1/*
2 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17 */
18#include "xfs.h"
19#include "xfs_fs.h"
20#include "xfs_types.h"
21#include "xfs_log.h"
22#include "xfs_inum.h"
23#include "xfs_trans.h"
24#include "xfs_sb.h"
25#include "xfs_dir.h"
26#include "xfs_dir2.h"
27#include "xfs_dmapi.h"
28#include "xfs_mount.h"
29#include "xfs_da_btree.h"
30#include "xfs_bmap_btree.h"
31#include "xfs_alloc_btree.h"
32#include "xfs_ialloc_btree.h"
33#include "xfs_alloc.h"
34#include "xfs_btree.h"
35#include "xfs_dir_sf.h"
36#include "xfs_dir2_sf.h"
37#include "xfs_attr_sf.h"
38#include "xfs_dinode.h"
39#include "xfs_inode.h"
40#include "xfs_bmap.h"
41#include "xfs_dir_leaf.h"
42#include "xfs_error.h"
43
44/*
45 * xfs_dir.c
46 *
47 * Provide the external interfaces to manage directories.
48 */
49
50/*========================================================================
51 * Function prototypes for the kernel.
52 *========================================================================*/
53
54/*
55 * Functions for the dirops interfaces.
56 */
57static void	xfs_dir_mount(struct xfs_mount *mp);
58
59static int	xfs_dir_isempty(struct xfs_inode *dp);
60
61static int	xfs_dir_init(struct xfs_trans *trans,
62			     struct xfs_inode *dir,
63			     struct xfs_inode *parent_dir);
64
65static int	xfs_dir_createname(struct xfs_trans *trans,
66				   struct xfs_inode *dp,
67				   char *name_string,
68				   int name_len,
69				   xfs_ino_t inode_number,
70				   xfs_fsblock_t *firstblock,
71				   xfs_bmap_free_t *flist,
72				   xfs_extlen_t total);
73
74static int	xfs_dir_lookup(struct xfs_trans *tp,
75			       struct xfs_inode *dp,
76			       char *name_string,
77			       int name_length,
78			       xfs_ino_t *inode_number);
79
80static int	xfs_dir_removename(struct xfs_trans *trans,
81				   struct xfs_inode *dp,
82				   char *name_string,
83				   int name_length,
84				   xfs_ino_t ino,
85				   xfs_fsblock_t *firstblock,
86				   xfs_bmap_free_t *flist,
87				   xfs_extlen_t total);
88
89static int	xfs_dir_getdents(struct xfs_trans *tp,
90				 struct xfs_inode *dp,
91				 struct uio *uiop,
92				 int *eofp);
93
94static int	xfs_dir_replace(struct xfs_trans *tp,
95				struct xfs_inode *dp,
96				char *name_string,
97				int name_length,
98				xfs_ino_t inode_number,
99				xfs_fsblock_t *firstblock,
100				xfs_bmap_free_t *flist,
101				xfs_extlen_t total);
102
103static int	xfs_dir_canenter(struct xfs_trans *tp,
104				 struct xfs_inode *dp,
105				 char *name_string,
106				 int name_length);
107
108static int	xfs_dir_shortform_validate_ondisk(xfs_mount_t *mp,
109						  xfs_dinode_t *dip);
110
111xfs_dirops_t xfsv1_dirops = {
112	.xd_mount			= xfs_dir_mount,
113	.xd_isempty			= xfs_dir_isempty,
114	.xd_init			= xfs_dir_init,
115	.xd_createname			= xfs_dir_createname,
116	.xd_lookup			= xfs_dir_lookup,
117	.xd_removename			= xfs_dir_removename,
118	.xd_getdents			= xfs_dir_getdents,
119	.xd_replace			= xfs_dir_replace,
120	.xd_canenter			= xfs_dir_canenter,
121	.xd_shortform_validate_ondisk	= xfs_dir_shortform_validate_ondisk,
122	.xd_shortform_to_single		= xfs_dir_shortform_to_leaf,
123};
124
125/*
126 * Internal routines when dirsize == XFS_LBSIZE(mp).
127 */
128STATIC int xfs_dir_leaf_lookup(xfs_da_args_t *args);
129STATIC int xfs_dir_leaf_removename(xfs_da_args_t *args, int *number_entries,
130						 int *total_namebytes);
131STATIC int xfs_dir_leaf_getdents(xfs_trans_t *trans, xfs_inode_t *dp,
132					     uio_t *uio, int *eofp,
133					     xfs_dirent_t *dbp,
134					     xfs_dir_put_t put);
135STATIC int xfs_dir_leaf_replace(xfs_da_args_t *args);
136
137/*
138 * Internal routines when dirsize > XFS_LBSIZE(mp).
139 */
140STATIC int xfs_dir_node_addname(xfs_da_args_t *args);
141STATIC int xfs_dir_node_lookup(xfs_da_args_t *args);
142STATIC int xfs_dir_node_removename(xfs_da_args_t *args);
143STATIC int xfs_dir_node_getdents(xfs_trans_t *trans, xfs_inode_t *dp,
144					     uio_t *uio, int *eofp,
145					     xfs_dirent_t *dbp,
146					     xfs_dir_put_t put);
147STATIC int xfs_dir_node_replace(xfs_da_args_t *args);
148
149#if defined(XFS_DIR_TRACE)
150ktrace_t *xfs_dir_trace_buf;
151#endif
152
153
154/*========================================================================
155 * Overall external interface routines.
156 *========================================================================*/
157
158xfs_dahash_t	xfs_dir_hash_dot, xfs_dir_hash_dotdot;
159
160/*
161 * One-time startup routine called from xfs_init().
162 */
163void
164xfs_dir_startup(void)
165{
166	xfs_dir_hash_dot = xfs_da_hashname(".", 1);
167	xfs_dir_hash_dotdot = xfs_da_hashname("..", 2);
168}
169
170/*
171 * Initialize directory-related fields in the mount structure.
172 */
173static void
174xfs_dir_mount(xfs_mount_t *mp)
175{
176	uint shortcount, leafcount, count;
177
178	mp->m_dirversion = 1;
179	if (!(mp->m_flags & XFS_MOUNT_ATTR2)) {
180		shortcount = (mp->m_attroffset -
181				(uint)sizeof(xfs_dir_sf_hdr_t)) /
182				 (uint)sizeof(xfs_dir_sf_entry_t);
183		leafcount = (XFS_LBSIZE(mp) -
184				(uint)sizeof(xfs_dir_leaf_hdr_t)) /
185				 ((uint)sizeof(xfs_dir_leaf_entry_t) +
186				  (uint)sizeof(xfs_dir_leaf_name_t));
187	} else {
188		shortcount = (XFS_BMDR_SPACE_CALC(MINABTPTRS) -
189			      (uint)sizeof(xfs_dir_sf_hdr_t)) /
190			       (uint)sizeof(xfs_dir_sf_entry_t);
191		leafcount = (XFS_LBSIZE(mp) -
192			    (uint)sizeof(xfs_dir_leaf_hdr_t)) /
193			     ((uint)sizeof(xfs_dir_leaf_entry_t) +
194			      (uint)sizeof(xfs_dir_leaf_name_t));
195	}
196	count = shortcount > leafcount ? shortcount : leafcount;
197	mp->m_dircook_elog = xfs_da_log2_roundup(count + 1);
198	ASSERT(mp->m_dircook_elog <= mp->m_sb.sb_blocklog);
199	mp->m_dir_node_ents = mp->m_attr_node_ents =
200		(XFS_LBSIZE(mp) - (uint)sizeof(xfs_da_node_hdr_t)) /
201		(uint)sizeof(xfs_da_node_entry_t);
202	mp->m_dir_magicpct = (XFS_LBSIZE(mp) * 37) / 100;
203	mp->m_dirblksize = mp->m_sb.sb_blocksize;
204	mp->m_dirblkfsbs = 1;
205}
206
207/*
208 * Return 1 if directory contains only "." and "..".
209 */
210static int
211xfs_dir_isempty(xfs_inode_t *dp)
212{
213	xfs_dir_sf_hdr_t *hdr;
214
215	ASSERT((dp->i_d.di_mode & S_IFMT) == S_IFDIR);
216	if (dp->i_d.di_size == 0)
217		return(1);
218	if (dp->i_d.di_size > XFS_IFORK_DSIZE(dp))
219		return(0);
220	hdr = (xfs_dir_sf_hdr_t *)dp->i_df.if_u1.if_data;
221	return(hdr->count == 0);
222}
223
224/*
225 * Initialize a directory with its "." and ".." entries.
226 */
227static int
228xfs_dir_init(xfs_trans_t *trans, xfs_inode_t *dir, xfs_inode_t *parent_dir)
229{
230	xfs_da_args_t args;
231	int error;
232
233	memset((char *)&args, 0, sizeof(args));
234	args.dp = dir;
235	args.trans = trans;
236
237	ASSERT((dir->i_d.di_mode & S_IFMT) == S_IFDIR);
238	if ((error = xfs_dir_ino_validate(trans->t_mountp, parent_dir->i_ino)))
239		return error;
240
241	return(xfs_dir_shortform_create(&args, parent_dir->i_ino));
242}
243
244/*
245 * Generic handler routine to add a name to a directory.
246 * Transitions directory from shortform to Btree as necessary.
247 */
248static int							/* error */
249xfs_dir_createname(xfs_trans_t *trans, xfs_inode_t *dp, char *name,
250		   int namelen, xfs_ino_t inum, xfs_fsblock_t *firstblock,
251		   xfs_bmap_free_t *flist, xfs_extlen_t total)
252{
253	xfs_da_args_t args;
254	int retval, newsize, done;
255
256	ASSERT((dp->i_d.di_mode & S_IFMT) == S_IFDIR);
257
258	if ((retval = xfs_dir_ino_validate(trans->t_mountp, inum)))
259		return (retval);
260
261	XFS_STATS_INC(xs_dir_create);
262	/*
263	 * Fill in the arg structure for this request.
264	 */
265	args.name = name;
266	args.namelen = namelen;
267	args.hashval = xfs_da_hashname(name, namelen);
268	args.inumber = inum;
269	args.dp = dp;
270	args.firstblock = firstblock;
271	args.flist = flist;
272	args.total = total;
273	args.whichfork = XFS_DATA_FORK;
274	args.trans = trans;
275	args.justcheck = 0;
276	args.addname = args.oknoent = 1;
277
278	/*
279	 * Decide on what work routines to call based on the inode size.
280	 */
281	done = 0;
282	if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
283		newsize = XFS_DIR_SF_ENTSIZE_BYNAME(args.namelen);
284		if ((dp->i_d.di_size + newsize) <= XFS_IFORK_DSIZE(dp)) {
285			retval = xfs_dir_shortform_addname(&args);
286			done = 1;
287		} else {
288			if (total == 0)
289				return XFS_ERROR(ENOSPC);
290			retval = xfs_dir_shortform_to_leaf(&args);
291			done = retval != 0;
292		}
293	}
294	if (!done && xfs_bmap_one_block(dp, XFS_DATA_FORK)) {
295		retval = xfs_dir_leaf_addname(&args);
296		done = retval != ENOSPC;
297		if (!done) {
298			if (total == 0)
299				return XFS_ERROR(ENOSPC);
300			retval = xfs_dir_leaf_to_node(&args);
301			done = retval != 0;
302		}
303	}
304	if (!done) {
305		retval = xfs_dir_node_addname(&args);
306	}
307	return(retval);
308}
309
310/*
311 * Generic handler routine to check if a name can be added to a directory,
312 * without adding any blocks to the directory.
313 */
314static int							/* error */
315xfs_dir_canenter(xfs_trans_t *trans, xfs_inode_t *dp, char *name, int namelen)
316{
317	xfs_da_args_t args;
318	int retval, newsize;
319
320	ASSERT((dp->i_d.di_mode & S_IFMT) == S_IFDIR);
321	/*
322	 * Fill in the arg structure for this request.
323	 */
324	args.name = name;
325	args.namelen = namelen;
326	args.hashval = xfs_da_hashname(name, namelen);
327	args.inumber = 0;
328	args.dp = dp;
329	args.firstblock = NULL;
330	args.flist = NULL;
331	args.total = 0;
332	args.whichfork = XFS_DATA_FORK;
333	args.trans = trans;
334	args.justcheck = args.addname = args.oknoent = 1;
335
336	/*
337	 * Decide on what work routines to call based on the inode size.
338	 */
339	if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
340		newsize = XFS_DIR_SF_ENTSIZE_BYNAME(args.namelen);
341		if ((dp->i_d.di_size + newsize) <= XFS_IFORK_DSIZE(dp))
342			retval = 0;
343		else
344			retval = XFS_ERROR(ENOSPC);
345	} else if (xfs_bmap_one_block(dp, XFS_DATA_FORK)) {
346		retval = xfs_dir_leaf_addname(&args);
347	} else {
348		retval = xfs_dir_node_addname(&args);
349	}
350	return(retval);
351}
352
353/*
354 * Generic handler routine to remove a name from a directory.
355 * Transitions directory from Btree to shortform as necessary.
356 */
357static int							/* error */
358xfs_dir_removename(xfs_trans_t *trans, xfs_inode_t *dp, char *name,
359		   int namelen, xfs_ino_t ino, xfs_fsblock_t *firstblock,
360		   xfs_bmap_free_t *flist, xfs_extlen_t total)
361{
362	xfs_da_args_t args;
363	int count, totallen, newsize, retval;
364
365	ASSERT((dp->i_d.di_mode & S_IFMT) == S_IFDIR);
366	XFS_STATS_INC(xs_dir_remove);
367	/*
368	 * Fill in the arg structure for this request.
369	 */
370	args.name = name;
371	args.namelen = namelen;
372	args.hashval = xfs_da_hashname(name, namelen);
373	args.inumber = ino;
374	args.dp = dp;
375	args.firstblock = firstblock;
376	args.flist = flist;
377	args.total = total;
378	args.whichfork = XFS_DATA_FORK;
379	args.trans = trans;
380	args.justcheck = args.addname = args.oknoent = 0;
381
382	/*
383	 * Decide on what work routines to call based on the inode size.
384	 */
385	if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
386		retval = xfs_dir_shortform_removename(&args);
387	} else if (xfs_bmap_one_block(dp, XFS_DATA_FORK)) {
388		count = totallen = 0;
389		retval = xfs_dir_leaf_removename(&args, &count, &totallen);
390		if (retval == 0) {
391			newsize = XFS_DIR_SF_ALLFIT(count, totallen);
392			if (newsize <= XFS_IFORK_DSIZE(dp)) {
393				retval = xfs_dir_leaf_to_shortform(&args);
394			}
395		}
396	} else {
397		retval = xfs_dir_node_removename(&args);
398	}
399	return(retval);
400}
401
402static int							/* error */
403xfs_dir_lookup(xfs_trans_t *trans, xfs_inode_t *dp, char *name, int namelen,
404				   xfs_ino_t *inum)
405{
406	xfs_da_args_t args;
407	int retval;
408
409	ASSERT((dp->i_d.di_mode & S_IFMT) == S_IFDIR);
410
411	XFS_STATS_INC(xs_dir_lookup);
412	/*
413	 * Fill in the arg structure for this request.
414	 */
415	args.name = name;
416	args.namelen = namelen;
417	args.hashval = xfs_da_hashname(name, namelen);
418	args.inumber = 0;
419	args.dp = dp;
420	args.firstblock = NULL;
421	args.flist = NULL;
422	args.total = 0;
423	args.whichfork = XFS_DATA_FORK;
424	args.trans = trans;
425	args.justcheck = args.addname = 0;
426	args.oknoent = 1;
427
428	/*
429	 * Decide on what work routines to call based on the inode size.
430	 */
431	if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
432		retval = xfs_dir_shortform_lookup(&args);
433	} else if (xfs_bmap_one_block(dp, XFS_DATA_FORK)) {
434		retval = xfs_dir_leaf_lookup(&args);
435	} else {
436		retval = xfs_dir_node_lookup(&args);
437	}
438	if (retval == EEXIST)
439		retval = 0;
440	*inum = args.inumber;
441	return(retval);
442}
443
444/*
445 * Implement readdir.
446 */
447static int							/* error */
448xfs_dir_getdents(xfs_trans_t *trans, xfs_inode_t *dp, uio_t *uio, int *eofp)
449{
450	xfs_dirent_t *dbp;
451	int  alignment, retval;
452	xfs_dir_put_t put;
453
454	XFS_STATS_INC(xs_dir_getdents);
455	ASSERT((dp->i_d.di_mode & S_IFMT) == S_IFDIR);
456
457	/*
458	 * If our caller has given us a single contiguous memory buffer,
459	 * just work directly within that buffer.  If it's in user memory,
460	 * lock it down first.
461	 */
462	alignment = sizeof(xfs_off_t) - 1;
463	if ((uio->uio_iovcnt == 1) &&
464	    (((__psint_t)uio->uio_iov[0].iov_base & alignment) == 0) &&
465	    ((uio->uio_iov[0].iov_len & alignment) == 0)) {
466		dbp = NULL;
467		put = xfs_dir_put_dirent64_direct;
468	} else {
469		dbp = kmem_alloc(sizeof(*dbp) + MAXNAMELEN, KM_SLEEP);
470		put = xfs_dir_put_dirent64_uio;
471	}
472
473	/*
474	 * Decide on what work routines to call based on the inode size.
475	 */
476	*eofp = 0;
477
478	if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
479		retval = xfs_dir_shortform_getdents(dp, uio, eofp, dbp, put);
480	} else if (xfs_bmap_one_block(dp, XFS_DATA_FORK)) {
481		retval = xfs_dir_leaf_getdents(trans, dp, uio, eofp, dbp, put);
482	} else {
483		retval = xfs_dir_node_getdents(trans, dp, uio, eofp, dbp, put);
484	}
485	if (dbp != NULL)
486		kmem_free(dbp, sizeof(*dbp) + MAXNAMELEN);
487
488	return(retval);
489}
490
491static int							/* error */
492xfs_dir_replace(xfs_trans_t *trans, xfs_inode_t *dp, char *name, int namelen,
493				    xfs_ino_t inum, xfs_fsblock_t *firstblock,
494				    xfs_bmap_free_t *flist, xfs_extlen_t total)
495{
496	xfs_da_args_t args;
497	int retval;
498
499	ASSERT((dp->i_d.di_mode & S_IFMT) == S_IFDIR);
500
501	if ((retval = xfs_dir_ino_validate(trans->t_mountp, inum)))
502		return retval;
503
504	/*
505	 * Fill in the arg structure for this request.
506	 */
507	args.name = name;
508	args.namelen = namelen;
509	args.hashval = xfs_da_hashname(name, namelen);
510	args.inumber = inum;
511	args.dp = dp;
512	args.firstblock = firstblock;
513	args.flist = flist;
514	args.total = total;
515	args.whichfork = XFS_DATA_FORK;
516	args.trans = trans;
517	args.justcheck = args.addname = args.oknoent = 0;
518
519	/*
520	 * Decide on what work routines to call based on the inode size.
521	 */
522	if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
523		retval = xfs_dir_shortform_replace(&args);
524	} else if (xfs_bmap_one_block(dp, XFS_DATA_FORK)) {
525		retval = xfs_dir_leaf_replace(&args);
526	} else {
527		retval = xfs_dir_node_replace(&args);
528	}
529
530	return(retval);
531}
532
533static int
534xfs_dir_shortform_validate_ondisk(xfs_mount_t *mp, xfs_dinode_t *dp)
535{
536	xfs_ino_t		ino;
537	int			namelen_sum;
538	int			count;
539	xfs_dir_shortform_t	*sf;
540	xfs_dir_sf_entry_t	*sfe;
541	int			i;
542
543
544
545	if ((INT_GET(dp->di_core.di_mode, ARCH_CONVERT) & S_IFMT) != S_IFDIR) {
546		return 0;
547	}
548	if (INT_GET(dp->di_core.di_format, ARCH_CONVERT) != XFS_DINODE_FMT_LOCAL) {
549		return 0;
550	}
551	if (INT_GET(dp->di_core.di_size, ARCH_CONVERT) < sizeof(sf->hdr)) {
552		xfs_fs_cmn_err(CE_WARN, mp, "Invalid shortform size: dp 0x%p",
553			dp);
554		return 1;
555	}
556	sf = (xfs_dir_shortform_t *)(&dp->di_u.di_dirsf);
557	ino = XFS_GET_DIR_INO8(sf->hdr.parent);
558	if (xfs_dir_ino_validate(mp, ino))
559		return 1;
560
561	count =	sf->hdr.count;
562	if ((count < 0) || ((count * 10) > XFS_LITINO(mp))) {
563		xfs_fs_cmn_err(CE_WARN, mp,
564			"Invalid shortform count: dp 0x%p", dp);
565		return(1);
566	}
567
568	if (count == 0) {
569		return 0;
570	}
571
572	namelen_sum = 0;
573	sfe = &sf->list[0];
574	for (i = sf->hdr.count - 1; i >= 0; i--) {
575		ino = XFS_GET_DIR_INO8(sfe->inumber);
576		xfs_dir_ino_validate(mp, ino);
577		if (sfe->namelen >= XFS_LITINO(mp)) {
578			xfs_fs_cmn_err(CE_WARN, mp,
579				"Invalid shortform namelen: dp 0x%p", dp);
580			return 1;
581		}
582		namelen_sum += sfe->namelen;
583		sfe = XFS_DIR_SF_NEXTENTRY(sfe);
584	}
585	if (namelen_sum >= XFS_LITINO(mp)) {
586		xfs_fs_cmn_err(CE_WARN, mp,
587			"Invalid shortform namelen: dp 0x%p", dp);
588		return 1;
589	}
590
591	return 0;
592}
593
594/*========================================================================
595 * External routines when dirsize == XFS_LBSIZE(dp->i_mount).
596 *========================================================================*/
597
598/*
599 * Add a name to the leaf directory structure
600 * This is the external routine.
601 */
602int
603xfs_dir_leaf_addname(xfs_da_args_t *args)
604{
605	int index, retval;
606	xfs_dabuf_t *bp;
607
608	retval = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp,
609					      XFS_DATA_FORK);
610	if (retval)
611		return(retval);
612	ASSERT(bp != NULL);
613
614	retval = xfs_dir_leaf_lookup_int(bp, args, &index);
615	if (retval == ENOENT)
616		retval = xfs_dir_leaf_add(bp, args, index);
617	xfs_da_buf_done(bp);
618	return(retval);
619}
620
621/*
622 * Remove a name from the leaf directory structure
623 * This is the external routine.
624 */
625STATIC int
626xfs_dir_leaf_removename(xfs_da_args_t *args, int *count, int *totallen)
627{
628	xfs_dir_leafblock_t *leaf;
629	int index, retval;
630	xfs_dabuf_t *bp;
631
632	retval = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp,
633					      XFS_DATA_FORK);
634	if (retval)
635		return(retval);
636	ASSERT(bp != NULL);
637	leaf = bp->data;
638	ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR_LEAF_MAGIC);
639	retval = xfs_dir_leaf_lookup_int(bp, args, &index);
640	if (retval == EEXIST) {
641		(void)xfs_dir_leaf_remove(args->trans, bp, index);
642		*count = INT_GET(leaf->hdr.count, ARCH_CONVERT);
643		*totallen = INT_GET(leaf->hdr.namebytes, ARCH_CONVERT);
644		retval = 0;
645	}
646	xfs_da_buf_done(bp);
647	return(retval);
648}
649
650/*
651 * Look up a name in a leaf directory structure.
652 * This is the external routine.
653 */
654STATIC int
655xfs_dir_leaf_lookup(xfs_da_args_t *args)
656{
657	int index, retval;
658	xfs_dabuf_t *bp;
659
660	retval = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp,
661					      XFS_DATA_FORK);
662	if (retval)
663		return(retval);
664	ASSERT(bp != NULL);
665	retval = xfs_dir_leaf_lookup_int(bp, args, &index);
666	xfs_da_brelse(args->trans, bp);
667	return(retval);
668}
669
670/*
671 * Copy out directory entries for getdents(), for leaf directories.
672 */
673STATIC int
674xfs_dir_leaf_getdents(xfs_trans_t *trans, xfs_inode_t *dp, uio_t *uio,
675				  int *eofp, xfs_dirent_t *dbp, xfs_dir_put_t put)
676{
677	xfs_dabuf_t *bp;
678	int retval, eob;
679
680	retval = xfs_da_read_buf(dp->i_transp, dp, 0, -1, &bp, XFS_DATA_FORK);
681	if (retval)
682		return(retval);
683	ASSERT(bp != NULL);
684	retval = xfs_dir_leaf_getdents_int(bp, dp, 0, uio, &eob, dbp, put, -1);
685	xfs_da_brelse(trans, bp);
686	*eofp = (eob == 0);
687	return(retval);
688}
689
690/*
691 * Look up a name in a leaf directory structure, replace the inode number.
692 * This is the external routine.
693 */
694STATIC int
695xfs_dir_leaf_replace(xfs_da_args_t *args)
696{
697	int index, retval;
698	xfs_dabuf_t *bp;
699	xfs_ino_t inum;
700	xfs_dir_leafblock_t *leaf;
701	xfs_dir_leaf_entry_t *entry;
702	xfs_dir_leaf_name_t *namest;
703
704	inum = args->inumber;
705	retval = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp,
706					      XFS_DATA_FORK);
707	if (retval)
708		return(retval);
709	ASSERT(bp != NULL);
710	retval = xfs_dir_leaf_lookup_int(bp, args, &index);
711	if (retval == EEXIST) {
712		leaf = bp->data;
713		entry = &leaf->entries[index];
714		namest = XFS_DIR_LEAF_NAMESTRUCT(leaf, INT_GET(entry->nameidx, ARCH_CONVERT));
715		/* XXX - replace assert? */
716		XFS_DIR_SF_PUT_DIRINO(&inum, &namest->inumber);
717		xfs_da_log_buf(args->trans, bp,
718		    XFS_DA_LOGRANGE(leaf, namest, sizeof(namest->inumber)));
719		xfs_da_buf_done(bp);
720		retval = 0;
721	} else
722		xfs_da_brelse(args->trans, bp);
723	return(retval);
724}
725
726
727/*========================================================================
728 * External routines when dirsize > XFS_LBSIZE(mp).
729 *========================================================================*/
730
731/*
732 * Add a name to a Btree-format directory.
733 *
734 * This will involve walking down the Btree, and may involve splitting
735 * leaf nodes and even splitting intermediate nodes up to and including
736 * the root node (a special case of an intermediate node).
737 */
738STATIC int
739xfs_dir_node_addname(xfs_da_args_t *args)
740{
741	xfs_da_state_t *state;
742	xfs_da_state_blk_t *blk;
743	int retval, error;
744
745	/*
746	 * Fill in bucket of arguments/results/context to carry around.
747	 */
748	state = xfs_da_state_alloc();
749	state->args = args;
750	state->mp = args->dp->i_mount;
751	state->blocksize = state->mp->m_sb.sb_blocksize;
752	state->node_ents = state->mp->m_dir_node_ents;
753
754	/*
755	 * Search to see if name already exists, and get back a pointer
756	 * to where it should go.
757	 */
758	error = xfs_da_node_lookup_int(state, &retval);
759	if (error)
760		retval = error;
761	if (retval != ENOENT)
762		goto error;
763	blk = &state->path.blk[ state->path.active-1 ];
764	ASSERT(blk->magic == XFS_DIR_LEAF_MAGIC);
765	retval = xfs_dir_leaf_add(blk->bp, args, blk->index);
766	if (retval == 0) {
767		/*
768		 * Addition succeeded, update Btree hashvals.
769		 */
770		if (!args->justcheck)
771			xfs_da_fixhashpath(state, &state->path);
772	} else {
773		/*
774		 * Addition failed, split as many Btree elements as required.
775		 */
776		if (args->total == 0) {
777			ASSERT(retval == ENOSPC);
778			goto error;
779		}
780		retval = xfs_da_split(state);
781	}
782error:
783	xfs_da_state_free(state);
784
785	return(retval);
786}
787
788/*
789 * Remove a name from a B-tree directory.
790 *
791 * This will involve walking down the Btree, and may involve joining
792 * leaf nodes and even joining intermediate nodes up to and including
793 * the root node (a special case of an intermediate node).
794 */
795STATIC int
796xfs_dir_node_removename(xfs_da_args_t *args)
797{
798	xfs_da_state_t *state;
799	xfs_da_state_blk_t *blk;
800	int retval, error;
801
802	state = xfs_da_state_alloc();
803	state->args = args;
804	state->mp = args->dp->i_mount;
805	state->blocksize = state->mp->m_sb.sb_blocksize;
806	state->node_ents = state->mp->m_dir_node_ents;
807
808	/*
809	 * Search to see if name exists, and get back a pointer to it.
810	 */
811	error = xfs_da_node_lookup_int(state, &retval);
812	if (error)
813		retval = error;
814	if (retval != EEXIST) {
815		xfs_da_state_free(state);
816		return(retval);
817	}
818
819	/*
820	 * Remove the name and update the hashvals in the tree.
821	 */
822	blk = &state->path.blk[ state->path.active-1 ];
823	ASSERT(blk->magic == XFS_DIR_LEAF_MAGIC);
824	retval = xfs_dir_leaf_remove(args->trans, blk->bp, blk->index);
825	xfs_da_fixhashpath(state, &state->path);
826
827	/*
828	 * Check to see if the tree needs to be collapsed.
829	 */
830	error = 0;
831	if (retval) {
832		error = xfs_da_join(state);
833	}
834
835	xfs_da_state_free(state);
836	if (error)
837		return(error);
838	return(0);
839}
840
841/*
842 * Look up a filename in a int directory.
843 * Use an internal routine to actually do all the work.
844 */
845STATIC int
846xfs_dir_node_lookup(xfs_da_args_t *args)
847{
848	xfs_da_state_t *state;
849	int retval, error, i;
850
851	state = xfs_da_state_alloc();
852	state->args = args;
853	state->mp = args->dp->i_mount;
854	state->blocksize = state->mp->m_sb.sb_blocksize;
855	state->node_ents = state->mp->m_dir_node_ents;
856
857	/*
858	 * Search to see if name exists,
859	 * and get back a pointer to it.
860	 */
861	error = xfs_da_node_lookup_int(state, &retval);
862	if (error) {
863		retval = error;
864	}
865
866	/*
867	 * If not in a transaction, we have to release all the buffers.
868	 */
869	for (i = 0; i < state->path.active; i++) {
870		xfs_da_brelse(args->trans, state->path.blk[i].bp);
871		state->path.blk[i].bp = NULL;
872	}
873
874	xfs_da_state_free(state);
875	return(retval);
876}
877
878STATIC int
879xfs_dir_node_getdents(xfs_trans_t *trans, xfs_inode_t *dp, uio_t *uio,
880				  int *eofp, xfs_dirent_t *dbp, xfs_dir_put_t put)
881{
882	xfs_da_intnode_t *node;
883	xfs_da_node_entry_t *btree;
884	xfs_dir_leafblock_t *leaf = NULL;
885	xfs_dablk_t bno, nextbno;
886	xfs_dahash_t cookhash;
887	xfs_mount_t *mp;
888	int error, eob, i;
889	xfs_dabuf_t *bp;
890	xfs_daddr_t nextda;
891
892	/*
893	 * Pick up our context.
894	 */
895	mp = dp->i_mount;
896	bp = NULL;
897	bno = XFS_DA_COOKIE_BNO(mp, uio->uio_offset);
898	cookhash = XFS_DA_COOKIE_HASH(mp, uio->uio_offset);
899
900	xfs_dir_trace_g_du("node: start", dp, uio);
901
902	/*
903	 * Re-find our place, even if we're confused about what our place is.
904	 *
905	 * First we check the block number from the magic cookie, it is a
906	 * cache of where we ended last time.  If we find a leaf block, and
907	 * the starting hashval in that block is less than our desired
908	 * hashval, then we run with it.
909	 */
910	if (bno > 0) {
911		error = xfs_da_read_buf(trans, dp, bno, -2, &bp, XFS_DATA_FORK);
912		if ((error != 0) && (error != EFSCORRUPTED))
913			return(error);
914		if (bp)
915			leaf = bp->data;
916		if (bp && be16_to_cpu(leaf->hdr.info.magic) != XFS_DIR_LEAF_MAGIC) {
917			xfs_dir_trace_g_dub("node: block not a leaf",
918						   dp, uio, bno);
919			xfs_da_brelse(trans, bp);
920			bp = NULL;
921		}
922		if (bp && INT_GET(leaf->entries[0].hashval, ARCH_CONVERT) > cookhash) {
923			xfs_dir_trace_g_dub("node: leaf hash too large",
924						   dp, uio, bno);
925			xfs_da_brelse(trans, bp);
926			bp = NULL;
927		}
928		if (bp &&
929		    cookhash > INT_GET(leaf->entries[INT_GET(leaf->hdr.count, ARCH_CONVERT) - 1].hashval, ARCH_CONVERT)) {
930			xfs_dir_trace_g_dub("node: leaf hash too small",
931						   dp, uio, bno);
932			xfs_da_brelse(trans, bp);
933			bp = NULL;
934		}
935	}
936
937	/*
938	 * If we did not find a leaf block from the blockno in the cookie,
939	 * or we there was no blockno in the cookie (eg: first time thru),
940	 * the we start at the top of the Btree and re-find our hashval.
941	 */
942	if (bp == NULL) {
943		xfs_dir_trace_g_du("node: start at root" , dp, uio);
944		bno = 0;
945		for (;;) {
946			error = xfs_da_read_buf(trans, dp, bno, -1, &bp,
947						       XFS_DATA_FORK);
948			if (error)
949				return(error);
950			if (bp == NULL)
951				return(XFS_ERROR(EFSCORRUPTED));
952			node = bp->data;
953			if (be16_to_cpu(node->hdr.info.magic) != XFS_DA_NODE_MAGIC)
954				break;
955			btree = &node->btree[0];
956			xfs_dir_trace_g_dun("node: node detail", dp, uio, node);
957			for (i = 0; i < be16_to_cpu(node->hdr.count); btree++, i++) {
958				if (be32_to_cpu(btree->hashval) >= cookhash) {
959					bno = be32_to_cpu(btree->before);
960					break;
961				}
962			}
963			if (i == be16_to_cpu(node->hdr.count)) {
964				xfs_da_brelse(trans, bp);
965				xfs_dir_trace_g_du("node: hash beyond EOF",
966							  dp, uio);
967				uio->uio_offset = XFS_DA_MAKE_COOKIE(mp, 0, 0,
968							     XFS_DA_MAXHASH);
969				*eofp = 1;
970				return(0);
971			}
972			xfs_dir_trace_g_dub("node: going to block",
973						   dp, uio, bno);
974			xfs_da_brelse(trans, bp);
975		}
976	}
977	ASSERT(cookhash != XFS_DA_MAXHASH);
978
979	/*
980	 * We've dropped down to the (first) leaf block that contains the
981	 * hashval we are interested in.  Continue rolling upward thru the
982	 * leaf blocks until we fill up our buffer.
983	 */
984	for (;;) {
985		leaf = bp->data;
986		if (unlikely(be16_to_cpu(leaf->hdr.info.magic) != XFS_DIR_LEAF_MAGIC)) {
987			xfs_dir_trace_g_dul("node: not a leaf", dp, uio, leaf);
988			xfs_da_brelse(trans, bp);
989			XFS_CORRUPTION_ERROR("xfs_dir_node_getdents(1)",
990					     XFS_ERRLEVEL_LOW, mp, leaf);
991			return XFS_ERROR(EFSCORRUPTED);
992		}
993		xfs_dir_trace_g_dul("node: leaf detail", dp, uio, leaf);
994		if ((nextbno = be32_to_cpu(leaf->hdr.info.forw))) {
995			nextda = xfs_da_reada_buf(trans, dp, nextbno,
996						  XFS_DATA_FORK);
997		} else
998			nextda = -1;
999		error = xfs_dir_leaf_getdents_int(bp, dp, bno, uio, &eob, dbp,
1000						  put, nextda);
1001		xfs_da_brelse(trans, bp);
1002		bno = nextbno;
1003		if (eob) {
1004			xfs_dir_trace_g_dub("node: E-O-B", dp, uio, bno);
1005			*eofp = 0;
1006			return(error);
1007		}
1008		if (bno == 0)
1009			break;
1010		error = xfs_da_read_buf(trans, dp, bno, nextda, &bp,
1011					XFS_DATA_FORK);
1012		if (error)
1013			return(error);
1014		if (unlikely(bp == NULL)) {
1015			XFS_ERROR_REPORT("xfs_dir_node_getdents(2)",
1016					 XFS_ERRLEVEL_LOW, mp);
1017			return(XFS_ERROR(EFSCORRUPTED));
1018		}
1019	}
1020	*eofp = 1;
1021	xfs_dir_trace_g_du("node: E-O-F", dp, uio);
1022	return(0);
1023}
1024
1025/*
1026 * Look up a filename in an int directory, replace the inode number.
1027 * Use an internal routine to actually do the lookup.
1028 */
1029STATIC int
1030xfs_dir_node_replace(xfs_da_args_t *args)
1031{
1032	xfs_da_state_t *state;
1033	xfs_da_state_blk_t *blk;
1034	xfs_dir_leafblock_t *leaf;
1035	xfs_dir_leaf_entry_t *entry;
1036	xfs_dir_leaf_name_t *namest;
1037	xfs_ino_t inum;
1038	int retval, error, i;
1039	xfs_dabuf_t *bp;
1040
1041	state = xfs_da_state_alloc();
1042	state->args = args;
1043	state->mp = args->dp->i_mount;
1044	state->blocksize = state->mp->m_sb.sb_blocksize;
1045	state->node_ents = state->mp->m_dir_node_ents;
1046	inum = args->inumber;
1047
1048	/*
1049	 * Search to see if name exists,
1050	 * and get back a pointer to it.
1051	 */
1052	error = xfs_da_node_lookup_int(state, &retval);
1053	if (error) {
1054		retval = error;
1055	}
1056
1057	if (retval == EEXIST) {
1058		blk = &state->path.blk[state->path.active - 1];
1059		ASSERT(blk->magic == XFS_DIR_LEAF_MAGIC);
1060		bp = blk->bp;
1061		leaf = bp->data;
1062		entry = &leaf->entries[blk->index];
1063		namest = XFS_DIR_LEAF_NAMESTRUCT(leaf, INT_GET(entry->nameidx, ARCH_CONVERT));
1064		/* XXX - replace assert ? */
1065		XFS_DIR_SF_PUT_DIRINO(&inum, &namest->inumber);
1066		xfs_da_log_buf(args->trans, bp,
1067		    XFS_DA_LOGRANGE(leaf, namest, sizeof(namest->inumber)));
1068		xfs_da_buf_done(bp);
1069		blk->bp = NULL;
1070		retval = 0;
1071	} else {
1072		i = state->path.active - 1;
1073		xfs_da_brelse(args->trans, state->path.blk[i].bp);
1074		state->path.blk[i].bp = NULL;
1075	}
1076	for (i = 0; i < state->path.active - 1; i++) {
1077		xfs_da_brelse(args->trans, state->path.blk[i].bp);
1078		state->path.blk[i].bp = NULL;
1079	}
1080
1081	xfs_da_state_free(state);
1082	return(retval);
1083}
1084
1085#if defined(XFS_DIR_TRACE)
1086/*
1087 * Add a trace buffer entry for an inode and a uio.
1088 */
1089void
1090xfs_dir_trace_g_du(char *where, xfs_inode_t *dp, uio_t *uio)
1091{
1092	xfs_dir_trace_enter(XFS_DIR_KTRACE_G_DU, where,
1093		     (void *)dp, (void *)dp->i_mount,
1094		     (void *)((unsigned long)(uio->uio_offset >> 32)),
1095		     (void *)((unsigned long)(uio->uio_offset & 0xFFFFFFFF)),
1096		     (void *)(unsigned long)uio->uio_resid,
1097		     NULL, NULL, NULL, NULL, NULL, NULL, NULL);
1098}
1099
1100/*
1101 * Add a trace buffer entry for an inode and a uio.
1102 */
1103void
1104xfs_dir_trace_g_dub(char *where, xfs_inode_t *dp, uio_t *uio, xfs_dablk_t bno)
1105{
1106	xfs_dir_trace_enter(XFS_DIR_KTRACE_G_DUB, where,
1107		     (void *)dp, (void *)dp->i_mount,
1108		     (void *)((unsigned long)(uio->uio_offset >> 32)),
1109		     (void *)((unsigned long)(uio->uio_offset & 0xFFFFFFFF)),
1110		     (void *)(unsigned long)uio->uio_resid,
1111		     (void *)(unsigned long)bno,
1112		     NULL, NULL, NULL, NULL, NULL, NULL);
1113}
1114
1115/*
1116 * Add a trace buffer entry for an inode and a uio.
1117 */
1118void
1119xfs_dir_trace_g_dun(char *where, xfs_inode_t *dp, uio_t *uio,
1120			xfs_da_intnode_t *node)
1121{
1122	int	last = be16_to_cpu(node->hdr.count) - 1;
1123
1124	xfs_dir_trace_enter(XFS_DIR_KTRACE_G_DUN, where,
1125		     (void *)dp, (void *)dp->i_mount,
1126		     (void *)((unsigned long)(uio->uio_offset >> 32)),
1127		     (void *)((unsigned long)(uio->uio_offset & 0xFFFFFFFF)),
1128		     (void *)(unsigned long)uio->uio_resid,
1129		     (void *)(unsigned long)be32_to_cpu(node->hdr.info.forw),
1130		     (void *)(unsigned long)
1131			be16_to_cpu(node->hdr.count),
1132		     (void *)(unsigned long)
1133			be32_to_cpu(node->btree[0].hashval),
1134		     (void *)(unsigned long)
1135			be32_to_cpu(node->btree[last].hashval),
1136		     NULL, NULL, NULL);
1137}
1138
1139/*
1140 * Add a trace buffer entry for an inode and a uio.
1141 */
1142void
1143xfs_dir_trace_g_dul(char *where, xfs_inode_t *dp, uio_t *uio,
1144			xfs_dir_leafblock_t *leaf)
1145{
1146	int	last = INT_GET(leaf->hdr.count, ARCH_CONVERT) - 1;
1147
1148	xfs_dir_trace_enter(XFS_DIR_KTRACE_G_DUL, where,
1149		     (void *)dp, (void *)dp->i_mount,
1150		     (void *)((unsigned long)(uio->uio_offset >> 32)),
1151		     (void *)((unsigned long)(uio->uio_offset & 0xFFFFFFFF)),
1152		     (void *)(unsigned long)uio->uio_resid,
1153		     (void *)(unsigned long)be32_to_cpu(leaf->hdr.info.forw),
1154		     (void *)(unsigned long)
1155			INT_GET(leaf->hdr.count, ARCH_CONVERT),
1156		     (void *)(unsigned long)
1157			INT_GET(leaf->entries[0].hashval, ARCH_CONVERT),
1158		     (void *)(unsigned long)
1159			INT_GET(leaf->entries[last].hashval, ARCH_CONVERT),
1160		     NULL, NULL, NULL);
1161}
1162
1163/*
1164 * Add a trace buffer entry for an inode and a uio.
1165 */
1166void
1167xfs_dir_trace_g_due(char *where, xfs_inode_t *dp, uio_t *uio,
1168			xfs_dir_leaf_entry_t *entry)
1169{
1170	xfs_dir_trace_enter(XFS_DIR_KTRACE_G_DUE, where,
1171		     (void *)dp, (void *)dp->i_mount,
1172		     (void *)((unsigned long)(uio->uio_offset >> 32)),
1173		     (void *)((unsigned long)(uio->uio_offset & 0xFFFFFFFF)),
1174		     (void *)(unsigned long)uio->uio_resid,
1175		     (void *)(unsigned long)
1176			INT_GET(entry->hashval, ARCH_CONVERT),
1177		     NULL, NULL, NULL, NULL, NULL, NULL);
1178}
1179
1180/*
1181 * Add a trace buffer entry for an inode and a uio.
1182 */
1183void
1184xfs_dir_trace_g_duc(char *where, xfs_inode_t *dp, uio_t *uio, xfs_off_t cookie)
1185{
1186	xfs_dir_trace_enter(XFS_DIR_KTRACE_G_DUC, where,
1187		     (void *)dp, (void *)dp->i_mount,
1188		     (void *)((unsigned long)(uio->uio_offset >> 32)),
1189		     (void *)((unsigned long)(uio->uio_offset & 0xFFFFFFFF)),
1190		     (void *)(unsigned long)uio->uio_resid,
1191		     (void *)((unsigned long)(cookie >> 32)),
1192		     (void *)((unsigned long)(cookie & 0xFFFFFFFF)),
1193		     NULL, NULL, NULL, NULL, NULL);
1194}
1195
1196/*
1197 * Add a trace buffer entry for the arguments given to the routine,
1198 * generic form.
1199 */
1200void
1201xfs_dir_trace_enter(int type, char *where,
1202			void * a0, void * a1,
1203			void * a2, void * a3,
1204			void * a4, void * a5,
1205			void * a6, void * a7,
1206			void * a8, void * a9,
1207			void * a10, void * a11)
1208{
1209	ASSERT(xfs_dir_trace_buf);
1210	ktrace_enter(xfs_dir_trace_buf, (void *)(unsigned long)type,
1211					(void *)where,
1212					(void *)a0, (void *)a1, (void *)a2,
1213					(void *)a3, (void *)a4, (void *)a5,
1214					(void *)a6, (void *)a7, (void *)a8,
1215					(void *)a9, (void *)a10, (void *)a11,
1216					NULL, NULL);
1217}
1218#endif	/* XFS_DIR_TRACE */
1219