xref: /linux/fs/xfs/libxfs/xfs_dir2_node.c (revision 36ec807b627b4c0a0a382f0ae48eac7187d14b2b)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
4  * Copyright (c) 2013 Red Hat, Inc.
5  * All Rights Reserved.
6  */
7 #include "xfs.h"
8 #include "xfs_fs.h"
9 #include "xfs_shared.h"
10 #include "xfs_format.h"
11 #include "xfs_log_format.h"
12 #include "xfs_trans_resv.h"
13 #include "xfs_mount.h"
14 #include "xfs_inode.h"
15 #include "xfs_bmap.h"
16 #include "xfs_dir2.h"
17 #include "xfs_dir2_priv.h"
18 #include "xfs_error.h"
19 #include "xfs_trace.h"
20 #include "xfs_trans.h"
21 #include "xfs_buf_item.h"
22 #include "xfs_log.h"
23 #include "xfs_health.h"
24 
25 /*
26  * Function declarations.
27  */
28 static int xfs_dir2_leafn_add(struct xfs_buf *bp, xfs_da_args_t *args,
29 			      int index);
30 static void xfs_dir2_leafn_rebalance(xfs_da_state_t *state,
31 				     xfs_da_state_blk_t *blk1,
32 				     xfs_da_state_blk_t *blk2);
33 static int xfs_dir2_leafn_remove(xfs_da_args_t *args, struct xfs_buf *bp,
34 				 int index, xfs_da_state_blk_t *dblk,
35 				 int *rval);
36 
37 /*
38  * Convert data space db to the corresponding free db.
39  */
40 static xfs_dir2_db_t
41 xfs_dir2_db_to_fdb(struct xfs_da_geometry *geo, xfs_dir2_db_t db)
42 {
43 	return xfs_dir2_byte_to_db(geo, XFS_DIR2_FREE_OFFSET) +
44 			(db / geo->free_max_bests);
45 }
46 
47 /*
48  * Convert data space db to the corresponding index in a free db.
49  */
50 static int
51 xfs_dir2_db_to_fdindex(struct xfs_da_geometry *geo, xfs_dir2_db_t db)
52 {
53 	return db % geo->free_max_bests;
54 }
55 
56 /*
57  * Check internal consistency of a leafn block.
58  */
59 #ifdef DEBUG
60 static xfs_failaddr_t
61 xfs_dir3_leafn_check(
62 	struct xfs_inode	*dp,
63 	struct xfs_buf		*bp)
64 {
65 	struct xfs_dir2_leaf	*leaf = bp->b_addr;
66 	struct xfs_dir3_icleaf_hdr leafhdr;
67 
68 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &leafhdr, leaf);
69 
70 	if (leafhdr.magic == XFS_DIR3_LEAFN_MAGIC) {
71 		struct xfs_dir3_leaf_hdr *leaf3 = bp->b_addr;
72 		if (be64_to_cpu(leaf3->info.blkno) != xfs_buf_daddr(bp))
73 			return __this_address;
74 	} else if (leafhdr.magic != XFS_DIR2_LEAFN_MAGIC)
75 		return __this_address;
76 
77 	return xfs_dir3_leaf_check_int(dp->i_mount, &leafhdr, leaf, false);
78 }
79 
80 static inline void
81 xfs_dir3_leaf_check(
82 	struct xfs_inode	*dp,
83 	struct xfs_buf		*bp)
84 {
85 	xfs_failaddr_t		fa;
86 
87 	fa = xfs_dir3_leafn_check(dp, bp);
88 	if (!fa)
89 		return;
90 	xfs_corruption_error(__func__, XFS_ERRLEVEL_LOW, dp->i_mount,
91 			bp->b_addr, BBTOB(bp->b_length), __FILE__, __LINE__,
92 			fa);
93 	ASSERT(0);
94 }
95 #else
96 #define	xfs_dir3_leaf_check(dp, bp)
97 #endif
98 
99 static xfs_failaddr_t
100 xfs_dir3_free_verify(
101 	struct xfs_buf		*bp)
102 {
103 	struct xfs_mount	*mp = bp->b_mount;
104 	struct xfs_dir2_free_hdr *hdr = bp->b_addr;
105 
106 	if (!xfs_verify_magic(bp, hdr->magic))
107 		return __this_address;
108 
109 	if (xfs_has_crc(mp)) {
110 		struct xfs_dir3_blk_hdr *hdr3 = bp->b_addr;
111 
112 		if (!uuid_equal(&hdr3->uuid, &mp->m_sb.sb_meta_uuid))
113 			return __this_address;
114 		if (be64_to_cpu(hdr3->blkno) != xfs_buf_daddr(bp))
115 			return __this_address;
116 		if (!xfs_log_check_lsn(mp, be64_to_cpu(hdr3->lsn)))
117 			return __this_address;
118 	}
119 
120 	/* XXX: should bounds check the xfs_dir3_icfree_hdr here */
121 
122 	return NULL;
123 }
124 
125 static void
126 xfs_dir3_free_read_verify(
127 	struct xfs_buf	*bp)
128 {
129 	struct xfs_mount	*mp = bp->b_mount;
130 	xfs_failaddr_t		fa;
131 
132 	if (xfs_has_crc(mp) &&
133 	    !xfs_buf_verify_cksum(bp, XFS_DIR3_FREE_CRC_OFF))
134 		xfs_verifier_error(bp, -EFSBADCRC, __this_address);
135 	else {
136 		fa = xfs_dir3_free_verify(bp);
137 		if (fa)
138 			xfs_verifier_error(bp, -EFSCORRUPTED, fa);
139 	}
140 }
141 
142 static void
143 xfs_dir3_free_write_verify(
144 	struct xfs_buf	*bp)
145 {
146 	struct xfs_mount	*mp = bp->b_mount;
147 	struct xfs_buf_log_item	*bip = bp->b_log_item;
148 	struct xfs_dir3_blk_hdr	*hdr3 = bp->b_addr;
149 	xfs_failaddr_t		fa;
150 
151 	fa = xfs_dir3_free_verify(bp);
152 	if (fa) {
153 		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
154 		return;
155 	}
156 
157 	if (!xfs_has_crc(mp))
158 		return;
159 
160 	if (bip)
161 		hdr3->lsn = cpu_to_be64(bip->bli_item.li_lsn);
162 
163 	xfs_buf_update_cksum(bp, XFS_DIR3_FREE_CRC_OFF);
164 }
165 
166 const struct xfs_buf_ops xfs_dir3_free_buf_ops = {
167 	.name = "xfs_dir3_free",
168 	.magic = { cpu_to_be32(XFS_DIR2_FREE_MAGIC),
169 		   cpu_to_be32(XFS_DIR3_FREE_MAGIC) },
170 	.verify_read = xfs_dir3_free_read_verify,
171 	.verify_write = xfs_dir3_free_write_verify,
172 	.verify_struct = xfs_dir3_free_verify,
173 };
174 
175 /* Everything ok in the free block header? */
176 static xfs_failaddr_t
177 xfs_dir3_free_header_check(
178 	struct xfs_buf		*bp,
179 	xfs_ino_t		owner,
180 	xfs_dablk_t		fbno)
181 {
182 	struct xfs_mount	*mp = bp->b_mount;
183 	int			maxbests = mp->m_dir_geo->free_max_bests;
184 	unsigned int		firstdb;
185 
186 	firstdb = (xfs_dir2_da_to_db(mp->m_dir_geo, fbno) -
187 		   xfs_dir2_byte_to_db(mp->m_dir_geo, XFS_DIR2_FREE_OFFSET)) *
188 			maxbests;
189 	if (xfs_has_crc(mp)) {
190 		struct xfs_dir3_free_hdr *hdr3 = bp->b_addr;
191 
192 		if (be32_to_cpu(hdr3->firstdb) != firstdb)
193 			return __this_address;
194 		if (be32_to_cpu(hdr3->nvalid) > maxbests)
195 			return __this_address;
196 		if (be32_to_cpu(hdr3->nvalid) < be32_to_cpu(hdr3->nused))
197 			return __this_address;
198 		if (be64_to_cpu(hdr3->hdr.owner) != owner)
199 			return __this_address;
200 	} else {
201 		struct xfs_dir2_free_hdr *hdr = bp->b_addr;
202 
203 		if (be32_to_cpu(hdr->firstdb) != firstdb)
204 			return __this_address;
205 		if (be32_to_cpu(hdr->nvalid) > maxbests)
206 			return __this_address;
207 		if (be32_to_cpu(hdr->nvalid) < be32_to_cpu(hdr->nused))
208 			return __this_address;
209 	}
210 	return NULL;
211 }
212 
213 static int
214 __xfs_dir3_free_read(
215 	struct xfs_trans	*tp,
216 	struct xfs_inode	*dp,
217 	xfs_ino_t		owner,
218 	xfs_dablk_t		fbno,
219 	unsigned int		flags,
220 	struct xfs_buf		**bpp)
221 {
222 	xfs_failaddr_t		fa;
223 	int			err;
224 
225 	err = xfs_da_read_buf(tp, dp, fbno, flags, bpp, XFS_DATA_FORK,
226 			&xfs_dir3_free_buf_ops);
227 	if (err || !*bpp)
228 		return err;
229 
230 	/* Check things that we can't do in the verifier. */
231 	fa = xfs_dir3_free_header_check(*bpp, owner, fbno);
232 	if (fa) {
233 		__xfs_buf_mark_corrupt(*bpp, fa);
234 		xfs_trans_brelse(tp, *bpp);
235 		*bpp = NULL;
236 		xfs_dirattr_mark_sick(dp, XFS_DATA_FORK);
237 		return -EFSCORRUPTED;
238 	}
239 
240 	/* try read returns without an error or *bpp if it lands in a hole */
241 	if (tp)
242 		xfs_trans_buf_set_type(tp, *bpp, XFS_BLFT_DIR_FREE_BUF);
243 
244 	return 0;
245 }
246 
247 void
248 xfs_dir2_free_hdr_from_disk(
249 	struct xfs_mount		*mp,
250 	struct xfs_dir3_icfree_hdr	*to,
251 	struct xfs_dir2_free		*from)
252 {
253 	if (xfs_has_crc(mp)) {
254 		struct xfs_dir3_free	*from3 = (struct xfs_dir3_free *)from;
255 
256 		to->magic = be32_to_cpu(from3->hdr.hdr.magic);
257 		to->firstdb = be32_to_cpu(from3->hdr.firstdb);
258 		to->nvalid = be32_to_cpu(from3->hdr.nvalid);
259 		to->nused = be32_to_cpu(from3->hdr.nused);
260 		to->bests = from3->bests;
261 
262 		ASSERT(to->magic == XFS_DIR3_FREE_MAGIC);
263 	} else {
264 		to->magic = be32_to_cpu(from->hdr.magic);
265 		to->firstdb = be32_to_cpu(from->hdr.firstdb);
266 		to->nvalid = be32_to_cpu(from->hdr.nvalid);
267 		to->nused = be32_to_cpu(from->hdr.nused);
268 		to->bests = from->bests;
269 
270 		ASSERT(to->magic == XFS_DIR2_FREE_MAGIC);
271 	}
272 }
273 
274 static void
275 xfs_dir2_free_hdr_to_disk(
276 	struct xfs_mount		*mp,
277 	struct xfs_dir2_free		*to,
278 	struct xfs_dir3_icfree_hdr	*from)
279 {
280 	if (xfs_has_crc(mp)) {
281 		struct xfs_dir3_free	*to3 = (struct xfs_dir3_free *)to;
282 
283 		ASSERT(from->magic == XFS_DIR3_FREE_MAGIC);
284 
285 		to3->hdr.hdr.magic = cpu_to_be32(from->magic);
286 		to3->hdr.firstdb = cpu_to_be32(from->firstdb);
287 		to3->hdr.nvalid = cpu_to_be32(from->nvalid);
288 		to3->hdr.nused = cpu_to_be32(from->nused);
289 	} else {
290 		ASSERT(from->magic == XFS_DIR2_FREE_MAGIC);
291 
292 		to->hdr.magic = cpu_to_be32(from->magic);
293 		to->hdr.firstdb = cpu_to_be32(from->firstdb);
294 		to->hdr.nvalid = cpu_to_be32(from->nvalid);
295 		to->hdr.nused = cpu_to_be32(from->nused);
296 	}
297 }
298 
299 int
300 xfs_dir2_free_read(
301 	struct xfs_trans	*tp,
302 	struct xfs_inode	*dp,
303 	xfs_ino_t		owner,
304 	xfs_dablk_t		fbno,
305 	struct xfs_buf		**bpp)
306 {
307 	return __xfs_dir3_free_read(tp, dp, owner, fbno, 0, bpp);
308 }
309 
310 static int
311 xfs_dir2_free_try_read(
312 	struct xfs_trans	*tp,
313 	struct xfs_inode	*dp,
314 	xfs_ino_t		owner,
315 	xfs_dablk_t		fbno,
316 	struct xfs_buf		**bpp)
317 {
318 	return __xfs_dir3_free_read(tp, dp, owner, fbno, XFS_DABUF_MAP_HOLE_OK,
319 			bpp);
320 }
321 
322 static int
323 xfs_dir3_free_get_buf(
324 	xfs_da_args_t		*args,
325 	xfs_dir2_db_t		fbno,
326 	struct xfs_buf		**bpp)
327 {
328 	struct xfs_trans	*tp = args->trans;
329 	struct xfs_inode	*dp = args->dp;
330 	struct xfs_mount	*mp = dp->i_mount;
331 	struct xfs_buf		*bp;
332 	int			error;
333 	struct xfs_dir3_icfree_hdr hdr;
334 
335 	error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(args->geo, fbno),
336 			&bp, XFS_DATA_FORK);
337 	if (error)
338 		return error;
339 
340 	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_DIR_FREE_BUF);
341 	bp->b_ops = &xfs_dir3_free_buf_ops;
342 
343 	/*
344 	 * Initialize the new block to be empty, and remember
345 	 * its first slot as our empty slot.
346 	 */
347 	memset(bp->b_addr, 0, sizeof(struct xfs_dir3_free_hdr));
348 	memset(&hdr, 0, sizeof(hdr));
349 
350 	if (xfs_has_crc(mp)) {
351 		struct xfs_dir3_free_hdr *hdr3 = bp->b_addr;
352 
353 		hdr.magic = XFS_DIR3_FREE_MAGIC;
354 
355 		hdr3->hdr.blkno = cpu_to_be64(xfs_buf_daddr(bp));
356 		hdr3->hdr.owner = cpu_to_be64(args->owner);
357 		uuid_copy(&hdr3->hdr.uuid, &mp->m_sb.sb_meta_uuid);
358 	} else
359 		hdr.magic = XFS_DIR2_FREE_MAGIC;
360 	xfs_dir2_free_hdr_to_disk(mp, bp->b_addr, &hdr);
361 	*bpp = bp;
362 	return 0;
363 }
364 
365 /*
366  * Log entries from a freespace block.
367  */
368 STATIC void
369 xfs_dir2_free_log_bests(
370 	struct xfs_da_args	*args,
371 	struct xfs_dir3_icfree_hdr *hdr,
372 	struct xfs_buf		*bp,
373 	int			first,		/* first entry to log */
374 	int			last)		/* last entry to log */
375 {
376 	struct xfs_dir2_free	*free = bp->b_addr;
377 
378 	ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC) ||
379 	       free->hdr.magic == cpu_to_be32(XFS_DIR3_FREE_MAGIC));
380 	xfs_trans_log_buf(args->trans, bp,
381 			  (char *)&hdr->bests[first] - (char *)free,
382 			  (char *)&hdr->bests[last] - (char *)free +
383 			   sizeof(hdr->bests[0]) - 1);
384 }
385 
386 /*
387  * Log header from a freespace block.
388  */
389 static void
390 xfs_dir2_free_log_header(
391 	struct xfs_da_args	*args,
392 	struct xfs_buf		*bp)
393 {
394 #ifdef DEBUG
395 	xfs_dir2_free_t		*free;		/* freespace structure */
396 
397 	free = bp->b_addr;
398 	ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC) ||
399 	       free->hdr.magic == cpu_to_be32(XFS_DIR3_FREE_MAGIC));
400 #endif
401 	xfs_trans_log_buf(args->trans, bp, 0,
402 			  args->geo->free_hdr_size - 1);
403 }
404 
405 /*
406  * Convert a leaf-format directory to a node-format directory.
407  * We need to change the magic number of the leaf block, and copy
408  * the freespace table out of the leaf block into its own block.
409  */
410 int						/* error */
411 xfs_dir2_leaf_to_node(
412 	xfs_da_args_t		*args,		/* operation arguments */
413 	struct xfs_buf		*lbp)		/* leaf buffer */
414 {
415 	xfs_inode_t		*dp;		/* incore directory inode */
416 	int			error;		/* error return value */
417 	struct xfs_buf		*fbp;		/* freespace buffer */
418 	xfs_dir2_db_t		fdb;		/* freespace block number */
419 	__be16			*from;		/* pointer to freespace entry */
420 	int			i;		/* leaf freespace index */
421 	xfs_dir2_leaf_t		*leaf;		/* leaf structure */
422 	xfs_dir2_leaf_tail_t	*ltp;		/* leaf tail structure */
423 	int			n;		/* count of live freespc ents */
424 	xfs_dir2_data_off_t	off;		/* freespace entry value */
425 	xfs_trans_t		*tp;		/* transaction pointer */
426 	struct xfs_dir3_icfree_hdr freehdr;
427 
428 	trace_xfs_dir2_leaf_to_node(args);
429 
430 	dp = args->dp;
431 	tp = args->trans;
432 	/*
433 	 * Add a freespace block to the directory.
434 	 */
435 	if ((error = xfs_dir2_grow_inode(args, XFS_DIR2_FREE_SPACE, &fdb))) {
436 		return error;
437 	}
438 	ASSERT(fdb == xfs_dir2_byte_to_db(args->geo, XFS_DIR2_FREE_OFFSET));
439 	/*
440 	 * Get the buffer for the new freespace block.
441 	 */
442 	error = xfs_dir3_free_get_buf(args, fdb, &fbp);
443 	if (error)
444 		return error;
445 
446 	xfs_dir2_free_hdr_from_disk(dp->i_mount, &freehdr, fbp->b_addr);
447 	leaf = lbp->b_addr;
448 	ltp = xfs_dir2_leaf_tail_p(args->geo, leaf);
449 	if (be32_to_cpu(ltp->bestcount) >
450 				(uint)dp->i_disk_size / args->geo->blksize) {
451 		xfs_buf_mark_corrupt(lbp);
452 		xfs_da_mark_sick(args);
453 		return -EFSCORRUPTED;
454 	}
455 
456 	/*
457 	 * Copy freespace entries from the leaf block to the new block.
458 	 * Count active entries.
459 	 */
460 	from = xfs_dir2_leaf_bests_p(ltp);
461 	for (i = n = 0; i < be32_to_cpu(ltp->bestcount); i++, from++) {
462 		off = be16_to_cpu(*from);
463 		if (off != NULLDATAOFF)
464 			n++;
465 		freehdr.bests[i] = cpu_to_be16(off);
466 	}
467 
468 	/*
469 	 * Now initialize the freespace block header.
470 	 */
471 	freehdr.nused = n;
472 	freehdr.nvalid = be32_to_cpu(ltp->bestcount);
473 
474 	xfs_dir2_free_hdr_to_disk(dp->i_mount, fbp->b_addr, &freehdr);
475 	xfs_dir2_free_log_bests(args, &freehdr, fbp, 0, freehdr.nvalid - 1);
476 	xfs_dir2_free_log_header(args, fbp);
477 
478 	/*
479 	 * Converting the leaf to a leafnode is just a matter of changing the
480 	 * magic number and the ops. Do the change directly to the buffer as
481 	 * it's less work (and less code) than decoding the header to host
482 	 * format and back again.
483 	 */
484 	if (leaf->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAF1_MAGIC))
485 		leaf->hdr.info.magic = cpu_to_be16(XFS_DIR2_LEAFN_MAGIC);
486 	else
487 		leaf->hdr.info.magic = cpu_to_be16(XFS_DIR3_LEAFN_MAGIC);
488 	lbp->b_ops = &xfs_dir3_leafn_buf_ops;
489 	xfs_trans_buf_set_type(tp, lbp, XFS_BLFT_DIR_LEAFN_BUF);
490 	xfs_dir3_leaf_log_header(args, lbp);
491 	xfs_dir3_leaf_check(dp, lbp);
492 	return 0;
493 }
494 
495 /*
496  * Add a leaf entry to a leaf block in a node-form directory.
497  * The other work necessary is done from the caller.
498  */
499 static int					/* error */
500 xfs_dir2_leafn_add(
501 	struct xfs_buf		*bp,		/* leaf buffer */
502 	struct xfs_da_args	*args,		/* operation arguments */
503 	int			index)		/* insertion pt for new entry */
504 {
505 	struct xfs_dir3_icleaf_hdr leafhdr;
506 	struct xfs_inode	*dp = args->dp;
507 	struct xfs_dir2_leaf	*leaf = bp->b_addr;
508 	struct xfs_dir2_leaf_entry *lep;
509 	struct xfs_dir2_leaf_entry *ents;
510 	int			compact;	/* compacting stale leaves */
511 	int			highstale = 0;	/* next stale entry */
512 	int			lfloghigh;	/* high leaf entry logging */
513 	int			lfloglow;	/* low leaf entry logging */
514 	int			lowstale = 0;	/* previous stale entry */
515 
516 	trace_xfs_dir2_leafn_add(args, index);
517 
518 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &leafhdr, leaf);
519 	ents = leafhdr.ents;
520 
521 	/*
522 	 * Quick check just to make sure we are not going to index
523 	 * into other peoples memory
524 	 */
525 	if (index < 0) {
526 		xfs_buf_mark_corrupt(bp);
527 		xfs_da_mark_sick(args);
528 		return -EFSCORRUPTED;
529 	}
530 
531 	/*
532 	 * If there are already the maximum number of leaf entries in
533 	 * the block, if there are no stale entries it won't fit.
534 	 * Caller will do a split.  If there are stale entries we'll do
535 	 * a compact.
536 	 */
537 
538 	if (leafhdr.count == args->geo->leaf_max_ents) {
539 		if (!leafhdr.stale)
540 			return -ENOSPC;
541 		compact = leafhdr.stale > 1;
542 	} else
543 		compact = 0;
544 	ASSERT(index == 0 || be32_to_cpu(ents[index - 1].hashval) <= args->hashval);
545 	ASSERT(index == leafhdr.count ||
546 	       be32_to_cpu(ents[index].hashval) >= args->hashval);
547 
548 	if (args->op_flags & XFS_DA_OP_JUSTCHECK)
549 		return 0;
550 
551 	/*
552 	 * Compact out all but one stale leaf entry.  Leaves behind
553 	 * the entry closest to index.
554 	 */
555 	if (compact)
556 		xfs_dir3_leaf_compact_x1(&leafhdr, ents, &index, &lowstale,
557 					 &highstale, &lfloglow, &lfloghigh);
558 	else if (leafhdr.stale) {
559 		/*
560 		 * Set impossible logging indices for this case.
561 		 */
562 		lfloglow = leafhdr.count;
563 		lfloghigh = -1;
564 	}
565 
566 	/*
567 	 * Insert the new entry, log everything.
568 	 */
569 	lep = xfs_dir3_leaf_find_entry(&leafhdr, ents, index, compact, lowstale,
570 				       highstale, &lfloglow, &lfloghigh);
571 
572 	lep->hashval = cpu_to_be32(args->hashval);
573 	lep->address = cpu_to_be32(xfs_dir2_db_off_to_dataptr(args->geo,
574 				args->blkno, args->index));
575 
576 	xfs_dir2_leaf_hdr_to_disk(dp->i_mount, leaf, &leafhdr);
577 	xfs_dir3_leaf_log_header(args, bp);
578 	xfs_dir3_leaf_log_ents(args, &leafhdr, bp, lfloglow, lfloghigh);
579 	xfs_dir3_leaf_check(dp, bp);
580 	return 0;
581 }
582 
583 #ifdef DEBUG
584 static void
585 xfs_dir2_free_hdr_check(
586 	struct xfs_inode *dp,
587 	struct xfs_buf	*bp,
588 	xfs_dir2_db_t	db)
589 {
590 	struct xfs_dir3_icfree_hdr hdr;
591 
592 	xfs_dir2_free_hdr_from_disk(dp->i_mount, &hdr, bp->b_addr);
593 
594 	ASSERT((hdr.firstdb % dp->i_mount->m_dir_geo->free_max_bests) == 0);
595 	ASSERT(hdr.firstdb <= db);
596 	ASSERT(db < hdr.firstdb + hdr.nvalid);
597 }
598 #else
599 #define xfs_dir2_free_hdr_check(dp, bp, db)
600 #endif	/* DEBUG */
601 
602 /*
603  * Return the last hash value in the leaf.
604  * Stale entries are ok.
605  */
606 xfs_dahash_t					/* hash value */
607 xfs_dir2_leaf_lasthash(
608 	struct xfs_inode *dp,
609 	struct xfs_buf	*bp,			/* leaf buffer */
610 	int		*count)			/* count of entries in leaf */
611 {
612 	struct xfs_dir3_icleaf_hdr leafhdr;
613 
614 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &leafhdr, bp->b_addr);
615 
616 	ASSERT(leafhdr.magic == XFS_DIR2_LEAFN_MAGIC ||
617 	       leafhdr.magic == XFS_DIR3_LEAFN_MAGIC ||
618 	       leafhdr.magic == XFS_DIR2_LEAF1_MAGIC ||
619 	       leafhdr.magic == XFS_DIR3_LEAF1_MAGIC);
620 
621 	if (count)
622 		*count = leafhdr.count;
623 	if (!leafhdr.count)
624 		return 0;
625 	return be32_to_cpu(leafhdr.ents[leafhdr.count - 1].hashval);
626 }
627 
628 /*
629  * Look up a leaf entry for space to add a name in a node-format leaf block.
630  * The extrablk in state is a freespace block.
631  */
632 STATIC int
633 xfs_dir2_leafn_lookup_for_addname(
634 	struct xfs_buf		*bp,		/* leaf buffer */
635 	xfs_da_args_t		*args,		/* operation arguments */
636 	int			*indexp,	/* out: leaf entry index */
637 	xfs_da_state_t		*state)		/* state to fill in */
638 {
639 	struct xfs_buf		*curbp = NULL;	/* current data/free buffer */
640 	xfs_dir2_db_t		curdb = -1;	/* current data block number */
641 	xfs_dir2_db_t		curfdb = -1;	/* current free block number */
642 	xfs_inode_t		*dp;		/* incore directory inode */
643 	int			error;		/* error return value */
644 	int			fi;		/* free entry index */
645 	xfs_dir2_free_t		*free = NULL;	/* free block structure */
646 	int			index;		/* leaf entry index */
647 	xfs_dir2_leaf_t		*leaf;		/* leaf structure */
648 	int			length;		/* length of new data entry */
649 	xfs_dir2_leaf_entry_t	*lep;		/* leaf entry */
650 	xfs_mount_t		*mp;		/* filesystem mount point */
651 	xfs_dir2_db_t		newdb;		/* new data block number */
652 	xfs_dir2_db_t		newfdb;		/* new free block number */
653 	xfs_trans_t		*tp;		/* transaction pointer */
654 	struct xfs_dir3_icleaf_hdr leafhdr;
655 
656 	dp = args->dp;
657 	tp = args->trans;
658 	mp = dp->i_mount;
659 	leaf = bp->b_addr;
660 	xfs_dir2_leaf_hdr_from_disk(mp, &leafhdr, leaf);
661 
662 	xfs_dir3_leaf_check(dp, bp);
663 	ASSERT(leafhdr.count > 0);
664 
665 	/*
666 	 * Look up the hash value in the leaf entries.
667 	 */
668 	index = xfs_dir2_leaf_search_hash(args, bp);
669 	/*
670 	 * Do we have a buffer coming in?
671 	 */
672 	if (state->extravalid) {
673 		/* If so, it's a free block buffer, get the block number. */
674 		curbp = state->extrablk.bp;
675 		curfdb = state->extrablk.blkno;
676 		free = curbp->b_addr;
677 		ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC) ||
678 		       free->hdr.magic == cpu_to_be32(XFS_DIR3_FREE_MAGIC));
679 	}
680 	length = xfs_dir2_data_entsize(mp, args->namelen);
681 	/*
682 	 * Loop over leaf entries with the right hash value.
683 	 */
684 	for (lep = &leafhdr.ents[index];
685 	     index < leafhdr.count && be32_to_cpu(lep->hashval) == args->hashval;
686 	     lep++, index++) {
687 		/*
688 		 * Skip stale leaf entries.
689 		 */
690 		if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
691 			continue;
692 		/*
693 		 * Pull the data block number from the entry.
694 		 */
695 		newdb = xfs_dir2_dataptr_to_db(args->geo,
696 					       be32_to_cpu(lep->address));
697 		/*
698 		 * For addname, we're looking for a place to put the new entry.
699 		 * We want to use a data block with an entry of equal
700 		 * hash value to ours if there is one with room.
701 		 *
702 		 * If this block isn't the data block we already have
703 		 * in hand, take a look at it.
704 		 */
705 		if (newdb != curdb) {
706 			struct xfs_dir3_icfree_hdr freehdr;
707 
708 			curdb = newdb;
709 			/*
710 			 * Convert the data block to the free block
711 			 * holding its freespace information.
712 			 */
713 			newfdb = xfs_dir2_db_to_fdb(args->geo, newdb);
714 			/*
715 			 * If it's not the one we have in hand, read it in.
716 			 */
717 			if (newfdb != curfdb) {
718 				/*
719 				 * If we had one before, drop it.
720 				 */
721 				if (curbp)
722 					xfs_trans_brelse(tp, curbp);
723 
724 				error = xfs_dir2_free_read(tp, dp, args->owner,
725 						xfs_dir2_db_to_da(args->geo,
726 								  newfdb),
727 						&curbp);
728 				if (error)
729 					return error;
730 				free = curbp->b_addr;
731 
732 				xfs_dir2_free_hdr_check(dp, curbp, curdb);
733 			}
734 			/*
735 			 * Get the index for our entry.
736 			 */
737 			fi = xfs_dir2_db_to_fdindex(args->geo, curdb);
738 			/*
739 			 * If it has room, return it.
740 			 */
741 			xfs_dir2_free_hdr_from_disk(mp, &freehdr, free);
742 			if (XFS_IS_CORRUPT(mp,
743 					   freehdr.bests[fi] ==
744 					   cpu_to_be16(NULLDATAOFF))) {
745 				if (curfdb != newfdb)
746 					xfs_trans_brelse(tp, curbp);
747 				xfs_da_mark_sick(args);
748 				return -EFSCORRUPTED;
749 			}
750 			curfdb = newfdb;
751 			if (be16_to_cpu(freehdr.bests[fi]) >= length)
752 				goto out;
753 		}
754 	}
755 	/* Didn't find any space */
756 	fi = -1;
757 out:
758 	ASSERT(args->op_flags & XFS_DA_OP_OKNOENT);
759 	if (curbp) {
760 		/* Giving back a free block. */
761 		state->extravalid = 1;
762 		state->extrablk.bp = curbp;
763 		state->extrablk.index = fi;
764 		state->extrablk.blkno = curfdb;
765 
766 		/*
767 		 * Important: this magic number is not in the buffer - it's for
768 		 * buffer type information and therefore only the free/data type
769 		 * matters here, not whether CRCs are enabled or not.
770 		 */
771 		state->extrablk.magic = XFS_DIR2_FREE_MAGIC;
772 	} else {
773 		state->extravalid = 0;
774 	}
775 	/*
776 	 * Return the index, that will be the insertion point.
777 	 */
778 	*indexp = index;
779 	return -ENOENT;
780 }
781 
782 /*
783  * Look up a leaf entry in a node-format leaf block.
784  * The extrablk in state a data block.
785  */
786 STATIC int
787 xfs_dir2_leafn_lookup_for_entry(
788 	struct xfs_buf		*bp,		/* leaf buffer */
789 	xfs_da_args_t		*args,		/* operation arguments */
790 	int			*indexp,	/* out: leaf entry index */
791 	xfs_da_state_t		*state)		/* state to fill in */
792 {
793 	struct xfs_buf		*curbp = NULL;	/* current data/free buffer */
794 	xfs_dir2_db_t		curdb = -1;	/* current data block number */
795 	xfs_dir2_data_entry_t	*dep;		/* data block entry */
796 	xfs_inode_t		*dp;		/* incore directory inode */
797 	int			error;		/* error return value */
798 	int			index;		/* leaf entry index */
799 	xfs_dir2_leaf_t		*leaf;		/* leaf structure */
800 	xfs_dir2_leaf_entry_t	*lep;		/* leaf entry */
801 	xfs_mount_t		*mp;		/* filesystem mount point */
802 	xfs_dir2_db_t		newdb;		/* new data block number */
803 	xfs_trans_t		*tp;		/* transaction pointer */
804 	enum xfs_dacmp		cmp;		/* comparison result */
805 	struct xfs_dir3_icleaf_hdr leafhdr;
806 
807 	dp = args->dp;
808 	tp = args->trans;
809 	mp = dp->i_mount;
810 	leaf = bp->b_addr;
811 	xfs_dir2_leaf_hdr_from_disk(mp, &leafhdr, leaf);
812 
813 	xfs_dir3_leaf_check(dp, bp);
814 	if (leafhdr.count <= 0) {
815 		xfs_buf_mark_corrupt(bp);
816 		xfs_da_mark_sick(args);
817 		return -EFSCORRUPTED;
818 	}
819 
820 	/*
821 	 * Look up the hash value in the leaf entries.
822 	 */
823 	index = xfs_dir2_leaf_search_hash(args, bp);
824 	/*
825 	 * Do we have a buffer coming in?
826 	 */
827 	if (state->extravalid) {
828 		curbp = state->extrablk.bp;
829 		curdb = state->extrablk.blkno;
830 	}
831 	/*
832 	 * Loop over leaf entries with the right hash value.
833 	 */
834 	for (lep = &leafhdr.ents[index];
835 	     index < leafhdr.count && be32_to_cpu(lep->hashval) == args->hashval;
836 	     lep++, index++) {
837 		/*
838 		 * Skip stale leaf entries.
839 		 */
840 		if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
841 			continue;
842 		/*
843 		 * Pull the data block number from the entry.
844 		 */
845 		newdb = xfs_dir2_dataptr_to_db(args->geo,
846 					       be32_to_cpu(lep->address));
847 		/*
848 		 * Not adding a new entry, so we really want to find
849 		 * the name given to us.
850 		 *
851 		 * If it's a different data block, go get it.
852 		 */
853 		if (newdb != curdb) {
854 			/*
855 			 * If we had a block before that we aren't saving
856 			 * for a CI name, drop it
857 			 */
858 			if (curbp && (args->cmpresult == XFS_CMP_DIFFERENT ||
859 						curdb != state->extrablk.blkno))
860 				xfs_trans_brelse(tp, curbp);
861 			/*
862 			 * If needing the block that is saved with a CI match,
863 			 * use it otherwise read in the new data block.
864 			 */
865 			if (args->cmpresult != XFS_CMP_DIFFERENT &&
866 					newdb == state->extrablk.blkno) {
867 				ASSERT(state->extravalid);
868 				curbp = state->extrablk.bp;
869 			} else {
870 				error = xfs_dir3_data_read(tp, dp, args->owner,
871 						xfs_dir2_db_to_da(args->geo,
872 								  newdb),
873 						0, &curbp);
874 				if (error)
875 					return error;
876 			}
877 			xfs_dir3_data_check(dp, curbp);
878 			curdb = newdb;
879 		}
880 		/*
881 		 * Point to the data entry.
882 		 */
883 		dep = (xfs_dir2_data_entry_t *)((char *)curbp->b_addr +
884 			xfs_dir2_dataptr_to_off(args->geo,
885 						be32_to_cpu(lep->address)));
886 		/*
887 		 * Compare the entry and if it's an exact match, return
888 		 * EEXIST immediately. If it's the first case-insensitive
889 		 * match, store the block & inode number and continue looking.
890 		 */
891 		cmp = xfs_dir2_compname(args, dep->name, dep->namelen);
892 		if (cmp != XFS_CMP_DIFFERENT && cmp != args->cmpresult) {
893 			/* If there is a CI match block, drop it */
894 			if (args->cmpresult != XFS_CMP_DIFFERENT &&
895 						curdb != state->extrablk.blkno)
896 				xfs_trans_brelse(tp, state->extrablk.bp);
897 			args->cmpresult = cmp;
898 			args->inumber = be64_to_cpu(dep->inumber);
899 			args->filetype = xfs_dir2_data_get_ftype(mp, dep);
900 			*indexp = index;
901 			state->extravalid = 1;
902 			state->extrablk.bp = curbp;
903 			state->extrablk.blkno = curdb;
904 			state->extrablk.index = (int)((char *)dep -
905 							(char *)curbp->b_addr);
906 			state->extrablk.magic = XFS_DIR2_DATA_MAGIC;
907 			curbp->b_ops = &xfs_dir3_data_buf_ops;
908 			xfs_trans_buf_set_type(tp, curbp, XFS_BLFT_DIR_DATA_BUF);
909 			if (cmp == XFS_CMP_EXACT)
910 				return -EEXIST;
911 		}
912 	}
913 	ASSERT(index == leafhdr.count || (args->op_flags & XFS_DA_OP_OKNOENT));
914 	if (curbp) {
915 		if (args->cmpresult == XFS_CMP_DIFFERENT) {
916 			/* Giving back last used data block. */
917 			state->extravalid = 1;
918 			state->extrablk.bp = curbp;
919 			state->extrablk.index = -1;
920 			state->extrablk.blkno = curdb;
921 			state->extrablk.magic = XFS_DIR2_DATA_MAGIC;
922 			curbp->b_ops = &xfs_dir3_data_buf_ops;
923 			xfs_trans_buf_set_type(tp, curbp, XFS_BLFT_DIR_DATA_BUF);
924 		} else {
925 			/* If the curbp is not the CI match block, drop it */
926 			if (state->extrablk.bp != curbp)
927 				xfs_trans_brelse(tp, curbp);
928 		}
929 	} else {
930 		state->extravalid = 0;
931 	}
932 	*indexp = index;
933 	return -ENOENT;
934 }
935 
936 /*
937  * Look up a leaf entry in a node-format leaf block.
938  * If this is an addname then the extrablk in state is a freespace block,
939  * otherwise it's a data block.
940  */
941 int
942 xfs_dir2_leafn_lookup_int(
943 	struct xfs_buf		*bp,		/* leaf buffer */
944 	xfs_da_args_t		*args,		/* operation arguments */
945 	int			*indexp,	/* out: leaf entry index */
946 	xfs_da_state_t		*state)		/* state to fill in */
947 {
948 	if (args->op_flags & XFS_DA_OP_ADDNAME)
949 		return xfs_dir2_leafn_lookup_for_addname(bp, args, indexp,
950 							state);
951 	return xfs_dir2_leafn_lookup_for_entry(bp, args, indexp, state);
952 }
953 
954 /*
955  * Move count leaf entries from source to destination leaf.
956  * Log entries and headers.  Stale entries are preserved.
957  */
958 static void
959 xfs_dir3_leafn_moveents(
960 	xfs_da_args_t			*args,	/* operation arguments */
961 	struct xfs_buf			*bp_s,	/* source */
962 	struct xfs_dir3_icleaf_hdr	*shdr,
963 	struct xfs_dir2_leaf_entry	*sents,
964 	int				start_s,/* source leaf index */
965 	struct xfs_buf			*bp_d,	/* destination */
966 	struct xfs_dir3_icleaf_hdr	*dhdr,
967 	struct xfs_dir2_leaf_entry	*dents,
968 	int				start_d,/* destination leaf index */
969 	int				count)	/* count of leaves to copy */
970 {
971 	int				stale;	/* count stale leaves copied */
972 
973 	trace_xfs_dir2_leafn_moveents(args, start_s, start_d, count);
974 
975 	/*
976 	 * Silently return if nothing to do.
977 	 */
978 	if (count == 0)
979 		return;
980 
981 	/*
982 	 * If the destination index is not the end of the current
983 	 * destination leaf entries, open up a hole in the destination
984 	 * to hold the new entries.
985 	 */
986 	if (start_d < dhdr->count) {
987 		memmove(&dents[start_d + count], &dents[start_d],
988 			(dhdr->count - start_d) * sizeof(xfs_dir2_leaf_entry_t));
989 		xfs_dir3_leaf_log_ents(args, dhdr, bp_d, start_d + count,
990 				       count + dhdr->count - 1);
991 	}
992 	/*
993 	 * If the source has stale leaves, count the ones in the copy range
994 	 * so we can update the header correctly.
995 	 */
996 	if (shdr->stale) {
997 		int	i;			/* temp leaf index */
998 
999 		for (i = start_s, stale = 0; i < start_s + count; i++) {
1000 			if (sents[i].address ==
1001 					cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
1002 				stale++;
1003 		}
1004 	} else
1005 		stale = 0;
1006 	/*
1007 	 * Copy the leaf entries from source to destination.
1008 	 */
1009 	memcpy(&dents[start_d], &sents[start_s],
1010 		count * sizeof(xfs_dir2_leaf_entry_t));
1011 	xfs_dir3_leaf_log_ents(args, dhdr, bp_d, start_d, start_d + count - 1);
1012 
1013 	/*
1014 	 * If there are source entries after the ones we copied,
1015 	 * delete the ones we copied by sliding the next ones down.
1016 	 */
1017 	if (start_s + count < shdr->count) {
1018 		memmove(&sents[start_s], &sents[start_s + count],
1019 			count * sizeof(xfs_dir2_leaf_entry_t));
1020 		xfs_dir3_leaf_log_ents(args, shdr, bp_s, start_s,
1021 				       start_s + count - 1);
1022 	}
1023 
1024 	/*
1025 	 * Update the headers and log them.
1026 	 */
1027 	shdr->count -= count;
1028 	shdr->stale -= stale;
1029 	dhdr->count += count;
1030 	dhdr->stale += stale;
1031 }
1032 
1033 /*
1034  * Determine the sort order of two leaf blocks.
1035  * Returns 1 if both are valid and leaf2 should be before leaf1, else 0.
1036  */
1037 int						/* sort order */
1038 xfs_dir2_leafn_order(
1039 	struct xfs_inode	*dp,
1040 	struct xfs_buf		*leaf1_bp,		/* leaf1 buffer */
1041 	struct xfs_buf		*leaf2_bp)		/* leaf2 buffer */
1042 {
1043 	struct xfs_dir2_leaf	*leaf1 = leaf1_bp->b_addr;
1044 	struct xfs_dir2_leaf	*leaf2 = leaf2_bp->b_addr;
1045 	struct xfs_dir2_leaf_entry *ents1;
1046 	struct xfs_dir2_leaf_entry *ents2;
1047 	struct xfs_dir3_icleaf_hdr hdr1;
1048 	struct xfs_dir3_icleaf_hdr hdr2;
1049 
1050 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &hdr1, leaf1);
1051 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &hdr2, leaf2);
1052 	ents1 = hdr1.ents;
1053 	ents2 = hdr2.ents;
1054 
1055 	if (hdr1.count > 0 && hdr2.count > 0 &&
1056 	    (be32_to_cpu(ents2[0].hashval) < be32_to_cpu(ents1[0].hashval) ||
1057 	     be32_to_cpu(ents2[hdr2.count - 1].hashval) <
1058 				be32_to_cpu(ents1[hdr1.count - 1].hashval)))
1059 		return 1;
1060 	return 0;
1061 }
1062 
1063 /*
1064  * Rebalance leaf entries between two leaf blocks.
1065  * This is actually only called when the second block is new,
1066  * though the code deals with the general case.
1067  * A new entry will be inserted in one of the blocks, and that
1068  * entry is taken into account when balancing.
1069  */
1070 static void
1071 xfs_dir2_leafn_rebalance(
1072 	xfs_da_state_t		*state,		/* btree cursor */
1073 	xfs_da_state_blk_t	*blk1,		/* first btree block */
1074 	xfs_da_state_blk_t	*blk2)		/* second btree block */
1075 {
1076 	xfs_da_args_t		*args;		/* operation arguments */
1077 	int			count;		/* count (& direction) leaves */
1078 	int			isleft;		/* new goes in left leaf */
1079 	xfs_dir2_leaf_t		*leaf1;		/* first leaf structure */
1080 	xfs_dir2_leaf_t		*leaf2;		/* second leaf structure */
1081 	int			mid;		/* midpoint leaf index */
1082 #if defined(DEBUG) || defined(XFS_WARN)
1083 	int			oldstale;	/* old count of stale leaves */
1084 #endif
1085 	int			oldsum;		/* old total leaf count */
1086 	int			swap_blocks;	/* swapped leaf blocks */
1087 	struct xfs_dir2_leaf_entry *ents1;
1088 	struct xfs_dir2_leaf_entry *ents2;
1089 	struct xfs_dir3_icleaf_hdr hdr1;
1090 	struct xfs_dir3_icleaf_hdr hdr2;
1091 	struct xfs_inode	*dp = state->args->dp;
1092 
1093 	args = state->args;
1094 	/*
1095 	 * If the block order is wrong, swap the arguments.
1096 	 */
1097 	swap_blocks = xfs_dir2_leafn_order(dp, blk1->bp, blk2->bp);
1098 	if (swap_blocks)
1099 		swap(blk1, blk2);
1100 
1101 	leaf1 = blk1->bp->b_addr;
1102 	leaf2 = blk2->bp->b_addr;
1103 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &hdr1, leaf1);
1104 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &hdr2, leaf2);
1105 	ents1 = hdr1.ents;
1106 	ents2 = hdr2.ents;
1107 
1108 	oldsum = hdr1.count + hdr2.count;
1109 #if defined(DEBUG) || defined(XFS_WARN)
1110 	oldstale = hdr1.stale + hdr2.stale;
1111 #endif
1112 	mid = oldsum >> 1;
1113 
1114 	/*
1115 	 * If the old leaf count was odd then the new one will be even,
1116 	 * so we need to divide the new count evenly.
1117 	 */
1118 	if (oldsum & 1) {
1119 		xfs_dahash_t	midhash;	/* middle entry hash value */
1120 
1121 		if (mid >= hdr1.count)
1122 			midhash = be32_to_cpu(ents2[mid - hdr1.count].hashval);
1123 		else
1124 			midhash = be32_to_cpu(ents1[mid].hashval);
1125 		isleft = args->hashval <= midhash;
1126 	}
1127 	/*
1128 	 * If the old count is even then the new count is odd, so there's
1129 	 * no preferred side for the new entry.
1130 	 * Pick the left one.
1131 	 */
1132 	else
1133 		isleft = 1;
1134 	/*
1135 	 * Calculate moved entry count.  Positive means left-to-right,
1136 	 * negative means right-to-left.  Then move the entries.
1137 	 */
1138 	count = hdr1.count - mid + (isleft == 0);
1139 	if (count > 0)
1140 		xfs_dir3_leafn_moveents(args, blk1->bp, &hdr1, ents1,
1141 					hdr1.count - count, blk2->bp,
1142 					&hdr2, ents2, 0, count);
1143 	else if (count < 0)
1144 		xfs_dir3_leafn_moveents(args, blk2->bp, &hdr2, ents2, 0,
1145 					blk1->bp, &hdr1, ents1,
1146 					hdr1.count, count);
1147 
1148 	ASSERT(hdr1.count + hdr2.count == oldsum);
1149 	ASSERT(hdr1.stale + hdr2.stale == oldstale);
1150 
1151 	/* log the changes made when moving the entries */
1152 	xfs_dir2_leaf_hdr_to_disk(dp->i_mount, leaf1, &hdr1);
1153 	xfs_dir2_leaf_hdr_to_disk(dp->i_mount, leaf2, &hdr2);
1154 	xfs_dir3_leaf_log_header(args, blk1->bp);
1155 	xfs_dir3_leaf_log_header(args, blk2->bp);
1156 
1157 	xfs_dir3_leaf_check(dp, blk1->bp);
1158 	xfs_dir3_leaf_check(dp, blk2->bp);
1159 
1160 	/*
1161 	 * Mark whether we're inserting into the old or new leaf.
1162 	 */
1163 	if (hdr1.count < hdr2.count)
1164 		state->inleaf = swap_blocks;
1165 	else if (hdr1.count > hdr2.count)
1166 		state->inleaf = !swap_blocks;
1167 	else
1168 		state->inleaf = swap_blocks ^ (blk1->index <= hdr1.count);
1169 	/*
1170 	 * Adjust the expected index for insertion.
1171 	 */
1172 	if (!state->inleaf)
1173 		blk2->index = blk1->index - hdr1.count;
1174 
1175 	/*
1176 	 * Finally sanity check just to make sure we are not returning a
1177 	 * negative index
1178 	 */
1179 	if (blk2->index < 0) {
1180 		state->inleaf = 1;
1181 		blk2->index = 0;
1182 		xfs_alert(dp->i_mount,
1183 	"%s: picked the wrong leaf? reverting original leaf: blk1->index %d",
1184 			__func__, blk1->index);
1185 	}
1186 }
1187 
1188 static int
1189 xfs_dir3_data_block_free(
1190 	xfs_da_args_t		*args,
1191 	struct xfs_dir2_data_hdr *hdr,
1192 	struct xfs_dir2_free	*free,
1193 	xfs_dir2_db_t		fdb,
1194 	int			findex,
1195 	struct xfs_buf		*fbp,
1196 	int			longest)
1197 {
1198 	int			logfree = 0;
1199 	struct xfs_dir3_icfree_hdr freehdr;
1200 	struct xfs_inode	*dp = args->dp;
1201 
1202 	xfs_dir2_free_hdr_from_disk(dp->i_mount, &freehdr, free);
1203 	if (hdr) {
1204 		/*
1205 		 * Data block is not empty, just set the free entry to the new
1206 		 * value.
1207 		 */
1208 		freehdr.bests[findex] = cpu_to_be16(longest);
1209 		xfs_dir2_free_log_bests(args, &freehdr, fbp, findex, findex);
1210 		return 0;
1211 	}
1212 
1213 	/* One less used entry in the free table. */
1214 	freehdr.nused--;
1215 
1216 	/*
1217 	 * If this was the last entry in the table, we can trim the table size
1218 	 * back.  There might be other entries at the end referring to
1219 	 * non-existent data blocks, get those too.
1220 	 */
1221 	if (findex == freehdr.nvalid - 1) {
1222 		int	i;		/* free entry index */
1223 
1224 		for (i = findex - 1; i >= 0; i--) {
1225 			if (freehdr.bests[i] != cpu_to_be16(NULLDATAOFF))
1226 				break;
1227 		}
1228 		freehdr.nvalid = i + 1;
1229 		logfree = 0;
1230 	} else {
1231 		/* Not the last entry, just punch it out.  */
1232 		freehdr.bests[findex] = cpu_to_be16(NULLDATAOFF);
1233 		logfree = 1;
1234 	}
1235 
1236 	xfs_dir2_free_hdr_to_disk(dp->i_mount, free, &freehdr);
1237 	xfs_dir2_free_log_header(args, fbp);
1238 
1239 	/*
1240 	 * If there are no useful entries left in the block, get rid of the
1241 	 * block if we can.
1242 	 */
1243 	if (!freehdr.nused) {
1244 		int error;
1245 
1246 		error = xfs_dir2_shrink_inode(args, fdb, fbp);
1247 		if (error == 0) {
1248 			fbp = NULL;
1249 			logfree = 0;
1250 		} else if (error != -ENOSPC || args->total != 0)
1251 			return error;
1252 		/*
1253 		 * It's possible to get ENOSPC if there is no
1254 		 * space reservation.  In this case some one
1255 		 * else will eventually get rid of this block.
1256 		 */
1257 	}
1258 
1259 	/* Log the free entry that changed, unless we got rid of it.  */
1260 	if (logfree)
1261 		xfs_dir2_free_log_bests(args, &freehdr, fbp, findex, findex);
1262 	return 0;
1263 }
1264 
1265 /*
1266  * Remove an entry from a node directory.
1267  * This removes the leaf entry and the data entry,
1268  * and updates the free block if necessary.
1269  */
1270 static int					/* error */
1271 xfs_dir2_leafn_remove(
1272 	xfs_da_args_t		*args,		/* operation arguments */
1273 	struct xfs_buf		*bp,		/* leaf buffer */
1274 	int			index,		/* leaf entry index */
1275 	xfs_da_state_blk_t	*dblk,		/* data block */
1276 	int			*rval)		/* resulting block needs join */
1277 {
1278 	struct xfs_da_geometry	*geo = args->geo;
1279 	xfs_dir2_data_hdr_t	*hdr;		/* data block header */
1280 	xfs_dir2_db_t		db;		/* data block number */
1281 	struct xfs_buf		*dbp;		/* data block buffer */
1282 	xfs_dir2_data_entry_t	*dep;		/* data block entry */
1283 	xfs_inode_t		*dp;		/* incore directory inode */
1284 	xfs_dir2_leaf_t		*leaf;		/* leaf structure */
1285 	xfs_dir2_leaf_entry_t	*lep;		/* leaf entry */
1286 	int			longest;	/* longest data free entry */
1287 	int			off;		/* data block entry offset */
1288 	int			needlog;	/* need to log data header */
1289 	int			needscan;	/* need to rescan data frees */
1290 	xfs_trans_t		*tp;		/* transaction pointer */
1291 	struct xfs_dir2_data_free *bf;		/* bestfree table */
1292 	struct xfs_dir3_icleaf_hdr leafhdr;
1293 
1294 	trace_xfs_dir2_leafn_remove(args, index);
1295 
1296 	dp = args->dp;
1297 	tp = args->trans;
1298 	leaf = bp->b_addr;
1299 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &leafhdr, leaf);
1300 
1301 	/*
1302 	 * Point to the entry we're removing.
1303 	 */
1304 	lep = &leafhdr.ents[index];
1305 
1306 	/*
1307 	 * Extract the data block and offset from the entry.
1308 	 */
1309 	db = xfs_dir2_dataptr_to_db(geo, be32_to_cpu(lep->address));
1310 	ASSERT(dblk->blkno == db);
1311 	off = xfs_dir2_dataptr_to_off(geo, be32_to_cpu(lep->address));
1312 	ASSERT(dblk->index == off);
1313 
1314 	/*
1315 	 * Kill the leaf entry by marking it stale.
1316 	 * Log the leaf block changes.
1317 	 */
1318 	leafhdr.stale++;
1319 	xfs_dir2_leaf_hdr_to_disk(dp->i_mount, leaf, &leafhdr);
1320 	xfs_dir3_leaf_log_header(args, bp);
1321 
1322 	lep->address = cpu_to_be32(XFS_DIR2_NULL_DATAPTR);
1323 	xfs_dir3_leaf_log_ents(args, &leafhdr, bp, index, index);
1324 
1325 	/*
1326 	 * Make the data entry free.  Keep track of the longest freespace
1327 	 * in the data block in case it changes.
1328 	 */
1329 	dbp = dblk->bp;
1330 	hdr = dbp->b_addr;
1331 	dep = (xfs_dir2_data_entry_t *)((char *)hdr + off);
1332 	bf = xfs_dir2_data_bestfree_p(dp->i_mount, hdr);
1333 	longest = be16_to_cpu(bf[0].length);
1334 	needlog = needscan = 0;
1335 	xfs_dir2_data_make_free(args, dbp, off,
1336 		xfs_dir2_data_entsize(dp->i_mount, dep->namelen), &needlog,
1337 		&needscan);
1338 	/*
1339 	 * Rescan the data block freespaces for bestfree.
1340 	 * Log the data block header if needed.
1341 	 */
1342 	if (needscan)
1343 		xfs_dir2_data_freescan(dp->i_mount, hdr, &needlog);
1344 	if (needlog)
1345 		xfs_dir2_data_log_header(args, dbp);
1346 	xfs_dir3_data_check(dp, dbp);
1347 	/*
1348 	 * If the longest data block freespace changes, need to update
1349 	 * the corresponding freeblock entry.
1350 	 */
1351 	if (longest < be16_to_cpu(bf[0].length)) {
1352 		int		error;		/* error return value */
1353 		struct xfs_buf	*fbp;		/* freeblock buffer */
1354 		xfs_dir2_db_t	fdb;		/* freeblock block number */
1355 		int		findex;		/* index in freeblock entries */
1356 		xfs_dir2_free_t	*free;		/* freeblock structure */
1357 
1358 		/*
1359 		 * Convert the data block number to a free block,
1360 		 * read in the free block.
1361 		 */
1362 		fdb = xfs_dir2_db_to_fdb(geo, db);
1363 		error = xfs_dir2_free_read(tp, dp, args->owner,
1364 				xfs_dir2_db_to_da(geo, fdb), &fbp);
1365 		if (error)
1366 			return error;
1367 		free = fbp->b_addr;
1368 #ifdef DEBUG
1369 	{
1370 		struct xfs_dir3_icfree_hdr freehdr;
1371 
1372 		xfs_dir2_free_hdr_from_disk(dp->i_mount, &freehdr, free);
1373 		ASSERT(freehdr.firstdb == geo->free_max_bests *
1374 			(fdb - xfs_dir2_byte_to_db(geo, XFS_DIR2_FREE_OFFSET)));
1375 	}
1376 #endif
1377 		/*
1378 		 * Calculate which entry we need to fix.
1379 		 */
1380 		findex = xfs_dir2_db_to_fdindex(geo, db);
1381 		longest = be16_to_cpu(bf[0].length);
1382 		/*
1383 		 * If the data block is now empty we can get rid of it
1384 		 * (usually).
1385 		 */
1386 		if (longest == geo->blksize - geo->data_entry_offset) {
1387 			/*
1388 			 * Try to punch out the data block.
1389 			 */
1390 			error = xfs_dir2_shrink_inode(args, db, dbp);
1391 			if (error == 0) {
1392 				dblk->bp = NULL;
1393 				hdr = NULL;
1394 			}
1395 			/*
1396 			 * We can get ENOSPC if there's no space reservation.
1397 			 * In this case just drop the buffer and some one else
1398 			 * will eventually get rid of the empty block.
1399 			 */
1400 			else if (!(error == -ENOSPC && args->total == 0))
1401 				return error;
1402 		}
1403 		/*
1404 		 * If we got rid of the data block, we can eliminate that entry
1405 		 * in the free block.
1406 		 */
1407 		error = xfs_dir3_data_block_free(args, hdr, free,
1408 						 fdb, findex, fbp, longest);
1409 		if (error)
1410 			return error;
1411 	}
1412 
1413 	xfs_dir3_leaf_check(dp, bp);
1414 	/*
1415 	 * Return indication of whether this leaf block is empty enough
1416 	 * to justify trying to join it with a neighbor.
1417 	 */
1418 	*rval = (geo->leaf_hdr_size +
1419 		 (uint)sizeof(leafhdr.ents) * (leafhdr.count - leafhdr.stale)) <
1420 		geo->magicpct;
1421 	return 0;
1422 }
1423 
1424 /*
1425  * Split the leaf entries in the old block into old and new blocks.
1426  */
1427 int						/* error */
1428 xfs_dir2_leafn_split(
1429 	xfs_da_state_t		*state,		/* btree cursor */
1430 	xfs_da_state_blk_t	*oldblk,	/* original block */
1431 	xfs_da_state_blk_t	*newblk)	/* newly created block */
1432 {
1433 	xfs_da_args_t		*args;		/* operation arguments */
1434 	xfs_dablk_t		blkno;		/* new leaf block number */
1435 	int			error;		/* error return value */
1436 	struct xfs_inode	*dp;
1437 
1438 	/*
1439 	 * Allocate space for a new leaf node.
1440 	 */
1441 	args = state->args;
1442 	dp = args->dp;
1443 	ASSERT(oldblk->magic == XFS_DIR2_LEAFN_MAGIC);
1444 	error = xfs_da_grow_inode(args, &blkno);
1445 	if (error) {
1446 		return error;
1447 	}
1448 	/*
1449 	 * Initialize the new leaf block.
1450 	 */
1451 	error = xfs_dir3_leaf_get_buf(args, xfs_dir2_da_to_db(args->geo, blkno),
1452 				      &newblk->bp, XFS_DIR2_LEAFN_MAGIC);
1453 	if (error)
1454 		return error;
1455 
1456 	newblk->blkno = blkno;
1457 	newblk->magic = XFS_DIR2_LEAFN_MAGIC;
1458 	/*
1459 	 * Rebalance the entries across the two leaves, link the new
1460 	 * block into the leaves.
1461 	 */
1462 	xfs_dir2_leafn_rebalance(state, oldblk, newblk);
1463 	error = xfs_da3_blk_link(state, oldblk, newblk);
1464 	if (error) {
1465 		return error;
1466 	}
1467 	/*
1468 	 * Insert the new entry in the correct block.
1469 	 */
1470 	if (state->inleaf)
1471 		error = xfs_dir2_leafn_add(oldblk->bp, args, oldblk->index);
1472 	else
1473 		error = xfs_dir2_leafn_add(newblk->bp, args, newblk->index);
1474 	/*
1475 	 * Update last hashval in each block since we added the name.
1476 	 */
1477 	oldblk->hashval = xfs_dir2_leaf_lasthash(dp, oldblk->bp, NULL);
1478 	newblk->hashval = xfs_dir2_leaf_lasthash(dp, newblk->bp, NULL);
1479 	xfs_dir3_leaf_check(dp, oldblk->bp);
1480 	xfs_dir3_leaf_check(dp, newblk->bp);
1481 	return error;
1482 }
1483 
1484 /*
1485  * Check a leaf block and its neighbors to see if the block should be
1486  * collapsed into one or the other neighbor.  Always keep the block
1487  * with the smaller block number.
1488  * If the current block is over 50% full, don't try to join it, return 0.
1489  * If the block is empty, fill in the state structure and return 2.
1490  * If it can be collapsed, fill in the state structure and return 1.
1491  * If nothing can be done, return 0.
1492  */
1493 int						/* error */
1494 xfs_dir2_leafn_toosmall(
1495 	xfs_da_state_t		*state,		/* btree cursor */
1496 	int			*action)	/* resulting action to take */
1497 {
1498 	xfs_da_state_blk_t	*blk;		/* leaf block */
1499 	xfs_dablk_t		blkno;		/* leaf block number */
1500 	struct xfs_buf		*bp;		/* leaf buffer */
1501 	int			bytes;		/* bytes in use */
1502 	int			count;		/* leaf live entry count */
1503 	int			error;		/* error return value */
1504 	int			forward;	/* sibling block direction */
1505 	int			i;		/* sibling counter */
1506 	xfs_dir2_leaf_t		*leaf;		/* leaf structure */
1507 	int			rval;		/* result from path_shift */
1508 	struct xfs_dir3_icleaf_hdr leafhdr;
1509 	struct xfs_dir2_leaf_entry *ents;
1510 	struct xfs_inode	*dp = state->args->dp;
1511 
1512 	/*
1513 	 * Check for the degenerate case of the block being over 50% full.
1514 	 * If so, it's not worth even looking to see if we might be able
1515 	 * to coalesce with a sibling.
1516 	 */
1517 	blk = &state->path.blk[state->path.active - 1];
1518 	leaf = blk->bp->b_addr;
1519 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &leafhdr, leaf);
1520 	ents = leafhdr.ents;
1521 	xfs_dir3_leaf_check(dp, blk->bp);
1522 
1523 	count = leafhdr.count - leafhdr.stale;
1524 	bytes = state->args->geo->leaf_hdr_size + count * sizeof(ents[0]);
1525 	if (bytes > (state->args->geo->blksize >> 1)) {
1526 		/*
1527 		 * Blk over 50%, don't try to join.
1528 		 */
1529 		*action = 0;
1530 		return 0;
1531 	}
1532 	/*
1533 	 * Check for the degenerate case of the block being empty.
1534 	 * If the block is empty, we'll simply delete it, no need to
1535 	 * coalesce it with a sibling block.  We choose (arbitrarily)
1536 	 * to merge with the forward block unless it is NULL.
1537 	 */
1538 	if (count == 0) {
1539 		/*
1540 		 * Make altpath point to the block we want to keep and
1541 		 * path point to the block we want to drop (this one).
1542 		 */
1543 		forward = (leafhdr.forw != 0);
1544 		memcpy(&state->altpath, &state->path, sizeof(state->path));
1545 		error = xfs_da3_path_shift(state, &state->altpath, forward, 0,
1546 			&rval);
1547 		if (error)
1548 			return error;
1549 		*action = rval ? 2 : 0;
1550 		return 0;
1551 	}
1552 	/*
1553 	 * Examine each sibling block to see if we can coalesce with
1554 	 * at least 25% free space to spare.  We need to figure out
1555 	 * whether to merge with the forward or the backward block.
1556 	 * We prefer coalescing with the lower numbered sibling so as
1557 	 * to shrink a directory over time.
1558 	 */
1559 	forward = leafhdr.forw < leafhdr.back;
1560 	for (i = 0, bp = NULL; i < 2; forward = !forward, i++) {
1561 		struct xfs_dir3_icleaf_hdr hdr2;
1562 
1563 		blkno = forward ? leafhdr.forw : leafhdr.back;
1564 		if (blkno == 0)
1565 			continue;
1566 		/*
1567 		 * Read the sibling leaf block.
1568 		 */
1569 		error = xfs_dir3_leafn_read(state->args->trans, dp,
1570 				state->args->owner, blkno, &bp);
1571 		if (error)
1572 			return error;
1573 
1574 		/*
1575 		 * Count bytes in the two blocks combined.
1576 		 */
1577 		count = leafhdr.count - leafhdr.stale;
1578 		bytes = state->args->geo->blksize -
1579 			(state->args->geo->blksize >> 2);
1580 
1581 		leaf = bp->b_addr;
1582 		xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &hdr2, leaf);
1583 		ents = hdr2.ents;
1584 		count += hdr2.count - hdr2.stale;
1585 		bytes -= count * sizeof(ents[0]);
1586 
1587 		/*
1588 		 * Fits with at least 25% to spare.
1589 		 */
1590 		if (bytes >= 0)
1591 			break;
1592 		xfs_trans_brelse(state->args->trans, bp);
1593 	}
1594 	/*
1595 	 * Didn't like either block, give up.
1596 	 */
1597 	if (i >= 2) {
1598 		*action = 0;
1599 		return 0;
1600 	}
1601 
1602 	/*
1603 	 * Make altpath point to the block we want to keep (the lower
1604 	 * numbered block) and path point to the block we want to drop.
1605 	 */
1606 	memcpy(&state->altpath, &state->path, sizeof(state->path));
1607 	if (blkno < blk->blkno)
1608 		error = xfs_da3_path_shift(state, &state->altpath, forward, 0,
1609 			&rval);
1610 	else
1611 		error = xfs_da3_path_shift(state, &state->path, forward, 0,
1612 			&rval);
1613 	if (error) {
1614 		return error;
1615 	}
1616 	*action = rval ? 0 : 1;
1617 	return 0;
1618 }
1619 
1620 /*
1621  * Move all the leaf entries from drop_blk to save_blk.
1622  * This is done as part of a join operation.
1623  */
1624 void
1625 xfs_dir2_leafn_unbalance(
1626 	xfs_da_state_t		*state,		/* cursor */
1627 	xfs_da_state_blk_t	*drop_blk,	/* dead block */
1628 	xfs_da_state_blk_t	*save_blk)	/* surviving block */
1629 {
1630 	xfs_da_args_t		*args;		/* operation arguments */
1631 	xfs_dir2_leaf_t		*drop_leaf;	/* dead leaf structure */
1632 	xfs_dir2_leaf_t		*save_leaf;	/* surviving leaf structure */
1633 	struct xfs_dir3_icleaf_hdr savehdr;
1634 	struct xfs_dir3_icleaf_hdr drophdr;
1635 	struct xfs_dir2_leaf_entry *sents;
1636 	struct xfs_dir2_leaf_entry *dents;
1637 	struct xfs_inode	*dp = state->args->dp;
1638 
1639 	args = state->args;
1640 	ASSERT(drop_blk->magic == XFS_DIR2_LEAFN_MAGIC);
1641 	ASSERT(save_blk->magic == XFS_DIR2_LEAFN_MAGIC);
1642 	drop_leaf = drop_blk->bp->b_addr;
1643 	save_leaf = save_blk->bp->b_addr;
1644 
1645 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &savehdr, save_leaf);
1646 	xfs_dir2_leaf_hdr_from_disk(dp->i_mount, &drophdr, drop_leaf);
1647 	sents = savehdr.ents;
1648 	dents = drophdr.ents;
1649 
1650 	/*
1651 	 * If there are any stale leaf entries, take this opportunity
1652 	 * to purge them.
1653 	 */
1654 	if (drophdr.stale)
1655 		xfs_dir3_leaf_compact(args, &drophdr, drop_blk->bp);
1656 	if (savehdr.stale)
1657 		xfs_dir3_leaf_compact(args, &savehdr, save_blk->bp);
1658 
1659 	/*
1660 	 * Move the entries from drop to the appropriate end of save.
1661 	 */
1662 	drop_blk->hashval = be32_to_cpu(dents[drophdr.count - 1].hashval);
1663 	if (xfs_dir2_leafn_order(dp, save_blk->bp, drop_blk->bp))
1664 		xfs_dir3_leafn_moveents(args, drop_blk->bp, &drophdr, dents, 0,
1665 					save_blk->bp, &savehdr, sents, 0,
1666 					drophdr.count);
1667 	else
1668 		xfs_dir3_leafn_moveents(args, drop_blk->bp, &drophdr, dents, 0,
1669 					save_blk->bp, &savehdr, sents,
1670 					savehdr.count, drophdr.count);
1671 	save_blk->hashval = be32_to_cpu(sents[savehdr.count - 1].hashval);
1672 
1673 	/* log the changes made when moving the entries */
1674 	xfs_dir2_leaf_hdr_to_disk(dp->i_mount, save_leaf, &savehdr);
1675 	xfs_dir2_leaf_hdr_to_disk(dp->i_mount, drop_leaf, &drophdr);
1676 	xfs_dir3_leaf_log_header(args, save_blk->bp);
1677 	xfs_dir3_leaf_log_header(args, drop_blk->bp);
1678 
1679 	xfs_dir3_leaf_check(dp, save_blk->bp);
1680 	xfs_dir3_leaf_check(dp, drop_blk->bp);
1681 }
1682 
1683 /*
1684  * Add a new data block to the directory at the free space index that the caller
1685  * has specified.
1686  */
1687 static int
1688 xfs_dir2_node_add_datablk(
1689 	struct xfs_da_args	*args,
1690 	struct xfs_da_state_blk	*fblk,
1691 	xfs_dir2_db_t		*dbno,
1692 	struct xfs_buf		**dbpp,
1693 	struct xfs_buf		**fbpp,
1694 	struct xfs_dir3_icfree_hdr *hdr,
1695 	int			*findex)
1696 {
1697 	struct xfs_inode	*dp = args->dp;
1698 	struct xfs_trans	*tp = args->trans;
1699 	struct xfs_mount	*mp = dp->i_mount;
1700 	struct xfs_dir2_data_free *bf;
1701 	xfs_dir2_db_t		fbno;
1702 	struct xfs_buf		*fbp;
1703 	struct xfs_buf		*dbp;
1704 	int			error;
1705 
1706 	/* Not allowed to allocate, return failure. */
1707 	if (args->total == 0)
1708 		return -ENOSPC;
1709 
1710 	/* Allocate and initialize the new data block.  */
1711 	error = xfs_dir2_grow_inode(args, XFS_DIR2_DATA_SPACE, dbno);
1712 	if (error)
1713 		return error;
1714 	error = xfs_dir3_data_init(args, *dbno, &dbp);
1715 	if (error)
1716 		return error;
1717 
1718 	/*
1719 	 * Get the freespace block corresponding to the data block
1720 	 * that was just allocated.
1721 	 */
1722 	fbno = xfs_dir2_db_to_fdb(args->geo, *dbno);
1723 	error = xfs_dir2_free_try_read(tp, dp, args->owner,
1724 			       xfs_dir2_db_to_da(args->geo, fbno), &fbp);
1725 	if (error)
1726 		return error;
1727 
1728 	/*
1729 	 * If there wasn't a freespace block, the read will
1730 	 * return a NULL fbp.  Allocate and initialize a new one.
1731 	 */
1732 	if (!fbp) {
1733 		error = xfs_dir2_grow_inode(args, XFS_DIR2_FREE_SPACE, &fbno);
1734 		if (error)
1735 			return error;
1736 
1737 		if (XFS_IS_CORRUPT(mp,
1738 				   xfs_dir2_db_to_fdb(args->geo, *dbno) !=
1739 				   fbno)) {
1740 			xfs_alert(mp,
1741 "%s: dir ino %llu needed freesp block %lld for data block %lld, got %lld",
1742 				__func__, (unsigned long long)dp->i_ino,
1743 				(long long)xfs_dir2_db_to_fdb(args->geo, *dbno),
1744 				(long long)*dbno, (long long)fbno);
1745 			if (fblk) {
1746 				xfs_alert(mp,
1747 			" fblk "PTR_FMT" blkno %llu index %d magic 0x%x",
1748 					fblk, (unsigned long long)fblk->blkno,
1749 					fblk->index, fblk->magic);
1750 			} else {
1751 				xfs_alert(mp, " ... fblk is NULL");
1752 			}
1753 			xfs_da_mark_sick(args);
1754 			return -EFSCORRUPTED;
1755 		}
1756 
1757 		/* Get a buffer for the new block. */
1758 		error = xfs_dir3_free_get_buf(args, fbno, &fbp);
1759 		if (error)
1760 			return error;
1761 		xfs_dir2_free_hdr_from_disk(mp, hdr, fbp->b_addr);
1762 
1763 		/* Remember the first slot as our empty slot. */
1764 		hdr->firstdb = (fbno - xfs_dir2_byte_to_db(args->geo,
1765 							XFS_DIR2_FREE_OFFSET)) *
1766 				args->geo->free_max_bests;
1767 	} else {
1768 		xfs_dir2_free_hdr_from_disk(mp, hdr, fbp->b_addr);
1769 	}
1770 
1771 	/* Set the freespace block index from the data block number. */
1772 	*findex = xfs_dir2_db_to_fdindex(args->geo, *dbno);
1773 
1774 	/* Extend the freespace table if the new data block is off the end. */
1775 	if (*findex >= hdr->nvalid) {
1776 		ASSERT(*findex < args->geo->free_max_bests);
1777 		hdr->nvalid = *findex + 1;
1778 		hdr->bests[*findex] = cpu_to_be16(NULLDATAOFF);
1779 	}
1780 
1781 	/*
1782 	 * If this entry was for an empty data block (this should always be
1783 	 * true) then update the header.
1784 	 */
1785 	if (hdr->bests[*findex] == cpu_to_be16(NULLDATAOFF)) {
1786 		hdr->nused++;
1787 		xfs_dir2_free_hdr_to_disk(mp, fbp->b_addr, hdr);
1788 		xfs_dir2_free_log_header(args, fbp);
1789 	}
1790 
1791 	/* Update the freespace value for the new block in the table. */
1792 	bf = xfs_dir2_data_bestfree_p(mp, dbp->b_addr);
1793 	hdr->bests[*findex] = bf[0].length;
1794 
1795 	*dbpp = dbp;
1796 	*fbpp = fbp;
1797 	return 0;
1798 }
1799 
1800 static int
1801 xfs_dir2_node_find_freeblk(
1802 	struct xfs_da_args	*args,
1803 	struct xfs_da_state_blk	*fblk,
1804 	xfs_dir2_db_t		*dbnop,
1805 	struct xfs_buf		**fbpp,
1806 	struct xfs_dir3_icfree_hdr *hdr,
1807 	int			*findexp,
1808 	int			length)
1809 {
1810 	struct xfs_inode	*dp = args->dp;
1811 	struct xfs_trans	*tp = args->trans;
1812 	struct xfs_buf		*fbp = NULL;
1813 	xfs_dir2_db_t		firstfbno;
1814 	xfs_dir2_db_t		lastfbno;
1815 	xfs_dir2_db_t		ifbno = -1;
1816 	xfs_dir2_db_t		dbno = -1;
1817 	xfs_dir2_db_t		fbno;
1818 	xfs_fileoff_t		fo;
1819 	int			findex = 0;
1820 	int			error;
1821 
1822 	/*
1823 	 * If we came in with a freespace block that means that lookup
1824 	 * found an entry with our hash value.  This is the freespace
1825 	 * block for that data entry.
1826 	 */
1827 	if (fblk) {
1828 		fbp = fblk->bp;
1829 		findex = fblk->index;
1830 		xfs_dir2_free_hdr_from_disk(dp->i_mount, hdr, fbp->b_addr);
1831 		if (findex >= 0) {
1832 			/* caller already found the freespace for us. */
1833 			ASSERT(findex < hdr->nvalid);
1834 			ASSERT(be16_to_cpu(hdr->bests[findex]) != NULLDATAOFF);
1835 			ASSERT(be16_to_cpu(hdr->bests[findex]) >= length);
1836 			dbno = hdr->firstdb + findex;
1837 			goto found_block;
1838 		}
1839 
1840 		/*
1841 		 * The data block looked at didn't have enough room.
1842 		 * We'll start at the beginning of the freespace entries.
1843 		 */
1844 		ifbno = fblk->blkno;
1845 		xfs_trans_brelse(tp, fbp);
1846 		fbp = NULL;
1847 		fblk->bp = NULL;
1848 	}
1849 
1850 	/*
1851 	 * If we don't have a data block yet, we're going to scan the freespace
1852 	 * data for a data block with enough free space in it.
1853 	 */
1854 	error = xfs_bmap_last_offset(dp, &fo, XFS_DATA_FORK);
1855 	if (error)
1856 		return error;
1857 	lastfbno = xfs_dir2_da_to_db(args->geo, (xfs_dablk_t)fo);
1858 	firstfbno = xfs_dir2_byte_to_db(args->geo, XFS_DIR2_FREE_OFFSET);
1859 
1860 	for (fbno = lastfbno - 1; fbno >= firstfbno; fbno--) {
1861 		/* If it's ifbno we already looked at it. */
1862 		if (fbno == ifbno)
1863 			continue;
1864 
1865 		/*
1866 		 * Read the block.  There can be holes in the freespace blocks,
1867 		 * so this might not succeed.  This should be really rare, so
1868 		 * there's no reason to avoid it.
1869 		 */
1870 		error = xfs_dir2_free_try_read(tp, dp, args->owner,
1871 				xfs_dir2_db_to_da(args->geo, fbno),
1872 				&fbp);
1873 		if (error)
1874 			return error;
1875 		if (!fbp)
1876 			continue;
1877 
1878 		xfs_dir2_free_hdr_from_disk(dp->i_mount, hdr, fbp->b_addr);
1879 
1880 		/* Scan the free entry array for a large enough free space. */
1881 		for (findex = hdr->nvalid - 1; findex >= 0; findex--) {
1882 			if (be16_to_cpu(hdr->bests[findex]) != NULLDATAOFF &&
1883 			    be16_to_cpu(hdr->bests[findex]) >= length) {
1884 				dbno = hdr->firstdb + findex;
1885 				goto found_block;
1886 			}
1887 		}
1888 
1889 		/* Didn't find free space, go on to next free block */
1890 		xfs_trans_brelse(tp, fbp);
1891 	}
1892 
1893 found_block:
1894 	*dbnop = dbno;
1895 	*fbpp = fbp;
1896 	*findexp = findex;
1897 	return 0;
1898 }
1899 
1900 /*
1901  * Add the data entry for a node-format directory name addition.
1902  * The leaf entry is added in xfs_dir2_leafn_add.
1903  * We may enter with a freespace block that the lookup found.
1904  */
1905 static int
1906 xfs_dir2_node_addname_int(
1907 	struct xfs_da_args	*args,		/* operation arguments */
1908 	struct xfs_da_state_blk	*fblk)		/* optional freespace block */
1909 {
1910 	struct xfs_dir2_data_unused *dup;	/* data unused entry pointer */
1911 	struct xfs_dir2_data_entry *dep;	/* data entry pointer */
1912 	struct xfs_dir2_data_hdr *hdr;		/* data block header */
1913 	struct xfs_dir2_data_free *bf;
1914 	struct xfs_trans	*tp = args->trans;
1915 	struct xfs_inode	*dp = args->dp;
1916 	struct xfs_dir3_icfree_hdr freehdr;
1917 	struct xfs_buf		*dbp;		/* data block buffer */
1918 	struct xfs_buf		*fbp;		/* freespace buffer */
1919 	xfs_dir2_data_aoff_t	aoff;
1920 	xfs_dir2_db_t		dbno;		/* data block number */
1921 	int			error;		/* error return value */
1922 	int			findex;		/* freespace entry index */
1923 	int			length;		/* length of the new entry */
1924 	int			logfree = 0;	/* need to log free entry */
1925 	int			needlog = 0;	/* need to log data header */
1926 	int			needscan = 0;	/* need to rescan data frees */
1927 	__be16			*tagp;		/* data entry tag pointer */
1928 
1929 	length = xfs_dir2_data_entsize(dp->i_mount, args->namelen);
1930 	error = xfs_dir2_node_find_freeblk(args, fblk, &dbno, &fbp, &freehdr,
1931 					   &findex, length);
1932 	if (error)
1933 		return error;
1934 
1935 	/*
1936 	 * Now we know if we must allocate blocks, so if we are checking whether
1937 	 * we can insert without allocation then we can return now.
1938 	 */
1939 	if (args->op_flags & XFS_DA_OP_JUSTCHECK) {
1940 		if (dbno == -1)
1941 			return -ENOSPC;
1942 		return 0;
1943 	}
1944 
1945 	/*
1946 	 * If we don't have a data block, we need to allocate one and make
1947 	 * the freespace entries refer to it.
1948 	 */
1949 	if (dbno == -1) {
1950 		/* we're going to have to log the free block index later */
1951 		logfree = 1;
1952 		error = xfs_dir2_node_add_datablk(args, fblk, &dbno, &dbp, &fbp,
1953 						  &freehdr, &findex);
1954 	} else {
1955 		/* Read the data block in. */
1956 		error = xfs_dir3_data_read(tp, dp, args->owner,
1957 				xfs_dir2_db_to_da(args->geo, dbno), 0, &dbp);
1958 	}
1959 	if (error)
1960 		return error;
1961 
1962 	/* setup for data block up now */
1963 	hdr = dbp->b_addr;
1964 	bf = xfs_dir2_data_bestfree_p(dp->i_mount, hdr);
1965 	ASSERT(be16_to_cpu(bf[0].length) >= length);
1966 
1967 	/* Point to the existing unused space. */
1968 	dup = (xfs_dir2_data_unused_t *)
1969 	      ((char *)hdr + be16_to_cpu(bf[0].offset));
1970 
1971 	/* Mark the first part of the unused space, inuse for us. */
1972 	aoff = (xfs_dir2_data_aoff_t)((char *)dup - (char *)hdr);
1973 	error = xfs_dir2_data_use_free(args, dbp, dup, aoff, length,
1974 			&needlog, &needscan);
1975 	if (error) {
1976 		xfs_trans_brelse(tp, dbp);
1977 		return error;
1978 	}
1979 
1980 	/* Fill in the new entry and log it. */
1981 	dep = (xfs_dir2_data_entry_t *)dup;
1982 	dep->inumber = cpu_to_be64(args->inumber);
1983 	dep->namelen = args->namelen;
1984 	memcpy(dep->name, args->name, dep->namelen);
1985 	xfs_dir2_data_put_ftype(dp->i_mount, dep, args->filetype);
1986 	tagp = xfs_dir2_data_entry_tag_p(dp->i_mount, dep);
1987 	*tagp = cpu_to_be16((char *)dep - (char *)hdr);
1988 	xfs_dir2_data_log_entry(args, dbp, dep);
1989 
1990 	/* Rescan the freespace and log the data block if needed. */
1991 	if (needscan)
1992 		xfs_dir2_data_freescan(dp->i_mount, hdr, &needlog);
1993 	if (needlog)
1994 		xfs_dir2_data_log_header(args, dbp);
1995 
1996 	/* If the freespace block entry is now wrong, update it. */
1997 	if (freehdr.bests[findex] != bf[0].length) {
1998 		freehdr.bests[findex] = bf[0].length;
1999 		logfree = 1;
2000 	}
2001 
2002 	/* Log the freespace entry if needed. */
2003 	if (logfree)
2004 		xfs_dir2_free_log_bests(args, &freehdr, fbp, findex, findex);
2005 
2006 	/* Return the data block and offset in args. */
2007 	args->blkno = (xfs_dablk_t)dbno;
2008 	args->index = be16_to_cpu(*tagp);
2009 	return 0;
2010 }
2011 
2012 /*
2013  * Top-level node form directory addname routine.
2014  */
2015 int						/* error */
2016 xfs_dir2_node_addname(
2017 	xfs_da_args_t		*args)		/* operation arguments */
2018 {
2019 	xfs_da_state_blk_t	*blk;		/* leaf block for insert */
2020 	int			error;		/* error return value */
2021 	int			rval;		/* sub-return value */
2022 	xfs_da_state_t		*state;		/* btree cursor */
2023 
2024 	trace_xfs_dir2_node_addname(args);
2025 
2026 	/*
2027 	 * Allocate and initialize the state (btree cursor).
2028 	 */
2029 	state = xfs_da_state_alloc(args);
2030 	/*
2031 	 * Look up the name.  We're not supposed to find it, but
2032 	 * this gives us the insertion point.
2033 	 */
2034 	error = xfs_da3_node_lookup_int(state, &rval);
2035 	if (error)
2036 		rval = error;
2037 	if (rval != -ENOENT) {
2038 		goto done;
2039 	}
2040 	/*
2041 	 * Add the data entry to a data block.
2042 	 * Extravalid is set to a freeblock found by lookup.
2043 	 */
2044 	rval = xfs_dir2_node_addname_int(args,
2045 		state->extravalid ? &state->extrablk : NULL);
2046 	if (rval) {
2047 		goto done;
2048 	}
2049 	blk = &state->path.blk[state->path.active - 1];
2050 	ASSERT(blk->magic == XFS_DIR2_LEAFN_MAGIC);
2051 	/*
2052 	 * Add the new leaf entry.
2053 	 */
2054 	rval = xfs_dir2_leafn_add(blk->bp, args, blk->index);
2055 	if (rval == 0) {
2056 		/*
2057 		 * It worked, fix the hash values up the btree.
2058 		 */
2059 		if (!(args->op_flags & XFS_DA_OP_JUSTCHECK))
2060 			xfs_da3_fixhashpath(state, &state->path);
2061 	} else {
2062 		/*
2063 		 * It didn't work, we need to split the leaf block.
2064 		 */
2065 		if (args->total == 0) {
2066 			ASSERT(rval == -ENOSPC);
2067 			goto done;
2068 		}
2069 		/*
2070 		 * Split the leaf block and insert the new entry.
2071 		 */
2072 		rval = xfs_da3_split(state);
2073 	}
2074 done:
2075 	xfs_da_state_free(state);
2076 	return rval;
2077 }
2078 
2079 /*
2080  * Lookup an entry in a node-format directory.
2081  * All the real work happens in xfs_da3_node_lookup_int.
2082  * The only real output is the inode number of the entry.
2083  */
2084 int						/* error */
2085 xfs_dir2_node_lookup(
2086 	xfs_da_args_t	*args)			/* operation arguments */
2087 {
2088 	int		error;			/* error return value */
2089 	int		i;			/* btree level */
2090 	int		rval;			/* operation return value */
2091 	xfs_da_state_t	*state;			/* btree cursor */
2092 
2093 	trace_xfs_dir2_node_lookup(args);
2094 
2095 	/*
2096 	 * Allocate and initialize the btree cursor.
2097 	 */
2098 	state = xfs_da_state_alloc(args);
2099 
2100 	/*
2101 	 * Fill in the path to the entry in the cursor.
2102 	 */
2103 	error = xfs_da3_node_lookup_int(state, &rval);
2104 	if (error)
2105 		rval = error;
2106 	else if (rval == -ENOENT && args->cmpresult == XFS_CMP_CASE) {
2107 		/* If a CI match, dup the actual name and return -EEXIST */
2108 		xfs_dir2_data_entry_t	*dep;
2109 
2110 		dep = (xfs_dir2_data_entry_t *)
2111 			((char *)state->extrablk.bp->b_addr +
2112 						 state->extrablk.index);
2113 		rval = xfs_dir_cilookup_result(args, dep->name, dep->namelen);
2114 	}
2115 	/*
2116 	 * Release the btree blocks and leaf block.
2117 	 */
2118 	for (i = 0; i < state->path.active; i++) {
2119 		xfs_trans_brelse(args->trans, state->path.blk[i].bp);
2120 		state->path.blk[i].bp = NULL;
2121 	}
2122 	/*
2123 	 * Release the data block if we have it.
2124 	 */
2125 	if (state->extravalid && state->extrablk.bp) {
2126 		xfs_trans_brelse(args->trans, state->extrablk.bp);
2127 		state->extrablk.bp = NULL;
2128 	}
2129 	xfs_da_state_free(state);
2130 	return rval;
2131 }
2132 
2133 /*
2134  * Remove an entry from a node-format directory.
2135  */
2136 int						/* error */
2137 xfs_dir2_node_removename(
2138 	struct xfs_da_args	*args)		/* operation arguments */
2139 {
2140 	struct xfs_da_state_blk	*blk;		/* leaf block */
2141 	int			error;		/* error return value */
2142 	int			rval;		/* operation return value */
2143 	struct xfs_da_state	*state;		/* btree cursor */
2144 
2145 	trace_xfs_dir2_node_removename(args);
2146 
2147 	/*
2148 	 * Allocate and initialize the btree cursor.
2149 	 */
2150 	state = xfs_da_state_alloc(args);
2151 
2152 	/* Look up the entry we're deleting, set up the cursor. */
2153 	error = xfs_da3_node_lookup_int(state, &rval);
2154 	if (error)
2155 		goto out_free;
2156 
2157 	/* Didn't find it, upper layer screwed up. */
2158 	if (rval != -EEXIST) {
2159 		error = rval;
2160 		goto out_free;
2161 	}
2162 
2163 	blk = &state->path.blk[state->path.active - 1];
2164 	ASSERT(blk->magic == XFS_DIR2_LEAFN_MAGIC);
2165 	ASSERT(state->extravalid);
2166 	/*
2167 	 * Remove the leaf and data entries.
2168 	 * Extrablk refers to the data block.
2169 	 */
2170 	error = xfs_dir2_leafn_remove(args, blk->bp, blk->index,
2171 		&state->extrablk, &rval);
2172 	if (error)
2173 		goto out_free;
2174 	/*
2175 	 * Fix the hash values up the btree.
2176 	 */
2177 	xfs_da3_fixhashpath(state, &state->path);
2178 	/*
2179 	 * If we need to join leaf blocks, do it.
2180 	 */
2181 	if (rval && state->path.active > 1)
2182 		error = xfs_da3_join(state);
2183 	/*
2184 	 * If no errors so far, try conversion to leaf format.
2185 	 */
2186 	if (!error)
2187 		error = xfs_dir2_node_to_leaf(state);
2188 out_free:
2189 	xfs_da_state_free(state);
2190 	return error;
2191 }
2192 
2193 /*
2194  * Replace an entry's inode number in a node-format directory.
2195  */
2196 int						/* error */
2197 xfs_dir2_node_replace(
2198 	xfs_da_args_t		*args)		/* operation arguments */
2199 {
2200 	xfs_da_state_blk_t	*blk;		/* leaf block */
2201 	xfs_dir2_data_hdr_t	*hdr;		/* data block header */
2202 	xfs_dir2_data_entry_t	*dep;		/* data entry changed */
2203 	int			error;		/* error return value */
2204 	int			i;		/* btree level */
2205 	xfs_ino_t		inum;		/* new inode number */
2206 	int			ftype;		/* new file type */
2207 	int			rval;		/* internal return value */
2208 	xfs_da_state_t		*state;		/* btree cursor */
2209 
2210 	trace_xfs_dir2_node_replace(args);
2211 
2212 	/*
2213 	 * Allocate and initialize the btree cursor.
2214 	 */
2215 	state = xfs_da_state_alloc(args);
2216 
2217 	/*
2218 	 * We have to save new inode number and ftype since
2219 	 * xfs_da3_node_lookup_int() is going to overwrite them
2220 	 */
2221 	inum = args->inumber;
2222 	ftype = args->filetype;
2223 
2224 	/*
2225 	 * Lookup the entry to change in the btree.
2226 	 */
2227 	error = xfs_da3_node_lookup_int(state, &rval);
2228 	if (error) {
2229 		rval = error;
2230 	}
2231 	/*
2232 	 * It should be found, since the vnodeops layer has looked it up
2233 	 * and locked it.  But paranoia is good.
2234 	 */
2235 	if (rval == -EEXIST) {
2236 		struct xfs_dir3_icleaf_hdr	leafhdr;
2237 
2238 		/*
2239 		 * Find the leaf entry.
2240 		 */
2241 		blk = &state->path.blk[state->path.active - 1];
2242 		ASSERT(blk->magic == XFS_DIR2_LEAFN_MAGIC);
2243 		ASSERT(state->extravalid);
2244 
2245 		xfs_dir2_leaf_hdr_from_disk(state->mp, &leafhdr,
2246 					    blk->bp->b_addr);
2247 		/*
2248 		 * Point to the data entry.
2249 		 */
2250 		hdr = state->extrablk.bp->b_addr;
2251 		ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
2252 		       hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC));
2253 		dep = (xfs_dir2_data_entry_t *)
2254 		      ((char *)hdr +
2255 		       xfs_dir2_dataptr_to_off(args->geo,
2256 				be32_to_cpu(leafhdr.ents[blk->index].address)));
2257 		ASSERT(inum != be64_to_cpu(dep->inumber));
2258 		/*
2259 		 * Fill in the new inode number and log the entry.
2260 		 */
2261 		dep->inumber = cpu_to_be64(inum);
2262 		xfs_dir2_data_put_ftype(state->mp, dep, ftype);
2263 		xfs_dir2_data_log_entry(args, state->extrablk.bp, dep);
2264 		rval = 0;
2265 	}
2266 	/*
2267 	 * Didn't find it, and we're holding a data block.  Drop it.
2268 	 */
2269 	else if (state->extravalid) {
2270 		xfs_trans_brelse(args->trans, state->extrablk.bp);
2271 		state->extrablk.bp = NULL;
2272 	}
2273 	/*
2274 	 * Release all the buffers in the cursor.
2275 	 */
2276 	for (i = 0; i < state->path.active; i++) {
2277 		xfs_trans_brelse(args->trans, state->path.blk[i].bp);
2278 		state->path.blk[i].bp = NULL;
2279 	}
2280 	xfs_da_state_free(state);
2281 	return rval;
2282 }
2283 
2284 /*
2285  * Trim off a trailing empty freespace block.
2286  * Return (in rvalp) 1 if we did it, 0 if not.
2287  */
2288 int						/* error */
2289 xfs_dir2_node_trim_free(
2290 	xfs_da_args_t		*args,		/* operation arguments */
2291 	xfs_fileoff_t		fo,		/* free block number */
2292 	int			*rvalp)		/* out: did something */
2293 {
2294 	struct xfs_buf		*bp;		/* freespace buffer */
2295 	xfs_inode_t		*dp;		/* incore directory inode */
2296 	int			error;		/* error return code */
2297 	xfs_dir2_free_t		*free;		/* freespace structure */
2298 	xfs_trans_t		*tp;		/* transaction pointer */
2299 	struct xfs_dir3_icfree_hdr freehdr;
2300 
2301 	dp = args->dp;
2302 	tp = args->trans;
2303 
2304 	*rvalp = 0;
2305 
2306 	/*
2307 	 * Read the freespace block.
2308 	 */
2309 	error = xfs_dir2_free_try_read(tp, dp, args->owner, fo, &bp);
2310 	if (error)
2311 		return error;
2312 	/*
2313 	 * There can be holes in freespace.  If fo is a hole, there's
2314 	 * nothing to do.
2315 	 */
2316 	if (!bp)
2317 		return 0;
2318 	free = bp->b_addr;
2319 	xfs_dir2_free_hdr_from_disk(dp->i_mount, &freehdr, free);
2320 
2321 	/*
2322 	 * If there are used entries, there's nothing to do.
2323 	 */
2324 	if (freehdr.nused > 0) {
2325 		xfs_trans_brelse(tp, bp);
2326 		return 0;
2327 	}
2328 	/*
2329 	 * Blow the block away.
2330 	 */
2331 	error = xfs_dir2_shrink_inode(args,
2332 			xfs_dir2_da_to_db(args->geo, (xfs_dablk_t)fo), bp);
2333 	if (error) {
2334 		/*
2335 		 * Can't fail with ENOSPC since that only happens with no
2336 		 * space reservation, when breaking up an extent into two
2337 		 * pieces.  This is the last block of an extent.
2338 		 */
2339 		ASSERT(error != -ENOSPC);
2340 		xfs_trans_brelse(tp, bp);
2341 		return error;
2342 	}
2343 	/*
2344 	 * Return that we succeeded.
2345 	 */
2346 	*rvalp = 1;
2347 	return 0;
2348 }
2349