1 /*
2  * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
3  *
4  * This program is free software; you can redistribute it and/or modify it
5  * under the terms of version 2 of the GNU General Public License as
6  * published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it would be useful, but
9  * WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
11  *
12  * Further, this software is distributed without any warranty that it is
13  * free of the rightful claim of any third person regarding infringement
14  * or the like.  Any license provided herein, whether implied or
15  * otherwise, applies only to this software file.  Patent licenses, if
16  * any, provided herein do not apply to combinations of this program with
17  * other software, or any other product whatsoever.
18  *
19  * You should have received a copy of the GNU General Public License along
20  * with this program; if not, write the Free Software Foundation, Inc., 59
21  * Temple Place - Suite 330, Boston MA 02111-1307, USA.
22  *
23  * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
24  * Mountain View, CA  94043, or:
25  *
26  * http://www.sgi.com
27  *
28  * For further information regarding this notice, see:
29  *
30  * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
31  */
32 
33 /*
34  * xfs_dir_leaf.c
35  *
36  * GROT: figure out how to recover gracefully when bmap returns ENOSPC.
37  */
38 
39 #include "xfs.h"
40 
41 #include "xfs_macros.h"
42 #include "xfs_types.h"
43 #include "xfs_inum.h"
44 #include "xfs_log.h"
45 #include "xfs_trans.h"
46 #include "xfs_sb.h"
47 #include "xfs_dir.h"
48 #include "xfs_dir2.h"
49 #include "xfs_dmapi.h"
50 #include "xfs_mount.h"
51 #include "xfs_alloc_btree.h"
52 #include "xfs_bmap_btree.h"
53 #include "xfs_ialloc_btree.h"
54 #include "xfs_alloc.h"
55 #include "xfs_btree.h"
56 #include "xfs_attr_sf.h"
57 #include "xfs_dir_sf.h"
58 #include "xfs_dir2_sf.h"
59 #include "xfs_dinode.h"
60 #include "xfs_inode_item.h"
61 #include "xfs_inode.h"
62 #include "xfs_bmap.h"
63 #include "xfs_da_btree.h"
64 #include "xfs_dir_leaf.h"
65 #include "xfs_error.h"
66 
67 /*
68  * xfs_dir_leaf.c
69  *
70  * Routines to implement leaf blocks of directories as Btrees of hashed names.
71  */
72 
73 /*========================================================================
74  * Function prototypes for the kernel.
75  *========================================================================*/
76 
77 /*
78  * Routines used for growing the Btree.
79  */
80 STATIC void xfs_dir_leaf_add_work(xfs_dabuf_t *leaf_buffer, xfs_da_args_t *args,
81 					      int insertion_index,
82 					      int freemap_index);
83 STATIC int xfs_dir_leaf_compact(xfs_trans_t *trans, xfs_dabuf_t *leaf_buffer,
84 					    int musthave, int justcheck);
85 STATIC void xfs_dir_leaf_rebalance(xfs_da_state_t *state,
86 						  xfs_da_state_blk_t *blk1,
87 						  xfs_da_state_blk_t *blk2);
88 STATIC int xfs_dir_leaf_figure_balance(xfs_da_state_t *state,
89 					  xfs_da_state_blk_t *leaf_blk_1,
90 					  xfs_da_state_blk_t *leaf_blk_2,
91 					  int *number_entries_in_blk1,
92 					  int *number_namebytes_in_blk1);
93 
94 /*
95  * Utility routines.
96  */
97 STATIC void xfs_dir_leaf_moveents(xfs_dir_leafblock_t *src_leaf,
98 					      int src_start,
99 					      xfs_dir_leafblock_t *dst_leaf,
100 					      int dst_start, int move_count,
101 					      xfs_mount_t *mp);
102 
103 
104 /*========================================================================
105  * External routines when dirsize < XFS_IFORK_DSIZE(dp).
106  *========================================================================*/
107 
108 
109 /*
110  * Validate a given inode number.
111  */
112 int
xfs_dir_ino_validate(xfs_mount_t * mp,xfs_ino_t ino)113 xfs_dir_ino_validate(xfs_mount_t *mp, xfs_ino_t ino)
114 {
115 	xfs_agblock_t	agblkno;
116 	xfs_agino_t	agino;
117 	xfs_agnumber_t	agno;
118 	int		ino_ok;
119 	int		ioff;
120 
121 	agno = XFS_INO_TO_AGNO(mp, ino);
122 	agblkno = XFS_INO_TO_AGBNO(mp, ino);
123 	ioff = XFS_INO_TO_OFFSET(mp, ino);
124 	agino = XFS_OFFBNO_TO_AGINO(mp, agblkno, ioff);
125 	ino_ok =
126 		agno < mp->m_sb.sb_agcount &&
127 		agblkno < mp->m_sb.sb_agblocks &&
128 		agblkno != 0 &&
129 		ioff < (1 << mp->m_sb.sb_inopblog) &&
130 		XFS_AGINO_TO_INO(mp, agno, agino) == ino;
131 	if (unlikely(XFS_TEST_ERROR(!ino_ok, mp, XFS_ERRTAG_DIR_INO_VALIDATE,
132 			XFS_RANDOM_DIR_INO_VALIDATE))) {
133 		xfs_fs_cmn_err(CE_WARN, mp, "Invalid inode number 0x%Lx",
134 				(unsigned long long) ino);
135 		XFS_ERROR_REPORT("xfs_dir_ino_validate", XFS_ERRLEVEL_LOW, mp);
136 		return XFS_ERROR(EFSCORRUPTED);
137 	}
138 	return 0;
139 }
140 
141 /*
142  * Create the initial contents of a shortform directory.
143  */
144 int
xfs_dir_shortform_create(xfs_da_args_t * args,xfs_ino_t parent)145 xfs_dir_shortform_create(xfs_da_args_t *args, xfs_ino_t parent)
146 {
147 	xfs_dir_sf_hdr_t *hdr;
148 	xfs_inode_t *dp;
149 
150 	dp = args->dp;
151 	ASSERT(dp != NULL);
152 	ASSERT(dp->i_d.di_size == 0);
153 	if (dp->i_d.di_format == XFS_DINODE_FMT_EXTENTS) {
154 		dp->i_df.if_flags &= ~XFS_IFEXTENTS;	/* just in case */
155 		dp->i_d.di_format = XFS_DINODE_FMT_LOCAL;
156 		xfs_trans_log_inode(args->trans, dp, XFS_ILOG_CORE);
157 		dp->i_df.if_flags |= XFS_IFINLINE;
158 	}
159 	ASSERT(dp->i_df.if_flags & XFS_IFINLINE);
160 	ASSERT(dp->i_df.if_bytes == 0);
161 	xfs_idata_realloc(dp, sizeof(*hdr), XFS_DATA_FORK);
162 	hdr = (xfs_dir_sf_hdr_t *)dp->i_df.if_u1.if_data;
163 	XFS_DIR_SF_PUT_DIRINO_ARCH(&parent, &hdr->parent, ARCH_CONVERT);
164 
165 	INT_ZERO(hdr->count, ARCH_CONVERT);
166 	dp->i_d.di_size = sizeof(*hdr);
167 	xfs_trans_log_inode(args->trans, dp, XFS_ILOG_CORE | XFS_ILOG_DDATA);
168 	return(0);
169 }
170 
171 /*
172  * Add a name to the shortform directory structure.
173  * Overflow from the inode has already been checked for.
174  */
175 int
xfs_dir_shortform_addname(xfs_da_args_t * args)176 xfs_dir_shortform_addname(xfs_da_args_t *args)
177 {
178 	xfs_dir_shortform_t *sf;
179 	xfs_dir_sf_entry_t *sfe;
180 	int i, offset, size;
181 	xfs_inode_t *dp;
182 
183 	dp = args->dp;
184 	ASSERT(dp->i_df.if_flags & XFS_IFINLINE);
185 	/*
186 	 * Catch the case where the conversion from shortform to leaf
187 	 * failed part way through.
188 	 */
189 	if (dp->i_d.di_size < sizeof(xfs_dir_sf_hdr_t)) {
190 		ASSERT(XFS_FORCED_SHUTDOWN(dp->i_mount));
191 		return XFS_ERROR(EIO);
192 	}
193 	ASSERT(dp->i_df.if_bytes == dp->i_d.di_size);
194 	ASSERT(dp->i_df.if_u1.if_data != NULL);
195 	sf = (xfs_dir_shortform_t *)dp->i_df.if_u1.if_data;
196 	sfe = &sf->list[0];
197 	for (i = INT_GET(sf->hdr.count, ARCH_CONVERT)-1; i >= 0; i--) {
198 		if (sfe->namelen == args->namelen &&
199 		    args->name[0] == sfe->name[0] &&
200 		    memcmp(args->name, sfe->name, args->namelen) == 0)
201 			return(XFS_ERROR(EEXIST));
202 		sfe = XFS_DIR_SF_NEXTENTRY(sfe);
203 	}
204 
205 	offset = (int)((char *)sfe - (char *)sf);
206 	size = XFS_DIR_SF_ENTSIZE_BYNAME(args->namelen);
207 	xfs_idata_realloc(dp, size, XFS_DATA_FORK);
208 	sf = (xfs_dir_shortform_t *)dp->i_df.if_u1.if_data;
209 	sfe = (xfs_dir_sf_entry_t *)((char *)sf + offset);
210 
211 	XFS_DIR_SF_PUT_DIRINO_ARCH(&args->inumber, &sfe->inumber, ARCH_CONVERT);
212 	sfe->namelen = args->namelen;
213 	memcpy(sfe->name, args->name, sfe->namelen);
214 	INT_MOD(sf->hdr.count, ARCH_CONVERT, +1);
215 
216 	dp->i_d.di_size += size;
217 	xfs_trans_log_inode(args->trans, dp, XFS_ILOG_CORE | XFS_ILOG_DDATA);
218 
219 	return(0);
220 }
221 
222 /*
223  * Remove a name from the shortform directory structure.
224  */
225 int
xfs_dir_shortform_removename(xfs_da_args_t * args)226 xfs_dir_shortform_removename(xfs_da_args_t *args)
227 {
228 	xfs_dir_shortform_t *sf;
229 	xfs_dir_sf_entry_t *sfe;
230 	int base, size = 0, i;
231 	xfs_inode_t *dp;
232 
233 	dp = args->dp;
234 	ASSERT(dp->i_df.if_flags & XFS_IFINLINE);
235 	/*
236 	 * Catch the case where the conversion from shortform to leaf
237 	 * failed part way through.
238 	 */
239 	if (dp->i_d.di_size < sizeof(xfs_dir_sf_hdr_t)) {
240 		ASSERT(XFS_FORCED_SHUTDOWN(dp->i_mount));
241 		return XFS_ERROR(EIO);
242 	}
243 	ASSERT(dp->i_df.if_bytes == dp->i_d.di_size);
244 	ASSERT(dp->i_df.if_u1.if_data != NULL);
245 	base = sizeof(xfs_dir_sf_hdr_t);
246 	sf = (xfs_dir_shortform_t *)dp->i_df.if_u1.if_data;
247 	sfe = &sf->list[0];
248 	for (i = INT_GET(sf->hdr.count, ARCH_CONVERT)-1; i >= 0; i--) {
249 		size = XFS_DIR_SF_ENTSIZE_BYENTRY(sfe);
250 		if (sfe->namelen == args->namelen &&
251 		    sfe->name[0] == args->name[0] &&
252 		    memcmp(sfe->name, args->name, args->namelen) == 0)
253 			break;
254 		base += size;
255 		sfe = XFS_DIR_SF_NEXTENTRY(sfe);
256 	}
257 	if (i < 0) {
258 		ASSERT(args->oknoent);
259 		return(XFS_ERROR(ENOENT));
260 	}
261 
262 	if ((base + size) != dp->i_d.di_size) {
263 		memmove(&((char *)sf)[base], &((char *)sf)[base+size],
264 					      dp->i_d.di_size - (base+size));
265 	}
266 	INT_MOD(sf->hdr.count, ARCH_CONVERT, -1);
267 
268 	xfs_idata_realloc(dp, -size, XFS_DATA_FORK);
269 	dp->i_d.di_size -= size;
270 	xfs_trans_log_inode(args->trans, dp, XFS_ILOG_CORE | XFS_ILOG_DDATA);
271 
272 	return(0);
273 }
274 
275 /*
276  * Look up a name in a shortform directory structure.
277  */
278 int
xfs_dir_shortform_lookup(xfs_da_args_t * args)279 xfs_dir_shortform_lookup(xfs_da_args_t *args)
280 {
281 	xfs_dir_shortform_t *sf;
282 	xfs_dir_sf_entry_t *sfe;
283 	int i;
284 	xfs_inode_t *dp;
285 
286 	dp = args->dp;
287 	ASSERT(dp->i_df.if_flags & XFS_IFINLINE);
288 	/*
289 	 * Catch the case where the conversion from shortform to leaf
290 	 * failed part way through.
291 	 */
292 	if (dp->i_d.di_size < sizeof(xfs_dir_sf_hdr_t)) {
293 		ASSERT(XFS_FORCED_SHUTDOWN(dp->i_mount));
294 		return XFS_ERROR(EIO);
295 	}
296 	ASSERT(dp->i_df.if_bytes == dp->i_d.di_size);
297 	ASSERT(dp->i_df.if_u1.if_data != NULL);
298 	sf = (xfs_dir_shortform_t *)dp->i_df.if_u1.if_data;
299 	if (args->namelen == 2 &&
300 	    args->name[0] == '.' && args->name[1] == '.') {
301 		XFS_DIR_SF_GET_DIRINO_ARCH(&sf->hdr.parent, &args->inumber, ARCH_CONVERT);
302 		return(XFS_ERROR(EEXIST));
303 	}
304 	if (args->namelen == 1 && args->name[0] == '.') {
305 		args->inumber = dp->i_ino;
306 		return(XFS_ERROR(EEXIST));
307 	}
308 	sfe = &sf->list[0];
309 	for (i = INT_GET(sf->hdr.count, ARCH_CONVERT)-1; i >= 0; i--) {
310 		if (sfe->namelen == args->namelen &&
311 		    sfe->name[0] == args->name[0] &&
312 		    memcmp(args->name, sfe->name, args->namelen) == 0) {
313 			XFS_DIR_SF_GET_DIRINO_ARCH(&sfe->inumber, &args->inumber, ARCH_CONVERT);
314 			return(XFS_ERROR(EEXIST));
315 		}
316 		sfe = XFS_DIR_SF_NEXTENTRY(sfe);
317 	}
318 	ASSERT(args->oknoent);
319 	return(XFS_ERROR(ENOENT));
320 }
321 
322 /*
323  * Convert from using the shortform to the leaf.
324  */
325 int
xfs_dir_shortform_to_leaf(xfs_da_args_t * iargs)326 xfs_dir_shortform_to_leaf(xfs_da_args_t *iargs)
327 {
328 	xfs_inode_t *dp;
329 	xfs_dir_shortform_t *sf;
330 	xfs_dir_sf_entry_t *sfe;
331 	xfs_da_args_t args;
332 	xfs_ino_t inumber;
333 	char *tmpbuffer;
334 	int retval, i, size;
335 	xfs_dablk_t blkno;
336 	xfs_dabuf_t *bp;
337 
338 	dp = iargs->dp;
339 	/*
340 	 * Catch the case where the conversion from shortform to leaf
341 	 * failed part way through.
342 	 */
343 	if (dp->i_d.di_size < sizeof(xfs_dir_sf_hdr_t)) {
344 		ASSERT(XFS_FORCED_SHUTDOWN(dp->i_mount));
345 		return XFS_ERROR(EIO);
346 	}
347 	ASSERT(dp->i_df.if_bytes == dp->i_d.di_size);
348 	ASSERT(dp->i_df.if_u1.if_data != NULL);
349 	size = dp->i_df.if_bytes;
350 	tmpbuffer = kmem_alloc(size, KM_SLEEP);
351 	ASSERT(tmpbuffer != NULL);
352 
353 	memcpy(tmpbuffer, dp->i_df.if_u1.if_data, size);
354 
355 	sf = (xfs_dir_shortform_t *)tmpbuffer;
356 	XFS_DIR_SF_GET_DIRINO_ARCH(&sf->hdr.parent, &inumber, ARCH_CONVERT);
357 
358 	xfs_idata_realloc(dp, -size, XFS_DATA_FORK);
359 	dp->i_d.di_size = 0;
360 	xfs_trans_log_inode(iargs->trans, dp, XFS_ILOG_CORE);
361 	retval = xfs_da_grow_inode(iargs, &blkno);
362 	if (retval)
363 		goto out;
364 
365 	ASSERT(blkno == 0);
366 	retval = xfs_dir_leaf_create(iargs, blkno, &bp);
367 	if (retval)
368 		goto out;
369 	xfs_da_buf_done(bp);
370 
371 	args.name = ".";
372 	args.namelen = 1;
373 	args.hashval = xfs_dir_hash_dot;
374 	args.inumber = dp->i_ino;
375 	args.dp = dp;
376 	args.firstblock = iargs->firstblock;
377 	args.flist = iargs->flist;
378 	args.total = iargs->total;
379 	args.whichfork = XFS_DATA_FORK;
380 	args.trans = iargs->trans;
381 	args.justcheck = 0;
382 	args.addname = args.oknoent = 1;
383 	retval = xfs_dir_leaf_addname(&args);
384 	if (retval)
385 		goto out;
386 
387 	args.name = "..";
388 	args.namelen = 2;
389 	args.hashval = xfs_dir_hash_dotdot;
390 	args.inumber = inumber;
391 	retval = xfs_dir_leaf_addname(&args);
392 	if (retval)
393 		goto out;
394 
395 	sfe = &sf->list[0];
396 	for (i = 0; i < INT_GET(sf->hdr.count, ARCH_CONVERT); i++) {
397 		args.name = (char *)(sfe->name);
398 		args.namelen = sfe->namelen;
399 		args.hashval = xfs_da_hashname((char *)(sfe->name),
400 					       sfe->namelen);
401 		XFS_DIR_SF_GET_DIRINO_ARCH(&sfe->inumber, &args.inumber, ARCH_CONVERT);
402 		retval = xfs_dir_leaf_addname(&args);
403 		if (retval)
404 			goto out;
405 		sfe = XFS_DIR_SF_NEXTENTRY(sfe);
406 	}
407 	retval = 0;
408 
409 out:
410 	kmem_free(tmpbuffer, size);
411 	return(retval);
412 }
413 
414 STATIC int
xfs_dir_shortform_compare(const void * a,const void * b)415 xfs_dir_shortform_compare(const void *a, const void *b)
416 {
417 	xfs_dir_sf_sort_t *sa, *sb;
418 
419 	sa = (xfs_dir_sf_sort_t *)a;
420 	sb = (xfs_dir_sf_sort_t *)b;
421 	if (sa->hash < sb->hash)
422 		return -1;
423 	else if (sa->hash > sb->hash)
424 		return 1;
425 	else
426 		return sa->entno - sb->entno;
427 }
428 
429 /*
430  * Copy out directory entries for getdents(), for shortform directories.
431  */
432 /*ARGSUSED*/
433 int
xfs_dir_shortform_getdents(xfs_inode_t * dp,uio_t * uio,int * eofp,xfs_dirent_t * dbp,xfs_dir_put_t put)434 xfs_dir_shortform_getdents(xfs_inode_t *dp, uio_t *uio, int *eofp,
435 				       xfs_dirent_t *dbp, xfs_dir_put_t put)
436 {
437 	xfs_dir_shortform_t *sf;
438 	xfs_dir_sf_entry_t *sfe;
439 	int retval, i, sbsize, nsbuf, lastresid=0, want_entno;
440 	xfs_mount_t *mp;
441 	xfs_dahash_t cookhash, hash;
442 	xfs_dir_put_args_t p;
443 	xfs_dir_sf_sort_t *sbuf, *sbp;
444 
445 	mp = dp->i_mount;
446 	sf = (xfs_dir_shortform_t *)dp->i_df.if_u1.if_data;
447 	cookhash = XFS_DA_COOKIE_HASH(mp, uio->uio_offset);
448 	want_entno = XFS_DA_COOKIE_ENTRY(mp, uio->uio_offset);
449 	nsbuf = INT_GET(sf->hdr.count, ARCH_CONVERT) + 2;
450 	sbsize = (nsbuf + 1) * sizeof(*sbuf);
451 	sbp = sbuf = kmem_alloc(sbsize, KM_SLEEP);
452 
453 	xfs_dir_trace_g_du("sf: start", dp, uio);
454 
455 	/*
456 	 * Collect all the entries into the buffer.
457 	 * Entry 0 is .
458 	 */
459 	sbp->entno = 0;
460 	sbp->seqno = 0;
461 	sbp->hash = xfs_dir_hash_dot;
462 	sbp->ino = dp->i_ino;
463 	sbp->name = ".";
464 	sbp->namelen = 1;
465 	sbp++;
466 
467 	/*
468 	 * Entry 1 is ..
469 	 */
470 	sbp->entno = 1;
471 	sbp->seqno = 0;
472 	sbp->hash = xfs_dir_hash_dotdot;
473 	sbp->ino = XFS_GET_DIR_INO_ARCH(mp, sf->hdr.parent, ARCH_CONVERT);
474 	sbp->name = "..";
475 	sbp->namelen = 2;
476 	sbp++;
477 
478 	/*
479 	 * Scan the directory data for the rest of the entries.
480 	 */
481 	for (i = 0, sfe = &sf->list[0];
482 			i < INT_GET(sf->hdr.count, ARCH_CONVERT); i++) {
483 
484 		if (unlikely(
485 		    ((char *)sfe < (char *)sf) ||
486 		    ((char *)sfe >= ((char *)sf + dp->i_df.if_bytes)))) {
487 			xfs_dir_trace_g_du("sf: corrupted", dp, uio);
488 			XFS_CORRUPTION_ERROR("xfs_dir_shortform_getdents",
489 					     XFS_ERRLEVEL_LOW, mp, sfe);
490 			kmem_free(sbuf, sbsize);
491 			return XFS_ERROR(EFSCORRUPTED);
492 		}
493 
494 		sbp->entno = i + 2;
495 		sbp->seqno = 0;
496 		sbp->hash = xfs_da_hashname((char *)sfe->name, sfe->namelen);
497 		sbp->ino = XFS_GET_DIR_INO_ARCH(mp, sfe->inumber, ARCH_CONVERT);
498 		sbp->name = (char *)sfe->name;
499 		sbp->namelen = sfe->namelen;
500 		sfe = XFS_DIR_SF_NEXTENTRY(sfe);
501 		sbp++;
502 	}
503 
504 	/*
505 	 * Sort the entries on hash then entno.
506 	 */
507 	qsort(sbuf, nsbuf, sizeof(*sbuf), xfs_dir_shortform_compare);
508 	/*
509 	 * Stuff in last entry.
510 	 */
511 	sbp->entno = nsbuf;
512 	sbp->hash = XFS_DA_MAXHASH;
513 	sbp->seqno = 0;
514 	/*
515 	 * Figure out the sequence numbers in case there's a hash duplicate.
516 	 */
517 	for (hash = sbuf->hash, sbp = sbuf + 1;
518 				sbp < &sbuf[nsbuf + 1]; sbp++) {
519 		if (sbp->hash == hash)
520 			sbp->seqno = sbp[-1].seqno + 1;
521 		else
522 			hash = sbp->hash;
523 	}
524 
525 	/*
526 	 * Set up put routine.
527 	 */
528 	p.dbp = dbp;
529 	p.put = put;
530 	p.uio = uio;
531 
532 	/*
533 	 * Find our place.
534 	 */
535 	for (sbp = sbuf; sbp < &sbuf[nsbuf + 1]; sbp++) {
536 		if (sbp->hash > cookhash ||
537 		    (sbp->hash == cookhash && sbp->seqno >= want_entno))
538 			break;
539 	}
540 
541 	/*
542 	 * Did we fail to find anything?  We stop at the last entry,
543 	 * the one we put maxhash into.
544 	 */
545 	if (sbp == &sbuf[nsbuf]) {
546 		kmem_free(sbuf, sbsize);
547 		xfs_dir_trace_g_du("sf: hash beyond end", dp, uio);
548 		uio->uio_offset = XFS_DA_MAKE_COOKIE(mp, 0, 0, XFS_DA_MAXHASH);
549 		*eofp = 1;
550 		return 0;
551 	}
552 
553 	/*
554 	 * Loop putting entries into the user buffer.
555 	 */
556 	while (sbp < &sbuf[nsbuf]) {
557 		/*
558 		 * Save the first resid in a run of equal-hashval entries
559 		 * so that we can back them out if they don't all fit.
560 		 */
561 		if (sbp->seqno == 0 || sbp == sbuf)
562 			lastresid = uio->uio_resid;
563 		XFS_PUT_COOKIE(p.cook, mp, 0, sbp[1].seqno, sbp[1].hash);
564 		p.ino = sbp->ino;
565 #if XFS_BIG_INUMS
566 		p.ino += mp->m_inoadd;
567 #endif
568 		p.name = sbp->name;
569 		p.namelen = sbp->namelen;
570 		retval = p.put(&p);
571 		if (!p.done) {
572 			uio->uio_offset =
573 				XFS_DA_MAKE_COOKIE(mp, 0, 0, sbp->hash);
574 			kmem_free(sbuf, sbsize);
575 			uio->uio_resid = lastresid;
576 			xfs_dir_trace_g_du("sf: E-O-B", dp, uio);
577 			return retval;
578 		}
579 		sbp++;
580 	}
581 	kmem_free(sbuf, sbsize);
582 	uio->uio_offset = p.cook.o;
583 	*eofp = 1;
584 	xfs_dir_trace_g_du("sf: E-O-F", dp, uio);
585 	return 0;
586 }
587 
588 /*
589  * Look up a name in a shortform directory structure, replace the inode number.
590  */
591 int
xfs_dir_shortform_replace(xfs_da_args_t * args)592 xfs_dir_shortform_replace(xfs_da_args_t *args)
593 {
594 	xfs_dir_shortform_t *sf;
595 	xfs_dir_sf_entry_t *sfe;
596 	xfs_inode_t *dp;
597 	int i;
598 
599 	dp = args->dp;
600 	ASSERT(dp->i_df.if_flags & XFS_IFINLINE);
601 	/*
602 	 * Catch the case where the conversion from shortform to leaf
603 	 * failed part way through.
604 	 */
605 	if (dp->i_d.di_size < sizeof(xfs_dir_sf_hdr_t)) {
606 		ASSERT(XFS_FORCED_SHUTDOWN(dp->i_mount));
607 		return XFS_ERROR(EIO);
608 	}
609 	ASSERT(dp->i_df.if_bytes == dp->i_d.di_size);
610 	ASSERT(dp->i_df.if_u1.if_data != NULL);
611 	sf = (xfs_dir_shortform_t *)dp->i_df.if_u1.if_data;
612 	if (args->namelen == 2 &&
613 	    args->name[0] == '.' && args->name[1] == '.') {
614 		/* XXX - replace assert? */
615 		XFS_DIR_SF_PUT_DIRINO_ARCH(&args->inumber, &sf->hdr.parent, ARCH_CONVERT);
616 		xfs_trans_log_inode(args->trans, dp, XFS_ILOG_DDATA);
617 		return(0);
618 	}
619 	ASSERT(args->namelen != 1 || args->name[0] != '.');
620 	sfe = &sf->list[0];
621 	for (i = INT_GET(sf->hdr.count, ARCH_CONVERT)-1; i >= 0; i--) {
622 		if (sfe->namelen == args->namelen &&
623 		    sfe->name[0] == args->name[0] &&
624 		    memcmp(args->name, sfe->name, args->namelen) == 0) {
625 			ASSERT(memcmp((char *)&args->inumber,
626 				(char *)&sfe->inumber, sizeof(xfs_ino_t)));
627 			XFS_DIR_SF_PUT_DIRINO_ARCH(&args->inumber, &sfe->inumber, ARCH_CONVERT);
628 			xfs_trans_log_inode(args->trans, dp, XFS_ILOG_DDATA);
629 			return(0);
630 		}
631 		sfe = XFS_DIR_SF_NEXTENTRY(sfe);
632 	}
633 	ASSERT(args->oknoent);
634 	return(XFS_ERROR(ENOENT));
635 }
636 
637 /*
638  * Convert a leaf directory to shortform structure
639  */
640 int
xfs_dir_leaf_to_shortform(xfs_da_args_t * iargs)641 xfs_dir_leaf_to_shortform(xfs_da_args_t *iargs)
642 {
643 	xfs_dir_leafblock_t *leaf;
644 	xfs_dir_leaf_hdr_t *hdr;
645 	xfs_dir_leaf_entry_t *entry;
646 	xfs_dir_leaf_name_t *namest;
647 	xfs_da_args_t args;
648 	xfs_inode_t *dp;
649 	xfs_ino_t parent;
650 	char *tmpbuffer;
651 	int retval, i;
652 	xfs_dabuf_t *bp;
653 
654 	dp = iargs->dp;
655 	tmpbuffer = kmem_alloc(XFS_LBSIZE(dp->i_mount), KM_SLEEP);
656 	ASSERT(tmpbuffer != NULL);
657 
658 	retval = xfs_da_read_buf(iargs->trans, iargs->dp, 0, -1, &bp,
659 					       XFS_DATA_FORK);
660 	if (retval)
661 		goto out;
662 	ASSERT(bp != NULL);
663 	memcpy(tmpbuffer, bp->data, XFS_LBSIZE(dp->i_mount));
664 	leaf = (xfs_dir_leafblock_t *)tmpbuffer;
665 	ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
666 	memset(bp->data, 0, XFS_LBSIZE(dp->i_mount));
667 
668 	/*
669 	 * Find and special case the parent inode number
670 	 */
671 	hdr = &leaf->hdr;
672 	entry = &leaf->entries[0];
673 	for (i = INT_GET(hdr->count, ARCH_CONVERT)-1; i >= 0; entry++, i--) {
674 		namest = XFS_DIR_LEAF_NAMESTRUCT(leaf, INT_GET(entry->nameidx, ARCH_CONVERT));
675 		if ((entry->namelen == 2) &&
676 		    (namest->name[0] == '.') &&
677 		    (namest->name[1] == '.')) {
678 			XFS_DIR_SF_GET_DIRINO_ARCH(&namest->inumber, &parent, ARCH_CONVERT);
679 			INT_ZERO(entry->nameidx, ARCH_CONVERT);
680 		} else if ((entry->namelen == 1) && (namest->name[0] == '.')) {
681 			INT_ZERO(entry->nameidx, ARCH_CONVERT);
682 		}
683 	}
684 	retval = xfs_da_shrink_inode(iargs, 0, bp);
685 	if (retval)
686 		goto out;
687 	retval = xfs_dir_shortform_create(iargs, parent);
688 	if (retval)
689 		goto out;
690 
691 	/*
692 	 * Copy the rest of the filenames
693 	 */
694 	entry = &leaf->entries[0];
695 	args.dp = dp;
696 	args.firstblock = iargs->firstblock;
697 	args.flist = iargs->flist;
698 	args.total = iargs->total;
699 	args.whichfork = XFS_DATA_FORK;
700 	args.trans = iargs->trans;
701 	args.justcheck = 0;
702 	args.addname = args.oknoent = 1;
703 	for (i = 0; i < INT_GET(hdr->count, ARCH_CONVERT); entry++, i++) {
704 		if (INT_ISZERO(entry->nameidx, ARCH_CONVERT))
705 			continue;
706 		namest = XFS_DIR_LEAF_NAMESTRUCT(leaf, INT_GET(entry->nameidx, ARCH_CONVERT));
707 		args.name = (char *)(namest->name);
708 		args.namelen = entry->namelen;
709 		args.hashval = INT_GET(entry->hashval, ARCH_CONVERT);
710 		XFS_DIR_SF_GET_DIRINO_ARCH(&namest->inumber, &args.inumber, ARCH_CONVERT);
711 		xfs_dir_shortform_addname(&args);
712 	}
713 
714 out:
715 	kmem_free(tmpbuffer, XFS_LBSIZE(dp->i_mount));
716 	return(retval);
717 }
718 
719 /*
720  * Convert from using a single leaf to a root node and a leaf.
721  */
722 int
xfs_dir_leaf_to_node(xfs_da_args_t * args)723 xfs_dir_leaf_to_node(xfs_da_args_t *args)
724 {
725 	xfs_dir_leafblock_t *leaf;
726 	xfs_da_intnode_t *node;
727 	xfs_inode_t *dp;
728 	xfs_dabuf_t *bp1, *bp2;
729 	xfs_dablk_t blkno;
730 	int retval;
731 
732 	dp = args->dp;
733 	retval = xfs_da_grow_inode(args, &blkno);
734 	ASSERT(blkno == 1);
735 	if (retval)
736 		return(retval);
737 	retval = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp1,
738 					      XFS_DATA_FORK);
739 	if (retval)
740 		return(retval);
741 	ASSERT(bp1 != NULL);
742 	retval = xfs_da_get_buf(args->trans, args->dp, 1, -1, &bp2,
743 					     XFS_DATA_FORK);
744 	if (retval) {
745 		xfs_da_buf_done(bp1);
746 		return(retval);
747 	}
748 	ASSERT(bp2 != NULL);
749 	memcpy(bp2->data, bp1->data, XFS_LBSIZE(dp->i_mount));
750 	xfs_da_buf_done(bp1);
751 	xfs_da_log_buf(args->trans, bp2, 0, XFS_LBSIZE(dp->i_mount) - 1);
752 
753 	/*
754 	 * Set up the new root node.
755 	 */
756 	retval = xfs_da_node_create(args, 0, 1, &bp1, XFS_DATA_FORK);
757 	if (retval) {
758 		xfs_da_buf_done(bp2);
759 		return(retval);
760 	}
761 	node = bp1->data;
762 	leaf = bp2->data;
763 	ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
764 	INT_SET(node->btree[0].hashval, ARCH_CONVERT, INT_GET(leaf->entries[ INT_GET(leaf->hdr.count, ARCH_CONVERT)-1 ].hashval, ARCH_CONVERT));
765 	xfs_da_buf_done(bp2);
766 	INT_SET(node->btree[0].before, ARCH_CONVERT, blkno);
767 	INT_SET(node->hdr.count, ARCH_CONVERT, 1);
768 	xfs_da_log_buf(args->trans, bp1,
769 		XFS_DA_LOGRANGE(node, &node->btree[0], sizeof(node->btree[0])));
770 	xfs_da_buf_done(bp1);
771 
772 	return(retval);
773 }
774 
775 
776 /*========================================================================
777  * Routines used for growing the Btree.
778  *========================================================================*/
779 
780 /*
781  * Create the initial contents of a leaf directory
782  * or a leaf in a node directory.
783  */
784 int
xfs_dir_leaf_create(xfs_da_args_t * args,xfs_dablk_t blkno,xfs_dabuf_t ** bpp)785 xfs_dir_leaf_create(xfs_da_args_t *args, xfs_dablk_t blkno, xfs_dabuf_t **bpp)
786 {
787 	xfs_dir_leafblock_t *leaf;
788 	xfs_dir_leaf_hdr_t *hdr;
789 	xfs_inode_t *dp;
790 	xfs_dabuf_t *bp;
791 	int retval;
792 
793 	dp = args->dp;
794 	ASSERT(dp != NULL);
795 	retval = xfs_da_get_buf(args->trans, dp, blkno, -1, &bp, XFS_DATA_FORK);
796 	if (retval)
797 		return(retval);
798 	ASSERT(bp != NULL);
799 	leaf = bp->data;
800 	memset((char *)leaf, 0, XFS_LBSIZE(dp->i_mount));
801 	hdr = &leaf->hdr;
802 	INT_SET(hdr->info.magic, ARCH_CONVERT, XFS_DIR_LEAF_MAGIC);
803 	INT_SET(hdr->firstused, ARCH_CONVERT, XFS_LBSIZE(dp->i_mount));
804 	if (INT_ISZERO(hdr->firstused, ARCH_CONVERT))
805 		INT_SET(hdr->firstused, ARCH_CONVERT, XFS_LBSIZE(dp->i_mount) - 1);
806 	INT_SET(hdr->freemap[0].base, ARCH_CONVERT, sizeof(xfs_dir_leaf_hdr_t));
807 	INT_SET(hdr->freemap[0].size, ARCH_CONVERT, INT_GET(hdr->firstused, ARCH_CONVERT) - INT_GET(hdr->freemap[0].base, ARCH_CONVERT));
808 
809 	xfs_da_log_buf(args->trans, bp, 0, XFS_LBSIZE(dp->i_mount) - 1);
810 
811 	*bpp = bp;
812 	return(0);
813 }
814 
815 /*
816  * Split the leaf node, rebalance, then add the new entry.
817  */
818 int
xfs_dir_leaf_split(xfs_da_state_t * state,xfs_da_state_blk_t * oldblk,xfs_da_state_blk_t * newblk)819 xfs_dir_leaf_split(xfs_da_state_t *state, xfs_da_state_blk_t *oldblk,
820 				  xfs_da_state_blk_t *newblk)
821 {
822 	xfs_dablk_t blkno;
823 	xfs_da_args_t *args;
824 	int error;
825 
826 	/*
827 	 * Allocate space for a new leaf node.
828 	 */
829 	args = state->args;
830 	ASSERT(args != NULL);
831 	ASSERT(oldblk->magic == XFS_DIR_LEAF_MAGIC);
832 	error = xfs_da_grow_inode(args, &blkno);
833 	if (error)
834 		return(error);
835 	error = xfs_dir_leaf_create(args, blkno, &newblk->bp);
836 	if (error)
837 		return(error);
838 	newblk->blkno = blkno;
839 	newblk->magic = XFS_DIR_LEAF_MAGIC;
840 
841 	/*
842 	 * Rebalance the entries across the two leaves.
843 	 */
844 	xfs_dir_leaf_rebalance(state, oldblk, newblk);
845 	error = xfs_da_blk_link(state, oldblk, newblk);
846 	if (error)
847 		return(error);
848 
849 	/*
850 	 * Insert the new entry in the correct block.
851 	 */
852 	if (state->inleaf) {
853 		error = xfs_dir_leaf_add(oldblk->bp, args, oldblk->index);
854 	} else {
855 		error = xfs_dir_leaf_add(newblk->bp, args, newblk->index);
856 	}
857 
858 	/*
859 	 * Update last hashval in each block since we added the name.
860 	 */
861 	oldblk->hashval = xfs_dir_leaf_lasthash(oldblk->bp, NULL);
862 	newblk->hashval = xfs_dir_leaf_lasthash(newblk->bp, NULL);
863 	return(error);
864 }
865 
866 /*
867  * Add a name to the leaf directory structure.
868  *
869  * Must take into account fragmented leaves and leaves where spacemap has
870  * lost some freespace information (ie: holes).
871  */
872 int
xfs_dir_leaf_add(xfs_dabuf_t * bp,xfs_da_args_t * args,int index)873 xfs_dir_leaf_add(xfs_dabuf_t *bp, xfs_da_args_t *args, int index)
874 {
875 	xfs_dir_leafblock_t *leaf;
876 	xfs_dir_leaf_hdr_t *hdr;
877 	xfs_dir_leaf_map_t *map;
878 	int tablesize, entsize, sum, i, tmp, error;
879 
880 	leaf = bp->data;
881 	ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
882 	ASSERT((index >= 0) && (index <= INT_GET(leaf->hdr.count, ARCH_CONVERT)));
883 	hdr = &leaf->hdr;
884 	entsize = XFS_DIR_LEAF_ENTSIZE_BYNAME(args->namelen);
885 
886 	/*
887 	 * Search through freemap for first-fit on new name length.
888 	 * (may need to figure in size of entry struct too)
889 	 */
890 	tablesize = (INT_GET(hdr->count, ARCH_CONVERT) + 1) * (uint)sizeof(xfs_dir_leaf_entry_t)
891 			+ (uint)sizeof(xfs_dir_leaf_hdr_t);
892 	map = &hdr->freemap[XFS_DIR_LEAF_MAPSIZE-1];
893 	for (sum = 0, i = XFS_DIR_LEAF_MAPSIZE-1; i >= 0; map--, i--) {
894 		if (tablesize > INT_GET(hdr->firstused, ARCH_CONVERT)) {
895 			sum += INT_GET(map->size, ARCH_CONVERT);
896 			continue;
897 		}
898 		if (INT_ISZERO(map->size, ARCH_CONVERT))
899 			continue;	/* no space in this map */
900 		tmp = entsize;
901 		if (INT_GET(map->base, ARCH_CONVERT) < INT_GET(hdr->firstused, ARCH_CONVERT))
902 			tmp += (uint)sizeof(xfs_dir_leaf_entry_t);
903 		if (INT_GET(map->size, ARCH_CONVERT) >= tmp) {
904 			if (!args->justcheck)
905 				xfs_dir_leaf_add_work(bp, args, index, i);
906 			return(0);
907 		}
908 		sum += INT_GET(map->size, ARCH_CONVERT);
909 	}
910 
911 	/*
912 	 * If there are no holes in the address space of the block,
913 	 * and we don't have enough freespace, then compaction will do us
914 	 * no good and we should just give up.
915 	 */
916 	if (!hdr->holes && (sum < entsize))
917 		return(XFS_ERROR(ENOSPC));
918 
919 	/*
920 	 * Compact the entries to coalesce free space.
921 	 * Pass the justcheck flag so the checking pass can return
922 	 * an error, without changing anything, if it won't fit.
923 	 */
924 	error = xfs_dir_leaf_compact(args->trans, bp,
925 			args->total == 0 ?
926 				entsize +
927 				(uint)sizeof(xfs_dir_leaf_entry_t) : 0,
928 			args->justcheck);
929 	if (error)
930 		return(error);
931 	/*
932 	 * After compaction, the block is guaranteed to have only one
933 	 * free region, in freemap[0].  If it is not big enough, give up.
934 	 */
935 	if (INT_GET(hdr->freemap[0].size, ARCH_CONVERT) <
936 	    (entsize + (uint)sizeof(xfs_dir_leaf_entry_t)))
937 		return(XFS_ERROR(ENOSPC));
938 
939 	if (!args->justcheck)
940 		xfs_dir_leaf_add_work(bp, args, index, 0);
941 	return(0);
942 }
943 
944 /*
945  * Add a name to a leaf directory structure.
946  */
947 STATIC void
xfs_dir_leaf_add_work(xfs_dabuf_t * bp,xfs_da_args_t * args,int index,int mapindex)948 xfs_dir_leaf_add_work(xfs_dabuf_t *bp, xfs_da_args_t *args, int index,
949 		      int mapindex)
950 {
951 	xfs_dir_leafblock_t *leaf;
952 	xfs_dir_leaf_hdr_t *hdr;
953 	xfs_dir_leaf_entry_t *entry;
954 	xfs_dir_leaf_name_t *namest;
955 	xfs_dir_leaf_map_t *map;
956 	/* REFERENCED */
957 	xfs_mount_t *mp;
958 	int tmp, i;
959 
960 	leaf = bp->data;
961 	ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
962 	hdr = &leaf->hdr;
963 	ASSERT((mapindex >= 0) && (mapindex < XFS_DIR_LEAF_MAPSIZE));
964 	ASSERT((index >= 0) && (index <= INT_GET(hdr->count, ARCH_CONVERT)));
965 
966 	/*
967 	 * Force open some space in the entry array and fill it in.
968 	 */
969 	entry = &leaf->entries[index];
970 	if (index < INT_GET(hdr->count, ARCH_CONVERT)) {
971 		tmp  = INT_GET(hdr->count, ARCH_CONVERT) - index;
972 		tmp *= (uint)sizeof(xfs_dir_leaf_entry_t);
973 		memmove(entry + 1, entry, tmp);
974 		xfs_da_log_buf(args->trans, bp,
975 		    XFS_DA_LOGRANGE(leaf, entry, tmp + (uint)sizeof(*entry)));
976 	}
977 	INT_MOD(hdr->count, ARCH_CONVERT, +1);
978 
979 	/*
980 	 * Allocate space for the new string (at the end of the run).
981 	 */
982 	map = &hdr->freemap[mapindex];
983 	mp = args->trans->t_mountp;
984 	ASSERT(INT_GET(map->base, ARCH_CONVERT) < XFS_LBSIZE(mp));
985 	ASSERT(INT_GET(map->size, ARCH_CONVERT) >= XFS_DIR_LEAF_ENTSIZE_BYNAME(args->namelen));
986 	ASSERT(INT_GET(map->size, ARCH_CONVERT) < XFS_LBSIZE(mp));
987 	INT_MOD(map->size, ARCH_CONVERT, -(XFS_DIR_LEAF_ENTSIZE_BYNAME(args->namelen)));
988 	INT_SET(entry->nameidx, ARCH_CONVERT, INT_GET(map->base, ARCH_CONVERT) + INT_GET(map->size, ARCH_CONVERT));
989 	INT_SET(entry->hashval, ARCH_CONVERT, args->hashval);
990 	entry->namelen = args->namelen;
991 	xfs_da_log_buf(args->trans, bp,
992 	    XFS_DA_LOGRANGE(leaf, entry, sizeof(*entry)));
993 
994 	/*
995 	 * Copy the string and inode number into the new space.
996 	 */
997 	namest = XFS_DIR_LEAF_NAMESTRUCT(leaf, INT_GET(entry->nameidx, ARCH_CONVERT));
998 	XFS_DIR_SF_PUT_DIRINO_ARCH(&args->inumber, &namest->inumber, ARCH_CONVERT);
999 	memcpy(namest->name, args->name, args->namelen);
1000 	xfs_da_log_buf(args->trans, bp,
1001 	    XFS_DA_LOGRANGE(leaf, namest, XFS_DIR_LEAF_ENTSIZE_BYENTRY(entry)));
1002 
1003 	/*
1004 	 * Update the control info for this leaf node
1005 	 */
1006 	if (INT_GET(entry->nameidx, ARCH_CONVERT) < INT_GET(hdr->firstused, ARCH_CONVERT))
1007 		INT_COPY(hdr->firstused, entry->nameidx, ARCH_CONVERT);
1008 	ASSERT(INT_GET(hdr->firstused, ARCH_CONVERT) >= ((INT_GET(hdr->count, ARCH_CONVERT)*sizeof(*entry))+sizeof(*hdr)));
1009 	tmp = (INT_GET(hdr->count, ARCH_CONVERT)-1) * (uint)sizeof(xfs_dir_leaf_entry_t)
1010 			+ (uint)sizeof(xfs_dir_leaf_hdr_t);
1011 	map = &hdr->freemap[0];
1012 	for (i = 0; i < XFS_DIR_LEAF_MAPSIZE; map++, i++) {
1013 		if (INT_GET(map->base, ARCH_CONVERT) == tmp) {
1014 			INT_MOD(map->base, ARCH_CONVERT, (uint)sizeof(xfs_dir_leaf_entry_t));
1015 			INT_MOD(map->size, ARCH_CONVERT, -((uint)sizeof(xfs_dir_leaf_entry_t)));
1016 		}
1017 	}
1018 	INT_MOD(hdr->namebytes, ARCH_CONVERT, args->namelen);
1019 	xfs_da_log_buf(args->trans, bp,
1020 		XFS_DA_LOGRANGE(leaf, hdr, sizeof(*hdr)));
1021 }
1022 
1023 /*
1024  * Garbage collect a leaf directory block by copying it to a new buffer.
1025  */
1026 STATIC int
xfs_dir_leaf_compact(xfs_trans_t * trans,xfs_dabuf_t * bp,int musthave,int justcheck)1027 xfs_dir_leaf_compact(xfs_trans_t *trans, xfs_dabuf_t *bp, int musthave,
1028 		     int justcheck)
1029 {
1030 	xfs_dir_leafblock_t *leaf_s, *leaf_d;
1031 	xfs_dir_leaf_hdr_t *hdr_s, *hdr_d;
1032 	xfs_mount_t *mp;
1033 	char *tmpbuffer;
1034 	char *tmpbuffer2=NULL;
1035 	int rval;
1036 	int lbsize;
1037 
1038 	mp = trans->t_mountp;
1039 	lbsize = XFS_LBSIZE(mp);
1040 	tmpbuffer = kmem_alloc(lbsize, KM_SLEEP);
1041 	ASSERT(tmpbuffer != NULL);
1042 	memcpy(tmpbuffer, bp->data, lbsize);
1043 
1044 	/*
1045 	 * Make a second copy in case xfs_dir_leaf_moveents()
1046 	 * below destroys the original.
1047 	 */
1048 	if (musthave || justcheck) {
1049 		tmpbuffer2 = kmem_alloc(lbsize, KM_SLEEP);
1050 		memcpy(tmpbuffer2, bp->data, lbsize);
1051 	}
1052 	memset(bp->data, 0, lbsize);
1053 
1054 	/*
1055 	 * Copy basic information
1056 	 */
1057 	leaf_s = (xfs_dir_leafblock_t *)tmpbuffer;
1058 	leaf_d = bp->data;
1059 	hdr_s = &leaf_s->hdr;
1060 	hdr_d = &leaf_d->hdr;
1061 	hdr_d->info = hdr_s->info;	/* struct copy */
1062 	INT_SET(hdr_d->firstused, ARCH_CONVERT, lbsize);
1063 	if (INT_ISZERO(hdr_d->firstused, ARCH_CONVERT))
1064 		INT_SET(hdr_d->firstused, ARCH_CONVERT, lbsize - 1);
1065 	INT_ZERO(hdr_d->namebytes, ARCH_CONVERT);
1066 	INT_ZERO(hdr_d->count, ARCH_CONVERT);
1067 	hdr_d->holes = 0;
1068 	INT_SET(hdr_d->freemap[0].base, ARCH_CONVERT, sizeof(xfs_dir_leaf_hdr_t));
1069 	INT_SET(hdr_d->freemap[0].size, ARCH_CONVERT, INT_GET(hdr_d->firstused, ARCH_CONVERT) - INT_GET(hdr_d->freemap[0].base, ARCH_CONVERT));
1070 
1071 	/*
1072 	 * Copy all entry's in the same (sorted) order,
1073 	 * but allocate filenames packed and in sequence.
1074 	 * This changes the source (leaf_s) as well.
1075 	 */
1076 	xfs_dir_leaf_moveents(leaf_s, 0, leaf_d, 0, (int)INT_GET(hdr_s->count, ARCH_CONVERT), mp);
1077 
1078 	if (musthave && INT_GET(hdr_d->freemap[0].size, ARCH_CONVERT) < musthave)
1079 		rval = XFS_ERROR(ENOSPC);
1080 	else
1081 		rval = 0;
1082 
1083 	if (justcheck || rval == ENOSPC) {
1084 		ASSERT(tmpbuffer2);
1085 		memcpy(bp->data, tmpbuffer2, lbsize);
1086 	} else {
1087 		xfs_da_log_buf(trans, bp, 0, lbsize - 1);
1088 	}
1089 
1090 	kmem_free(tmpbuffer, lbsize);
1091 	if (musthave || justcheck)
1092 		kmem_free(tmpbuffer2, lbsize);
1093 	return(rval);
1094 }
1095 
1096 /*
1097  * Redistribute the directory entries between two leaf nodes,
1098  * taking into account the size of the new entry.
1099  *
1100  * NOTE: if new block is empty, then it will get the upper half of old block.
1101  */
1102 STATIC void
xfs_dir_leaf_rebalance(xfs_da_state_t * state,xfs_da_state_blk_t * blk1,xfs_da_state_blk_t * blk2)1103 xfs_dir_leaf_rebalance(xfs_da_state_t *state, xfs_da_state_blk_t *blk1,
1104 				      xfs_da_state_blk_t *blk2)
1105 {
1106 	xfs_da_state_blk_t *tmp_blk;
1107 	xfs_dir_leafblock_t *leaf1, *leaf2;
1108 	xfs_dir_leaf_hdr_t *hdr1, *hdr2;
1109 	int count, totallen, max, space, swap;
1110 
1111 	/*
1112 	 * Set up environment.
1113 	 */
1114 	ASSERT(blk1->magic == XFS_DIR_LEAF_MAGIC);
1115 	ASSERT(blk2->magic == XFS_DIR_LEAF_MAGIC);
1116 	leaf1 = blk1->bp->data;
1117 	leaf2 = blk2->bp->data;
1118 	ASSERT(INT_GET(leaf1->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1119 	ASSERT(INT_GET(leaf2->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1120 
1121 	/*
1122 	 * Check ordering of blocks, reverse if it makes things simpler.
1123 	 */
1124 	swap = 0;
1125 	if (xfs_dir_leaf_order(blk1->bp, blk2->bp)) {
1126 		tmp_blk = blk1;
1127 		blk1 = blk2;
1128 		blk2 = tmp_blk;
1129 		leaf1 = blk1->bp->data;
1130 		leaf2 = blk2->bp->data;
1131 		swap = 1;
1132 	}
1133 	hdr1 = &leaf1->hdr;
1134 	hdr2 = &leaf2->hdr;
1135 
1136 	/*
1137 	 * Examine entries until we reduce the absolute difference in
1138 	 * byte usage between the two blocks to a minimum.  Then get
1139 	 * the direction to copy and the number of elements to move.
1140 	 */
1141 	state->inleaf = xfs_dir_leaf_figure_balance(state, blk1, blk2,
1142 							   &count, &totallen);
1143 	if (swap)
1144 		state->inleaf = !state->inleaf;
1145 
1146 	/*
1147 	 * Move any entries required from leaf to leaf:
1148 	 */
1149 	if (count < INT_GET(hdr1->count, ARCH_CONVERT)) {
1150 		/*
1151 		 * Figure the total bytes to be added to the destination leaf.
1152 		 */
1153 		count = INT_GET(hdr1->count, ARCH_CONVERT) - count;	/* number entries being moved */
1154 		space  = INT_GET(hdr1->namebytes, ARCH_CONVERT) - totallen;
1155 		space += count * ((uint)sizeof(xfs_dir_leaf_name_t)-1);
1156 		space += count * (uint)sizeof(xfs_dir_leaf_entry_t);
1157 
1158 		/*
1159 		 * leaf2 is the destination, compact it if it looks tight.
1160 		 */
1161 		max  = INT_GET(hdr2->firstused, ARCH_CONVERT) - (uint)sizeof(xfs_dir_leaf_hdr_t);
1162 		max -= INT_GET(hdr2->count, ARCH_CONVERT) * (uint)sizeof(xfs_dir_leaf_entry_t);
1163 		if (space > max) {
1164 			xfs_dir_leaf_compact(state->args->trans, blk2->bp,
1165 								 0, 0);
1166 		}
1167 
1168 		/*
1169 		 * Move high entries from leaf1 to low end of leaf2.
1170 		 */
1171 		xfs_dir_leaf_moveents(leaf1, INT_GET(hdr1->count, ARCH_CONVERT) - count,
1172 					     leaf2, 0, count, state->mp);
1173 
1174 		xfs_da_log_buf(state->args->trans, blk1->bp, 0,
1175 						   state->blocksize-1);
1176 		xfs_da_log_buf(state->args->trans, blk2->bp, 0,
1177 						   state->blocksize-1);
1178 
1179 	} else if (count > INT_GET(hdr1->count, ARCH_CONVERT)) {
1180 		/*
1181 		 * Figure the total bytes to be added to the destination leaf.
1182 		 */
1183 		count -= INT_GET(hdr1->count, ARCH_CONVERT);		/* number entries being moved */
1184 		space  = totallen - INT_GET(hdr1->namebytes, ARCH_CONVERT);
1185 		space += count * ((uint)sizeof(xfs_dir_leaf_name_t)-1);
1186 		space += count * (uint)sizeof(xfs_dir_leaf_entry_t);
1187 
1188 		/*
1189 		 * leaf1 is the destination, compact it if it looks tight.
1190 		 */
1191 		max  = INT_GET(hdr1->firstused, ARCH_CONVERT) - (uint)sizeof(xfs_dir_leaf_hdr_t);
1192 		max -= INT_GET(hdr1->count, ARCH_CONVERT) * (uint)sizeof(xfs_dir_leaf_entry_t);
1193 		if (space > max) {
1194 			xfs_dir_leaf_compact(state->args->trans, blk1->bp,
1195 								 0, 0);
1196 		}
1197 
1198 		/*
1199 		 * Move low entries from leaf2 to high end of leaf1.
1200 		 */
1201 		xfs_dir_leaf_moveents(leaf2, 0, leaf1, (int)INT_GET(hdr1->count, ARCH_CONVERT),
1202 					     count, state->mp);
1203 
1204 		xfs_da_log_buf(state->args->trans, blk1->bp, 0,
1205 						   state->blocksize-1);
1206 		xfs_da_log_buf(state->args->trans, blk2->bp, 0,
1207 						   state->blocksize-1);
1208 	}
1209 
1210 	/*
1211 	 * Copy out last hashval in each block for B-tree code.
1212 	 */
1213 	blk1->hashval = INT_GET(leaf1->entries[ INT_GET(leaf1->hdr.count, ARCH_CONVERT)-1 ].hashval, ARCH_CONVERT);
1214 	blk2->hashval = INT_GET(leaf2->entries[ INT_GET(leaf2->hdr.count, ARCH_CONVERT)-1 ].hashval, ARCH_CONVERT);
1215 
1216 	/*
1217 	 * Adjust the expected index for insertion.
1218 	 * GROT: this doesn't work unless blk2 was originally empty.
1219 	 */
1220 	if (!state->inleaf) {
1221 		blk2->index = blk1->index - INT_GET(leaf1->hdr.count, ARCH_CONVERT);
1222 	}
1223 }
1224 
1225 /*
1226  * Examine entries until we reduce the absolute difference in
1227  * byte usage between the two blocks to a minimum.
1228  * GROT: Is this really necessary?  With other than a 512 byte blocksize,
1229  * GROT: there will always be enough room in either block for a new entry.
1230  * GROT: Do a double-split for this case?
1231  */
1232 STATIC int
xfs_dir_leaf_figure_balance(xfs_da_state_t * state,xfs_da_state_blk_t * blk1,xfs_da_state_blk_t * blk2,int * countarg,int * namebytesarg)1233 xfs_dir_leaf_figure_balance(xfs_da_state_t *state,
1234 					   xfs_da_state_blk_t *blk1,
1235 					   xfs_da_state_blk_t *blk2,
1236 					   int *countarg, int *namebytesarg)
1237 {
1238 	xfs_dir_leafblock_t *leaf1, *leaf2;
1239 	xfs_dir_leaf_hdr_t *hdr1, *hdr2;
1240 	xfs_dir_leaf_entry_t *entry;
1241 	int count, max, totallen, half;
1242 	int lastdelta, foundit, tmp;
1243 
1244 	/*
1245 	 * Set up environment.
1246 	 */
1247 	leaf1 = blk1->bp->data;
1248 	leaf2 = blk2->bp->data;
1249 	hdr1 = &leaf1->hdr;
1250 	hdr2 = &leaf2->hdr;
1251 	foundit = 0;
1252 	totallen = 0;
1253 
1254 	/*
1255 	 * Examine entries until we reduce the absolute difference in
1256 	 * byte usage between the two blocks to a minimum.
1257 	 */
1258 	max = INT_GET(hdr1->count, ARCH_CONVERT) + INT_GET(hdr2->count, ARCH_CONVERT);
1259 	half  = (max+1) * (uint)(sizeof(*entry)+sizeof(xfs_dir_leaf_entry_t)-1);
1260 	half += INT_GET(hdr1->namebytes, ARCH_CONVERT) + INT_GET(hdr2->namebytes, ARCH_CONVERT) + state->args->namelen;
1261 	half /= 2;
1262 	lastdelta = state->blocksize;
1263 	entry = &leaf1->entries[0];
1264 	for (count = 0; count < max; entry++, count++) {
1265 
1266 #define XFS_DIR_ABS(A)	(((A) < 0) ? -(A) : (A))
1267 		/*
1268 		 * The new entry is in the first block, account for it.
1269 		 */
1270 		if (count == blk1->index) {
1271 			tmp = totallen + (uint)sizeof(*entry)
1272 				+ XFS_DIR_LEAF_ENTSIZE_BYNAME(state->args->namelen);
1273 			if (XFS_DIR_ABS(half - tmp) > lastdelta)
1274 				break;
1275 			lastdelta = XFS_DIR_ABS(half - tmp);
1276 			totallen = tmp;
1277 			foundit = 1;
1278 		}
1279 
1280 		/*
1281 		 * Wrap around into the second block if necessary.
1282 		 */
1283 		if (count == INT_GET(hdr1->count, ARCH_CONVERT)) {
1284 			leaf1 = leaf2;
1285 			entry = &leaf1->entries[0];
1286 		}
1287 
1288 		/*
1289 		 * Figure out if next leaf entry would be too much.
1290 		 */
1291 		tmp = totallen + (uint)sizeof(*entry)
1292 				+ XFS_DIR_LEAF_ENTSIZE_BYENTRY(entry);
1293 		if (XFS_DIR_ABS(half - tmp) > lastdelta)
1294 			break;
1295 		lastdelta = XFS_DIR_ABS(half - tmp);
1296 		totallen = tmp;
1297 #undef XFS_DIR_ABS
1298 	}
1299 
1300 	/*
1301 	 * Calculate the number of namebytes that will end up in lower block.
1302 	 * If new entry not in lower block, fix up the count.
1303 	 */
1304 	totallen -=
1305 		count * (uint)(sizeof(*entry)+sizeof(xfs_dir_leaf_entry_t)-1);
1306 	if (foundit) {
1307 		totallen -= (sizeof(*entry)+sizeof(xfs_dir_leaf_entry_t)-1) +
1308 			    state->args->namelen;
1309 	}
1310 
1311 	*countarg = count;
1312 	*namebytesarg = totallen;
1313 	return(foundit);
1314 }
1315 
1316 /*========================================================================
1317  * Routines used for shrinking the Btree.
1318  *========================================================================*/
1319 
1320 /*
1321  * Check a leaf block and its neighbors to see if the block should be
1322  * collapsed into one or the other neighbor.  Always keep the block
1323  * with the smaller block number.
1324  * If the current block is over 50% full, don't try to join it, return 0.
1325  * If the block is empty, fill in the state structure and return 2.
1326  * If it can be collapsed, fill in the state structure and return 1.
1327  * If nothing can be done, return 0.
1328  */
1329 int
xfs_dir_leaf_toosmall(xfs_da_state_t * state,int * action)1330 xfs_dir_leaf_toosmall(xfs_da_state_t *state, int *action)
1331 {
1332 	xfs_dir_leafblock_t *leaf;
1333 	xfs_da_state_blk_t *blk;
1334 	xfs_da_blkinfo_t *info;
1335 	int count, bytes, forward, error, retval, i;
1336 	xfs_dablk_t blkno;
1337 	xfs_dabuf_t *bp;
1338 
1339 	/*
1340 	 * Check for the degenerate case of the block being over 50% full.
1341 	 * If so, it's not worth even looking to see if we might be able
1342 	 * to coalesce with a sibling.
1343 	 */
1344 	blk = &state->path.blk[ state->path.active-1 ];
1345 	info = blk->bp->data;
1346 	ASSERT(INT_GET(info->magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1347 	leaf = (xfs_dir_leafblock_t *)info;
1348 	count = INT_GET(leaf->hdr.count, ARCH_CONVERT);
1349 	bytes = (uint)sizeof(xfs_dir_leaf_hdr_t) +
1350 		count * (uint)sizeof(xfs_dir_leaf_entry_t) +
1351 		count * ((uint)sizeof(xfs_dir_leaf_name_t)-1) +
1352 		INT_GET(leaf->hdr.namebytes, ARCH_CONVERT);
1353 	if (bytes > (state->blocksize >> 1)) {
1354 		*action = 0;	/* blk over 50%, don't try to join */
1355 		return(0);
1356 	}
1357 
1358 	/*
1359 	 * Check for the degenerate case of the block being empty.
1360 	 * If the block is empty, we'll simply delete it, no need to
1361 	 * coalesce it with a sibling block.  We choose (aribtrarily)
1362 	 * to merge with the forward block unless it is NULL.
1363 	 */
1364 	if (count == 0) {
1365 		/*
1366 		 * Make altpath point to the block we want to keep and
1367 		 * path point to the block we want to drop (this one).
1368 		 */
1369 		forward = !INT_ISZERO(info->forw, ARCH_CONVERT);
1370 		memcpy(&state->altpath, &state->path, sizeof(state->path));
1371 		error = xfs_da_path_shift(state, &state->altpath, forward,
1372 						 0, &retval);
1373 		if (error)
1374 			return(error);
1375 		if (retval) {
1376 			*action = 0;
1377 		} else {
1378 			*action = 2;
1379 		}
1380 		return(0);
1381 	}
1382 
1383 	/*
1384 	 * Examine each sibling block to see if we can coalesce with
1385 	 * at least 25% free space to spare.  We need to figure out
1386 	 * whether to merge with the forward or the backward block.
1387 	 * We prefer coalescing with the lower numbered sibling so as
1388 	 * to shrink a directory over time.
1389 	 */
1390 	forward = (INT_GET(info->forw, ARCH_CONVERT) < INT_GET(info->back, ARCH_CONVERT));	/* start with smaller blk num */
1391 	for (i = 0; i < 2; forward = !forward, i++) {
1392 		if (forward)
1393 			blkno = INT_GET(info->forw, ARCH_CONVERT);
1394 		else
1395 			blkno = INT_GET(info->back, ARCH_CONVERT);
1396 		if (blkno == 0)
1397 			continue;
1398 		error = xfs_da_read_buf(state->args->trans, state->args->dp,
1399 							    blkno, -1, &bp,
1400 							    XFS_DATA_FORK);
1401 		if (error)
1402 			return(error);
1403 		ASSERT(bp != NULL);
1404 
1405 		leaf = (xfs_dir_leafblock_t *)info;
1406 		count  = INT_GET(leaf->hdr.count, ARCH_CONVERT);
1407 		bytes  = state->blocksize - (state->blocksize>>2);
1408 		bytes -= INT_GET(leaf->hdr.namebytes, ARCH_CONVERT);
1409 		leaf = bp->data;
1410 		ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1411 		count += INT_GET(leaf->hdr.count, ARCH_CONVERT);
1412 		bytes -= INT_GET(leaf->hdr.namebytes, ARCH_CONVERT);
1413 		bytes -= count * ((uint)sizeof(xfs_dir_leaf_name_t) - 1);
1414 		bytes -= count * (uint)sizeof(xfs_dir_leaf_entry_t);
1415 		bytes -= (uint)sizeof(xfs_dir_leaf_hdr_t);
1416 		if (bytes >= 0)
1417 			break;	/* fits with at least 25% to spare */
1418 
1419 		xfs_da_brelse(state->args->trans, bp);
1420 	}
1421 	if (i >= 2) {
1422 		*action = 0;
1423 		return(0);
1424 	}
1425 	xfs_da_buf_done(bp);
1426 
1427 	/*
1428 	 * Make altpath point to the block we want to keep (the lower
1429 	 * numbered block) and path point to the block we want to drop.
1430 	 */
1431 	memcpy(&state->altpath, &state->path, sizeof(state->path));
1432 	if (blkno < blk->blkno) {
1433 		error = xfs_da_path_shift(state, &state->altpath, forward,
1434 						 0, &retval);
1435 	} else {
1436 		error = xfs_da_path_shift(state, &state->path, forward,
1437 						 0, &retval);
1438 	}
1439 	if (error)
1440 		return(error);
1441 	if (retval) {
1442 		*action = 0;
1443 	} else {
1444 		*action = 1;
1445 	}
1446 	return(0);
1447 }
1448 
1449 /*
1450  * Remove a name from the leaf directory structure.
1451  *
1452  * Return 1 if leaf is less than 37% full, 0 if >= 37% full.
1453  * If two leaves are 37% full, when combined they will leave 25% free.
1454  */
1455 int
xfs_dir_leaf_remove(xfs_trans_t * trans,xfs_dabuf_t * bp,int index)1456 xfs_dir_leaf_remove(xfs_trans_t *trans, xfs_dabuf_t *bp, int index)
1457 {
1458 	xfs_dir_leafblock_t *leaf;
1459 	xfs_dir_leaf_hdr_t *hdr;
1460 	xfs_dir_leaf_map_t *map;
1461 	xfs_dir_leaf_entry_t *entry;
1462 	xfs_dir_leaf_name_t *namest;
1463 	int before, after, smallest, entsize;
1464 	int tablesize, tmp, i;
1465 	xfs_mount_t *mp;
1466 
1467 	leaf = bp->data;
1468 	ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1469 	hdr = &leaf->hdr;
1470 	mp = trans->t_mountp;
1471 	ASSERT((INT_GET(hdr->count, ARCH_CONVERT) > 0) && (INT_GET(hdr->count, ARCH_CONVERT) < (XFS_LBSIZE(mp)/8)));
1472 	ASSERT((index >= 0) && (index < INT_GET(hdr->count, ARCH_CONVERT)));
1473 	ASSERT(INT_GET(hdr->firstused, ARCH_CONVERT) >= ((INT_GET(hdr->count, ARCH_CONVERT)*sizeof(*entry))+sizeof(*hdr)));
1474 	entry = &leaf->entries[index];
1475 	ASSERT(INT_GET(entry->nameidx, ARCH_CONVERT) >= INT_GET(hdr->firstused, ARCH_CONVERT));
1476 	ASSERT(INT_GET(entry->nameidx, ARCH_CONVERT) < XFS_LBSIZE(mp));
1477 
1478 	/*
1479 	 * Scan through free region table:
1480 	 *    check for adjacency of free'd entry with an existing one,
1481 	 *    find smallest free region in case we need to replace it,
1482 	 *    adjust any map that borders the entry table,
1483 	 */
1484 	tablesize = INT_GET(hdr->count, ARCH_CONVERT) * (uint)sizeof(xfs_dir_leaf_entry_t)
1485 			+ (uint)sizeof(xfs_dir_leaf_hdr_t);
1486 	map = &hdr->freemap[0];
1487 	tmp = INT_GET(map->size, ARCH_CONVERT);
1488 	before = after = -1;
1489 	smallest = XFS_DIR_LEAF_MAPSIZE - 1;
1490 	entsize = XFS_DIR_LEAF_ENTSIZE_BYENTRY(entry);
1491 	for (i = 0; i < XFS_DIR_LEAF_MAPSIZE; map++, i++) {
1492 		ASSERT(INT_GET(map->base, ARCH_CONVERT) < XFS_LBSIZE(mp));
1493 		ASSERT(INT_GET(map->size, ARCH_CONVERT) < XFS_LBSIZE(mp));
1494 		if (INT_GET(map->base, ARCH_CONVERT) == tablesize) {
1495 			INT_MOD(map->base, ARCH_CONVERT, -((uint)sizeof(xfs_dir_leaf_entry_t)));
1496 			INT_MOD(map->size, ARCH_CONVERT, (uint)sizeof(xfs_dir_leaf_entry_t));
1497 		}
1498 
1499 		if ((INT_GET(map->base, ARCH_CONVERT) + INT_GET(map->size, ARCH_CONVERT)) == INT_GET(entry->nameidx, ARCH_CONVERT)) {
1500 			before = i;
1501 		} else if (INT_GET(map->base, ARCH_CONVERT) == (INT_GET(entry->nameidx, ARCH_CONVERT) + entsize)) {
1502 			after = i;
1503 		} else if (INT_GET(map->size, ARCH_CONVERT) < tmp) {
1504 			tmp = INT_GET(map->size, ARCH_CONVERT);
1505 			smallest = i;
1506 		}
1507 	}
1508 
1509 	/*
1510 	 * Coalesce adjacent freemap regions,
1511 	 * or replace the smallest region.
1512 	 */
1513 	if ((before >= 0) || (after >= 0)) {
1514 		if ((before >= 0) && (after >= 0)) {
1515 			map = &hdr->freemap[before];
1516 			INT_MOD(map->size, ARCH_CONVERT, entsize);
1517 			INT_MOD(map->size, ARCH_CONVERT, INT_GET(hdr->freemap[after].size, ARCH_CONVERT));
1518 			INT_ZERO(hdr->freemap[after].base, ARCH_CONVERT);
1519 			INT_ZERO(hdr->freemap[after].size, ARCH_CONVERT);
1520 		} else if (before >= 0) {
1521 			map = &hdr->freemap[before];
1522 			INT_MOD(map->size, ARCH_CONVERT, entsize);
1523 		} else {
1524 			map = &hdr->freemap[after];
1525 			INT_COPY(map->base, entry->nameidx, ARCH_CONVERT);
1526 			INT_MOD(map->size, ARCH_CONVERT, entsize);
1527 		}
1528 	} else {
1529 		/*
1530 		 * Replace smallest region (if it is smaller than free'd entry)
1531 		 */
1532 		map = &hdr->freemap[smallest];
1533 		if (INT_GET(map->size, ARCH_CONVERT) < entsize) {
1534 			INT_COPY(map->base, entry->nameidx, ARCH_CONVERT);
1535 			INT_SET(map->size, ARCH_CONVERT, entsize);
1536 		}
1537 	}
1538 
1539 	/*
1540 	 * Did we remove the first entry?
1541 	 */
1542 	if (INT_GET(entry->nameidx, ARCH_CONVERT) == INT_GET(hdr->firstused, ARCH_CONVERT))
1543 		smallest = 1;
1544 	else
1545 		smallest = 0;
1546 
1547 	/*
1548 	 * Compress the remaining entries and zero out the removed stuff.
1549 	 */
1550 	namest = XFS_DIR_LEAF_NAMESTRUCT(leaf, INT_GET(entry->nameidx, ARCH_CONVERT));
1551 	memset((char *)namest, 0, entsize);
1552 	xfs_da_log_buf(trans, bp, XFS_DA_LOGRANGE(leaf, namest, entsize));
1553 
1554 	INT_MOD(hdr->namebytes, ARCH_CONVERT, -(entry->namelen));
1555 	tmp = (INT_GET(hdr->count, ARCH_CONVERT) - index) * (uint)sizeof(xfs_dir_leaf_entry_t);
1556 	memmove(entry, entry + 1, tmp);
1557 	INT_MOD(hdr->count, ARCH_CONVERT, -1);
1558 	xfs_da_log_buf(trans, bp,
1559 	    XFS_DA_LOGRANGE(leaf, entry, tmp + (uint)sizeof(*entry)));
1560 	entry = &leaf->entries[INT_GET(hdr->count, ARCH_CONVERT)];
1561 	memset((char *)entry, 0, sizeof(xfs_dir_leaf_entry_t));
1562 
1563 	/*
1564 	 * If we removed the first entry, re-find the first used byte
1565 	 * in the name area.  Note that if the entry was the "firstused",
1566 	 * then we don't have a "hole" in our block resulting from
1567 	 * removing the name.
1568 	 */
1569 	if (smallest) {
1570 		tmp = XFS_LBSIZE(mp);
1571 		entry = &leaf->entries[0];
1572 		for (i = INT_GET(hdr->count, ARCH_CONVERT)-1; i >= 0; entry++, i--) {
1573 			ASSERT(INT_GET(entry->nameidx, ARCH_CONVERT) >= INT_GET(hdr->firstused, ARCH_CONVERT));
1574 			ASSERT(INT_GET(entry->nameidx, ARCH_CONVERT) < XFS_LBSIZE(mp));
1575 			if (INT_GET(entry->nameidx, ARCH_CONVERT) < tmp)
1576 				tmp = INT_GET(entry->nameidx, ARCH_CONVERT);
1577 		}
1578 		INT_SET(hdr->firstused, ARCH_CONVERT, tmp);
1579 		if (INT_ISZERO(hdr->firstused, ARCH_CONVERT))
1580 			INT_SET(hdr->firstused, ARCH_CONVERT, tmp - 1);
1581 	} else {
1582 		hdr->holes = 1;		/* mark as needing compaction */
1583 	}
1584 
1585 	xfs_da_log_buf(trans, bp, XFS_DA_LOGRANGE(leaf, hdr, sizeof(*hdr)));
1586 
1587 	/*
1588 	 * Check if leaf is less than 50% full, caller may want to
1589 	 * "join" the leaf with a sibling if so.
1590 	 */
1591 	tmp  = (uint)sizeof(xfs_dir_leaf_hdr_t);
1592 	tmp += INT_GET(leaf->hdr.count, ARCH_CONVERT) * (uint)sizeof(xfs_dir_leaf_entry_t);
1593 	tmp += INT_GET(leaf->hdr.count, ARCH_CONVERT) * ((uint)sizeof(xfs_dir_leaf_name_t) - 1);
1594 	tmp += INT_GET(leaf->hdr.namebytes, ARCH_CONVERT);
1595 	if (tmp < mp->m_dir_magicpct)
1596 		return(1);			/* leaf is < 37% full */
1597 	return(0);
1598 }
1599 
1600 /*
1601  * Move all the directory entries from drop_leaf into save_leaf.
1602  */
1603 void
xfs_dir_leaf_unbalance(xfs_da_state_t * state,xfs_da_state_blk_t * drop_blk,xfs_da_state_blk_t * save_blk)1604 xfs_dir_leaf_unbalance(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
1605 				      xfs_da_state_blk_t *save_blk)
1606 {
1607 	xfs_dir_leafblock_t *drop_leaf, *save_leaf, *tmp_leaf;
1608 	xfs_dir_leaf_hdr_t *drop_hdr, *save_hdr, *tmp_hdr;
1609 	xfs_mount_t *mp;
1610 	char *tmpbuffer;
1611 
1612 	/*
1613 	 * Set up environment.
1614 	 */
1615 	mp = state->mp;
1616 	ASSERT(drop_blk->magic == XFS_DIR_LEAF_MAGIC);
1617 	ASSERT(save_blk->magic == XFS_DIR_LEAF_MAGIC);
1618 	drop_leaf = drop_blk->bp->data;
1619 	save_leaf = save_blk->bp->data;
1620 	ASSERT(INT_GET(drop_leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1621 	ASSERT(INT_GET(save_leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1622 	drop_hdr = &drop_leaf->hdr;
1623 	save_hdr = &save_leaf->hdr;
1624 
1625 	/*
1626 	 * Save last hashval from dying block for later Btree fixup.
1627 	 */
1628 	drop_blk->hashval = INT_GET(drop_leaf->entries[ drop_leaf->hdr.count-1 ].hashval, ARCH_CONVERT);
1629 
1630 	/*
1631 	 * Check if we need a temp buffer, or can we do it in place.
1632 	 * Note that we don't check "leaf" for holes because we will
1633 	 * always be dropping it, toosmall() decided that for us already.
1634 	 */
1635 	if (save_hdr->holes == 0) {
1636 		/*
1637 		 * dest leaf has no holes, so we add there.  May need
1638 		 * to make some room in the entry array.
1639 		 */
1640 		if (xfs_dir_leaf_order(save_blk->bp, drop_blk->bp)) {
1641 			xfs_dir_leaf_moveents(drop_leaf, 0, save_leaf, 0,
1642 						 (int)INT_GET(drop_hdr->count, ARCH_CONVERT), mp);
1643 		} else {
1644 			xfs_dir_leaf_moveents(drop_leaf, 0,
1645 					      save_leaf, INT_GET(save_hdr->count, ARCH_CONVERT),
1646 					      (int)INT_GET(drop_hdr->count, ARCH_CONVERT), mp);
1647 		}
1648 	} else {
1649 		/*
1650 		 * Destination has holes, so we make a temporary copy
1651 		 * of the leaf and add them both to that.
1652 		 */
1653 		tmpbuffer = kmem_alloc(state->blocksize, KM_SLEEP);
1654 		ASSERT(tmpbuffer != NULL);
1655 		memset(tmpbuffer, 0, state->blocksize);
1656 		tmp_leaf = (xfs_dir_leafblock_t *)tmpbuffer;
1657 		tmp_hdr = &tmp_leaf->hdr;
1658 		tmp_hdr->info = save_hdr->info;	/* struct copy */
1659 		INT_ZERO(tmp_hdr->count, ARCH_CONVERT);
1660 		INT_SET(tmp_hdr->firstused, ARCH_CONVERT, state->blocksize);
1661 		if (INT_ISZERO(tmp_hdr->firstused, ARCH_CONVERT))
1662 			INT_SET(tmp_hdr->firstused, ARCH_CONVERT, state->blocksize - 1);
1663 		INT_ZERO(tmp_hdr->namebytes, ARCH_CONVERT);
1664 		if (xfs_dir_leaf_order(save_blk->bp, drop_blk->bp)) {
1665 			xfs_dir_leaf_moveents(drop_leaf, 0, tmp_leaf, 0,
1666 						 (int)INT_GET(drop_hdr->count, ARCH_CONVERT), mp);
1667 			xfs_dir_leaf_moveents(save_leaf, 0,
1668 					      tmp_leaf, INT_GET(tmp_leaf->hdr.count, ARCH_CONVERT),
1669 					      (int)INT_GET(save_hdr->count, ARCH_CONVERT), mp);
1670 		} else {
1671 			xfs_dir_leaf_moveents(save_leaf, 0, tmp_leaf, 0,
1672 						 (int)INT_GET(save_hdr->count, ARCH_CONVERT), mp);
1673 			xfs_dir_leaf_moveents(drop_leaf, 0,
1674 					      tmp_leaf, INT_GET(tmp_leaf->hdr.count, ARCH_CONVERT),
1675 					      (int)INT_GET(drop_hdr->count, ARCH_CONVERT), mp);
1676 		}
1677 		memcpy(save_leaf, tmp_leaf, state->blocksize);
1678 		kmem_free(tmpbuffer, state->blocksize);
1679 	}
1680 
1681 	xfs_da_log_buf(state->args->trans, save_blk->bp, 0,
1682 					   state->blocksize - 1);
1683 
1684 	/*
1685 	 * Copy out last hashval in each block for B-tree code.
1686 	 */
1687 	save_blk->hashval = INT_GET(save_leaf->entries[ INT_GET(save_leaf->hdr.count, ARCH_CONVERT)-1 ].hashval, ARCH_CONVERT);
1688 }
1689 
1690 /*========================================================================
1691  * Routines used for finding things in the Btree.
1692  *========================================================================*/
1693 
1694 /*
1695  * Look up a name in a leaf directory structure.
1696  * This is the internal routine, it uses the caller's buffer.
1697  *
1698  * Note that duplicate keys are allowed, but only check within the
1699  * current leaf node.  The Btree code must check in adjacent leaf nodes.
1700  *
1701  * Return in *index the index into the entry[] array of either the found
1702  * entry, or where the entry should have been (insert before that entry).
1703  *
1704  * Don't change the args->inumber unless we find the filename.
1705  */
1706 int
xfs_dir_leaf_lookup_int(xfs_dabuf_t * bp,xfs_da_args_t * args,int * index)1707 xfs_dir_leaf_lookup_int(xfs_dabuf_t *bp, xfs_da_args_t *args, int *index)
1708 {
1709 	xfs_dir_leafblock_t *leaf;
1710 	xfs_dir_leaf_entry_t *entry;
1711 	xfs_dir_leaf_name_t *namest;
1712 	int probe, span;
1713 	xfs_dahash_t hashval;
1714 
1715 	leaf = bp->data;
1716 	ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1717 	ASSERT(INT_GET(leaf->hdr.count, ARCH_CONVERT) < (XFS_LBSIZE(args->dp->i_mount)/8));
1718 
1719 	/*
1720 	 * Binary search.  (note: small blocks will skip this loop)
1721 	 */
1722 	hashval = args->hashval;
1723 	probe = span = INT_GET(leaf->hdr.count, ARCH_CONVERT) / 2;
1724 	for (entry = &leaf->entries[probe]; span > 4;
1725 		   entry = &leaf->entries[probe]) {
1726 		span /= 2;
1727 		if (INT_GET(entry->hashval, ARCH_CONVERT) < hashval)
1728 			probe += span;
1729 		else if (INT_GET(entry->hashval, ARCH_CONVERT) > hashval)
1730 			probe -= span;
1731 		else
1732 			break;
1733 	}
1734 	ASSERT((probe >= 0) && \
1735 	       ((INT_ISZERO(leaf->hdr.count, ARCH_CONVERT)) || (probe < INT_GET(leaf->hdr.count, ARCH_CONVERT))));
1736 	ASSERT((span <= 4) || (INT_GET(entry->hashval, ARCH_CONVERT) == hashval));
1737 
1738 	/*
1739 	 * Since we may have duplicate hashval's, find the first matching
1740 	 * hashval in the leaf.
1741 	 */
1742 	while ((probe > 0) && (INT_GET(entry->hashval, ARCH_CONVERT) >= hashval)) {
1743 		entry--;
1744 		probe--;
1745 	}
1746 	while ((probe < INT_GET(leaf->hdr.count, ARCH_CONVERT)) && (INT_GET(entry->hashval, ARCH_CONVERT) < hashval)) {
1747 		entry++;
1748 		probe++;
1749 	}
1750 	if ((probe == INT_GET(leaf->hdr.count, ARCH_CONVERT)) || (INT_GET(entry->hashval, ARCH_CONVERT) != hashval)) {
1751 		*index = probe;
1752 		ASSERT(args->oknoent);
1753 		return(XFS_ERROR(ENOENT));
1754 	}
1755 
1756 	/*
1757 	 * Duplicate keys may be present, so search all of them for a match.
1758 	 */
1759 	while ((probe < INT_GET(leaf->hdr.count, ARCH_CONVERT)) && (INT_GET(entry->hashval, ARCH_CONVERT) == hashval)) {
1760 		namest = XFS_DIR_LEAF_NAMESTRUCT(leaf, INT_GET(entry->nameidx, ARCH_CONVERT));
1761 		if (entry->namelen == args->namelen &&
1762 		    namest->name[0] == args->name[0] &&
1763 		    memcmp(args->name, namest->name, args->namelen) == 0) {
1764 			XFS_DIR_SF_GET_DIRINO_ARCH(&namest->inumber, &args->inumber, ARCH_CONVERT);
1765 			*index = probe;
1766 			return(XFS_ERROR(EEXIST));
1767 		}
1768 		entry++;
1769 		probe++;
1770 	}
1771 	*index = probe;
1772 	ASSERT(probe == INT_GET(leaf->hdr.count, ARCH_CONVERT) || args->oknoent);
1773 	return(XFS_ERROR(ENOENT));
1774 }
1775 
1776 /*========================================================================
1777  * Utility routines.
1778  *========================================================================*/
1779 
1780 /*
1781  * Move the indicated entries from one leaf to another.
1782  * NOTE: this routine modifies both source and destination leaves.
1783  */
1784 /* ARGSUSED */
1785 STATIC void
xfs_dir_leaf_moveents(xfs_dir_leafblock_t * leaf_s,int start_s,xfs_dir_leafblock_t * leaf_d,int start_d,int count,xfs_mount_t * mp)1786 xfs_dir_leaf_moveents(xfs_dir_leafblock_t *leaf_s, int start_s,
1787 		      xfs_dir_leafblock_t *leaf_d, int start_d,
1788 		      int count, xfs_mount_t *mp)
1789 {
1790 	xfs_dir_leaf_hdr_t *hdr_s, *hdr_d;
1791 	xfs_dir_leaf_entry_t *entry_s, *entry_d;
1792 	int tmp, i;
1793 
1794 	/*
1795 	 * Check for nothing to do.
1796 	 */
1797 	if (count == 0)
1798 		return;
1799 
1800 	/*
1801 	 * Set up environment.
1802 	 */
1803 	ASSERT(INT_GET(leaf_s->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1804 	ASSERT(INT_GET(leaf_d->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1805 	hdr_s = &leaf_s->hdr;
1806 	hdr_d = &leaf_d->hdr;
1807 	ASSERT((INT_GET(hdr_s->count, ARCH_CONVERT) > 0) && (INT_GET(hdr_s->count, ARCH_CONVERT) < (XFS_LBSIZE(mp)/8)));
1808 	ASSERT(INT_GET(hdr_s->firstused, ARCH_CONVERT) >=
1809 		((INT_GET(hdr_s->count, ARCH_CONVERT)*sizeof(*entry_s))+sizeof(*hdr_s)));
1810 	ASSERT(INT_GET(hdr_d->count, ARCH_CONVERT) < (XFS_LBSIZE(mp)/8));
1811 	ASSERT(INT_GET(hdr_d->firstused, ARCH_CONVERT) >=
1812 		((INT_GET(hdr_d->count, ARCH_CONVERT)*sizeof(*entry_d))+sizeof(*hdr_d)));
1813 
1814 	ASSERT(start_s < INT_GET(hdr_s->count, ARCH_CONVERT));
1815 	ASSERT(start_d <= INT_GET(hdr_d->count, ARCH_CONVERT));
1816 	ASSERT(count <= INT_GET(hdr_s->count, ARCH_CONVERT));
1817 
1818 	/*
1819 	 * Move the entries in the destination leaf up to make a hole?
1820 	 */
1821 	if (start_d < INT_GET(hdr_d->count, ARCH_CONVERT)) {
1822 		tmp  = INT_GET(hdr_d->count, ARCH_CONVERT) - start_d;
1823 		tmp *= (uint)sizeof(xfs_dir_leaf_entry_t);
1824 		entry_s = &leaf_d->entries[start_d];
1825 		entry_d = &leaf_d->entries[start_d + count];
1826 		memcpy(entry_d, entry_s, tmp);
1827 	}
1828 
1829 	/*
1830 	 * Copy all entry's in the same (sorted) order,
1831 	 * but allocate filenames packed and in sequence.
1832 	 */
1833 	entry_s = &leaf_s->entries[start_s];
1834 	entry_d = &leaf_d->entries[start_d];
1835 	for (i = 0; i < count; entry_s++, entry_d++, i++) {
1836 		ASSERT(INT_GET(entry_s->nameidx, ARCH_CONVERT) >= INT_GET(hdr_s->firstused, ARCH_CONVERT));
1837 		tmp = XFS_DIR_LEAF_ENTSIZE_BYENTRY(entry_s);
1838 		INT_MOD(hdr_d->firstused, ARCH_CONVERT, -(tmp));
1839 		entry_d->hashval = entry_s->hashval; /* INT_: direct copy */
1840 		INT_COPY(entry_d->nameidx, hdr_d->firstused, ARCH_CONVERT);
1841 		entry_d->namelen = entry_s->namelen;
1842 		ASSERT(INT_GET(entry_d->nameidx, ARCH_CONVERT) + tmp <= XFS_LBSIZE(mp));
1843 		memcpy(XFS_DIR_LEAF_NAMESTRUCT(leaf_d, INT_GET(entry_d->nameidx, ARCH_CONVERT)),
1844 		       XFS_DIR_LEAF_NAMESTRUCT(leaf_s, INT_GET(entry_s->nameidx, ARCH_CONVERT)), tmp);
1845 		ASSERT(INT_GET(entry_s->nameidx, ARCH_CONVERT) + tmp <= XFS_LBSIZE(mp));
1846 		memset((char *)XFS_DIR_LEAF_NAMESTRUCT(leaf_s, INT_GET(entry_s->nameidx, ARCH_CONVERT)),
1847 		      0, tmp);
1848 		INT_MOD(hdr_s->namebytes, ARCH_CONVERT, -(entry_d->namelen));
1849 		INT_MOD(hdr_d->namebytes, ARCH_CONVERT, entry_d->namelen);
1850 		INT_MOD(hdr_s->count, ARCH_CONVERT, -1);
1851 		INT_MOD(hdr_d->count, ARCH_CONVERT, +1);
1852 		tmp  = INT_GET(hdr_d->count, ARCH_CONVERT) * (uint)sizeof(xfs_dir_leaf_entry_t)
1853 				+ (uint)sizeof(xfs_dir_leaf_hdr_t);
1854 		ASSERT(INT_GET(hdr_d->firstused, ARCH_CONVERT) >= tmp);
1855 
1856 	}
1857 
1858 	/*
1859 	 * Zero out the entries we just copied.
1860 	 */
1861 	if (start_s == INT_GET(hdr_s->count, ARCH_CONVERT)) {
1862 		tmp = count * (uint)sizeof(xfs_dir_leaf_entry_t);
1863 		entry_s = &leaf_s->entries[start_s];
1864 		ASSERT((char *)entry_s + tmp <= (char *)leaf_s + XFS_LBSIZE(mp));
1865 		memset((char *)entry_s, 0, tmp);
1866 	} else {
1867 		/*
1868 		 * Move the remaining entries down to fill the hole,
1869 		 * then zero the entries at the top.
1870 		 */
1871 		tmp  = INT_GET(hdr_s->count, ARCH_CONVERT) - count;
1872 		tmp *= (uint)sizeof(xfs_dir_leaf_entry_t);
1873 		entry_s = &leaf_s->entries[start_s + count];
1874 		entry_d = &leaf_s->entries[start_s];
1875 		memcpy(entry_d, entry_s, tmp);
1876 
1877 		tmp = count * (uint)sizeof(xfs_dir_leaf_entry_t);
1878 		entry_s = &leaf_s->entries[INT_GET(hdr_s->count, ARCH_CONVERT)];
1879 		ASSERT((char *)entry_s + tmp <= (char *)leaf_s + XFS_LBSIZE(mp));
1880 		memset((char *)entry_s, 0, tmp);
1881 	}
1882 
1883 	/*
1884 	 * Fill in the freemap information
1885 	 */
1886 	INT_SET(hdr_d->freemap[0].base, ARCH_CONVERT, (uint)sizeof(xfs_dir_leaf_hdr_t));
1887 	INT_MOD(hdr_d->freemap[0].base, ARCH_CONVERT, INT_GET(hdr_d->count, ARCH_CONVERT) * (uint)sizeof(xfs_dir_leaf_entry_t));
1888 	INT_SET(hdr_d->freemap[0].size, ARCH_CONVERT, INT_GET(hdr_d->firstused, ARCH_CONVERT) - INT_GET(hdr_d->freemap[0].base, ARCH_CONVERT));
1889 	INT_SET(hdr_d->freemap[1].base, ARCH_CONVERT, INT_ZERO(hdr_d->freemap[2].base, ARCH_CONVERT));
1890 	INT_SET(hdr_d->freemap[1].size, ARCH_CONVERT, INT_ZERO(hdr_d->freemap[2].size, ARCH_CONVERT));
1891 	hdr_s->holes = 1;	/* leaf may not be compact */
1892 }
1893 
1894 /*
1895  * Compare two leaf blocks "order".
1896  */
1897 int
xfs_dir_leaf_order(xfs_dabuf_t * leaf1_bp,xfs_dabuf_t * leaf2_bp)1898 xfs_dir_leaf_order(xfs_dabuf_t *leaf1_bp, xfs_dabuf_t *leaf2_bp)
1899 {
1900 	xfs_dir_leafblock_t *leaf1, *leaf2;
1901 
1902 	leaf1 = leaf1_bp->data;
1903 	leaf2 = leaf2_bp->data;
1904 	ASSERT((INT_GET(leaf1->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC) &&
1905 	       (INT_GET(leaf2->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC));
1906 	if ((INT_GET(leaf1->hdr.count, ARCH_CONVERT) > 0) && (INT_GET(leaf2->hdr.count, ARCH_CONVERT) > 0) &&
1907 	    ((INT_GET(leaf2->entries[ 0 ].hashval, ARCH_CONVERT) <
1908 	      INT_GET(leaf1->entries[ 0 ].hashval, ARCH_CONVERT)) ||
1909 	     (INT_GET(leaf2->entries[ INT_GET(leaf2->hdr.count, ARCH_CONVERT)-1 ].hashval, ARCH_CONVERT) <
1910 	      INT_GET(leaf1->entries[ INT_GET(leaf1->hdr.count, ARCH_CONVERT)-1 ].hashval, ARCH_CONVERT)))) {
1911 		return(1);
1912 	}
1913 	return(0);
1914 }
1915 
1916 /*
1917  * Pick up the last hashvalue from a leaf block.
1918  */
1919 xfs_dahash_t
xfs_dir_leaf_lasthash(xfs_dabuf_t * bp,int * count)1920 xfs_dir_leaf_lasthash(xfs_dabuf_t *bp, int *count)
1921 {
1922 	xfs_dir_leafblock_t *leaf;
1923 
1924 	leaf = bp->data;
1925 	ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC);
1926 	if (count)
1927 		*count = INT_GET(leaf->hdr.count, ARCH_CONVERT);
1928 	if (INT_ISZERO(leaf->hdr.count, ARCH_CONVERT))
1929 		return(0);
1930 	return(INT_GET(leaf->entries[ INT_GET(leaf->hdr.count, ARCH_CONVERT)-1 ].hashval, ARCH_CONVERT));
1931 }
1932 
1933 /*
1934  * Copy out directory entries for getdents(), for leaf directories.
1935  */
1936 int
xfs_dir_leaf_getdents_int(xfs_dabuf_t * bp,xfs_inode_t * dp,xfs_dablk_t bno,uio_t * uio,int * eobp,xfs_dirent_t * dbp,xfs_dir_put_t put,xfs_daddr_t nextda)1937 xfs_dir_leaf_getdents_int(
1938 	xfs_dabuf_t	*bp,
1939 	xfs_inode_t	*dp,
1940 	xfs_dablk_t	bno,
1941 	uio_t		*uio,
1942 	int		*eobp,
1943 	xfs_dirent_t	*dbp,
1944 	xfs_dir_put_t	put,
1945 	xfs_daddr_t		nextda)
1946 {
1947 	xfs_dir_leafblock_t	*leaf;
1948 	xfs_dir_leaf_entry_t	*entry;
1949 	xfs_dir_leaf_name_t	*namest;
1950 	int			entno, want_entno, i, nextentno;
1951 	xfs_mount_t		*mp;
1952 	xfs_dahash_t		cookhash;
1953 	xfs_dahash_t		nexthash = 0;
1954 #if (BITS_PER_LONG == 32)
1955 	xfs_dahash_t		lasthash = XFS_DA_MAXHASH;
1956 #endif
1957 	xfs_dir_put_args_t	p;
1958 
1959 	mp = dp->i_mount;
1960 	leaf = bp->data;
1961 	if (INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) != XFS_DIR_LEAF_MAGIC) {
1962 		*eobp = 1;
1963 		return(XFS_ERROR(ENOENT));	/* XXX wrong code */
1964 	}
1965 
1966 	want_entno = XFS_DA_COOKIE_ENTRY(mp, uio->uio_offset);
1967 
1968 	cookhash = XFS_DA_COOKIE_HASH(mp, uio->uio_offset);
1969 
1970 	xfs_dir_trace_g_dul("leaf: start", dp, uio, leaf);
1971 
1972 	/*
1973 	 * Re-find our place.
1974 	 */
1975 	for (i = entno = 0, entry = &leaf->entries[0];
1976 		     i < INT_GET(leaf->hdr.count, ARCH_CONVERT);
1977 			     entry++, i++) {
1978 
1979 		namest = XFS_DIR_LEAF_NAMESTRUCT(leaf,
1980 				    INT_GET(entry->nameidx, ARCH_CONVERT));
1981 
1982 		if (unlikely(
1983 		    ((char *)namest < (char *)leaf) ||
1984 		    ((char *)namest >= (char *)leaf + XFS_LBSIZE(mp)))) {
1985 			XFS_CORRUPTION_ERROR("xfs_dir_leaf_getdents_int(1)",
1986 					     XFS_ERRLEVEL_LOW, mp, leaf);
1987 			xfs_dir_trace_g_du("leaf: corrupted", dp, uio);
1988 			return XFS_ERROR(EFSCORRUPTED);
1989 		}
1990 		if (INT_GET(entry->hashval, ARCH_CONVERT) >= cookhash) {
1991 			if (   entno < want_entno
1992 			    && INT_GET(entry->hashval, ARCH_CONVERT)
1993 							== cookhash) {
1994 				/*
1995 				 * Trying to get to a particular offset in a
1996 				 * run of equal-hashval entries.
1997 				 */
1998 				entno++;
1999 			} else if (   want_entno > 0
2000 				   && entno == want_entno
2001 				   && INT_GET(entry->hashval, ARCH_CONVERT)
2002 							== cookhash) {
2003 				break;
2004 			} else {
2005 				entno = 0;
2006 				break;
2007 			}
2008 		}
2009 	}
2010 
2011 	if (i == INT_GET(leaf->hdr.count, ARCH_CONVERT)) {
2012 		xfs_dir_trace_g_du("leaf: hash not found", dp, uio);
2013 		if (!INT_GET(leaf->hdr.info.forw, ARCH_CONVERT))
2014 			uio->uio_offset =
2015 				XFS_DA_MAKE_COOKIE(mp, 0, 0, XFS_DA_MAXHASH);
2016 		/*
2017 		 * Don't set uio_offset if there's another block:
2018 		 * the node code will be setting uio_offset anyway.
2019 		 */
2020 		*eobp = 0;
2021 		return(0);
2022 	}
2023 	xfs_dir_trace_g_due("leaf: hash found", dp, uio, entry);
2024 
2025 	p.dbp = dbp;
2026 	p.put = put;
2027 	p.uio = uio;
2028 
2029 	/*
2030 	 * We're synchronized, start copying entries out to the user.
2031 	 */
2032 	for (; entno >= 0 && i < INT_GET(leaf->hdr.count, ARCH_CONVERT);
2033 			     entry++, i++, (entno = nextentno)) {
2034 		int lastresid=0, retval;
2035 		xfs_dircook_t lastoffset;
2036 		xfs_dahash_t thishash;
2037 
2038 		/*
2039 		 * Check for a damaged directory leaf block and pick up
2040 		 * the inode number from this entry.
2041 		 */
2042 		namest = XFS_DIR_LEAF_NAMESTRUCT(leaf,
2043 				    INT_GET(entry->nameidx, ARCH_CONVERT));
2044 
2045 		if (unlikely(
2046 		    ((char *)namest < (char *)leaf) ||
2047 		    ((char *)namest >= (char *)leaf + XFS_LBSIZE(mp)))) {
2048 			XFS_CORRUPTION_ERROR("xfs_dir_leaf_getdents_int(2)",
2049 					     XFS_ERRLEVEL_LOW, mp, leaf);
2050 			xfs_dir_trace_g_du("leaf: corrupted", dp, uio);
2051 			return XFS_ERROR(EFSCORRUPTED);
2052 		}
2053 
2054 		xfs_dir_trace_g_duc("leaf: middle cookie  ",
2055 						   dp, uio, p.cook.o);
2056 
2057 		if (i < (INT_GET(leaf->hdr.count, ARCH_CONVERT) - 1)) {
2058 			nexthash = INT_GET(entry[1].hashval, ARCH_CONVERT);
2059 
2060 			if (nexthash == INT_GET(entry->hashval, ARCH_CONVERT))
2061 				nextentno = entno + 1;
2062 			else
2063 				nextentno = 0;
2064 			XFS_PUT_COOKIE(p.cook, mp, bno, nextentno, nexthash);
2065 			xfs_dir_trace_g_duc("leaf: middle cookie  ",
2066 						   dp, uio, p.cook.o);
2067 
2068 		} else if ((thishash = INT_GET(leaf->hdr.info.forw,
2069 							ARCH_CONVERT))) {
2070 			xfs_dabuf_t *bp2;
2071 			xfs_dir_leafblock_t *leaf2;
2072 
2073 			ASSERT(nextda != -1);
2074 
2075 			retval = xfs_da_read_buf(dp->i_transp, dp, thishash,
2076 						 nextda, &bp2, XFS_DATA_FORK);
2077 			if (retval)
2078 				return(retval);
2079 
2080 			ASSERT(bp2 != NULL);
2081 
2082 			leaf2 = bp2->data;
2083 
2084 			if (unlikely(
2085 			       (INT_GET(leaf2->hdr.info.magic, ARCH_CONVERT)
2086 						!= XFS_DIR_LEAF_MAGIC)
2087 			    || (INT_GET(leaf2->hdr.info.back, ARCH_CONVERT)
2088 						!= bno))) {	/* GROT */
2089 				XFS_CORRUPTION_ERROR("xfs_dir_leaf_getdents_int(3)",
2090 						     XFS_ERRLEVEL_LOW, mp,
2091 						     leaf2);
2092 				xfs_da_brelse(dp->i_transp, bp2);
2093 
2094 				return(XFS_ERROR(EFSCORRUPTED));
2095 			}
2096 
2097 			nexthash = INT_GET(leaf2->entries[0].hashval,
2098 								ARCH_CONVERT);
2099 			nextentno = -1;
2100 			XFS_PUT_COOKIE(p.cook, mp, thishash, 0, nexthash);
2101 			xfs_da_brelse(dp->i_transp, bp2);
2102 			xfs_dir_trace_g_duc("leaf: next blk cookie",
2103 						   dp, uio, p.cook.o);
2104 		} else {
2105 			nextentno = -1;
2106 			XFS_PUT_COOKIE(p.cook, mp, 0, 0, XFS_DA_MAXHASH);
2107 		}
2108 
2109 		/*
2110 		 * Save off the cookie so we can fall back should the
2111 		 * 'put' into the outgoing buffer fails.  To handle a run
2112 		 * of equal-hashvals, the off_t structure on 64bit
2113 		 * builds has entno built into the cookie to ID the
2114 		 * entry.  On 32bit builds, we only have space for the
2115 		 * hashval so we can't ID specific entries within a group
2116 		 * of same hashval entries.   For this, lastoffset is set
2117 		 * to the first in the run of equal hashvals so we don't
2118 		 * include any entries unless we can include all entries
2119 		 * that share the same hashval.  Hopefully the buffer
2120 		 * provided is big enough to handle it (see pv763517).
2121 		 */
2122 #if (BITS_PER_LONG == 32)
2123 		if ((thishash = INT_GET(entry->hashval, ARCH_CONVERT))
2124 								!= lasthash) {
2125 			XFS_PUT_COOKIE(lastoffset, mp, bno, entno, thishash);
2126 			lastresid = uio->uio_resid;
2127 			lasthash = thishash;
2128 		} else {
2129 			xfs_dir_trace_g_duc("leaf: DUP COOKIES, skipped",
2130 						   dp, uio, p.cook.o);
2131 		}
2132 #else
2133 		thishash = INT_GET(entry->hashval, ARCH_CONVERT);
2134 		XFS_PUT_COOKIE(lastoffset, mp, bno, entno, thishash);
2135 		lastresid = uio->uio_resid;
2136 #endif /* BITS_PER_LONG == 32 */
2137 
2138 		/*
2139 		 * Put the current entry into the outgoing buffer.  If we fail
2140 		 * then restore the UIO to the first entry in the current
2141 		 * run of equal-hashval entries (probably one 1 entry long).
2142 		 */
2143 		p.ino = XFS_GET_DIR_INO_ARCH(mp, namest->inumber, ARCH_CONVERT);
2144 #if XFS_BIG_INUMS
2145 		p.ino += mp->m_inoadd;
2146 #endif
2147 		p.name = (char *)namest->name;
2148 		p.namelen = entry->namelen;
2149 
2150 		retval = p.put(&p);
2151 
2152 		if (!p.done) {
2153 			uio->uio_offset = lastoffset.o;
2154 			uio->uio_resid = lastresid;
2155 
2156 			*eobp = 1;
2157 
2158 			xfs_dir_trace_g_du("leaf: E-O-B", dp, uio);
2159 
2160 			return(retval);
2161 		}
2162 	}
2163 
2164 	uio->uio_offset = p.cook.o;
2165 
2166 	*eobp = 0;
2167 
2168 	xfs_dir_trace_g_du("leaf: E-O-F", dp, uio);
2169 
2170 	return(0);
2171 }
2172 
2173 /*
2174  * Format a dirent64 structure and copy it out the the user's buffer.
2175  */
2176 int
xfs_dir_put_dirent64_direct(xfs_dir_put_args_t * pa)2177 xfs_dir_put_dirent64_direct(xfs_dir_put_args_t *pa)
2178 {
2179 	iovec_t *iovp;
2180 	int reclen, namelen;
2181 	xfs_dirent_t *idbp;
2182 	uio_t *uio;
2183 
2184 	namelen = pa->namelen;
2185 	reclen = DIRENTSIZE(namelen);
2186 	uio = pa->uio;
2187 	if (reclen > uio->uio_resid) {
2188 		pa->done = 0;
2189 		return 0;
2190 	}
2191 	iovp = uio->uio_iov;
2192 	idbp = (xfs_dirent_t *)iovp->iov_base;
2193 	iovp->iov_base = (char *)idbp + reclen;
2194 	iovp->iov_len -= reclen;
2195 	uio->uio_resid -= reclen;
2196 	idbp->d_reclen = reclen;
2197 	idbp->d_ino = pa->ino;
2198 	idbp->d_off = pa->cook.o;
2199 	idbp->d_name[namelen] = '\0';
2200 	pa->done = 1;
2201 	memcpy(idbp->d_name, pa->name, namelen);
2202 	return 0;
2203 }
2204 
2205 /*
2206  * Format a dirent64 structure and copy it out the the user's buffer.
2207  */
2208 int
xfs_dir_put_dirent64_uio(xfs_dir_put_args_t * pa)2209 xfs_dir_put_dirent64_uio(xfs_dir_put_args_t *pa)
2210 {
2211 	int		retval, reclen, namelen;
2212 	xfs_dirent_t	*idbp;
2213 	uio_t		*uio;
2214 
2215 	namelen = pa->namelen;
2216 	reclen = DIRENTSIZE(namelen);
2217 	uio = pa->uio;
2218 	if (reclen > uio->uio_resid) {
2219 		pa->done = 0;
2220 		return 0;
2221 	}
2222 	idbp = pa->dbp;
2223 	idbp->d_reclen = reclen;
2224 	idbp->d_ino = pa->ino;
2225 	idbp->d_off = pa->cook.o;
2226 	idbp->d_name[namelen] = '\0';
2227 	memcpy(idbp->d_name, pa->name, namelen);
2228 	retval = uio_read((caddr_t)idbp, reclen, uio);
2229 	pa->done = (retval == 0);
2230 	return retval;
2231 }
2232