_entries_num;
106 }
107 
108 static uint32_t
ext2fs_htree_get_hash(struct ext2fs_htree_entry * ep)109 ext2fs_htree_get_hash(struct ext2fs_htree_entry *ep)
110 {
111 	return ep->h_hash;
112 }
113 
114 
115 static void
ext2fs_htree_set_block(struct ext2fs_htree_entry * ep,uint32_t blk)116 ext2fs_htree_set_block(struct ext2fs_htree_entry *ep, uint32_t blk)
117 {
118 	ep->h_blk = blk;
119 }
120 
121 static void
ext2fs_htree_set_count(struct ext2fs_htree_entry * ep,uint16_t cnt)122 ext2fs_htree_set_count(struct ext2fs_htree_entry *ep, uint16_t cnt)
123 {
124 	((struct ext2fs_htree_count *)(ep))->h_entries_num = cnt;
125 }
126 
127 static void
ext2fs_htree_set_hash(struct ext2fs_htree_entry * ep,uint32_t hash)128 ext2fs_htree_set_hash(struct ext2fs_htree_entry *ep, uint32_t hash)
129 {
130 	ep->h_hash = hash;
131 }
132 
133 static void
ext2fs_htree_set_limit(struct ext2fs_htree_entry * ep,uint16_t limit)134 ext2fs_htree_set_limit(struct ext2fs_htree_entry *ep, uint16_t limit)
135 {
136 	((struct ext2fs_htree_count *)(ep))->h_entries_max = limit;
137 }
138 
139 static uint32_t
ext2fs_htree_node_limit(struct inode * ip)140 ext2fs_htree_node_limit(struct inode *ip)
141 {
142 	struct m_ext2fs *fs;
143 	uint32_t space;
144 
145 	fs = ip->i_e2fs;
146 	space = fs->e2fs_bsize - EXT2_DIR_REC_LEN(0);
147 
148 	return space / sizeof(struct ext2fs_htree_entry);
149 }
150 
151 static int
ext2fs_htree_append_block(struct vnode * vp,char * data,struct componentname * cnp,uint32_t blksize)152 ext2fs_htree_append_block(struct vnode *vp, char *data,
153     struct componentname *cnp, uint32_t blksize)
154 {
155 	struct iovec aiov;
156 	struct uio auio;
157 	struct inode *dp = VTOI(vp);
158 	uint64_t cursize, newsize;
159 	int error;
160 
161 	cursize = roundup(dp->i_size, blksize);
162 	newsize = cursize + blksize;
163 
164 	auio.uio_offset = cursize;
165 	auio.uio_resid = blksize;
166 	aiov.iov_len = blksize;
167 	aiov.iov_base = data;
168 	auio.uio_iov = &aiov;
169 	auio.uio_iovcnt = 1;
170 	auio.uio_rw = UIO_WRITE;
171 	auio.uio_vmspace = vmspace_kernel();
172 	error = VOP_WRITE(vp, &auio, IO_SYNC, cnp->cn_cred);
173 	if (!error)
174 		dp->i_size = newsize;
175 
176 	return error;
177 }
178 
179 static int
ext2fs_htree_writebuf(struct ext2fs_htree_lookup_info * info)180 ext2fs_htree_writebuf(struct ext2fs_htree_lookup_info *info)
181 {
182 	int i, error;
183 
184 	for (i = 0; i < info->h_levels_num; i++) {
185 		struct buf *bp = info->h_levels[i].h_bp;
186 		error = bwrite(bp);
187 		if (error)
188 			return error;
189 	}
190 
191 	return 0;
192 }
193 
194 static void
ext2fs_htree_insert_entry_to_level(struct ext2fs_htree_lookup_level * level,uint32_t hash,uint32_t blk)195 ext2fs_htree_insert_entry_to_level(struct ext2fs_htree_lookup_level *level,
196     uint32_t hash, uint32_t blk)
197 {
198 	struct ext2fs_htree_entry *target;
199 	int entries_num;
200 
201 	target = level->h_entry + 1;
202 	entries_num = ext2fs_htree_get_count(level->h_entries);
203 
204 	memmove(target + 1, target, (char *)(level->h_entries + entries_num) -
205 	    (char *)target);
206 	ext2fs_htree_set_block(target, blk);
207 	ext2fs_htree_set_hash(target, hash);
208 	ext2fs_htree_set_count(level->h_entries, entries_num + 1);
209 }
210 
211 /*
212  * Insert an index entry to the index node.
213  */
214 static void
ext2fs_htree_insert_entry(struct ext2fs_htree_lookup_info * info,uint32_t hash,uint32_t blk)215 ext2fs_htree_insert_entry(struct ext2fs_htree_lookup_info *info,
216     uint32_t hash, uint32_t blk)
217 {
218 	struct ext2fs_htree_lookup_level *level;
219 
220 	level = &info->h_levels[info->h_levels_num - 1];
221 	ext2fs_htree_insert_entry_to_level(level, hash, blk);
222 }
223 
224 /*
225  * Compare two entry sort descriptors by name hash value.
226  * This is used together with qsort.
227  */
228 static int
ext2fs_htree_cmp_sort_entry(const void * e1,const void * e2)229 ext2fs_htree_cmp_sort_entry(const void *e1, const void *e2)
230 {
231 	const struct ext2fs_htree_sort_entry *entry1, *entry2;
232 
233 	entry1 = (const struct ext2fs_htree_sort_entry *)e1;
234 	entry2 = (const struct ext2fs_htree_sort_entry *)e2;
235 
236 	if (entry1->h_hash < entry2->h_hash)
237 		return -1;
238 	if (entry1->h_hash > entry2->h_hash)
239 		return 1;
240 	return 0;
241 }
242 
243 /*
244  * Append an entry to the end of the directory block.
245  */
246 static void
ext2fs_append_entry(char * block,uint32_t blksize,struct ext2fs_direct * last_entry,struct ext2fs_direct * new_entry)247 ext2fs_append_entry(char *block, uint32_t blksize,
248     struct ext2fs_direct *last_entry, struct ext2fs_direct *new_entry)
249 {
250 	uint16_t entry_len;
251 
252 	entry_len = EXT2_DIR_REC_LEN(last_entry->e2d_namlen);
253 	last_entry->e2d_reclen = entry_len;
254 	last_entry = (struct ext2fs_direct *)((char *)last_entry + entry_len);
255 	new_entry->e2d_reclen = block + blksize - (char *)last_entry;
256 	memcpy(last_entry, new_entry, EXT2_DIR_REC_LEN(new_entry->e2d_namlen));
257 }
258 
259 /*
260  * Move half of entries from the old directory block to the new one.
261  */
262 static int
ext2fs_htree_split_dirblock(char * block1,char * block2,uint32_t blksize,uint32_t * hash_seed,uint8_t hash_version,uint32_t * split_hash,struct ext2fs_direct * entry)263 ext2fs_htree_split_dirblock(char *block1, char *block2, uint32_t blksize,
264     uint32_t *hash_seed, uint8_t hash_version,
265     uint32_t *split_hash, struct ext2fs_direct *entry)
266 {
267 	int entry_cnt = 0;
268 	int size = 0;
269 	int i, k;
270 	uint32_t offset;
271 	uint16_t entry_len = 0;
272 	uint32_t entry_hash;
273 	struct ext2fs_direct *ep, *last;
274 	char *dest;
275 	struct ext2fs_htree_sort_entry *sort_info, dummy;
276 
277 	ep = (struct ext2fs_direct *)block1;
278 	dest = block2;
279 	sort_info = (struct ext2fs_htree_sort_entry *)
280 	    ((char *)block2 + blksize);
281 
282 	/*
283 	 * Calculate name hash value for the entry which is to be added.
284 	 */
285 	ext2fs_htree_hash(entry->e2d_name, entry->e2d_namlen, hash_seed,
286 	    hash_version, &entry_hash, NULL);
287 
288 	/*
289 	 * Fill in directory entry sort descriptors.
290 	 */
291 	while ((char *)ep < block1 + blksize) {
292 		if (ep->e2d_ino && ep->e2d_namlen) {
293 			entry_cnt++;
294 			sort_info--;
295 			sort_info->h_size = ep->e2d_reclen;
296 			sort_info->h_offset = (char *)ep - block1;
297 			ext2fs_htree_hash(ep->e2d_name, ep->e2d_namlen,
298 			    hash_seed, hash_version,
299 			    &sort_info->h_hash, NULL);
300 		}
301 		ep = (struct ext2fs_direct *)
302 		    ((char *)ep + ep->e2d_reclen);
303 	}
304 
305 	/*
306 	 * Sort directory entry descriptors by name hash value.
307 	 */
308 	kheapsort(sort_info, entry_cnt, sizeof(struct ext2fs_htree_sort_entry),
309 	    ext2fs_htree_cmp_sort_entry, &dummy);
310 
311 	/*
312 	 * Count the number of entries to move to directory block 2.
313 	 */
314 	for (i = entry_cnt - 1; i >= 0; i--) {
315 		if (sort_info[i].h_size + size > blksize / 2)
316 			break;
317 		size += sort_info[i].h_size;
318 	}
319 
320 	*split_hash = sort_info[i + 1].h_hash;
321 
322 	/*
323 	 * Set collision bit.
324 	 */
325 	if (*split_hash == sort_info[i].h_hash)
326 		*split_hash += 1;
327 
328 	/*
329 	 * Move half of directory entries from block 1 to block 2.
330 	 */
331 	for (k = i + 1; k < entry_cnt; k++) {
332 		ep = (struct ext2fs_direct *)((char *)block1 +
333 		    sort_info[k].h_offset);
334 		entry_len = EXT2_DIR_REC_LEN(ep->e2d_namlen);
335 		memcpy(dest, ep, entry_len);
336 		((struct ext2fs_direct *)dest)->e2d_reclen = entry_len;
337 		/* Mark directory entry as unused. */
338 		ep->e2d_ino = 0;
339 		dest += entry_len;
340 	}
341 	dest -= entry_len;
342 
343 	/* Shrink directory entries in block 1. */
344 	last = (struct ext2fs_direct *)block1;
345 	entry_len = 0;
346 	for (offset = 0; offset < blksize;) {
347 		ep = (struct ext2fs_direct *)(block1 + offset);
348 		offset += ep->e2d_reclen;
349 		if (ep->e2d_ino) {
350 			last = (struct ext2fs_direct *)
351 			   ((char *)last + entry_len);
352 			entry_len = EXT2_DIR_REC_LEN(ep->e2d_namlen);
353 			memcpy((void *)last, (void *)ep, entry_len);
354 			last->e2d_reclen = entry_len;
355 		}
356 	}
357 
358 	if (entry_hash >= *split_hash) {
359 		/* Add entry to block 2. */
360 		ext2fs_append_entry(block2, blksize,
361 		    (struct ext2fs_direct *)dest, entry);
362 
363 		/* Adjust length field of last entry of block 1. */
364 		last->e2d_reclen = block1 + blksize - (char *)last;
365 	} else {
366 		/* Add entry to block 1. */
367 		ext2fs_append_entry(block1, blksize, last, entry);
368 
369 		/* Adjust length field of last entry of block 2. */
370 		((struct ext2fs_direct *)dest)->e2d_reclen =
371 		    block2 + blksize - dest;
372 	}
373 
374 	return 0;
375 }
376 
377 /*
378  * Create an HTree index for a directory having entries which are no more
379  * accommodable in a single dir-block.
380  */
381 int
ext2fs_htree_create_index(struct vnode * vp,struct componentname * cnp,struct ext2fs_direct * new_entry)382 ext2fs_htree_create_index(struct vnode *vp, struct componentname *cnp,
383     struct ext2fs_direct *new_entry)
384 {
385 	struct buf *bp = NULL;
386 	struct inode *dp;
387 	struct ext2fs *fs;
388 	struct m_ext2fs *m_fs;
389 	struct ext2fs_direct *ep, *dotdot;
390 	struct ext2fs_htree_root *root;
391 	struct ext2fs_htree_lookup_info info;
392 	uint32_t blksize, dirlen, split_hash;
393 	uint8_t hash_version;
394 	char *buf1 = NULL;
395 	char *buf2 = NULL;
396 	int error = 0;
397 
398 	dp = VTOI(vp);
399 	fs = &(dp->i_e2fs->e2fs);
400 	m_fs = dp->i_e2fs;
401 	blksize = m_fs->e2fs_bsize;
402 
403 	buf1 = malloc(blksize, M_TEMP, M_WAITOK | M_ZERO);
404 	buf2 = malloc(blksize, M_TEMP, M_WAITOK | M_ZERO);
405 
406 	if ((error = ext2fs_blkatoff(vp, 0, NULL, &bp)) != 0)
407 		goto out;
408 
409 	root = (struct ext2fs_htree_root *)bp->b_data;
410 	dotdot = (struct ext2fs_direct *)((char *)&(root->h_dotdot));
411 	ep = (struct ext2fs_direct *)((char *)dotdot + dotdot->e2d_reclen);
412 	dirlen = (char *)root + blksize - (char *)ep;
413 	memcpy(buf1, ep, dirlen);
414 	ep = (struct ext2fs_direct *)buf1;
415 	while ((char *)ep < buf1 + dirlen)
416 		ep = (struct ext2fs_direct *)((char *)ep + ep->e2d_reclen);
417 	ep->e2d_reclen = buf1 + blksize - (char *)ep;
418 	/* XXX It should be made dp->i_flag |= IN_E3INDEX; */
419 	dp->i_din.e2fs_din->e2di_flags |= EXT2_INDEX;
420 
421 	/*
422 	 * Initialize index root.
423 	 */
424 	dotdot->e2d_reclen = blksize - EXT2_DIR_REC_LEN(1);
425 	memset(&root->h_info, 0, sizeof(root->h_info));
426 	root->h_info.h_hash_version = fs->e3fs_def_hash_version;
427 	root->h_info.h_info_len = sizeof(root->h_info);
428 	ext2fs_htree_set_block(root->h_entries, 1);
429 	ext2fs_htree_set_count(root->h_entries, 1);
430 	ext2fs_htree_set_limit(root->h_entries,
431 	    ext2fs_htree_root_limit(dp, sizeof(root->h_info)));
432 
433 	memset(&info, 0, sizeof(info));
434 	info.h_levels_num = 1;
435 	info.h_levels[0].h_entries = root->h_entries;
436 	info.h_levels[0].h_entry = root->h_entries;
437 
438 	hash_version = root->h_info.h_hash_version;
439 	if (hash_version <= EXT2_HTREE_TEA)
440 		hash_version += m_fs->e2fs_uhash;
441 	ext2fs_htree_split_dirblock(buf1, buf2, blksize, fs->e3fs_hash_seed,
442 	    hash_version, &split_hash, new_entry);
443 	ext2fs_htree_insert_entry(&info, split_hash, 2);
444 
445 	/*
446 	 * Write directory block 0.
447 	 */
448 	if ( (vp)->v_mount->mnt_iflag & IO_SYNC)
449 		(void)bwrite(bp);
450 	else
451 		bdwrite(bp);
452 
453 	dp->i_flag |= IN_CHANGE | IN_UPDATE;
454 
455 	/*
456 	 * Write directory block 1.
457 	 */
458 	error = ext2fs_htree_append_block(vp, buf1, cnp, blksize);
459 	if (error)
460 		goto out1;
461 
462 	/*
463 	 * Write directory block 2.
464 	 */
465 	error = ext2fs_htree_append_block(vp, buf2, cnp, blksize);
466 	goto out1;
467 out:
468 	if (bp != NULL)
469 		brelse(bp, 0);
470 out1:
471 	free(buf1, M_TEMP);
472 	free(buf2, M_TEMP);
473 	return error;
474 }
475 
476 /*
477  * Add an entry to the directory using htree index.
478  */
479 int
ext2fs_htree_add_entry(struct vnode * dvp,struct ext2fs_direct * entry,struct componentname * cnp,size_t newentrysize)480 ext2fs_htree_add_entry(struct vnode *dvp, struct ext2fs_direct *entry,
481     struct componentname *cnp, size_t newentrysize)
482 {
483 	struct ext2fs_htree_entry *entries, *leaf_node;
484 	struct ext2fs_htree_lookup_info info;
485 	struct buf *bp = NULL;
486 	struct ext2fs *fs;
487 	struct m_ext2fs *m_fs;
488 	struct inode *ip;
489 	uint16_t ent_num;
490 	uint32_t dirhash, split_hash;
491 	uint32_t blksize, blknum;
492 	uint64_t cursize, dirsize;
493 	uint8_t hash_version;
494 	char *newdirblock = NULL;
495 	char *newidxblock = NULL;
496 	struct ext2fs_htree_node *dst_node;
497 	struct ext2fs_htree_entry *dst_entries;
498 	struct ext2fs_htree_entry *root_entries;
499 	struct buf *dst_bp = NULL;
500 	int error, write_bp = 0, write_dst_bp = 0, write_info = 0;
501 
502 	ip = VTOI(dvp);
503 	m_fs = ip->i_e2fs;
504 	fs = &(m_fs->e2fs);
505 	blksize = m_fs->e2fs_bsize;
506 
507 	if (ip->i_crap.ulr_count != 0)
508 		return ext2fs_add_entry(dvp, entry, &(ip->i_crap), newentrysize);
509 
510 	/* Target directory block is full, split it */
511 	memset(&info, 0, sizeof(info));
512 	error = ext2fs_htree_find_leaf(ip, entry->e2d_name, entry->e2d_namlen,
513 	    &dirhash, &hash_version, &info);
514 	if (error)
515 		return error;
516 	entries = info.h_levels[info.h_levels_num - 1].h_entries;
517 	ent_num = ext2fs_htree_get_count(entries);
518 	if (ent_num == ext2fs_htree_get_limit(entries)) {
519 		/* Split the index node. */
520 		root_entries = info.h_levels[0].h_entries;
521 		newidxblock = malloc(blksize, M_TEMP, M_WAITOK | M_ZERO);
522 		dst_node = (struct ext2fs_htree_node *)newidxblock;
523 		dst_entries = dst_node->h_entries;
524 		memset(&dst_node->h_fake_dirent, 0,
525 		    sizeof(dst_node->h_fake_dirent));
526 		dst_node->h_fake_dirent.e2d_reclen = blksize;
527 
528 		cursize = roundup(ip->i_size, blksize);
529 		dirsize = cursize + blksize;
530 		blknum = dirsize / blksize - 1;
531 
532 		error = ext2fs_htree_append_block(dvp, newidxblock,
533 		    cnp, blksize);
534 		if (error)
535 			goto finish;
536 		error = ext2fs_blkatoff(dvp, cursize, NULL, &dst_bp);
537 		if (error)
538 			goto finish;
539 
540 		dst_node = (struct ext2fs_htree_node *)dst_bp->b_data;
541 		dst_entries = dst_node->h_entries;
542 
543 		if (info.h_levels_num == 2) {
544 			uint16_t src_ent_num, dst_ent_num;
545 
546 			if (ext2fs_htree_get_count(root_entries) ==
547 			    ext2fs_htree_get_limit(root_entries)) {
548 				/* Directory index is full */
549 				error = EIO;
550 				goto finish;
551 			}
552 
553 			src_ent_num = ent_num / 2;
554 			dst_ent_num = ent_num - src_ent_num;
555 			split_hash = ext2fs_htree_get_hash(entries + src_ent_num);
556 
557 			/* Move half of index entries to the new index node */
558 			memcpy(dst_entries, entries + src_ent_num,
559 			    dst_ent_num * sizeof(struct ext2fs_htree_entry));
560 			ext2fs_htree_set_count(entries, src_ent_num);
561 			ext2fs_htree_set_count(dst_entries, dst_ent_num);
562 			ext2fs_htree_set_limit(dst_entries,
563 			    ext2fs_htree_node_limit(ip));
564 
565 			if (info.h_levels[1].h_entry >= entries + src_ent_num) {
566 				struct buf *tmp = info.h_levels[1].h_bp;
567 				info.h_levels[1].h_bp = dst_bp;
568 				dst_bp = tmp;
569 
570 				info.h_levels[1].h_entry =
571 				    info.h_levels[1].h_entry -
572 				    (entries + src_ent_num) +
573 				    dst_entries;
574 				info.h_levels[1].h_entries = dst_entries;
575 			}
576 			ext2fs_htree_insert_entry_to_level(&info.h_levels[0],
577 			    split_hash, blknum);
578 
579 			/* Write new index node to disk */
580 			error = bwrite(dst_bp);
581 			ip->i_flag |= IN_CHANGE | IN_UPDATE;
582 			if (error)
583 				goto finish;
584 
585 			write_dst_bp = 1;
586 		} else {
587 			/* Create second level for htree index */
588 
589 			struct ext2fs_htree_root *idx_root;
590 
591 			memcpy(dst_entries, entries,
592 			    ent_num * sizeof(struct ext2fs_htree_entry));
593 			ext2fs_htree_set_limit(dst_entries,
594 			    ext2fs_htree_node_limit(ip));
595 
596 			idx_root = (struct ext2fs_htree_root *)
597 			    info.h_levels[0].h_bp->b_data;
598 			idx_root->h_info.h_ind_levels = 1;
599 
600 			ext2fs_htree_set_count(entries, 1);
601 			ext2fs_htree_set_block(entries, blknum);
602 
603 			info.h_levels_num = 2;
604 			info.h_levels[1].h_entries = dst_entries;
605 			info.h_levels[1].h_entry = info.h_levels[0].h_entry -
606 			    info.h_levels[0].h_entries + dst_entries;
607 			info.h_levels[1].h_bp = dst_bp;
608 			dst_bp = NULL;
609 		}
610 	}
611 
612 	leaf_node = info.h_levels[info.h_levels_num - 1].h_entry;
613 	blknum = ext2fs_htree_get_block(leaf_node);
614 	error = ext2fs_blkatoff(dvp, blknum * blksize, NULL, &bp);
615 	if (error)
616 		goto finish;
617 
618 	/* Split target directory block */
619 	newdirblock = malloc(blksize, M_TEMP, M_WAITOK | M_ZERO);
620 	ext2fs_htree_split_dirblock((char *)bp->b_data, newdirblock, blksize,
621 	    fs->e3fs_hash_seed, hash_version, &split_hash, entry);
622 	cursize = roundup(ip->i_size, blksize);
623 	dirsize = cursize + blksize;
624 	blknum = dirsize / blksize - 1;
625 
626 	/* Add index entry for the new directory block */
627 	ext2fs_htree_insert_entry(&info, split_hash, blknum);
628 
629 	/* Write the new directory block to the end of the directory */
630 	error = ext2fs_htree_append_block(dvp, newdirblock, cnp, blksize);
631 
632 	if (error)
633 		goto finish;
634 
635 	/* Write the target directory block */
636 	error = bwrite(bp);
637 	ip->i_flag |= IN_CHANGE | IN_UPDATE;
638 
639 	if (error)
640 		goto finish;
641 	write_bp = 1;
642 
643 	/* Write the index block */
644 	error = ext2fs_htree_writebuf(&info);
645 	if (!error)
646 		write_info = 1;
647 
648 finish:
649 	if (dst_bp != NULL && !write_dst_bp)
650 		brelse(dst_bp, 0);
651 	if (bp != NULL && !write_bp)
652 		brelse(bp, 0);
653 	if (newdirblock != NULL)
654 		free(newdirblock, M_TEMP);
655 	if (newidxblock != NULL)
656 		free(newidxblock, M_TEMP);
657 	if (!write_info)
658 		ext2fs_htree_release(&info);
659 	return error;
660 }
661 
662 static int
ext2fs_htree_check_next(struct inode * ip,uint32_t hash,const char * name,struct ext2fs_htree_lookup_info * info)663 ext2fs_htree_check_next(struct inode *ip, uint32_t hash, const char *name,
664     struct ext2fs_htree_lookup_info *info)
665 {
666 	struct vnode *vp = ITOV(ip);
667 	struct ext2fs_htree_lookup_level *level;
668 	struct buf *bp;
669 	uint32_t next_hash;
670 	int idx = info->h_levels_num - 1;
671 	int levels = 0;
672 
673 	for (;;) {
674 		level = &info->h_levels[idx];
675 		level->h_entry++;
676 		if (level->h_entry < level->h_entries +
677 		    ext2fs_htree_get_count(level->h_entries))
678 			break;
679 		if (idx == 0)
680 			return 0;
681 		idx--;
682 		levels++;
683 	}
684 
685 	next_hash = ext2fs_htree_get_hash(level->h_entry);
686 	if ((hash & 1) == 0) {
687 		if (hash != (next_hash & ~1))
688 			return 0;
689 	}
690 
691 	while (levels > 0) {
692 		levels--;
693 		if (ext2fs_blkatoff(vp, ext2fs_htree_get_block(level->h_entry) *
694 		    ip->i_e2fs->e2fs_bsize, NULL, &bp) != 0)
695 			return 0;
696 		level = &info->h_levels[idx + 1];
697 		brelse(level->h_bp, 0);
698 		level->h_bp = bp;
699 		level->h_entry = level->h_entries =
700 		    ((struct ext2fs_htree_node *)bp->b_data)->h_entries;
701 	}
702 
703 	return 1;
704 }
705 
706 static int
ext2fs_htree_find_leaf(struct inode * ip,const char * name,int namelen,uint32_t * hash,uint8_t * hash_ver,struct ext2fs_htree_lookup_info * info)707 ext2fs_htree_find_leaf(struct inode *ip, const char *name, int namelen,
708     uint32_t *hash, uint8_t *hash_ver,
709     struct ext2fs_htree_lookup_info *info)
710 {
711 	struct vnode *vp;
712 	struct ext2fs *fs;/* F, G, and H are MD4 functions */
713 	struct m_ext2fs *m_fs;
714 	struct buf *bp = NULL;
715 	struct ext2fs_htree_root *rootp;
716 	struct ext2fs_htree_entry *entp, *start, *end, *middle, *found;
717 	struct ext2fs_htree_lookup_level *level_info;
718 	uint32_t hash_major = 0, hash_minor = 0;
719 	uint32_t levels, cnt;
720 	uint8_t hash_version;
721 
722 	if (name == NULL || info == NULL)
723 		return -1;
724 
725 	vp = ITOV(ip);
726 	fs = &(ip->i_e2fs->e2fs);
727 	m_fs = ip->i_e2fs;
728 
729 	if (ext2fs_blkatoff(vp, 0, NULL, &bp) != 0)
730 		return -1;
731 
732 	info->h_levels_num = 1;
733 	info->h_levels[0].h_bp = bp;
734 	rootp = (struct ext2fs_htree_root *)bp->b_data;
735 	if (rootp->h_info.h_hash_version != EXT2_HTREE_LEGACY &&
736 	    rootp->h_info.h_hash_version != EXT2_HTREE_HALF_MD4 &&
737 	    rootp->h_info.h_hash_version != EXT2_HTREE_TEA)
738 		goto error;
739 
740 	hash_version = rootp->h_info.h_hash_version;
741 	if (hash_version <= EXT2_HTREE_TEA)
742 		hash_version += m_fs->e2fs_uhash;
743 	*hash_ver = hash_version;
744 
745 	ext2fs_htree_hash(name, namelen, fs->e3fs_hash_seed,
746 	    hash_version, &hash_major, &hash_minor);
747 	*hash = hash_major;
748 
749 	if ((levels = rootp->h_info.h_ind_levels) > 1)
750 		goto error;
751 
752 	entp = (struct ext2fs_htree_entry *)(((char *)&rootp->h_info) +
753 	    rootp->h_info.h_info_len);
754 
755 	if (ext2fs_htree_get_limit(entp) !=
756 	    ext2fs_htree_root_limit(ip, rootp->h_info.h_info_len))
757 		goto error;
758 
759 	for (;;) {
760 		cnt = ext2fs_htree_get_count(entp);
761 		if (cnt == 0 || cnt > ext2fs_htree_get_limit(entp))
762 			goto error;
763 
764 		start = entp + 1;
765 		end = entp + cnt - 1;
766 		while (start <= end) {
767 			middle = start + (end - start) / 2;
768 			if (ext2fs_htree_get_hash(middle) > hash_major)
769 				end = middle - 1;
770 			else
771 				start = middle + 1;
772 		}
773 		found = start - 1;
774 
775 		level_info = &(info->h_levels[info->h_levels_num - 1]);
776 		level_info->h_bp = bp;
777 		level_info->h_entries = entp;
778 		level_info->h_entry = found;
779 		if (levels == 0)
780 			return 0;
781 		levels--;
782 		if (ext2fs_blkatoff(vp,
783 		    ext2fs_htree_get_block(found) * m_fs->e2fs_bsize,
784 		    NULL, &bp) != 0)
785 			goto error;
786 		entp = ((struct ext2fs_htree_node *)bp->b_data)->h_entries;
787 		info->h_levels_num++;
788 		info->h_levels[info->h_levels_num - 1].h_bp = bp;
789 	}
790 
791 error:
792 	ext2fs_htree_release(info);
793 	return -1;
794 }
795 
796 /*
797  * Try to lookup a directory entry in HTree index
798  */
799 int
ext2fs_htree_lookup(struct inode * ip,const char * name,int namelen,struct buf ** bpp,int * entryoffp,doff_t * offp,doff_t * prevoffp,doff_t * endusefulp,struct ext2fs_searchslot * ss)800 ext2fs_htree_lookup(struct inode *ip, const char *name, int namelen,
801     struct buf **bpp, int *entryoffp, doff_t *offp,
802     doff_t *prevoffp, doff_t *endusefulp, struct ext2fs_searchslot *ss)
803 {
804 	struct vnode *vp;
805 	struct ext2fs_htree_lookup_info info;
806 	struct ext2fs_htree_entry *leaf_node;
807 	struct m_ext2fs *m_fs;
808 	struct buf *bp;
809 	uint32_t blk;
810 	uint32_t dirhash;
811 	uint32_t bsize;
812 	uint8_t hash_version;
813 	int search_next;
814 
815 	m_fs = ip->i_e2fs;
816 	bsize = m_fs->e2fs_bsize;
817 	vp = ITOV(ip);
818 
819 	/* TODO: print error msg because we don't lookup '.' and '..' */
820 
821 	memset(&info, 0, sizeof(info));
822 	if (ext2fs_htree_find_leaf(ip, name, namelen, &dirhash,
823 	    &hash_version, &info)) {
824 		return -1;
825 	}
826 
827 	do {
828 		leaf_node = info.h_levels[info.h_levels_num - 1].h_entry;
829 		blk = ext2fs_htree_get_block(leaf_node);
830 		if (ext2fs_blkatoff(vp, blk * bsize, NULL, &bp) != 0) {
831 			ext2fs_htree_release(&info);
832 			return -1;
833 		}
834 
835 		*offp = blk * bsize;
836 		*entryoffp = 0;
837 		*prevoffp = blk * bsize;
838 		*endusefulp = blk * bsize;
839 
840 		if (ss->slotstatus == NONE) {
841 			ss->slotoffset = -1;
842 			ss->slotfreespace = 0;
843 		}
844 
845 		int found;
846 		if (ext2fs_search_dirblock(ip, bp->b_data, &found,
847 		    name, namelen, entryoffp, offp, prevoffp,
848 		    endusefulp, ss) != 0) {
849 			brelse(bp, 0);
850 			ext2fs_htree_release(&info);
851 			return -1;
852 		}
853 
854 		if (found) {
855 			*bpp = bp;
856 			ext2fs_htree_release(&info);
857 			return 0;
858 		}
859 
860 		brelse(bp, 0);
861 		search_next = ext2fs_htree_check_next(ip, dirhash, name, &info);
862 	} while (search_next);
863 
864 	ext2fs_htree_release(&info);
865 	return ENOENT;
866 }
867