jfs_dtree.c 103 KB
Newer Older
Dave Kleikamp's avatar
Dave Kleikamp committed
1
/*
Dave Kleikamp's avatar
Dave Kleikamp committed
2
 *   Copyright (C) International Business Machines Corp., 2000-2003
Dave Kleikamp's avatar
Dave Kleikamp committed
3 4 5 6 7 8 9 10 11 12 13 14 15 16
 *
 *   This program is free software;  you can redistribute it and/or modify
 *   it under the terms of the GNU General Public License as published by
 *   the Free Software Foundation; either version 2 of the License, or 
 *   (at your option) any later version.
 * 
 *   This program is distributed in the hope that it will be useful,
 *   but WITHOUT ANY WARRANTY;  without even the implied warranty of
 *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
 *   the GNU General Public License for more details.
 *
 *   You should have received a copy of the GNU General Public License
 *   along with this program;  if not, write to the Free Software 
 *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
 */
Dave Kleikamp's avatar
Dave Kleikamp committed
18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112

/*
 *	jfs_dtree.c: directory B+-tree manager
 *
 * B+-tree with variable length key directory:
 *
 * each directory page is structured as an array of 32-byte
 * directory entry slots initialized as a freelist
 * to avoid search/compaction of free space at insertion.
 * when an entry is inserted, a number of slots are allocated
 * from the freelist as required to store variable length data
 * of the entry; when the entry is deleted, slots of the entry
 * are returned to freelist.
 *
 * leaf entry stores full name as key and file serial number
 * (aka inode number) as data.
 * internal/router entry stores sufffix compressed name
 * as key and simple extent descriptor as data.
 *
 * each directory page maintains a sorted entry index table
 * which stores the start slot index of sorted entries
 * to allow binary search on the table.
 *
 * directory starts as a root/leaf page in on-disk inode
 * inline data area.
 * when it becomes full, it starts a leaf of a external extent
 * of length of 1 block. each time the first leaf becomes full,
 * it is extended rather than split (its size is doubled),
 * until its length becoms 4 KBytes, from then the extent is split
 * with new 4 Kbyte extent when it becomes full
 * to reduce external fragmentation of small directories.
 *
 * blah, blah, blah, for linear scan of directory in pieces by
 * readdir().
 *
 *
 *	case-insensitive directory file system
 *
 * names are stored in case-sensitive way in leaf entry.
 * but stored, searched and compared in case-insensitive (uppercase) order
 * (i.e., both search key and entry key are folded for search/compare):
 * (note that case-sensitive order is BROKEN in storage, e.g.,
 *  sensitive: Ad, aB, aC, aD -> insensitive: aB, aC, aD, Ad
 *
 *  entries which folds to the same key makes up a equivalent class
 *  whose members are stored as contiguous cluster (may cross page boundary)
 *  but whose order is arbitrary and acts as duplicate, e.g.,
 *  abc, Abc, aBc, abC)
 *
 * once match is found at leaf, requires scan forward/backward
 * either for, in case-insensitive search, duplicate
 * or for, in case-sensitive search, for exact match
 *
 * router entry must be created/stored in case-insensitive way
 * in internal entry:
 * (right most key of left page and left most key of right page
 * are folded, and its suffix compression is propagated as router
 * key in parent)
 * (e.g., if split occurs <abc> and <aBd>, <ABD> trather than <aB>
 * should be made the router key for the split)
 *
 * case-insensitive search:
 *
 * 	fold search key;
 *
 *	case-insensitive search of B-tree:
 *	for internal entry, router key is already folded;
 *	for leaf entry, fold the entry key before comparison.
 *
 *	if (leaf entry case-insensitive match found)
 *		if (next entry satisfies case-insensitive match)
 *			return EDUPLICATE;
 *		if (prev entry satisfies case-insensitive match)
 *			return EDUPLICATE;
 *		return match;
 *	else
 *		return no match;
 *
 * 	serialization:
 * target directory inode lock is being held on entry/exit
 * of all main directory service routines.
 *
 *	log based recovery:
 */

#include <linux/fs.h>
#include "jfs_incore.h"
#include "jfs_superblock.h"
#include "jfs_filsys.h"
#include "jfs_metapage.h"
#include "jfs_dmap.h"
#include "jfs_unicode.h"
#include "jfs_debug.h"

/* dtree split parameter */
113 114
struct dtsplit {
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
115 116
	s16 index;
	s16 nslot;
117
	struct component_name *key;
Dave Kleikamp's avatar
Dave Kleikamp committed
118
	ddata_t *data;
119 120
	struct pxdlist *pxdlist;
};
Dave Kleikamp's avatar
Dave Kleikamp committed
121 122 123 124 125 126 127 128 129 130 131 132 133

#define DT_PAGE(IP, MP) BT_PAGE(IP, MP, dtpage_t, i_dtroot)

/* get page buffer for specified block address */
#define DT_GETPAGE(IP, BN, MP, SIZE, P, RC)\
{\
	BT_GETPAGE(IP, BN, MP, dtpage_t, SIZE, P, RC, i_dtroot)\
	if (!(RC))\
	{\
		if (((P)->header.nextindex > (((BN)==0)?DTROOTMAXSLOT:(P)->header.maxslot)) ||\
		    ((BN) && ((P)->header.maxslot > DTPAGEMAXSLOT)))\
		{\
			BT_PUTPAGE(MP);\
Dave Kleikamp's avatar
Dave Kleikamp committed
134
			jfs_error((IP)->i_sb, "DT_GETPAGE: dtree page corrupt");\
Dave Kleikamp's avatar
Dave Kleikamp committed
135
			MP = NULL;\
136
			RC = -EIO;\
Dave Kleikamp's avatar
Dave Kleikamp committed
137 138 139 140 141 142 143 144 145 146 147 148 149 150
		}\
	}\
}

/* for consistency */
#define DT_PUTPAGE(MP) BT_PUTPAGE(MP)

#define DT_GETSEARCH(IP, LEAF, BN, MP, P, INDEX) \
	BT_GETSEARCH(IP, LEAF, BN, MP, dtpage_t, P, INDEX, i_dtroot)

/*
 * forward references
 */
static int dtSplitUp(tid_t tid, struct inode *ip,
151
		     struct dtsplit * split, struct btstack * btstack);
Dave Kleikamp's avatar
Dave Kleikamp committed
152

153 154
static int dtSplitPage(tid_t tid, struct inode *ip, struct dtsplit * split,
		       struct metapage ** rmpp, dtpage_t ** rpp, pxd_t * rxdp);
Dave Kleikamp's avatar
Dave Kleikamp committed
155 156

static int dtExtendPage(tid_t tid, struct inode *ip,
157
			struct dtsplit * split, struct btstack * btstack);
Dave Kleikamp's avatar
Dave Kleikamp committed
158 159

static int dtSplitRoot(tid_t tid, struct inode *ip,
160
		       struct dtsplit * split, struct metapage ** rmpp);
Dave Kleikamp's avatar
Dave Kleikamp committed
161

162 163
static int dtDeleteUp(tid_t tid, struct inode *ip, struct metapage * fmp,
		      dtpage_t * fp, struct btstack * btstack);
Dave Kleikamp's avatar
Dave Kleikamp committed
164 165

static int dtSearchNode(struct inode *ip,
166
			s64 lmxaddr, pxd_t * kpxd, struct btstack * btstack);
Dave Kleikamp's avatar
Dave Kleikamp committed
167 168 169

static int dtRelink(tid_t tid, struct inode *ip, dtpage_t * p);

170
static int dtReadFirst(struct inode *ip, struct btstack * btstack);
Dave Kleikamp's avatar
Dave Kleikamp committed
171 172

static int dtReadNext(struct inode *ip,
173
		      loff_t * offset, struct btstack * btstack);
Dave Kleikamp's avatar
Dave Kleikamp committed
174

175
static int dtCompare(struct component_name * key, dtpage_t * p, int si);
Dave Kleikamp's avatar
Dave Kleikamp committed
176

177 178
static int ciCompare(struct component_name * key, dtpage_t * p, int si,
		     int flag);
Dave Kleikamp's avatar
Dave Kleikamp committed
179

180 181
static void dtGetKey(dtpage_t * p, int i, struct component_name * key,
		     int flag);
Dave Kleikamp's avatar
Dave Kleikamp committed
182 183

static void ciGetLeafPrefixKey(dtpage_t * lp, int li, dtpage_t * rp,
184
			       int ri, struct component_name * key, int flag);
Dave Kleikamp's avatar
Dave Kleikamp committed
185

186 187
static void dtInsertEntry(dtpage_t * p, int index, struct component_name * key,
			  ddata_t * data, struct dt_lock **);
Dave Kleikamp's avatar
Dave Kleikamp committed
188 189

static void dtMoveEntry(dtpage_t * sp, int si, dtpage_t * dp,
190
			struct dt_lock ** sdtlock, struct dt_lock ** ddtlock,
Dave Kleikamp's avatar
Dave Kleikamp committed
191 192
			int do_index);

193
static void dtDeleteEntry(dtpage_t * p, int fi, struct dt_lock ** dtlock);
Dave Kleikamp's avatar
Dave Kleikamp committed
194

195
static void dtTruncateEntry(dtpage_t * p, int ti, struct dt_lock ** dtlock);
Dave Kleikamp's avatar
Dave Kleikamp committed
196

197
static void dtLinelockFreelist(dtpage_t * p, int m, struct dt_lock ** dtlock);
Dave Kleikamp's avatar
Dave Kleikamp committed
198 199 200

#define ciToUpper(c)	UniStrupr((c)->name)

201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222
/*
 *	read_index_page()
 *
 *	Reads a page of a directory's index table.
 *	Having metadata mapped into the directory inode's address space
 *	presents a multitude of problems.  We avoid this by mapping to
 *	the absolute address space outside of the *_metapage routines
 */
static struct metapage *read_index_page(struct inode *inode, s64 blkno)
{
	int rc;
	s64 xaddr;
	int xflag;
	s32 xlen;

	rc = xtLookup(inode, blkno, 1, &xflag, &xaddr, &xlen, 1);
	if (rc || (xlen == 0))
		return NULL;

	return read_metapage(inode, xaddr, PSIZE, 1);
}

223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
/*
 *	get_index_page()
 *
 *	Same as get_index_page(), but get's a new page without reading
 */
static struct metapage *get_index_page(struct inode *inode, s64 blkno)
{
	int rc;
	s64 xaddr;
	int xflag;
	s32 xlen;

	rc = xtLookup(inode, blkno, 1, &xflag, &xaddr, &xlen, 1);
	if (rc || (xlen == 0))
		return NULL;

	return get_metapage(inode, xaddr, PSIZE, 1);
}

Dave Kleikamp's avatar
Dave Kleikamp committed
242 243 244 245 246 247 248 249
/*
 *	find_index()
 *
 *	Returns dtree page containing directory table entry for specified
 *	index and pointer to its entry.
 *
 *	mp must be released by caller.
 */
250
static struct dir_table_slot *find_index(struct inode *ip, u32 index,
251
					 struct metapage ** mp, s64 *lblock)
Dave Kleikamp's avatar
Dave Kleikamp committed
252 253 254 255 256
{
	struct jfs_inode_info *jfs_ip = JFS_IP(ip);
	s64 blkno;
	s64 offset;
	int page_offset;
257
	struct dir_table_slot *slot;
Dave Kleikamp's avatar
Dave Kleikamp committed
258 259 260 261
	static int maxWarnings = 10;

	if (index < 2) {
		if (maxWarnings) {
262
			jfs_warn("find_entry called with index = %d", index);
Dave Kleikamp's avatar
Dave Kleikamp committed
263 264 265 266 267 268
			maxWarnings--;
		}
		return 0;
	}

	if (index >= jfs_ip->next_index) {
269
		jfs_warn("find_entry called with index >= next_index");
Dave Kleikamp's avatar
Dave Kleikamp committed
270 271 272 273 274 275 276 277 278 279
		return 0;
	}

	if (jfs_ip->next_index <= (MAX_INLINE_DIRTABLE_ENTRY + 1)) {
		/*
		 * Inline directory table
		 */
		*mp = 0;
		slot = &jfs_ip->i_dirtable[index - 2];
	} else {
280
		offset = (index - 2) * sizeof(struct dir_table_slot);
Dave Kleikamp's avatar
Dave Kleikamp committed
281 282 283 284
		page_offset = offset & (PSIZE - 1);
		blkno = ((offset + 1) >> L2PSIZE) <<
		    JFS_SBI(ip->i_sb)->l2nbperpage;

285
		if (*mp && (*lblock != blkno)) {
Dave Kleikamp's avatar
Dave Kleikamp committed
286 287 288
			release_metapage(*mp);
			*mp = 0;
		}
289 290 291 292
		if (*mp == 0) {
			*lblock = blkno;
			*mp = read_index_page(ip, blkno);
		}
Dave Kleikamp's avatar
Dave Kleikamp committed
293
		if (*mp == 0) {
294
			jfs_err("free_index: error reading directory table");
Dave Kleikamp's avatar
Dave Kleikamp committed
295 296 297 298
			return 0;
		}

		slot =
299 300
		    (struct dir_table_slot *) ((char *) (*mp)->data +
					       page_offset);
Dave Kleikamp's avatar
Dave Kleikamp committed
301 302 303 304
	}
	return slot;
}

305
static inline void lock_index(tid_t tid, struct inode *ip, struct metapage * mp,
Dave Kleikamp's avatar
Dave Kleikamp committed
306 307
			      u32 index)
{
308 309 310
	struct tlock *tlck;
	struct linelock *llck;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
311 312

	tlck = txLock(tid, ip, mp, tlckDATA);
313
	llck = (struct linelock *) tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340

	if (llck->index >= llck->maxcnt)
		llck = txLinelock(llck);
	lv = &llck->lv[llck->index];

	/*
	 *      Linelock slot size is twice the size of directory table
	 *      slot size.  512 entries per page.
	 */
	lv->offset = ((index - 2) & 511) >> 1;
	lv->length = 1;
	llck->index++;
}

/*
 *	add_index()
 *
 *	Adds an entry to the directory index table.  This is used to provide
 *	each directory entry with a persistent index in which to resume
 *	directory traversals
 */
static u32 add_index(tid_t tid, struct inode *ip, s64 bn, int slot)
{
	struct super_block *sb = ip->i_sb;
	struct jfs_sb_info *sbi = JFS_SBI(sb);
	struct jfs_inode_info *jfs_ip = JFS_IP(ip);
	u64 blkno;
341
	struct dir_table_slot *dirtab_slot;
Dave Kleikamp's avatar
Dave Kleikamp committed
342
	u32 index;
343 344 345
	struct linelock *llck;
	struct lv *lv;
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
346 347 348
	s64 offset;
	uint page_offset;
	int rc;
349
	struct tlock *tlck;
Dave Kleikamp's avatar
Dave Kleikamp committed
350 351 352 353 354
	s64 xaddr;

	ASSERT(DO_INDEX(ip));

	if (jfs_ip->next_index < 2) {
355 356
		jfs_warn("add_index: next_index = %d.  Resetting!",
			   jfs_ip->next_index);
Dave Kleikamp's avatar
Dave Kleikamp committed
357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389
		jfs_ip->next_index = 2;
	}

	index = jfs_ip->next_index++;

	if (index <= MAX_INLINE_DIRTABLE_ENTRY) {
		/*
		 * i_size reflects size of index table, or 8 bytes per entry.
		 */
		ip->i_size = (loff_t) (index - 1) << 3;

		/*
		 * dir table fits inline within inode
		 */
		dirtab_slot = &jfs_ip->i_dirtable[index-2];
		dirtab_slot->flag = DIR_INDEX_VALID;
		dirtab_slot->slot = slot;
		DTSaddress(dirtab_slot, bn);

		set_cflag(COMMIT_Dirtable, ip);

		return index;
	}
	if (index == (MAX_INLINE_DIRTABLE_ENTRY + 1)) {
		/*
		 * It's time to move the inline table to an external
		 * page and begin to build the xtree
		 */

		/*
		 * Save the table, we're going to overwrite it with the
		 * xtree root
		 */
390
		struct dir_table_slot temp_table[12];
Dave Kleikamp's avatar
Dave Kleikamp committed
391 392 393 394 395 396 397 398 399 400 401 402 403 404
		memcpy(temp_table, &jfs_ip->i_dirtable, sizeof(temp_table));

		/*
		 * Initialize empty x-tree
		 */
		xtInitRoot(tid, ip);

		/*
		 * Allocate the first block & add it to the xtree
		 */
		xaddr = 0;
		if ((rc =
		     xtInsert(tid, ip, 0, 0, sbi->nbperpage,
			      &xaddr, 0))) {
405
			jfs_warn("add_index: xtInsert failed!");
406
			return -EPERM;
Dave Kleikamp's avatar
Dave Kleikamp committed
407 408 409 410
		}
		ip->i_size = PSIZE;
		ip->i_blocks += LBLK2PBLK(sb, sbi->nbperpage);

411
		if ((mp = get_index_page(ip, 0)) == 0) {
412
			jfs_err("add_index: get_metapage failed!");
Dave Kleikamp's avatar
Dave Kleikamp committed
413
			xtTruncate(tid, ip, 0, COMMIT_PWMAP);
414
			return -EPERM;
Dave Kleikamp's avatar
Dave Kleikamp committed
415 416
		}
		tlck = txLock(tid, ip, mp, tlckDATA);
417
		llck = (struct linelock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435
		ASSERT(llck->index == 0);
		lv = &llck->lv[0];

		lv->offset = 0;
		lv->length = 6;	/* tlckDATA slot size is 16 bytes */
		llck->index++;

		memcpy(mp->data, temp_table, sizeof(temp_table));

		mark_metapage_dirty(mp);
		release_metapage(mp);

		/*
		 * Logging is now directed by xtree tlocks
		 */
		clear_cflag(COMMIT_Dirtable, ip);
	}

436
	offset = (index - 2) * sizeof(struct dir_table_slot);
Dave Kleikamp's avatar
Dave Kleikamp committed
437 438 439 440 441 442 443 444 445 446
	page_offset = offset & (PSIZE - 1);
	blkno = ((offset + 1) >> L2PSIZE) << sbi->l2nbperpage;
	if (page_offset == 0) {
		/*
		 * This will be the beginning of a new page
		 */
		xaddr = 0;
		if ((rc =
		     xtInsert(tid, ip, 0, blkno, sbi->nbperpage,
			      &xaddr, 0))) {
447
			jfs_warn("add_index: xtInsert failed!");
Dave Kleikamp's avatar
Dave Kleikamp committed
448
			jfs_ip->next_index--;
449
			return -EPERM;
Dave Kleikamp's avatar
Dave Kleikamp committed
450 451 452 453
		}
		ip->i_size += PSIZE;
		ip->i_blocks += LBLK2PBLK(sb, sbi->nbperpage);

454
		if ((mp = get_index_page(ip, blkno)))
Dave Kleikamp's avatar
Dave Kleikamp committed
455 456 457 458
			memset(mp->data, 0, PSIZE);	/* Just looks better */
		else
			xtTruncate(tid, ip, offset, COMMIT_PWMAP);
	} else
459
		mp = read_index_page(ip, blkno);
Dave Kleikamp's avatar
Dave Kleikamp committed
460 461

	if (mp == 0) {
462
		jfs_err("add_index: get/read_metapage failed!");
463
		return -EPERM;
Dave Kleikamp's avatar
Dave Kleikamp committed
464 465 466 467 468
	}

	lock_index(tid, ip, mp, index);

	dirtab_slot =
469
	    (struct dir_table_slot *) ((char *) mp->data + page_offset);
Dave Kleikamp's avatar
Dave Kleikamp committed
470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486
	dirtab_slot->flag = DIR_INDEX_VALID;
	dirtab_slot->slot = slot;
	DTSaddress(dirtab_slot, bn);

	mark_metapage_dirty(mp);
	release_metapage(mp);

	return index;
}

/*
 *	free_index()
 *
 *	Marks an entry to the directory index table as free.
 */
static void free_index(tid_t tid, struct inode *ip, u32 index, u32 next)
{
487
	struct dir_table_slot *dirtab_slot;
488
	s64 lblock;
489
	struct metapage *mp = 0;
Dave Kleikamp's avatar
Dave Kleikamp committed
490

491
	dirtab_slot = find_index(ip, index, &mp, &lblock);
Dave Kleikamp's avatar
Dave Kleikamp committed
492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513

	if (dirtab_slot == 0)
		return;

	dirtab_slot->flag = DIR_INDEX_FREE;
	dirtab_slot->slot = dirtab_slot->addr1 = 0;
	dirtab_slot->addr2 = cpu_to_le32(next);

	if (mp) {
		lock_index(tid, ip, mp, index);
		mark_metapage_dirty(mp);
		release_metapage(mp);
	} else
		set_cflag(COMMIT_Dirtable, ip);
}

/*
 *	modify_index()
 *
 *	Changes an entry in the directory index table
 */
static void modify_index(tid_t tid, struct inode *ip, u32 index, s64 bn,
514
			 int slot, struct metapage ** mp, u64 *lblock)
Dave Kleikamp's avatar
Dave Kleikamp committed
515
{
516
	struct dir_table_slot *dirtab_slot;
Dave Kleikamp's avatar
Dave Kleikamp committed
517

518
	dirtab_slot = find_index(ip, index, mp, lblock);
Dave Kleikamp's avatar
Dave Kleikamp committed
519 520 521 522 523 524 525 526 527 528 529 530 531 532 533

	if (dirtab_slot == 0)
		return;

	DTSaddress(dirtab_slot, bn);
	dirtab_slot->slot = slot;

	if (*mp) {
		lock_index(tid, ip, *mp, index);
		mark_metapage_dirty(*mp);
	} else
		set_cflag(COMMIT_Dirtable, ip);
}

/*
534
 *	read_index()
Dave Kleikamp's avatar
Dave Kleikamp committed
535 536 537
 *
 *	reads a directory table slot
 */
538
static int read_index(struct inode *ip, u32 index,
539
		     struct dir_table_slot * dirtab_slot)
Dave Kleikamp's avatar
Dave Kleikamp committed
540
{
541
	s64 lblock;
542 543
	struct metapage *mp = 0;
	struct dir_table_slot *slot;
Dave Kleikamp's avatar
Dave Kleikamp committed
544

545
	slot = find_index(ip, index, &mp, &lblock);
Dave Kleikamp's avatar
Dave Kleikamp committed
546 547 548 549
	if (slot == 0) {
		return -EIO;
	}

550
	memcpy(dirtab_slot, slot, sizeof(struct dir_table_slot));
Dave Kleikamp's avatar
Dave Kleikamp committed
551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568

	if (mp)
		release_metapage(mp);

	return 0;
}

/*
 *	dtSearch()
 *
 * function:
 *	Search for the entry with specified key
 *
 * parameter:
 *
 * return: 0 - search result on stack, leaf page pinned;
 *	   errno - I/O error
 */
569 570
int dtSearch(struct inode *ip, struct component_name * key, ino_t * data,
	     struct btstack * btstack, int flag)
Dave Kleikamp's avatar
Dave Kleikamp committed
571 572 573 574
{
	int rc = 0;
	int cmp = 1;		/* init for empty page */
	s64 bn;
575
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
576 577 578
	dtpage_t *p;
	s8 *stbl;
	int base, index, lim;
579
	struct btframe *btsp;
Dave Kleikamp's avatar
Dave Kleikamp committed
580 581 582
	pxd_t *pxd;
	int psize = 288;	/* initial in-line directory */
	ino_t inumber;
583
	struct component_name ciKey;
Dave Kleikamp's avatar
Dave Kleikamp committed
584 585 586 587 588 589
	struct super_block *sb = ip->i_sb;

	ciKey.name =
	    (wchar_t *) kmalloc((JFS_NAME_MAX + 1) * sizeof(wchar_t),
				GFP_NOFS);
	if (ciKey.name == 0) {
590
		rc = -ENOMEM;
Dave Kleikamp's avatar
Dave Kleikamp committed
591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659
		goto dtSearch_Exit2;
	}


	/* uppercase search key for c-i directory */
	UniStrcpy(ciKey.name, key->name);
	ciKey.namlen = key->namlen;

	/* only uppercase if case-insensitive support is on */
	if ((JFS_SBI(sb)->mntflag & JFS_OS2) == JFS_OS2) {
		ciToUpper(&ciKey);
	}
	BT_CLR(btstack);	/* reset stack */

	/* init level count for max pages to split */
	btstack->nsplit = 1;

	/*
	 *      search down tree from root:
	 *
	 * between two consecutive entries of <Ki, Pi> and <Kj, Pj> of
	 * internal page, child page Pi contains entry with k, Ki <= K < Kj.
	 *
	 * if entry with search key K is not found
	 * internal page search find the entry with largest key Ki
	 * less than K which point to the child page to search;
	 * leaf page search find the entry with smallest key Kj
	 * greater than K so that the returned index is the position of
	 * the entry to be shifted right for insertion of new entry.
	 * for empty tree, search key is greater than any key of the tree.
	 *
	 * by convention, root bn = 0.
	 */
	for (bn = 0;;) {
		/* get/pin the page to search */
		DT_GETPAGE(ip, bn, mp, psize, p, rc);
		if (rc)
			goto dtSearch_Exit1;

		/* get sorted entry table of the page */
		stbl = DT_GETSTBL(p);

		/*
		 * binary search with search key K on the current page.
		 */
		for (base = 0, lim = p->header.nextindex; lim; lim >>= 1) {
			index = base + (lim >> 1);

			if (p->header.flag & BT_LEAF) {
				/* uppercase leaf name to compare */
				cmp =
				    ciCompare(&ciKey, p, stbl[index],
					      JFS_SBI(sb)->mntflag);
			} else {
				/* router key is in uppercase */

				cmp = dtCompare(&ciKey, p, stbl[index]);


			}
			if (cmp == 0) {
				/*
				 *      search hit
				 */
				/* search hit - leaf page:
				 * return the entry found
				 */
				if (p->header.flag & BT_LEAF) {
					inumber = le32_to_cpu(
660
			((struct ldtentry *) & p->slot[stbl[index]])->inumber);
Dave Kleikamp's avatar
Dave Kleikamp committed
661 662 663 664 665 666 667 668 669 670 671 672 673 674 675

					/*
					 * search for JFS_LOOKUP
					 */
					if (flag == JFS_LOOKUP) {
						*data = inumber;
						rc = 0;
						goto out;
					}

					/*
					 * search for JFS_CREATE
					 */
					if (flag == JFS_CREATE) {
						*data = inumber;
676
						rc = -EEXIST;
Dave Kleikamp's avatar
Dave Kleikamp committed
677 678 679 680 681 682 683 684 685
						goto out;
					}

					/*
					 * search for JFS_REMOVE or JFS_RENAME
					 */
					if ((flag == JFS_REMOVE ||
					     flag == JFS_RENAME) &&
					    *data != inumber) {
686
						rc = -ESTALE;
Dave Kleikamp's avatar
Dave Kleikamp committed
687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733
						goto out;
					}

					/*
					 * JFS_REMOVE|JFS_FINDDIR|JFS_RENAME
					 */
					/* save search result */
					*data = inumber;
					btsp = btstack->top;
					btsp->bn = bn;
					btsp->index = index;
					btsp->mp = mp;

					rc = 0;
					goto dtSearch_Exit1;
				}

				/* search hit - internal page:
				 * descend/search its child page
				 */
				goto getChild;
			}

			if (cmp > 0) {
				base = index + 1;
				--lim;
			}
		}

		/*
		 *      search miss
		 *
		 * base is the smallest index with key (Kj) greater than
		 * search key (K) and may be zero or (maxindex + 1) index.
		 */
		/*
		 * search miss - leaf page
		 *
		 * return location of entry (base) where new entry with
		 * search key K is to be inserted.
		 */
		if (p->header.flag & BT_LEAF) {
			/*
			 * search for JFS_LOOKUP, JFS_REMOVE, or JFS_RENAME
			 */
			if (flag == JFS_LOOKUP || flag == JFS_REMOVE ||
			    flag == JFS_RENAME) {
734
				rc = -ENOENT;
Dave Kleikamp's avatar
Dave Kleikamp committed
735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769
				goto out;
			}

			/*
			 * search for JFS_CREATE|JFS_FINDDIR:
			 *
			 * save search result
			 */
			*data = 0;
			btsp = btstack->top;
			btsp->bn = bn;
			btsp->index = base;
			btsp->mp = mp;

			rc = 0;
			goto dtSearch_Exit1;
		}

		/*
		 * search miss - internal page
		 *
		 * if base is non-zero, decrement base by one to get the parent
		 * entry of the child page to search.
		 */
		index = base ? base - 1 : base;

		/*
		 * go down to child page
		 */
	      getChild:
		/* update max. number of pages to split */
		if (btstack->nsplit >= 8) {
			/* Something's corrupted, mark filesytem dirty so
			 * chkdsk will fix it.
			 */
Dave Kleikamp's avatar
Dave Kleikamp committed
770
			jfs_error(sb, "stack overrun in dtSearch!");
771
			rc = -EIO;
Dave Kleikamp's avatar
Dave Kleikamp committed
772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811
			goto out;
		}
		btstack->nsplit++;

		/* push (bn, index) of the parent page/entry */
		BT_PUSH(btstack, bn, index);

		/* get the child page block number */
		pxd = (pxd_t *) & p->slot[stbl[index]];
		bn = addressPXD(pxd);
		psize = lengthPXD(pxd) << JFS_SBI(ip->i_sb)->l2bsize;

		/* unpin the parent page */
		DT_PUTPAGE(mp);
	}

      out:
	DT_PUTPAGE(mp);

      dtSearch_Exit1:

	kfree(ciKey.name);

      dtSearch_Exit2:

	return rc;
}


/*
 *	dtInsert()
 *
 * function: insert an entry to directory tree
 *
 * parameter:
 *
 * return: 0 - success;
 *	   errno - failure;
 */
int dtInsert(tid_t tid, struct inode *ip,
812
	 struct component_name * name, ino_t * fsn, struct btstack * btstack)
Dave Kleikamp's avatar
Dave Kleikamp committed
813 814
{
	int rc = 0;
815
	struct metapage *mp;	/* meta-page buffer */
Dave Kleikamp's avatar
Dave Kleikamp committed
816 817 818
	dtpage_t *p;		/* base B+-tree index page */
	s64 bn;
	int index;
819
	struct dtsplit split;	/* split information */
Dave Kleikamp's avatar
Dave Kleikamp committed
820
	ddata_t data;
821
	struct dt_lock *dtlck;
Dave Kleikamp's avatar
Dave Kleikamp committed
822
	int n;
823 824
	struct tlock *tlck;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
825 826 827 828 829 830 831 832 833 834 835 836 837 838

	/*
	 *      retrieve search result
	 *
	 * dtSearch() returns (leaf page pinned, index at which to insert).
	 * n.b. dtSearch() may return index of (maxindex + 1) of
	 * the full page.
	 */
	DT_GETSEARCH(ip, btstack->top, bn, mp, p, index);

	/*
	 *      insert entry for new key
	 */
	if (DO_INDEX(ip)) {
839
		if (JFS_IP(ip)->next_index == DIREND) {
Dave Kleikamp's avatar
Dave Kleikamp committed
840
			DT_PUTPAGE(mp);
841
			return -EMLINK;
Dave Kleikamp's avatar
Dave Kleikamp committed
842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878
		}
		n = NDTLEAF(name->namlen);
		data.leaf.tid = tid;
		data.leaf.ip = ip;
	} else {
		n = NDTLEAF_LEGACY(name->namlen);
		data.leaf.ip = 0;	/* signifies legacy directory format */
	}
	data.leaf.ino = cpu_to_le32(*fsn);

	/*
	 *      leaf page does not have enough room for new entry:
	 *
	 *      extend/split the leaf page;
	 *
	 * dtSplitUp() will insert the entry and unpin the leaf page.
	 */
	if (n > p->header.freecnt) {
		split.mp = mp;
		split.index = index;
		split.nslot = n;
		split.key = name;
		split.data = &data;
		rc = dtSplitUp(tid, ip, &split, btstack);
		return rc;
	}

	/*
	 *      leaf page does have enough room for new entry:
	 *
	 *      insert the new data entry into the leaf page;
	 */
	BT_MARK_DIRTY(mp, ip);
	/*
	 * acquire a transaction lock on the leaf page
	 */
	tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
879
	dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
880
	ASSERT(dtlck->index == 0);
881
	lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
882 883 884 885 886 887 888 889 890 891 892

	/* linelock header */
	lv->offset = 0;
	lv->length = 1;
	dtlck->index++;

	dtInsertEntry(p, index, name, &data, &dtlck);

	/* linelock stbl of non-root leaf page */
	if (!(p->header.flag & BT_ROOT)) {
		if (dtlck->index >= dtlck->maxcnt)
893 894
			dtlck = (struct dt_lock *) txLinelock(dtlck);
		lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920
		n = index >> L2DTSLOTSIZE;
		lv->offset = p->header.stblindex + n;
		lv->length =
		    ((p->header.nextindex - 1) >> L2DTSLOTSIZE) - n + 1;
		dtlck->index++;
	}

	/* unpin the leaf page */
	DT_PUTPAGE(mp);

	return 0;
}


/*
 *	dtSplitUp()
 *
 * function: propagate insertion bottom up;
 *
 * parameter:
 *
 * return: 0 - success;
 *	   errno - failure;
 * 	leaf page unpinned;
 */
static int dtSplitUp(tid_t tid,
921
	  struct inode *ip, struct dtsplit * split, struct btstack * btstack)
Dave Kleikamp's avatar
Dave Kleikamp committed
922 923 924
{
	struct jfs_sb_info *sbi = JFS_SBI(ip->i_sb);
	int rc = 0;
925
	struct metapage *smp;
Dave Kleikamp's avatar
Dave Kleikamp committed
926
	dtpage_t *sp;		/* split page */
927
	struct metapage *rmp;
Dave Kleikamp's avatar
Dave Kleikamp committed
928 929
	dtpage_t *rp;		/* new right page split from sp */
	pxd_t rpxd;		/* new right page extent descriptor */
930
	struct metapage *lmp;
Dave Kleikamp's avatar
Dave Kleikamp committed
931 932
	dtpage_t *lp;		/* left child page */
	int skip;		/* index of entry of insertion */
933
	struct btframe *parent;	/* parent page entry on traverse stack */
Dave Kleikamp's avatar
Dave Kleikamp committed
934 935
	s64 xaddr, nxaddr;
	int xlen, xsize;
936
	struct pxdlist pxdlist;
Dave Kleikamp's avatar
Dave Kleikamp committed
937
	pxd_t *pxd;
938
	struct component_name key = { 0, 0 };
Dave Kleikamp's avatar
Dave Kleikamp committed
939 940
	ddata_t *data = split->data;
	int n;
941 942 943
	struct dt_lock *dtlck;
	struct tlock *tlck;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
944 945 946 947 948 949 950 951 952 953

	/* get split page */
	smp = split->mp;
	sp = DT_PAGE(ip, smp);

	key.name =
	    (wchar_t *) kmalloc((JFS_NAME_MAX + 2) * sizeof(wchar_t),
				GFP_NOFS);
	if (key.name == 0) {
		DT_PUTPAGE(smp);
954
		rc = -ENOMEM;
Dave Kleikamp's avatar
Dave Kleikamp committed
955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179
		goto dtSplitUp_Exit;
	}

	/*
	 *      split leaf page
	 *
	 * The split routines insert the new entry, and
	 * acquire txLock as appropriate.
	 */
	/*
	 *      split root leaf page:
	 */
	if (sp->header.flag & BT_ROOT) {
		/*
		 * allocate a single extent child page
		 */
		xlen = 1;
		n = sbi->bsize >> L2DTSLOTSIZE;
		n -= (n + 31) >> L2DTSLOTSIZE;	/* stbl size */
		n -= DTROOTMAXSLOT - sp->header.freecnt; /* header + entries */
		if (n <= split->nslot)
			xlen++;
		if ((rc = dbAlloc(ip, 0, (s64) xlen, &xaddr)))
			goto freeKeyName;

		pxdlist.maxnpxd = 1;
		pxdlist.npxd = 0;
		pxd = &pxdlist.pxd[0];
		PXDaddress(pxd, xaddr);
		PXDlength(pxd, xlen);
		split->pxdlist = &pxdlist;
		rc = dtSplitRoot(tid, ip, split, &rmp);

		DT_PUTPAGE(rmp);
		DT_PUTPAGE(smp);

		goto freeKeyName;
	}

	/*
	 *      extend first leaf page
	 *
	 * extend the 1st extent if less than buffer page size
	 * (dtExtendPage() reurns leaf page unpinned)
	 */
	pxd = &sp->header.self;
	xlen = lengthPXD(pxd);
	xsize = xlen << sbi->l2bsize;
	if (xsize < PSIZE) {
		xaddr = addressPXD(pxd);
		n = xsize >> L2DTSLOTSIZE;
		n -= (n + 31) >> L2DTSLOTSIZE;	/* stbl size */
		if ((n + sp->header.freecnt) <= split->nslot)
			n = xlen + (xlen << 1);
		else
			n = xlen;
		if ((rc = dbReAlloc(sbi->ipbmap, xaddr, (s64) xlen,
				    (s64) n, &nxaddr)))
			goto extendOut;

		pxdlist.maxnpxd = 1;
		pxdlist.npxd = 0;
		pxd = &pxdlist.pxd[0];
		PXDaddress(pxd, nxaddr)
		    PXDlength(pxd, xlen + n);
		split->pxdlist = &pxdlist;
		if ((rc = dtExtendPage(tid, ip, split, btstack))) {
			nxaddr = addressPXD(pxd);
			if (xaddr != nxaddr) {
				/* free relocated extent */
				xlen = lengthPXD(pxd);
				dbFree(ip, nxaddr, (s64) xlen);
			} else {
				/* free extended delta */
				xlen = lengthPXD(pxd) - n;
				xaddr = addressPXD(pxd) + xlen;
				dbFree(ip, xaddr, (s64) n);
			}
		}

	      extendOut:
		DT_PUTPAGE(smp);
		goto freeKeyName;
	}

	/*
	 *      split leaf page <sp> into <sp> and a new right page <rp>.
	 *
	 * return <rp> pinned and its extent descriptor <rpxd>
	 */
	/*
	 * allocate new directory page extent and
	 * new index page(s) to cover page split(s)
	 *
	 * allocation hint: ?
	 */
	n = btstack->nsplit;
	pxdlist.maxnpxd = pxdlist.npxd = 0;
	xlen = sbi->nbperpage;
	for (pxd = pxdlist.pxd; n > 0; n--, pxd++) {
		if ((rc = dbAlloc(ip, 0, (s64) xlen, &xaddr)) == 0) {
			PXDaddress(pxd, xaddr);
			PXDlength(pxd, xlen);
			pxdlist.maxnpxd++;
			continue;
		}

		DT_PUTPAGE(smp);

		/* undo allocation */
		goto splitOut;
	}

	split->pxdlist = &pxdlist;
	if ((rc = dtSplitPage(tid, ip, split, &rmp, &rp, &rpxd))) {
		DT_PUTPAGE(smp);

		/* undo allocation */
		goto splitOut;
	}

	/*
	 * propagate up the router entry for the leaf page just split
	 *
	 * insert a router entry for the new page into the parent page,
	 * propagate the insert/split up the tree by walking back the stack
	 * of (bn of parent page, index of child page entry in parent page)
	 * that were traversed during the search for the page that split.
	 *
	 * the propagation of insert/split up the tree stops if the root
	 * splits or the page inserted into doesn't have to split to hold
	 * the new entry.
	 *
	 * the parent entry for the split page remains the same, and
	 * a new entry is inserted at its right with the first key and
	 * block number of the new right page.
	 *
	 * There are a maximum of 4 pages pinned at any time:
	 * two children, left parent and right parent (when the parent splits).
	 * keep the child pages pinned while working on the parent.
	 * make sure that all pins are released at exit.
	 */
	while ((parent = BT_POP(btstack)) != NULL) {
		/* parent page specified by stack frame <parent> */

		/* keep current child pages (<lp>, <rp>) pinned */
		lmp = smp;
		lp = sp;

		/*
		 * insert router entry in parent for new right child page <rp>
		 */
		/* get the parent page <sp> */
		DT_GETPAGE(ip, parent->bn, smp, PSIZE, sp, rc);
		if (rc) {
			DT_PUTPAGE(lmp);
			DT_PUTPAGE(rmp);
			goto splitOut;
		}

		/*
		 * The new key entry goes ONE AFTER the index of parent entry,
		 * because the split was to the right.
		 */
		skip = parent->index + 1;

		/*
		 * compute the key for the router entry
		 *
		 * key suffix compression:
		 * for internal pages that have leaf pages as children,
		 * retain only what's needed to distinguish between
		 * the new entry and the entry on the page to its left.
		 * If the keys compare equal, retain the entire key.
		 *
		 * note that compression is performed only at computing
		 * router key at the lowest internal level.
		 * further compression of the key between pairs of higher
		 * level internal pages loses too much information and
		 * the search may fail.
		 * (e.g., two adjacent leaf pages of {a, ..., x} {xx, ...,}
		 * results in two adjacent parent entries (a)(xx).
		 * if split occurs between these two entries, and
		 * if compression is applied, the router key of parent entry
		 * of right page (x) will divert search for x into right
		 * subtree and miss x in the left subtree.)
		 *
		 * the entire key must be retained for the next-to-leftmost
		 * internal key at any level of the tree, or search may fail
		 * (e.g., ?)
		 */
		switch (rp->header.flag & BT_TYPE) {
		case BT_LEAF:
			/*
			 * compute the length of prefix for suffix compression
			 * between last entry of left page and first entry
			 * of right page
			 */
			if ((sp->header.flag & BT_ROOT && skip > 1) ||
			    sp->header.prev != 0 || skip > 1) {
				/* compute uppercase router prefix key */
				ciGetLeafPrefixKey(lp,
						   lp->header.nextindex - 1,
						   rp, 0, &key, sbi->mntflag);
			} else {
				/* next to leftmost entry of
				   lowest internal level */

				/* compute uppercase router key */
				dtGetKey(rp, 0, &key, sbi->mntflag);
				key.name[key.namlen] = 0;

				if ((sbi->mntflag & JFS_OS2) == JFS_OS2)
					ciToUpper(&key);
			}

			n = NDTINTERNAL(key.namlen);
			break;

		case BT_INTERNAL:
			dtGetKey(rp, 0, &key, sbi->mntflag);
			n = NDTINTERNAL(key.namlen);
			break;

		default:
1180
			jfs_err("dtSplitUp(): UFO!");
Dave Kleikamp's avatar
Dave Kleikamp committed
1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
			break;
		}

		/* unpin left child page */
		DT_PUTPAGE(lmp);

		/*
		 * compute the data for the router entry
		 */
		data->xd = rpxd;	/* child page xd */

		/*
		 * parent page is full - split the parent page
		 */
		if (n > sp->header.freecnt) {
			/* init for parent page split */
			split->mp = smp;
			split->index = skip;	/* index at insert */
			split->nslot = n;
			split->key = &key;
			/* split->data = data; */

			/* unpin right child page */
			DT_PUTPAGE(rmp);

			/* The split routines insert the new entry,
			 * acquire txLock as appropriate.
			 * return <rp> pinned and its block number <rbn>.
			 */
			rc = (sp->header.flag & BT_ROOT) ?
			    dtSplitRoot(tid, ip, split, &rmp) :
			    dtSplitPage(tid, ip, split, &rmp, &rp, &rpxd);
			if (rc) {
				DT_PUTPAGE(smp);
				goto splitOut;
			}

			/* smp and rmp are pinned */
		}
		/*
		 * parent page is not full - insert router entry in parent page
		 */
		else {
			BT_MARK_DIRTY(smp, ip);
			/*
			 * acquire a transaction lock on the parent page
			 */
			tlck = txLock(tid, ip, smp, tlckDTREE | tlckENTRY);
1229
			dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
1230
			ASSERT(dtlck->index == 0);
1231
			lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288

			/* linelock header */
			lv->offset = 0;
			lv->length = 1;
			dtlck->index++;

			/* linelock stbl of non-root parent page */
			if (!(sp->header.flag & BT_ROOT)) {
				lv++;
				n = skip >> L2DTSLOTSIZE;
				lv->offset = sp->header.stblindex + n;
				lv->length =
				    ((sp->header.nextindex -
				      1) >> L2DTSLOTSIZE) - n + 1;
				dtlck->index++;
			}

			dtInsertEntry(sp, skip, &key, data, &dtlck);

			/* exit propagate up */
			break;
		}
	}

	/* unpin current split and its right page */
	DT_PUTPAGE(smp);
	DT_PUTPAGE(rmp);

	/*
	 * free remaining extents allocated for split
	 */
      splitOut:
	n = pxdlist.npxd;
	pxd = &pxdlist.pxd[n];
	for (; n < pxdlist.maxnpxd; n++, pxd++)
		dbFree(ip, addressPXD(pxd), (s64) lengthPXD(pxd));

      freeKeyName:
	kfree(key.name);

      dtSplitUp_Exit:

	return rc;
}


/*
 *	dtSplitPage()
 *
 * function: Split a non-root page of a btree.
 *
 * parameter:
 *
 * return: 0 - success;
 *	   errno - failure;
 *	return split and new page pinned;
 */
1289 1290
static int dtSplitPage(tid_t tid, struct inode *ip, struct dtsplit * split,
	    struct metapage ** rmpp, dtpage_t ** rpp, pxd_t * rpxdp)
Dave Kleikamp's avatar
Dave Kleikamp committed
1291 1292 1293
{
	struct super_block *sb = ip->i_sb;
	int rc = 0;
1294
	struct metapage *smp;
Dave Kleikamp's avatar
Dave Kleikamp committed
1295
	dtpage_t *sp;
1296
	struct metapage *rmp;
Dave Kleikamp's avatar
Dave Kleikamp committed
1297 1298
	dtpage_t *rp;		/* new right page allocated */
	s64 rbn;		/* new right page block number */
1299
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
1300 1301
	dtpage_t *p;
	s64 nextbn;
1302
	struct pxdlist *pxdlist;
Dave Kleikamp's avatar
Dave Kleikamp committed
1303 1304
	pxd_t *pxd;
	int skip, nextindex, half, left, nxt, off, si;
1305 1306
	struct ldtentry *ldtentry;
	struct idtentry *idtentry;
Dave Kleikamp's avatar
Dave Kleikamp committed
1307
	u8 *stbl;
1308
	struct dtslot *f;
Dave Kleikamp's avatar
Dave Kleikamp committed
1309 1310
	int fsi, stblsize;
	int n;
1311 1312 1313 1314
	struct dt_lock *sdtlck, *rdtlck;
	struct tlock *tlck;
	struct dt_lock *dtlck;
	struct lv *slv, *rlv, *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328

	/* get split page */
	smp = split->mp;
	sp = DT_PAGE(ip, smp);

	/*
	 * allocate the new right page for the split
	 */
	pxdlist = split->pxdlist;
	pxd = &pxdlist->pxd[pxdlist->npxd];
	pxdlist->npxd++;
	rbn = addressPXD(pxd);
	rmp = get_metapage(ip, rbn, PSIZE, 1);
	if (rmp == NULL)
1329
		return -EIO;
Dave Kleikamp's avatar
Dave Kleikamp committed
1330

1331
	jfs_info("dtSplitPage: ip:0x%p smp:0x%p rmp:0x%p", ip, smp, rmp);
Dave Kleikamp's avatar
Dave Kleikamp committed
1332 1333 1334 1335 1336 1337

	BT_MARK_DIRTY(rmp, ip);
	/*
	 * acquire a transaction lock on the new right page
	 */
	tlck = txLock(tid, ip, rmp, tlckDTREE | tlckNEW);
1338
	rdtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350

	rp = (dtpage_t *) rmp->data;
	*rpp = rp;
	rp->header.self = *pxd;

	BT_MARK_DIRTY(smp, ip);
	/*
	 * acquire a transaction lock on the split page
	 *
	 * action:
	 */
	tlck = txLock(tid, ip, smp, tlckDTREE | tlckENTRY);
1351
	sdtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
1352 1353 1354

	/* linelock header of split page */
	ASSERT(sdtlck->index == 0);
1355
	slv = & sdtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400
	slv->offset = 0;
	slv->length = 1;
	sdtlck->index++;

	/*
	 * initialize/update sibling pointers between sp and rp
	 */
	nextbn = le64_to_cpu(sp->header.next);
	rp->header.next = cpu_to_le64(nextbn);
	rp->header.prev = cpu_to_le64(addressPXD(&sp->header.self));
	sp->header.next = cpu_to_le64(rbn);

	/*
	 * initialize new right page
	 */
	rp->header.flag = sp->header.flag;

	/* compute sorted entry table at start of extent data area */
	rp->header.nextindex = 0;
	rp->header.stblindex = 1;

	n = PSIZE >> L2DTSLOTSIZE;
	rp->header.maxslot = n;
	stblsize = (n + 31) >> L2DTSLOTSIZE;	/* in unit of slot */

	/* init freelist */
	fsi = rp->header.stblindex + stblsize;
	rp->header.freelist = fsi;
	rp->header.freecnt = rp->header.maxslot - fsi;

	/*
	 *      sequential append at tail: append without split
	 *
	 * If splitting the last page on a level because of appending
	 * a entry to it (skip is maxentry), it's likely that the access is
	 * sequential. Adding an empty page on the side of the level is less
	 * work and can push the fill factor much higher than normal.
	 * If we're wrong it's no big deal, we'll just do the split the right
	 * way next time.
	 * (It may look like it's equally easy to do a similar hack for
	 * reverse sorted data, that is, split the tree left,
	 * but it's not. Be my guest.)
	 */
	if (nextbn == 0 && split->index == sp->header.nextindex) {
		/* linelock header + stbl (first slot) of new page */
1401
		rlv = & rdtlck->lv[rdtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
		rlv->offset = 0;
		rlv->length = 2;
		rdtlck->index++;

		/*
		 * initialize freelist of new right page
		 */
		f = &rp->slot[fsi];
		for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
			f->next = fsi;
		f->next = -1;

		/* insert entry at the first entry of the new right page */
		dtInsertEntry(rp, 0, split->key, split->data, &rdtlck);

		goto out;
	}

	/*
	 *      non-sequential insert (at possibly middle page)
	 */

	/*
	 * update prev pointer of previous right sibling page;
	 */
	if (nextbn != 0) {
		DT_GETPAGE(ip, nextbn, mp, PSIZE, p, rc);
		if (rc)
			return rc;

		BT_MARK_DIRTY(mp, ip);
		/*
		 * acquire a transaction lock on the next page
		 */
		tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
1437 1438
		jfs_info("dtSplitPage: tlck = 0x%p, ip = 0x%p, mp=0x%p",
			tlck, ip, mp);
1439
		dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
1440 1441

		/* linelock header of previous right sibling page */
1442
		lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474
		lv->offset = 0;
		lv->length = 1;
		dtlck->index++;

		p->header.prev = cpu_to_le64(rbn);

		DT_PUTPAGE(mp);
	}

	/*
	 * split the data between the split and right pages.
	 */
	skip = split->index;
	half = (PSIZE >> L2DTSLOTSIZE) >> 1;	/* swag */
	left = 0;

	/*
	 *      compute fill factor for split pages
	 *
	 * <nxt> traces the next entry to move to rp
	 * <off> traces the next entry to stay in sp
	 */
	stbl = (u8 *) & sp->slot[sp->header.stblindex];
	nextindex = sp->header.nextindex;
	for (nxt = off = 0; nxt < nextindex; ++off) {
		if (off == skip)
			/* check for fill factor with new entry size */
			n = split->nslot;
		else {
			si = stbl[nxt];
			switch (sp->header.flag & BT_TYPE) {
			case BT_LEAF:
1475
				ldtentry = (struct ldtentry *) & sp->slot[si];
Dave Kleikamp's avatar
Dave Kleikamp committed
1476 1477 1478 1479 1480 1481 1482 1483
				if (DO_INDEX(ip))
					n = NDTLEAF(ldtentry->namlen);
				else
					n = NDTLEAF_LEGACY(ldtentry->
							   namlen);
				break;

			case BT_INTERNAL:
1484
				idtentry = (struct idtentry *) & sp->slot[si];
Dave Kleikamp's avatar
Dave Kleikamp committed
1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510
				n = NDTINTERNAL(idtentry->namlen);
				break;

			default:
				break;
			}

			++nxt;	/* advance to next entry to move in sp */
		}

		left += n;
		if (left >= half)
			break;
	}

	/* <nxt> poins to the 1st entry to move */

	/*
	 *      move entries to right page
	 *
	 * dtMoveEntry() initializes rp and reserves entry for insertion
	 *
	 * split page moved out entries are linelocked;
	 * new/right page moved in entries are linelocked;
	 */
	/* linelock header + stbl of new right page */
1511
	rlv = & rdtlck->lv[rdtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532
	rlv->offset = 0;
	rlv->length = 5;
	rdtlck->index++;

	dtMoveEntry(sp, nxt, rp, &sdtlck, &rdtlck, DO_INDEX(ip));

	sp->header.nextindex = nxt;

	/*
	 * finalize freelist of new right page
	 */
	fsi = rp->header.freelist;
	f = &rp->slot[fsi];
	for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
		f->next = fsi;
	f->next = -1;

	/*
	 * Update directory index table for entries now in right page
	 */
	if ((rp->header.flag & BT_LEAF) && DO_INDEX(ip)) {
1533 1534
		s64 lblock;

Dave Kleikamp's avatar
Dave Kleikamp committed
1535 1536 1537
		mp = 0;
		stbl = DT_GETSTBL(rp);
		for (n = 0; n < rp->header.nextindex; n++) {
1538
			ldtentry = (struct ldtentry *) & rp->slot[stbl[n]];
Dave Kleikamp's avatar
Dave Kleikamp committed
1539
			modify_index(tid, ip, le32_to_cpu(ldtentry->index),
1540
				     rbn, n, &mp, &lblock);
Dave Kleikamp's avatar
Dave Kleikamp committed
1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554
		}
		if (mp)
			release_metapage(mp);
	}

	/*
	 * the skipped index was on the left page,
	 */
	if (skip <= off) {
		/* insert the new entry in the split page */
		dtInsertEntry(sp, skip, split->key, split->data, &sdtlck);

		/* linelock stbl of split page */
		if (sdtlck->index >= sdtlck->maxcnt)
1555 1556
			sdtlck = (struct dt_lock *) txLinelock(sdtlck);
		slv = & sdtlck->lv[sdtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579
		n = skip >> L2DTSLOTSIZE;
		slv->offset = sp->header.stblindex + n;
		slv->length =
		    ((sp->header.nextindex - 1) >> L2DTSLOTSIZE) - n + 1;
		sdtlck->index++;
	}
	/*
	 * the skipped index was on the right page,
	 */
	else {
		/* adjust the skip index to reflect the new position */
		skip -= nxt;

		/* insert the new entry in the right page */
		dtInsertEntry(rp, skip, split->key, split->data, &rdtlck);
	}

      out:
	*rmpp = rmp;
	*rpxdp = *pxd;

	ip->i_blocks += LBLK2PBLK(sb, lengthPXD(pxd));

1580
	return rc;
Dave Kleikamp's avatar
Dave Kleikamp committed
1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595
}


/*
 *	dtExtendPage()
 *
 * function: extend 1st/only directory leaf page
 *
 * parameter:
 *
 * return: 0 - success;
 *	   errno - failure;
 *	return extended page pinned;
 */
static int dtExtendPage(tid_t tid,
1596
	     struct inode *ip, struct dtsplit * split, struct btstack * btstack)
Dave Kleikamp's avatar
Dave Kleikamp committed
1597 1598 1599
{
	struct super_block *sb = ip->i_sb;
	int rc;
1600
	struct metapage *smp, *pmp, *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
1601
	dtpage_t *sp, *pp;
1602
	struct pxdlist *pxdlist;
Dave Kleikamp's avatar
Dave Kleikamp committed
1603 1604 1605 1606 1607
	pxd_t *pxd, *tpxd;
	int xlen, xsize;
	int newstblindex, newstblsize;
	int oldstblindex, oldstblsize;
	int fsi, last;
1608 1609
	struct dtslot *f;
	struct btframe *parent;
Dave Kleikamp's avatar
Dave Kleikamp committed
1610
	int n;
1611
	struct dt_lock *dtlck;
Dave Kleikamp's avatar
Dave Kleikamp committed
1612
	s64 xaddr, txaddr;
1613 1614 1615
	struct tlock *tlck;
	struct pxd_lock *pxdlock;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
1616
	uint type;
1617
	struct ldtentry *ldtentry;
Dave Kleikamp's avatar
Dave Kleikamp committed
1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649
	u8 *stbl;

	/* get page to extend */
	smp = split->mp;
	sp = DT_PAGE(ip, smp);

	/* get parent/root page */
	parent = BT_POP(btstack);
	DT_GETPAGE(ip, parent->bn, pmp, PSIZE, pp, rc);
	if (rc)
		return (rc);

	/*
	 *      extend the extent
	 */
	pxdlist = split->pxdlist;
	pxd = &pxdlist->pxd[pxdlist->npxd];
	pxdlist->npxd++;

	xaddr = addressPXD(pxd);
	tpxd = &sp->header.self;
	txaddr = addressPXD(tpxd);
	/* in-place extension */
	if (xaddr == txaddr) {
		type = tlckEXTEND;
	}
	/* relocation */
	else {
		type = tlckNEW;

		/* save moved extent descriptor for later free */
		tlck = txMaplock(tid, ip, tlckDTREE | tlckRELOCATE);
1650
		pxdlock = (struct pxd_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
1651 1652 1653 1654 1655 1656 1657 1658
		pxdlock->flag = mlckFREEPXD;
		pxdlock->pxd = sp->header.self;
		pxdlock->index = 1;

		/*
		 * Update directory index table to reflect new page address
		 */
		if (DO_INDEX(ip)) {
1659 1660
			s64 lblock;

Dave Kleikamp's avatar
Dave Kleikamp committed
1661 1662 1663 1664
			mp = 0;
			stbl = DT_GETSTBL(sp);
			for (n = 0; n < sp->header.nextindex; n++) {
				ldtentry =
1665
				    (struct ldtentry *) & sp->slot[stbl[n]];
Dave Kleikamp's avatar
Dave Kleikamp committed
1666 1667
				modify_index(tid, ip,
					     le32_to_cpu(ldtentry->index),
1668
					     xaddr, n, &mp, &lblock);
Dave Kleikamp's avatar
Dave Kleikamp committed
1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
			}
			if (mp)
				release_metapage(mp);
		}
	}

	/*
	 *      extend the page
	 */
	sp->header.self = *pxd;

1680
	jfs_info("dtExtendPage: ip:0x%p smp:0x%p sp:0x%p", ip, smp, sp);
Dave Kleikamp's avatar
Dave Kleikamp committed
1681 1682 1683 1684 1685 1686

	BT_MARK_DIRTY(smp, ip);
	/*
	 * acquire a transaction lock on the extended/leaf page
	 */
	tlck = txLock(tid, ip, smp, tlckDTREE | type);
1687 1688
	dtlck = (struct dt_lock *) & tlck->lock;
	lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795

	/* update buffer extent descriptor of extended page */
	xlen = lengthPXD(pxd);
	xsize = xlen << JFS_SBI(sb)->l2bsize;
#ifdef _STILL_TO_PORT
	bmSetXD(smp, xaddr, xsize);
#endif				/*  _STILL_TO_PORT */

	/*
	 * copy old stbl to new stbl at start of extended area
	 */
	oldstblindex = sp->header.stblindex;
	oldstblsize = (sp->header.maxslot + 31) >> L2DTSLOTSIZE;
	newstblindex = sp->header.maxslot;
	n = xsize >> L2DTSLOTSIZE;
	newstblsize = (n + 31) >> L2DTSLOTSIZE;
	memcpy(&sp->slot[newstblindex], &sp->slot[oldstblindex],
	       sp->header.nextindex);

	/*
	 * in-line extension: linelock old area of extended page
	 */
	if (type == tlckEXTEND) {
		/* linelock header */
		lv->offset = 0;
		lv->length = 1;
		dtlck->index++;
		lv++;

		/* linelock new stbl of extended page */
		lv->offset = newstblindex;
		lv->length = newstblsize;
	}
	/*
	 * relocation: linelock whole relocated area
	 */
	else {
		lv->offset = 0;
		lv->length = sp->header.maxslot + newstblsize;
	}

	dtlck->index++;

	sp->header.maxslot = n;
	sp->header.stblindex = newstblindex;
	/* sp->header.nextindex remains the same */

	/*
	 * add old stbl region at head of freelist
	 */
	fsi = oldstblindex;
	f = &sp->slot[fsi];
	last = sp->header.freelist;
	for (n = 0; n < oldstblsize; n++, fsi++, f++) {
		f->next = last;
		last = fsi;
	}
	sp->header.freelist = last;
	sp->header.freecnt += oldstblsize;

	/*
	 * append free region of newly extended area at tail of freelist
	 */
	/* init free region of newly extended area */
	fsi = n = newstblindex + newstblsize;
	f = &sp->slot[fsi];
	for (fsi++; fsi < sp->header.maxslot; f++, fsi++)
		f->next = fsi;
	f->next = -1;

	/* append new free region at tail of old freelist */
	fsi = sp->header.freelist;
	if (fsi == -1)
		sp->header.freelist = n;
	else {
		do {
			f = &sp->slot[fsi];
			fsi = f->next;
		} while (fsi != -1);

		f->next = n;
	}

	sp->header.freecnt += sp->header.maxslot - n;

	/*
	 * insert the new entry
	 */
	dtInsertEntry(sp, split->index, split->key, split->data, &dtlck);

	BT_MARK_DIRTY(pmp, ip);
	/*
	 * linelock any freeslots residing in old extent
	 */
	if (type == tlckEXTEND) {
		n = sp->header.maxslot >> 2;
		if (sp->header.freelist < n)
			dtLinelockFreelist(sp, n, &dtlck);
	}

	/*
	 *      update parent entry on the parent/root page
	 */
	/*
	 * acquire a transaction lock on the parent/root page
	 */
	tlck = txLock(tid, ip, pmp, tlckDTREE | tlckENTRY);
1796 1797
	dtlck = (struct dt_lock *) & tlck->lock;
	lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841

	/* linelock parent entry - 1st slot */
	lv->offset = 1;
	lv->length = 1;
	dtlck->index++;

	/* update the parent pxd for page extension */
	tpxd = (pxd_t *) & pp->slot[1];
	*tpxd = *pxd;

	/* Since the directory might have an EA and/or ACL associated with it
	 * we need to make sure we take that into account when setting the
	 * i_nblocks
	 */
	ip->i_blocks = LBLK2PBLK(ip->i_sb, xlen +
				 ((JFS_IP(ip)->ea.flag & DXD_EXTENT) ?
				  lengthDXD(&JFS_IP(ip)->ea) : 0) +
				 ((JFS_IP(ip)->acl.flag & DXD_EXTENT) ?
				  lengthDXD(&JFS_IP(ip)->acl) : 0));

	DT_PUTPAGE(pmp);
	return 0;
}


/*
 *	dtSplitRoot()
 *
 * function:
 *	split the full root page into
 *	original/root/split page and new right page
 *	i.e., root remains fixed in tree anchor (inode) and
 *	the root is copied to a single new right child page
 *	since root page << non-root page, and
 *	the split root page contains a single entry for the
 *	new right child page.
 *
 * parameter:
 *
 * return: 0 - success;
 *	   errno - failure;
 *	return new page pinned;
 */
static int dtSplitRoot(tid_t tid,
1842
	    struct inode *ip, struct dtsplit * split, struct metapage ** rmpp)
Dave Kleikamp's avatar
Dave Kleikamp committed
1843 1844
{
	struct super_block *sb = ip->i_sb;
1845
	struct metapage *smp;
Dave Kleikamp's avatar
Dave Kleikamp committed
1846
	dtroot_t *sp;
1847
	struct metapage *rmp;
Dave Kleikamp's avatar
Dave Kleikamp committed
1848 1849 1850 1851
	dtpage_t *rp;
	s64 rbn;
	int xlen;
	int xsize;
1852
	struct dtslot *f;
Dave Kleikamp's avatar
Dave Kleikamp committed
1853 1854
	s8 *stbl;
	int fsi, stblsize, n;
1855
	struct idtentry *s;
Dave Kleikamp's avatar
Dave Kleikamp committed
1856
	pxd_t *ppxd;
1857
	struct pxdlist *pxdlist;
Dave Kleikamp's avatar
Dave Kleikamp committed
1858
	pxd_t *pxd;
1859 1860 1861
	struct dt_lock *dtlck;
	struct tlock *tlck;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886

	/* get split root page */
	smp = split->mp;
	sp = &JFS_IP(ip)->i_dtroot;

	/*
	 *      allocate/initialize a single (right) child page
	 *
	 * N.B. at first split, a one (or two) block to fit new entry
	 * is allocated; at subsequent split, a full page is allocated;
	 */
	pxdlist = split->pxdlist;
	pxd = &pxdlist->pxd[pxdlist->npxd];
	pxdlist->npxd++;
	rbn = addressPXD(pxd);
	xlen = lengthPXD(pxd);
	xsize = xlen << JFS_SBI(sb)->l2bsize;
	rmp = get_metapage(ip, rbn, xsize, 1);
	rp = rmp->data;

	BT_MARK_DIRTY(rmp, ip);
	/*
	 * acquire a transaction lock on the new right page
	 */
	tlck = txLock(tid, ip, rmp, tlckDTREE | tlckNEW);
1887
	dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901

	rp->header.flag =
	    (sp->header.flag & BT_LEAF) ? BT_LEAF : BT_INTERNAL;
	rp->header.self = *pxd;

	/* initialize sibling pointers */
	rp->header.next = 0;
	rp->header.prev = 0;

	/*
	 *      move in-line root page into new right page extent
	 */
	/* linelock header + copied entries + new stbl (1st slot) in new page */
	ASSERT(dtlck->index == 0);
1902
	lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950
	lv->offset = 0;
	lv->length = 10;	/* 1 + 8 + 1 */
	dtlck->index++;

	n = xsize >> L2DTSLOTSIZE;
	rp->header.maxslot = n;
	stblsize = (n + 31) >> L2DTSLOTSIZE;

	/* copy old stbl to new stbl at start of extended area */
	rp->header.stblindex = DTROOTMAXSLOT;
	stbl = (s8 *) & rp->slot[DTROOTMAXSLOT];
	memcpy(stbl, sp->header.stbl, sp->header.nextindex);
	rp->header.nextindex = sp->header.nextindex;

	/* copy old data area to start of new data area */
	memcpy(&rp->slot[1], &sp->slot[1], IDATASIZE);

	/*
	 * append free region of newly extended area at tail of freelist
	 */
	/* init free region of newly extended area */
	fsi = n = DTROOTMAXSLOT + stblsize;
	f = &rp->slot[fsi];
	for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
		f->next = fsi;
	f->next = -1;

	/* append new free region at tail of old freelist */
	fsi = sp->header.freelist;
	if (fsi == -1)
		rp->header.freelist = n;
	else {
		rp->header.freelist = fsi;

		do {
			f = &rp->slot[fsi];
			fsi = f->next;
		} while (fsi != -1);

		f->next = n;
	}

	rp->header.freecnt = sp->header.freecnt + rp->header.maxslot - n;

	/*
	 * Update directory index table for entries now in right page
	 */
	if ((rp->header.flag & BT_LEAF) && DO_INDEX(ip)) {
1951
		s64 lblock;
1952 1953
		struct metapage *mp = 0;
		struct ldtentry *ldtentry;
Dave Kleikamp's avatar
Dave Kleikamp committed
1954 1955 1956

		stbl = DT_GETSTBL(rp);
		for (n = 0; n < rp->header.nextindex; n++) {
1957
			ldtentry = (struct ldtentry *) & rp->slot[stbl[n]];
Dave Kleikamp's avatar
Dave Kleikamp committed
1958
			modify_index(tid, ip, le32_to_cpu(ldtentry->index),
1959
				     rbn, n, &mp, &lblock);
Dave Kleikamp's avatar
Dave Kleikamp committed
1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983
		}
		if (mp)
			release_metapage(mp);
	}
	/*
	 * insert the new entry into the new right/child page
	 * (skip index in the new right page will not change)
	 */
	dtInsertEntry(rp, split->index, split->key, split->data, &dtlck);

	/*
	 *      reset parent/root page
	 *
	 * set the 1st entry offset to 0, which force the left-most key
	 * at any level of the tree to be less than any search key.
	 *
	 * The btree comparison code guarantees that the left-most key on any
	 * level of the tree is never used, so it doesn't need to be filled in.
	 */
	BT_MARK_DIRTY(smp, ip);
	/*
	 * acquire a transaction lock on the root page (in-memory inode)
	 */
	tlck = txLock(tid, ip, smp, tlckDTREE | tlckNEW | tlckBTROOT);
1984
	dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
1985 1986 1987

	/* linelock root */
	ASSERT(dtlck->index == 0);
1988
	lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
	lv->offset = 0;
	lv->length = DTROOTMAXSLOT;
	dtlck->index++;

	/* update page header of root */
	if (sp->header.flag & BT_LEAF) {
		sp->header.flag &= ~BT_LEAF;
		sp->header.flag |= BT_INTERNAL;
	}

	/* init the first entry */
2000
	s = (struct idtentry *) & sp->slot[DTENTRYSTART];
Dave Kleikamp's avatar
Dave Kleikamp committed
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038
	ppxd = (pxd_t *) s;
	*ppxd = *pxd;
	s->next = -1;
	s->namlen = 0;

	stbl = sp->header.stbl;
	stbl[0] = DTENTRYSTART;
	sp->header.nextindex = 1;

	/* init freelist */
	fsi = DTENTRYSTART + 1;
	f = &sp->slot[fsi];

	/* init free region of remaining area */
	for (fsi++; fsi < DTROOTMAXSLOT; f++, fsi++)
		f->next = fsi;
	f->next = -1;

	sp->header.freelist = DTENTRYSTART + 1;
	sp->header.freecnt = DTROOTMAXSLOT - (DTENTRYSTART + 1);

	*rmpp = rmp;

	ip->i_blocks += LBLK2PBLK(ip->i_sb, lengthPXD(pxd));
	return 0;
}


/*
 *	dtDelete()
 *
 * function: delete the entry(s) referenced by a key.
 *
 * parameter:
 *
 * return:
 */
int dtDelete(tid_t tid,
2039
	 struct inode *ip, struct component_name * key, ino_t * ino, int flag)
Dave Kleikamp's avatar
Dave Kleikamp committed
2040 2041 2042
{
	int rc = 0;
	s64 bn;
2043
	struct metapage *mp, *imp;
Dave Kleikamp's avatar
Dave Kleikamp committed
2044 2045
	dtpage_t *p;
	int index;
2046 2047 2048 2049
	struct btstack btstack;
	struct dt_lock *dtlck;
	struct tlock *tlck;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
2050
	int i;
2051
	struct ldtentry *ldtentry;
Dave Kleikamp's avatar
Dave Kleikamp committed
2052 2053
	u8 *stbl;
	u32 table_index, next_index;
2054
	struct metapage *nmp;
Dave Kleikamp's avatar
Dave Kleikamp committed
2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074
	dtpage_t *np;

	/*
	 *      search for the entry to delete:
	 *
	 * dtSearch() returns (leaf page pinned, index at which to delete).
	 */
	if ((rc = dtSearch(ip, key, ino, &btstack, flag)))
		return rc;

	/* retrieve search result */
	DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);

	/*
	 * We need to find put the index of the next entry into the
	 * directory index table in order to resume a readdir from this
	 * entry.
	 */
	if (DO_INDEX(ip)) {
		stbl = DT_GETSTBL(p);
2075
		ldtentry = (struct ldtentry *) & p->slot[stbl[index]];
Dave Kleikamp's avatar
Dave Kleikamp committed
2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092
		table_index = le32_to_cpu(ldtentry->index);
		if (index == (p->header.nextindex - 1)) {
			/*
			 * Last entry in this leaf page
			 */
			if ((p->header.flag & BT_ROOT)
			    || (p->header.next == 0))
				next_index = -1;
			else {
				/* Read next leaf page */
				DT_GETPAGE(ip, le64_to_cpu(p->header.next),
					   nmp, PSIZE, np, rc);
				if (rc)
					next_index = -1;
				else {
					stbl = DT_GETSTBL(np);
					ldtentry =
2093
					    (struct ldtentry *) & np->
Dave Kleikamp's avatar
Dave Kleikamp committed
2094 2095 2096 2097 2098 2099 2100 2101
					    slot[stbl[0]];
					next_index =
					    le32_to_cpu(ldtentry->index);
					DT_PUTPAGE(nmp);
				}
			}
		} else {
			ldtentry =
2102
			    (struct ldtentry *) & p->slot[stbl[index + 1]];
Dave Kleikamp's avatar
Dave Kleikamp committed
2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124
			next_index = le32_to_cpu(ldtentry->index);
		}
		free_index(tid, ip, table_index, next_index);
	}
	/*
	 * the leaf page becomes empty, delete the page
	 */
	if (p->header.nextindex == 1) {
		/* delete empty page */
		rc = dtDeleteUp(tid, ip, mp, p, &btstack);
	}
	/*
	 * the leaf page has other entries remaining:
	 *
	 * delete the entry from the leaf page.
	 */
	else {
		BT_MARK_DIRTY(mp, ip);
		/*
		 * acquire a transaction lock on the leaf page
		 */
		tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
2125
		dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2126 2127 2128 2129 2130 2131 2132 2133 2134

		/*
		 * Do not assume that dtlck->index will be zero.  During a
		 * rename within a directory, this transaction may have
		 * modified this page already when adding the new entry.
		 */

		/* linelock header */
		if (dtlck->index >= dtlck->maxcnt)
2135 2136
			dtlck = (struct dt_lock *) txLinelock(dtlck);
		lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
2137 2138 2139 2140 2141 2142 2143
		lv->offset = 0;
		lv->length = 1;
		dtlck->index++;

		/* linelock stbl of non-root leaf page */
		if (!(p->header.flag & BT_ROOT)) {
			if (dtlck->index >= dtlck->maxcnt)
2144 2145
				dtlck = (struct dt_lock *) txLinelock(dtlck);
			lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160
			i = index >> L2DTSLOTSIZE;
			lv->offset = p->header.stblindex + i;
			lv->length =
			    ((p->header.nextindex - 1) >> L2DTSLOTSIZE) -
			    i + 1;
			dtlck->index++;
		}

		/* free the leaf entry */
		dtDeleteEntry(p, index, &dtlck);

		/*
		 * Update directory index table for entries moved in stbl
		 */
		if (DO_INDEX(ip) && index < p->header.nextindex) {
2161 2162
			s64 lblock;

Dave Kleikamp's avatar
Dave Kleikamp committed
2163 2164 2165 2166
			imp = 0;
			stbl = DT_GETSTBL(p);
			for (i = index; i < p->header.nextindex; i++) {
				ldtentry =
2167
				    (struct ldtentry *) & p->slot[stbl[i]];
Dave Kleikamp's avatar
Dave Kleikamp committed
2168 2169
				modify_index(tid, ip,
					     le32_to_cpu(ldtentry->index),
2170
					     bn, i, &imp, &lblock);
Dave Kleikamp's avatar
Dave Kleikamp committed
2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193
			}
			if (imp)
				release_metapage(imp);
		}

		DT_PUTPAGE(mp);
	}

	return rc;
}


/*
 *	dtDeleteUp()
 *
 * function:
 *	free empty pages as propagating deletion up the tree
 *
 * parameter:
 *
 * return:
 */
static int dtDeleteUp(tid_t tid, struct inode *ip,
2194
	   struct metapage * fmp, dtpage_t * fp, struct btstack * btstack)
Dave Kleikamp's avatar
Dave Kleikamp committed
2195 2196
{
	int rc = 0;
2197
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
2198 2199 2200
	dtpage_t *p;
	int index, nextindex;
	int xlen;
2201 2202 2203 2204 2205
	struct btframe *parent;
	struct dt_lock *dtlck;
	struct tlock *tlck;
	struct lv *lv;
	struct pxd_lock *pxdlock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234
	int i;

	/*
	 *      keep the root leaf page which has become empty
	 */
	if (BT_IS_ROOT(fmp)) {
		/*
		 * reset the root
		 *
		 * dtInitRoot() acquires txlock on the root
		 */
		dtInitRoot(tid, ip, PARENT(ip));

		DT_PUTPAGE(fmp);

		return 0;
	}

	/*
	 *      free the non-root leaf page
	 */
	/*
	 * acquire a transaction lock on the page
	 *
	 * write FREEXTENT|NOREDOPAGE log record
	 * N.B. linelock is overlaid as freed extent descriptor, and
	 * the buffer page is freed;
	 */
	tlck = txMaplock(tid, ip, tlckDTREE | tlckFREE);
2235
	pxdlock = (struct pxd_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306
	pxdlock->flag = mlckFREEPXD;
	pxdlock->pxd = fp->header.self;
	pxdlock->index = 1;

	/* update sibling pointers */
	if ((rc = dtRelink(tid, ip, fp)))
		return rc;

	xlen = lengthPXD(&fp->header.self);
	ip->i_blocks -= LBLK2PBLK(ip->i_sb, xlen);

	/* free/invalidate its buffer page */
	discard_metapage(fmp);

	/*
	 *      propagate page deletion up the directory tree
	 *
	 * If the delete from the parent page makes it empty,
	 * continue all the way up the tree.
	 * stop if the root page is reached (which is never deleted) or
	 * if the entry deletion does not empty the page.
	 */
	while ((parent = BT_POP(btstack)) != NULL) {
		/* pin the parent page <sp> */
		DT_GETPAGE(ip, parent->bn, mp, PSIZE, p, rc);
		if (rc)
			return rc;

		/*
		 * free the extent of the child page deleted
		 */
		index = parent->index;

		/*
		 * delete the entry for the child page from parent
		 */
		nextindex = p->header.nextindex;

		/*
		 * the parent has the single entry being deleted:
		 *
		 * free the parent page which has become empty.
		 */
		if (nextindex == 1) {
			/*
			 * keep the root internal page which has become empty
			 */
			if (p->header.flag & BT_ROOT) {
				/*
				 * reset the root
				 *
				 * dtInitRoot() acquires txlock on the root
				 */
				dtInitRoot(tid, ip, PARENT(ip));

				DT_PUTPAGE(mp);

				return 0;
			}
			/*
			 * free the parent page
			 */
			else {
				/*
				 * acquire a transaction lock on the page
				 *
				 * write FREEXTENT|NOREDOPAGE log record
				 */
				tlck =
				    txMaplock(tid, ip,
					      tlckDTREE | tlckFREE);
2307
				pxdlock = (struct pxd_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338
				pxdlock->flag = mlckFREEPXD;
				pxdlock->pxd = p->header.self;
				pxdlock->index = 1;

				/* update sibling pointers */
				if ((rc = dtRelink(tid, ip, p)))
					return rc;

				xlen = lengthPXD(&p->header.self);
				ip->i_blocks -= LBLK2PBLK(ip->i_sb, xlen);

				/* free/invalidate its buffer page */
				discard_metapage(mp);

				/* propagate up */
				continue;
			}
		}

		/*
		 * the parent has other entries remaining:
		 *
		 * delete the router entry from the parent page.
		 */
		BT_MARK_DIRTY(mp, ip);
		/*
		 * acquire a transaction lock on the page
		 *
		 * action: router entry deletion
		 */
		tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
2339
		dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2340 2341 2342

		/* linelock header */
		if (dtlck->index >= dtlck->maxcnt)
2343 2344
			dtlck = (struct dt_lock *) txLinelock(dtlck);
		lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
2345 2346 2347 2348 2349 2350 2351 2352 2353
		lv->offset = 0;
		lv->length = 1;
		dtlck->index++;

		/* linelock stbl of non-root leaf page */
		if (!(p->header.flag & BT_ROOT)) {
			if (dtlck->index < dtlck->maxcnt)
				lv++;
			else {
2354 2355
				dtlck = (struct dt_lock *) txLinelock(dtlck);
				lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393
			}
			i = index >> L2DTSLOTSIZE;
			lv->offset = p->header.stblindex + i;
			lv->length =
			    ((p->header.nextindex - 1) >> L2DTSLOTSIZE) -
			    i + 1;
			dtlck->index++;
		}

		/* free the router entry */
		dtDeleteEntry(p, index, &dtlck);

		/* reset key of new leftmost entry of level (for consistency) */
		if (index == 0 &&
		    ((p->header.flag & BT_ROOT) || p->header.prev == 0))
			dtTruncateEntry(p, 0, &dtlck);

		/* unpin the parent page */
		DT_PUTPAGE(mp);

		/* exit propagation up */
		break;
	}

	return 0;
}


/*
 * NAME:        dtRelocate()
 *
 * FUNCTION:    relocate dtpage (internal or leaf) of directory;
 *              This function is mainly used by defragfs utility.
 */
int dtRelocate(tid_t tid, struct inode *ip, s64 lmxaddr, pxd_t * opxd,
	       s64 nxaddr)
{
	int rc = 0;
2394
	struct metapage *mp, *pmp, *lmp, *rmp;
Dave Kleikamp's avatar
Dave Kleikamp committed
2395 2396 2397
	dtpage_t *p, *pp, *rp = 0, *lp= 0;
	s64 bn;
	int index;
2398
	struct btstack btstack;
Dave Kleikamp's avatar
Dave Kleikamp committed
2399 2400 2401
	pxd_t *pxd;
	s64 oxaddr, nextbn, prevbn;
	int xlen, xsize;
2402 2403 2404
	struct tlock *tlck;
	struct dt_lock *dtlck;
	struct pxd_lock *pxdlock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2405
	s8 *stbl;
2406
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
2407 2408 2409 2410

	oxaddr = addressPXD(opxd);
	xlen = lengthPXD(opxd);

2411
	jfs_info("dtRelocate: lmxaddr:%Ld xaddr:%Ld:%Ld xlen:%d",
2412
		   (long long)lmxaddr, (long long)oxaddr, (long long)nxaddr,
2413
		   xlen);
Dave Kleikamp's avatar
Dave Kleikamp committed
2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424

	/*
	 *      1. get the internal parent dtpage covering
	 *      router entry for the tartget page to be relocated;
	 */
	rc = dtSearchNode(ip, lmxaddr, opxd, &btstack);
	if (rc)
		return rc;

	/* retrieve search result */
	DT_GETSEARCH(ip, btstack.top, bn, pmp, pp, index);
2425
	jfs_info("dtRelocate: parent router entry validated.");
Dave Kleikamp's avatar
Dave Kleikamp committed
2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471

	/*
	 *      2. relocate the target dtpage
	 */
	/* read in the target page from src extent */
	DT_GETPAGE(ip, oxaddr, mp, PSIZE, p, rc);
	if (rc) {
		/* release the pinned parent page */
		DT_PUTPAGE(pmp);
		return rc;
	}

	/*
	 * read in sibling pages if any to update sibling pointers;
	 */
	rmp = NULL;
	if (p->header.next) {
		nextbn = le64_to_cpu(p->header.next);
		DT_GETPAGE(ip, nextbn, rmp, PSIZE, rp, rc);
		if (rc) {
			DT_PUTPAGE(mp);
			DT_PUTPAGE(pmp);
			return (rc);
		}
	}

	lmp = NULL;
	if (p->header.prev) {
		prevbn = le64_to_cpu(p->header.prev);
		DT_GETPAGE(ip, prevbn, lmp, PSIZE, lp, rc);
		if (rc) {
			DT_PUTPAGE(mp);
			DT_PUTPAGE(pmp);
			if (rmp)
				DT_PUTPAGE(rmp);
			return (rc);
		}
	}

	/* at this point, all xtpages to be updated are in memory */

	/*
	 * update sibling pointers of sibling dtpages if any;
	 */
	if (lmp) {
		tlck = txLock(tid, ip, lmp, tlckDTREE | tlckRELINK);
2472
		dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2473 2474
		/* linelock header */
		ASSERT(dtlck->index == 0);
2475
		lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
2476 2477 2478 2479 2480 2481 2482 2483 2484 2485
		lv->offset = 0;
		lv->length = 1;
		dtlck->index++;

		lp->header.next = cpu_to_le64(nxaddr);
		DT_PUTPAGE(lmp);
	}

	if (rmp) {
		tlck = txLock(tid, ip, rmp, tlckDTREE | tlckRELINK);
2486
		dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2487 2488
		/* linelock header */
		ASSERT(dtlck->index == 0);
2489
		lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507
		lv->offset = 0;
		lv->length = 1;
		dtlck->index++;

		rp->header.prev = cpu_to_le64(nxaddr);
		DT_PUTPAGE(rmp);
	}

	/*
	 * update the target dtpage to be relocated
	 *
	 * write LOG_REDOPAGE of LOG_NEW type for dst page
	 * for the whole target page (logredo() will apply
	 * after image and update bmap for allocation of the
	 * dst extent), and update bmap for allocation of
	 * the dst extent;
	 */
	tlck = txLock(tid, ip, mp, tlckDTREE | tlckNEW);
2508
	dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2509 2510
	/* linelock header */
	ASSERT(dtlck->index == 0);
2511
	lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530

	/* update the self address in the dtpage header */
	pxd = &p->header.self;
	PXDaddress(pxd, nxaddr);

	/* the dst page is the same as the src page, i.e.,
	 * linelock for afterimage of the whole page;
	 */
	lv->offset = 0;
	lv->length = p->header.maxslot;
	dtlck->index++;

	/* update the buffer extent descriptor of the dtpage */
	xsize = xlen << JFS_SBI(ip->i_sb)->l2bsize;
#ifdef _STILL_TO_PORT
	bmSetXD(mp, nxaddr, xsize);
#endif /* _STILL_TO_PORT */
	/* unpin the relocated page */
	DT_PUTPAGE(mp);
2531
	jfs_info("dtRelocate: target dtpage relocated.");
Dave Kleikamp's avatar
Dave Kleikamp committed
2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546

	/* the moved extent is dtpage, then a LOG_NOREDOPAGE log rec
	 * needs to be written (in logredo(), the LOG_NOREDOPAGE log rec
	 * will also force a bmap update ).
	 */

	/*
	 *      3. acquire maplock for the source extent to be freed;
	 */
	/* for dtpage relocation, write a LOG_NOREDOPAGE record
	 * for the source dtpage (logredo() will init NoRedoPage
	 * filter and will also update bmap for free of the source
	 * dtpage), and upadte bmap for free of the source dtpage;
	 */
	tlck = txMaplock(tid, ip, tlckDTREE | tlckFREE);
2547
	pxdlock = (struct pxd_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558
	pxdlock->flag = mlckFREEPXD;
	PXDaddress(&pxdlock->pxd, oxaddr);
	PXDlength(&pxdlock->pxd, xlen);
	pxdlock->index = 1;

	/*
	 *      4. update the parent router entry for relocation;
	 *
	 * acquire tlck for the parent entry covering the target dtpage;
	 * write LOG_REDOPAGE to apply after image only;
	 */
2559
	jfs_info("dtRelocate: update parent router entry.");
Dave Kleikamp's avatar
Dave Kleikamp committed
2560
	tlck = txLock(tid, ip, pmp, tlckDTREE | tlckENTRY);
2561 2562
	dtlck = (struct dt_lock *) & tlck->lock;
	lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590

	/* update the PXD with the new address */
	stbl = DT_GETSTBL(pp);
	pxd = (pxd_t *) & pp->slot[stbl[index]];
	PXDaddress(pxd, nxaddr);
	lv->offset = stbl[index];
	lv->length = 1;
	dtlck->index++;

	/* unpin the parent dtpage */
	DT_PUTPAGE(pmp);

	return rc;
}


/*
 * NAME:	dtSearchNode()
 *
 * FUNCTION:	Search for an dtpage containing a specified address
 *              This function is mainly used by defragfs utility.
 *
 * NOTE:	Search result on stack, the found page is pinned at exit.
 *		The result page must be an internal dtpage.
 *		lmxaddr give the address of the left most page of the
 *		dtree level, in which the required dtpage resides.
 */
static int dtSearchNode(struct inode *ip, s64 lmxaddr, pxd_t * kpxd,
2591
			struct btstack * btstack)
Dave Kleikamp's avatar
Dave Kleikamp committed
2592 2593 2594
{
	int rc = 0;
	s64 bn;
2595
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
2596 2597 2598 2599 2600
	dtpage_t *p;
	int psize = 288;	/* initial in-line directory */
	s8 *stbl;
	int i;
	pxd_t *pxd;
2601
	struct btframe *btsp;
Dave Kleikamp's avatar
Dave Kleikamp committed
2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628

	BT_CLR(btstack);	/* reset stack */

	/*
	 *      descend tree to the level with specified leftmost page
	 *
	 *  by convention, root bn = 0.
	 */
	for (bn = 0;;) {
		/* get/pin the page to search */
		DT_GETPAGE(ip, bn, mp, psize, p, rc);
		if (rc)
			return rc;

		/* does the xaddr of leftmost page of the levevl
		 * matches levevl search key ?
		 */
		if (p->header.flag & BT_ROOT) {
			if (lmxaddr == 0)
				break;
		} else if (addressPXD(&p->header.self) == lmxaddr)
			break;

		/*
		 * descend down to leftmost child page
		 */
		if (p->header.flag & BT_LEAF)
2629
			return -ESTALE;
Dave Kleikamp's avatar
Dave Kleikamp committed
2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666

		/* get the leftmost entry */
		stbl = DT_GETSTBL(p);
		pxd = (pxd_t *) & p->slot[stbl[0]];

		/* get the child page block address */
		bn = addressPXD(pxd);
		psize = lengthPXD(pxd) << JFS_SBI(ip->i_sb)->l2bsize;
		/* unpin the parent page */
		DT_PUTPAGE(mp);
	}

	/*
	 *      search each page at the current levevl
	 */
      loop:
	stbl = DT_GETSTBL(p);
	for (i = 0; i < p->header.nextindex; i++) {
		pxd = (pxd_t *) & p->slot[stbl[i]];

		/* found the specified router entry */
		if (addressPXD(pxd) == addressPXD(kpxd) &&
		    lengthPXD(pxd) == lengthPXD(kpxd)) {
			btsp = btstack->top;
			btsp->bn = bn;
			btsp->index = i;
			btsp->mp = mp;

			return 0;
		}
	}

	/* get the right sibling page if any */
	if (p->header.next)
		bn = le64_to_cpu(p->header.next);
	else {
		DT_PUTPAGE(mp);
2667
		return -ESTALE;
Dave Kleikamp's avatar
Dave Kleikamp committed
2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695
	}

	/* unpin current page */
	DT_PUTPAGE(mp);

	/* get the right sibling page */
	DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
	if (rc)
		return rc;

	goto loop;
}


/*
 *	dtRelink()
 *
 * function:
 *	link around a freed page.
 *
 * parameter:
 *	fp:	page to be freed
 *
 * return:
 */
static int dtRelink(tid_t tid, struct inode *ip, dtpage_t * p)
{
	int rc;
2696
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
2697
	s64 nextbn, prevbn;
2698 2699 2700
	struct tlock *tlck;
	struct dt_lock *dtlck;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717

	nextbn = le64_to_cpu(p->header.next);
	prevbn = le64_to_cpu(p->header.prev);

	/* update prev pointer of the next page */
	if (nextbn != 0) {
		DT_GETPAGE(ip, nextbn, mp, PSIZE, p, rc);
		if (rc)
			return rc;

		BT_MARK_DIRTY(mp, ip);
		/*
		 * acquire a transaction lock on the next page
		 *
		 * action: update prev pointer;
		 */
		tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
2718 2719
		jfs_info("dtRelink nextbn: tlck = 0x%p, ip = 0x%p, mp=0x%p",
			tlck, ip, mp);
2720
		dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2721 2722 2723

		/* linelock header */
		if (dtlck->index >= dtlck->maxcnt)
2724 2725
			dtlck = (struct dt_lock *) txLinelock(dtlck);
		lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746
		lv->offset = 0;
		lv->length = 1;
		dtlck->index++;

		p->header.prev = cpu_to_le64(prevbn);
		DT_PUTPAGE(mp);
	}

	/* update next pointer of the previous page */
	if (prevbn != 0) {
		DT_GETPAGE(ip, prevbn, mp, PSIZE, p, rc);
		if (rc)
			return rc;

		BT_MARK_DIRTY(mp, ip);
		/*
		 * acquire a transaction lock on the prev page
		 *
		 * action: update next pointer;
		 */
		tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
2747 2748
		jfs_info("dtRelink prevbn: tlck = 0x%p, ip = 0x%p, mp=0x%p",
			tlck, ip, mp);
2749
		dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2750 2751 2752

		/* linelock header */
		if (dtlck->index >= dtlck->maxcnt)
2753 2754
			dtlck = (struct dt_lock *) txLinelock(dtlck);
		lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776
		lv->offset = 0;
		lv->length = 1;
		dtlck->index++;

		p->header.next = cpu_to_le64(nextbn);
		DT_PUTPAGE(mp);
	}

	return 0;
}


/*
 *	dtInitRoot()
 *
 * initialize directory root (inline in inode)
 */
void dtInitRoot(tid_t tid, struct inode *ip, u32 idotdot)
{
	struct jfs_inode_info *jfs_ip = JFS_IP(ip);
	dtroot_t *p;
	int fsi;
2777 2778 2779 2780
	struct dtslot *f;
	struct tlock *tlck;
	struct dt_lock *dtlck;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
2781 2782 2783 2784 2785 2786 2787 2788
	u16 xflag_save;

	/*
	 * If this was previously an non-empty directory, we need to remove
	 * the old directory table.
	 */
	if (DO_INDEX(ip)) {
		if (jfs_ip->next_index > (MAX_INLINE_DIRTABLE_ENTRY + 1)) {
2789
			struct tblock *tblk = tid_to_tblock(tid);
Dave Kleikamp's avatar
Dave Kleikamp committed
2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821
			/*
			 * We're playing games with the tid's xflag.  If
			 * we're removing a regular file, the file's xtree
			 * is committed with COMMIT_PMAP, but we always
			 * commit the directories xtree with COMMIT_PWMAP.
			 */
			xflag_save = tblk->xflag;
			tblk->xflag = 0;
			/*
			 * xtTruncate isn't guaranteed to fully truncate
			 * the xtree.  The caller needs to check i_size
			 * after committing the transaction to see if
			 * additional truncation is needed.  The
			 * COMMIT_Stale flag tells caller that we
			 * initiated the truncation.
			 */
			xtTruncate(tid, ip, 0, COMMIT_PWMAP);
			set_cflag(COMMIT_Stale, ip);

			tblk->xflag = xflag_save;
		} else
			ip->i_size = 1;

		jfs_ip->next_index = 2;
	} else
		ip->i_size = IDATASIZE;

	/*
	 * acquire a transaction lock on the root
	 *
	 * action: directory initialization;
	 */
2822
	tlck = txLock(tid, ip, (struct metapage *) & jfs_ip->bxflag,
Dave Kleikamp's avatar
Dave Kleikamp committed
2823
		      tlckDTREE | tlckENTRY | tlckBTROOT);
2824
	dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
2825 2826 2827

	/* linelock root */
	ASSERT(dtlck->index == 0);
2828
	lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864
	lv->offset = 0;
	lv->length = DTROOTMAXSLOT;
	dtlck->index++;

	p = &jfs_ip->i_dtroot;

	p->header.flag = DXD_INDEX | BT_ROOT | BT_LEAF;

	p->header.nextindex = 0;

	/* init freelist */
	fsi = 1;
	f = &p->slot[fsi];

	/* init data area of root */
	for (fsi++; fsi < DTROOTMAXSLOT; f++, fsi++)
		f->next = fsi;
	f->next = -1;

	p->header.freelist = 1;
	p->header.freecnt = 8;

	/* init '..' entry */
	p->header.idotdot = cpu_to_le32(idotdot);

#if 0
	ip->i_blocks = LBLK2PBLK(ip->i_sb,
				 ((jfs_ip->ea.flag & DXD_EXTENT) ?
				  lengthDXD(&jfs_ip->ea) : 0) +
				 ((jfs_ip->acl.flag & DXD_EXTENT) ?
				  lengthDXD(&jfs_ip->acl) : 0));
#endif

	return;
}

2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908
/*
 *	add_missing_indices()
 *
 * function: Fix dtree page in which one or more entries has an invalid index.
 *	     fsck.jfs should really fix this, but it currently does not.
 *	     Called from jfs_readdir when bad index is detected.
 */
static void add_missing_indices(struct inode *inode, s64 bn)
{
	struct ldtentry *d;
	struct dt_lock *dtlck;
	int i;
	uint index;
	struct lv *lv;
	struct metapage *mp;
	dtpage_t *p;
	int rc;
	s8 *stbl;
	tid_t tid;
	struct tlock *tlck;

	tid = txBegin(inode->i_sb, 0);

	DT_GETPAGE(inode, bn, mp, PSIZE, p, rc);

	if (rc) {
		printk(KERN_ERR "DT_GETPAGE failed!\n");
		goto end;
	}
	BT_MARK_DIRTY(mp, inode);

	ASSERT(p->header.flag & BT_LEAF);

	tlck = txLock(tid, inode, mp, tlckDTREE | tlckENTRY);
	dtlck = (struct dt_lock *) &tlck->lock;

	stbl = DT_GETSTBL(p);
	for (i = 0; i < p->header.nextindex; i++) {
		d = (struct ldtentry *) &p->slot[stbl[i]];
		index = le32_to_cpu(d->index);
		if ((index < 2) || (index >= JFS_IP(inode)->next_index)) {
			d->index = cpu_to_le32(add_index(tid, inode, bn, i));
			if (dtlck->index >= dtlck->maxcnt)
				dtlck = (struct dt_lock *) txLinelock(dtlck);
2909
			lv = &dtlck->lv[dtlck->index];
2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925
			lv->offset = stbl[i];
			lv->length = 1;
			dtlck->index++;
		}
	}

	DT_PUTPAGE(mp);
	(void) txCommit(tid, 1, &inode, 0);
end:
	txEnd(tid);
}

/*
 * Buffer to hold directory entry info while traversing a dtree page
 * before being fed to the filldir function
 */
Dave Kleikamp's avatar
Dave Kleikamp committed
2926 2927 2928 2929 2930 2931 2932
struct jfs_dirent {
	loff_t position;
	int ino;
	u16 name_len;
	char name[0];
};

2933 2934 2935
/*
 * function to determine next variable-sized jfs_dirent in buffer
 */
Dave Kleikamp's avatar
Dave Kleikamp committed
2936 2937 2938 2939 2940 2941 2942 2943 2944
inline struct jfs_dirent *next_jfs_dirent(struct jfs_dirent *dirent)
{
	return (struct jfs_dirent *)
		((char *)dirent +
		 ((sizeof (struct jfs_dirent) + dirent->name_len + 1 +
		   sizeof (loff_t) - 1) &
		  ~(sizeof (loff_t) - 1)));
}

Dave Kleikamp's avatar
Dave Kleikamp committed
2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960
/*
 *	jfs_readdir()
 *
 * function: read directory entries sequentially
 *	from the specified entry offset
 *
 * parameter:
 *
 * return: offset = (pn, index) of start entry
 *	of next jfs_readdir()/dtRead()
 */
int jfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
	struct inode *ip = filp->f_dentry->d_inode;
	struct nls_table *codepage = JFS_SBI(ip->i_sb)->nls_tab;
	int rc = 0;
Dave Kleikamp's avatar
Dave Kleikamp committed
2961
	loff_t dtpos;	/* legacy OS/2 style position */
Dave Kleikamp's avatar
Dave Kleikamp committed
2962 2963 2964 2965
	struct dtoffset {
		s16 pn;
		s16 index;
		s32 unused;
Dave Kleikamp's avatar
Dave Kleikamp committed
2966
	} *dtoffset = (struct dtoffset *) &dtpos;
Dave Kleikamp's avatar
Dave Kleikamp committed
2967
	s64 bn;
2968
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
2969 2970 2971
	dtpage_t *p;
	int index;
	s8 *stbl;
2972
	struct btstack btstack;
Dave Kleikamp's avatar
Dave Kleikamp committed
2973
	int i, next;
2974 2975
	struct ldtentry *d;
	struct dtslot *t;
Dave Kleikamp's avatar
Dave Kleikamp committed
2976 2977 2978
	int d_namleft, len, outlen;
	unsigned long dirent_buf;
	char *name_ptr;
Dave Kleikamp's avatar
Dave Kleikamp committed
2979 2980 2981
	u32 dir_index;
	int do_index = 0;
	uint loop_count = 0;
Dave Kleikamp's avatar
Dave Kleikamp committed
2982 2983
	struct jfs_dirent *jfs_dirent;
	int jfs_dirents;
2984 2985
	int overflow, fix_page, page_fixed = 0;
	static int unique_pos = 2;	/* If we can't fix broken index */
Dave Kleikamp's avatar
Dave Kleikamp committed
2986

2987
	if (filp->f_pos == DIREND)
Dave Kleikamp's avatar
Dave Kleikamp committed
2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001
		return 0;

	if (DO_INDEX(ip)) {
		/*
		 * persistent index is stored in directory entries.
		 * Special cases:        0 = .
		 *                       1 = ..
		 *                      -1 = End of directory
		 */
		do_index = 1;

		dir_index = (u32) filp->f_pos;

		if (dir_index > 1) {
3002
			struct dir_table_slot dirtab_slot;
Dave Kleikamp's avatar
Dave Kleikamp committed
3003

Dave Kleikamp's avatar
Dave Kleikamp committed
3004 3005 3006
			if (dtEmpty(ip) ||
			    (dir_index >= JFS_IP(ip)->next_index)) {
				/* Stale position.  Directory has shrunk */
3007
				filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3008 3009 3010
				return 0;
			}
		      repeat:
3011
			rc = read_index(ip, dir_index, &dirtab_slot);
Dave Kleikamp's avatar
Dave Kleikamp committed
3012
			if (rc) {
3013
				filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3014 3015 3016 3017
				return rc;
			}
			if (dirtab_slot.flag == DIR_INDEX_FREE) {
				if (loop_count++ > JFS_IP(ip)->next_index) {
3018 3019
					jfs_err("jfs_readdir detected "
						   "infinite loop!");
3020
					filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3021 3022 3023 3024
					return 0;
				}
				dir_index = le32_to_cpu(dirtab_slot.addr2);
				if (dir_index == -1) {
3025
					filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3026 3027 3028 3029 3030 3031 3032 3033
					return 0;
				}
				goto repeat;
			}
			bn = addressDTS(&dirtab_slot);
			index = dirtab_slot.slot;
			DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
			if (rc) {
3034
				filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3035 3036 3037
				return 0;
			}
			if (p->header.flag & BT_INTERNAL) {
3038
				jfs_err("jfs_readdir: bad index table");
Dave Kleikamp's avatar
Dave Kleikamp committed
3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049
				DT_PUTPAGE(mp);
				filp->f_pos = -1;
				return 0;
			}
		} else {
			if (dir_index == 0) {
				/*
				 * self "."
				 */
				filp->f_pos = 0;
				if (filldir(dirent, ".", 1, 0, ip->i_ino,
3050
					    DT_DIR))
Dave Kleikamp's avatar
Dave Kleikamp committed
3051 3052 3053 3054 3055 3056
					return 0;
			}
			/*
			 * parent ".."
			 */
			filp->f_pos = 1;
3057
			if (filldir(dirent, "..", 2, 1, PARENT(ip), DT_DIR))
Dave Kleikamp's avatar
Dave Kleikamp committed
3058 3059 3060 3061 3062 3063
				return 0;

			/*
			 * Find first entry of left-most leaf
			 */
			if (dtEmpty(ip)) {
3064
				filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3065 3066 3067
				return 0;
			}

3068
			if ((rc = dtReadFirst(ip, &btstack)))
3069
				return rc;
Dave Kleikamp's avatar
Dave Kleikamp committed
3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081

			DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
		}
	} else {
		/*
		 * Legacy filesystem - OS/2 & Linux JFS < 0.3.6
		 *
		 * pn = index = 0:      First entry "."
		 * pn = 0; index = 1:   Second entry ".."
		 * pn > 0:              Real entries, pn=1 -> leftmost page
		 * pn = index = -1:     No more entries
		 */
Dave Kleikamp's avatar
Dave Kleikamp committed
3082 3083
		dtpos = filp->f_pos;
		if (dtpos == 0) {
Dave Kleikamp's avatar
Dave Kleikamp committed
3084 3085 3086
			/* build "." entry */

			if (filldir(dirent, ".", 1, filp->f_pos, ip->i_ino,
3087
				    DT_DIR))
Dave Kleikamp's avatar
Dave Kleikamp committed
3088 3089
				return 0;
			dtoffset->index = 1;
Dave Kleikamp's avatar
Dave Kleikamp committed
3090
			filp->f_pos = dtpos;
Dave Kleikamp's avatar
Dave Kleikamp committed
3091 3092 3093 3094 3095 3096 3097
		}

		if (dtoffset->pn == 0) {
			if (dtoffset->index == 1) {
				/* build ".." entry */

				if (filldir(dirent, "..", 2, filp->f_pos,
3098
					    PARENT(ip), DT_DIR))
Dave Kleikamp's avatar
Dave Kleikamp committed
3099 3100
					return 0;
			} else {
3101 3102
				jfs_err("jfs_readdir called with "
					"invalid offset!");
Dave Kleikamp's avatar
Dave Kleikamp committed
3103 3104 3105
			}
			dtoffset->pn = 1;
			dtoffset->index = 0;
Dave Kleikamp's avatar
Dave Kleikamp committed
3106
			filp->f_pos = dtpos;
Dave Kleikamp's avatar
Dave Kleikamp committed
3107 3108 3109
		}

		if (dtEmpty(ip)) {
3110
			filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3111 3112 3113 3114
			return 0;
		}

		if ((rc = dtReadNext(ip, &filp->f_pos, &btstack))) {
3115 3116
			jfs_err("jfs_readdir: unexpected rc = %d "
				"from dtReadNext", rc);
3117
			filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3118 3119 3120 3121 3122 3123 3124
			return 0;
		}
		/* get start leaf page and index */
		DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);

		/* offset beyond directory eof ? */
		if (bn < 0) {
3125
			filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3126 3127 3128 3129
			return 0;
		}
	}

Dave Kleikamp's avatar
Dave Kleikamp committed
3130 3131
	dirent_buf = __get_free_page(GFP_KERNEL);
	if (dirent_buf == 0) {
Dave Kleikamp's avatar
Dave Kleikamp committed
3132
		DT_PUTPAGE(mp);
3133
		jfs_warn("jfs_readdir: __get_free_page failed!");
3134
		filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3135
		return -ENOMEM;
Dave Kleikamp's avatar
Dave Kleikamp committed
3136
	}
Dave Kleikamp's avatar
Dave Kleikamp committed
3137

Dave Kleikamp's avatar
Dave Kleikamp committed
3138
	while (1) {
Dave Kleikamp's avatar
Dave Kleikamp committed
3139 3140
		jfs_dirent = (struct jfs_dirent *) dirent_buf;
		jfs_dirents = 0;
3141
		overflow = fix_page = 0;
Dave Kleikamp's avatar
Dave Kleikamp committed
3142

Dave Kleikamp's avatar
Dave Kleikamp committed
3143 3144 3145
		stbl = DT_GETSTBL(p);

		for (i = index; i < p->header.nextindex; i++) {
3146
			d = (struct ldtentry *) & p->slot[stbl[i]];
Dave Kleikamp's avatar
Dave Kleikamp committed
3147

Dave Kleikamp's avatar
Dave Kleikamp committed
3148 3149 3150 3151 3152 3153 3154 3155
			if (((long) jfs_dirent + d->namlen + 1) >
			    (dirent_buf + PSIZE)) {
				/* DBCS codepages could overrun dirent_buf */
				index = i;
				overflow = 1;
				break;
			}

Dave Kleikamp's avatar
Dave Kleikamp committed
3156
			d_namleft = d->namlen;
Dave Kleikamp's avatar
Dave Kleikamp committed
3157 3158
			name_ptr = jfs_dirent->name;
			jfs_dirent->ino = le32_to_cpu(d->inumber);
Dave Kleikamp's avatar
Dave Kleikamp committed
3159 3160 3161

			if (do_index) {
				len = min(d_namleft, DTLHDRDATALEN);
3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186
				jfs_dirent->position = le32_to_cpu(d->index);
				/*
				 * d->index should always be valid, but it
				 * isn't.  fsck.jfs doesn't create the
				 * directory index for the lost+found
				 * directory.  Rather than let it go,
				 * we can try to fix it.
				 */
				if ((jfs_dirent->position < 2) ||
				    (jfs_dirent->position >=
				     JFS_IP(ip)->next_index)) {
					if (!page_fixed && !isReadOnly(ip)) {
						fix_page = 1;
						/*
						 * setting overflow and setting
						 * index to i will cause the
						 * same page to be processed
						 * again starting here
						 */
						overflow = 1;
						index = i;
						break;
					}
					jfs_dirent->position = unique_pos++;
				}
Dave Kleikamp's avatar
Dave Kleikamp committed
3187 3188
			} else {
				jfs_dirent->position = dtpos;
Dave Kleikamp's avatar
Dave Kleikamp committed
3189
				len = min(d_namleft, DTLHDRDATALEN_LEGACY);
Dave Kleikamp's avatar
Dave Kleikamp committed
3190
			}
Dave Kleikamp's avatar
Dave Kleikamp committed
3191 3192 3193 3194

			/* copy the name of head/only segment */
			outlen = jfs_strfromUCS_le(name_ptr, d->name, len,
						   codepage);
Dave Kleikamp's avatar
Dave Kleikamp committed
3195
			jfs_dirent->name_len = outlen;
Dave Kleikamp's avatar
Dave Kleikamp committed
3196 3197 3198 3199

			/* copy name in the additional segment(s) */
			next = d->next;
			while (next >= 0) {
3200
				t = (struct dtslot *) & p->slot[next];
Dave Kleikamp's avatar
Dave Kleikamp committed
3201 3202
				name_ptr += outlen;
				d_namleft -= len;
3203 3204
				/* Sanity Check */
				if (d_namleft == 0) {
Dave Kleikamp's avatar
Dave Kleikamp committed
3205 3206 3207 3208 3209 3210
					jfs_error(ip->i_sb,
						  "JFS:Dtree error: ino = "
						  "%ld, bn=%Ld, index = %d",
						  (long)ip->i_ino,
						  (long long)bn,
						  i);
3211 3212
					goto skip_one;
				}
Dave Kleikamp's avatar
Dave Kleikamp committed
3213 3214 3215
				len = min(d_namleft, DTSLOTDATALEN);
				outlen = jfs_strfromUCS_le(name_ptr, t->name,
							   len, codepage);
Dave Kleikamp's avatar
Dave Kleikamp committed
3216
				jfs_dirent->name_len += outlen;
Dave Kleikamp's avatar
Dave Kleikamp committed
3217 3218 3219 3220

				next = t->next;
			}

Dave Kleikamp's avatar
Dave Kleikamp committed
3221 3222
			jfs_dirents++;
			jfs_dirent = next_jfs_dirent(jfs_dirent);
3223
skip_one:
Dave Kleikamp's avatar
Dave Kleikamp committed
3224 3225 3226 3227
			if (!do_index)
				dtoffset->index++;
		}

Dave Kleikamp's avatar
Dave Kleikamp committed
3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240
		if (!overflow) {
			/* Point to next leaf page */
			if (p->header.flag & BT_ROOT)
				bn = 0;
			else {
				bn = le64_to_cpu(p->header.next);
				index = 0;
				/* update offset (pn:index) for new page */
				if (!do_index) {
					dtoffset->pn++;
					dtoffset->index = 0;
				}
			}
3241
			page_fixed = 0;
Dave Kleikamp's avatar
Dave Kleikamp committed
3242
		}
Dave Kleikamp's avatar
Dave Kleikamp committed
3243

Dave Kleikamp's avatar
Dave Kleikamp committed
3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254
		/* unpin previous leaf page */
		DT_PUTPAGE(mp);

		jfs_dirent = (struct jfs_dirent *) dirent_buf;
		while (jfs_dirents--) {
			filp->f_pos = jfs_dirent->position;
			if (filldir(dirent, jfs_dirent->name,
				    jfs_dirent->name_len, filp->f_pos,
				    jfs_dirent->ino, DT_UNKNOWN))
				goto out;
			jfs_dirent = next_jfs_dirent(jfs_dirent);
Dave Kleikamp's avatar
Dave Kleikamp committed
3255 3256
		}

3257 3258 3259 3260 3261
		if (fix_page) {
			add_missing_indices(ip, bn);
			page_fixed = 1;
		}

Dave Kleikamp's avatar
Dave Kleikamp committed
3262
		if (!overflow && (bn == 0)) {
3263
			filp->f_pos = DIREND;
Dave Kleikamp's avatar
Dave Kleikamp committed
3264 3265 3266 3267 3268
			break;
		}

		DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
		if (rc) {
Dave Kleikamp's avatar
Dave Kleikamp committed
3269
			free_page(dirent_buf);
3270
			return rc;
Dave Kleikamp's avatar
Dave Kleikamp committed
3271 3272 3273 3274
		}
	}

      out:
Dave Kleikamp's avatar
Dave Kleikamp committed
3275
	free_page(dirent_buf);
3276

Dave Kleikamp's avatar
Dave Kleikamp committed
3277 3278 3279 3280 3281 3282 3283 3284 3285
	return rc;
}


/*
 *	dtReadFirst()
 *
 * function: get the leftmost page of the directory
 */
3286
static int dtReadFirst(struct inode *ip, struct btstack * btstack)
Dave Kleikamp's avatar
Dave Kleikamp committed
3287 3288 3289 3290
{
	int rc = 0;
	s64 bn;
	int psize = 288;	/* initial in-line directory */
3291
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
3292 3293
	dtpage_t *p;
	s8 *stbl;
3294
	struct btframe *btsp;
Dave Kleikamp's avatar
Dave Kleikamp committed
3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351
	pxd_t *xd;

	BT_CLR(btstack);	/* reset stack */

	/*
	 *      descend leftmost path of the tree
	 *
	 * by convention, root bn = 0.
	 */
	for (bn = 0;;) {
		DT_GETPAGE(ip, bn, mp, psize, p, rc);
		if (rc)
			return rc;

		/*
		 * leftmost leaf page
		 */
		if (p->header.flag & BT_LEAF) {
			/* return leftmost entry */
			btsp = btstack->top;
			btsp->bn = bn;
			btsp->index = 0;
			btsp->mp = mp;

			return 0;
		}

		/*
		 * descend down to leftmost child page
		 */
		/* push (bn, index) of the parent page/entry */
		BT_PUSH(btstack, bn, 0);

		/* get the leftmost entry */
		stbl = DT_GETSTBL(p);
		xd = (pxd_t *) & p->slot[stbl[0]];

		/* get the child page block address */
		bn = addressPXD(xd);
		psize = lengthPXD(xd) << JFS_SBI(ip->i_sb)->l2bsize;

		/* unpin the parent page */
		DT_PUTPAGE(mp);
	}
}


/*
 *	dtReadNext()
 *
 * function: get the page of the specified offset (pn:index)
 *
 * return: if (offset > eof), bn = -1;
 *
 * note: if index > nextindex of the target leaf page,
 * start with 1st entry of next leaf page;
 */
3352 3353
static int dtReadNext(struct inode *ip, loff_t * offset,
		      struct btstack * btstack)
Dave Kleikamp's avatar
Dave Kleikamp committed
3354 3355 3356 3357 3358 3359 3360 3361
{
	int rc = 0;
	struct dtoffset {
		s16 pn;
		s16 index;
		s32 unused;
	} *dtoffset = (struct dtoffset *) offset;
	s64 bn;
3362
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
3363 3364 3365 3366
	dtpage_t *p;
	int index;
	int pn;
	s8 *stbl;
3367
	struct btframe *btsp, *parent;
Dave Kleikamp's avatar
Dave Kleikamp committed
3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527
	pxd_t *xd;

	/*
	 * get leftmost leaf page pinned
	 */
	if ((rc = dtReadFirst(ip, btstack)))
		return rc;

	/* get leaf page */
	DT_GETSEARCH(ip, btstack->top, bn, mp, p, index);

	/* get the start offset (pn:index) */
	pn = dtoffset->pn - 1;	/* Now pn = 0 represents leftmost leaf */
	index = dtoffset->index;

	/* start at leftmost page ? */
	if (pn == 0) {
		/* offset beyond eof ? */
		if (index < p->header.nextindex)
			goto out;

		if (p->header.flag & BT_ROOT) {
			bn = -1;
			goto out;
		}

		/* start with 1st entry of next leaf page */
		dtoffset->pn++;
		dtoffset->index = index = 0;
		goto a;
	}

	/* start at non-leftmost page: scan parent pages for large pn */
	if (p->header.flag & BT_ROOT) {
		bn = -1;
		goto out;
	}

	/* start after next leaf page ? */
	if (pn > 1)
		goto b;

	/* get leaf page pn = 1 */
      a:
	bn = le64_to_cpu(p->header.next);

	/* unpin leaf page */
	DT_PUTPAGE(mp);

	/* offset beyond eof ? */
	if (bn == 0) {
		bn = -1;
		goto out;
	}

	goto c;

	/*
	 * scan last internal page level to get target leaf page
	 */
      b:
	/* unpin leftmost leaf page */
	DT_PUTPAGE(mp);

	/* get left most parent page */
	btsp = btstack->top;
	parent = btsp - 1;
	bn = parent->bn;
	DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
	if (rc)
		return rc;

	/* scan parent pages at last internal page level */
	while (pn >= p->header.nextindex) {
		pn -= p->header.nextindex;

		/* get next parent page address */
		bn = le64_to_cpu(p->header.next);

		/* unpin current parent page */
		DT_PUTPAGE(mp);

		/* offset beyond eof ? */
		if (bn == 0) {
			bn = -1;
			goto out;
		}

		/* get next parent page */
		DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
		if (rc)
			return rc;

		/* update parent page stack frame */
		parent->bn = bn;
	}

	/* get leaf page address */
	stbl = DT_GETSTBL(p);
	xd = (pxd_t *) & p->slot[stbl[pn]];
	bn = addressPXD(xd);

	/* unpin parent page */
	DT_PUTPAGE(mp);

	/*
	 * get target leaf page
	 */
      c:
	DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
	if (rc)
		return rc;

	/*
	 * leaf page has been completed:
	 * start with 1st entry of next leaf page
	 */
	if (index >= p->header.nextindex) {
		bn = le64_to_cpu(p->header.next);

		/* unpin leaf page */
		DT_PUTPAGE(mp);

		/* offset beyond eof ? */
		if (bn == 0) {
			bn = -1;
			goto out;
		}

		/* get next leaf page */
		DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
		if (rc)
			return rc;

		/* start with 1st entry of next leaf page */
		dtoffset->pn++;
		dtoffset->index = 0;
	}

      out:
	/* return target leaf page pinned */
	btsp = btstack->top;
	btsp->bn = bn;
	btsp->index = dtoffset->index;
	btsp->mp = mp;

	return 0;
}


/*
 *	dtCompare()
 *
 * function: compare search key with an internal entry
 *
 * return:
 *	< 0 if k is < record
 *	= 0 if k is = record
 *	> 0 if k is > record
 */
3528
static int dtCompare(struct component_name * key,	/* search key */
Dave Kleikamp's avatar
Dave Kleikamp committed
3529 3530 3531
		     dtpage_t * p,	/* directory page */
		     int si)
{				/* entry slot index */
3532 3533
	wchar_t *kname, *name;
	int klen, namlen, len, rc;
3534 3535
	struct idtentry *ih;
	struct dtslot *t;
Dave Kleikamp's avatar
Dave Kleikamp committed
3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555

	/*
	 * force the left-most key on internal pages, at any level of
	 * the tree, to be less than any search key.
	 * this obviates having to update the leftmost key on an internal
	 * page when the user inserts a new key in the tree smaller than
	 * anything that has been stored.
	 *
	 * (? if/when dtSearch() narrows down to 1st entry (index = 0),
	 * at any internal page at any level of the tree,
	 * it descends to child of the entry anyway -
	 * ? make the entry as min size dummy entry)
	 *
	 * if (e->index == 0 && h->prevpg == P_INVALID && !(h->flags & BT_LEAF))
	 * return (1);
	 */

	kname = key->name;
	klen = key->namlen;

3556
	ih = (struct idtentry *) & p->slot[si];
Dave Kleikamp's avatar
Dave Kleikamp committed
3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573
	si = ih->next;
	name = ih->name;
	namlen = ih->namlen;
	len = min(namlen, DTIHDRDATALEN);

	/* compare with head/only segment */
	len = min(klen, len);
	if ((rc = UniStrncmp_le(kname, name, len)))
		return rc;

	klen -= len;
	namlen -= len;

	/* compare with additional segment(s) */
	kname += len;
	while (klen > 0 && namlen > 0) {
		/* compare with next name segment */
3574
		t = (struct dtslot *) & p->slot[si];
Dave Kleikamp's avatar
Dave Kleikamp committed
3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602
		len = min(namlen, DTSLOTDATALEN);
		len = min(klen, len);
		name = t->name;
		if ((rc = UniStrncmp_le(kname, name, len)))
			return rc;

		klen -= len;
		namlen -= len;
		kname += len;
		si = t->next;
	}

	return (klen - namlen);
}




/*
 *	ciCompare()
 *
 * function: compare search key with an (leaf/internal) entry
 *
 * return:
 *	< 0 if k is < record
 *	= 0 if k is = record
 *	> 0 if k is > record
 */
3603
static int ciCompare(struct component_name * key,	/* search key */
Dave Kleikamp's avatar
Dave Kleikamp committed
3604 3605 3606 3607
		     dtpage_t * p,	/* directory page */
		     int si,	/* entry slot index */
		     int flag)
{
3608 3609
	wchar_t *kname, *name, x;
	int klen, namlen, len, rc;
3610 3611 3612
	struct ldtentry *lh;
	struct idtentry *ih;
	struct dtslot *t;
Dave Kleikamp's avatar
Dave Kleikamp committed
3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637
	int i;

	/*
	 * force the left-most key on internal pages, at any level of
	 * the tree, to be less than any search key.
	 * this obviates having to update the leftmost key on an internal
	 * page when the user inserts a new key in the tree smaller than
	 * anything that has been stored.
	 *
	 * (? if/when dtSearch() narrows down to 1st entry (index = 0),
	 * at any internal page at any level of the tree,
	 * it descends to child of the entry anyway -
	 * ? make the entry as min size dummy entry)
	 *
	 * if (e->index == 0 && h->prevpg == P_INVALID && !(h->flags & BT_LEAF))
	 * return (1);
	 */

	kname = key->name;
	klen = key->namlen;

	/*
	 * leaf page entry
	 */
	if (p->header.flag & BT_LEAF) {
3638
		lh = (struct ldtentry *) & p->slot[si];
Dave Kleikamp's avatar
Dave Kleikamp committed
3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650
		si = lh->next;
		name = lh->name;
		namlen = lh->namlen;
		if (flag & JFS_DIR_INDEX)
			len = min(namlen, DTLHDRDATALEN);
		else
			len = min(namlen, DTLHDRDATALEN_LEGACY);
	}
	/*
	 * internal page entry
	 */
	else {
3651
		ih = (struct idtentry *) & p->slot[si];
Dave Kleikamp's avatar
Dave Kleikamp committed
3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675
		si = ih->next;
		name = ih->name;
		namlen = ih->namlen;
		len = min(namlen, DTIHDRDATALEN);
	}

	/* compare with head/only segment */
	len = min(klen, len);
	for (i = 0; i < len; i++, kname++, name++) {
		/* only uppercase if case-insensitive support is on */
		if ((flag & JFS_OS2) == JFS_OS2)
			x = UniToupper(le16_to_cpu(*name));
		else
			x = le16_to_cpu(*name);
		if ((rc = *kname - x))
			return rc;
	}

	klen -= len;
	namlen -= len;

	/* compare with additional segment(s) */
	while (klen > 0 && namlen > 0) {
		/* compare with next name segment */
3676
		t = (struct dtslot *) & p->slot[si];
Dave Kleikamp's avatar
Dave Kleikamp committed
3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710
		len = min(namlen, DTSLOTDATALEN);
		len = min(klen, len);
		name = t->name;
		for (i = 0; i < len; i++, kname++, name++) {
			/* only uppercase if case-insensitive support is on */
			if ((flag & JFS_OS2) == JFS_OS2)
				x = UniToupper(le16_to_cpu(*name));
			else
				x = le16_to_cpu(*name);

			if ((rc = *kname - x))
				return rc;
		}

		klen -= len;
		namlen -= len;
		si = t->next;
	}

	return (klen - namlen);
}


/*
 *	ciGetLeafPrefixKey()
 *
 * function: compute prefix of suffix compression
 *	     from two adjacent leaf entries
 *	     across page boundary
 *
 * return:
 *	Number of prefix bytes needed to distinguish b from a.
 */
static void ciGetLeafPrefixKey(dtpage_t * lp, int li, dtpage_t * rp,
3711
			       int ri, struct component_name * key, int flag)
Dave Kleikamp's avatar
Dave Kleikamp committed
3712
{
3713 3714
	int klen, namlen;
	wchar_t *pl, *pr, *kname;
Dave Kleikamp's avatar
Dave Kleikamp committed
3715
	wchar_t lname[JFS_NAME_MAX + 1];
3716
	struct component_name lkey = { 0, lname };
Dave Kleikamp's avatar
Dave Kleikamp committed
3717
	wchar_t rname[JFS_NAME_MAX + 1];
3718
	struct component_name rkey = { 0, rname };
Dave Kleikamp's avatar
Dave Kleikamp committed
3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764

	/* get left and right key */
	dtGetKey(lp, li, &lkey, flag);
	lkey.name[lkey.namlen] = 0;

	if ((flag & JFS_OS2) == JFS_OS2)
		ciToUpper(&lkey);

	dtGetKey(rp, ri, &rkey, flag);
	rkey.name[rkey.namlen] = 0;


	if ((flag & JFS_OS2) == JFS_OS2)
		ciToUpper(&rkey);

	/* compute prefix */
	klen = 0;
	kname = key->name;
	namlen = min(lkey.namlen, rkey.namlen);
	for (pl = lkey.name, pr = rkey.name;
	     namlen; pl++, pr++, namlen--, klen++, kname++) {
		*kname = *pr;
		if (*pl != *pr) {
			key->namlen = klen + 1;
			return;
		}
	}

	/* l->namlen <= r->namlen since l <= r */
	if (lkey.namlen < rkey.namlen) {
		*kname = *pr;
		key->namlen = klen + 1;
	} else			/* l->namelen == r->namelen */
		key->namlen = klen;

	return;
}



/*
 *	dtGetKey()
 *
 * function: get key of the entry
 */
static void dtGetKey(dtpage_t * p, int i,	/* entry index */
3765
		     struct component_name * key, int flag)
Dave Kleikamp's avatar
Dave Kleikamp committed
3766 3767 3768
{
	int si;
	s8 *stbl;
3769 3770 3771
	struct ldtentry *lh;
	struct idtentry *ih;
	struct dtslot *t;
Dave Kleikamp's avatar
Dave Kleikamp committed
3772 3773 3774 3775 3776 3777 3778
	int namlen, len;
	wchar_t *name, *kname;

	/* get entry */
	stbl = DT_GETSTBL(p);
	si = stbl[i];
	if (p->header.flag & BT_LEAF) {
3779
		lh = (struct ldtentry *) & p->slot[si];
Dave Kleikamp's avatar
Dave Kleikamp committed
3780 3781 3782 3783 3784 3785 3786 3787
		si = lh->next;
		namlen = lh->namlen;
		name = lh->name;
		if (flag & JFS_DIR_INDEX)
			len = min(namlen, DTLHDRDATALEN);
		else
			len = min(namlen, DTLHDRDATALEN_LEGACY);
	} else {
3788
		ih = (struct idtentry *) & p->slot[si];
Dave Kleikamp's avatar
Dave Kleikamp committed
3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826
		si = ih->next;
		namlen = ih->namlen;
		name = ih->name;
		len = min(namlen, DTIHDRDATALEN);
	}

	key->namlen = namlen;
	kname = key->name;

	/*
	 * move head/only segment
	 */
	UniStrncpy_le(kname, name, len);

	/*
	 * move additional segment(s)
	 */
	while (si >= 0) {
		/* get next segment */
		t = &p->slot[si];
		kname += len;
		namlen -= len;
		len = min(namlen, DTSLOTDATALEN);
		UniStrncpy_le(kname, t->name, len);

		si = t->next;
	}
}


/*
 *	dtInsertEntry()
 *
 * function: allocate free slot(s) and
 *	     write a leaf/internal entry
 *
 * return: entry slot index
 */
3827 3828
static void dtInsertEntry(dtpage_t * p, int index, struct component_name * key,
			  ddata_t * data, struct dt_lock ** dtlock)
Dave Kleikamp's avatar
Dave Kleikamp committed
3829
{
3830 3831 3832
	struct dtslot *h, *t;
	struct ldtentry *lh = 0;
	struct idtentry *ih = 0;
Dave Kleikamp's avatar
Dave Kleikamp committed
3833 3834 3835 3836
	int hsi, fsi, klen, len, nextindex;
	wchar_t *kname, *name;
	s8 *stbl;
	pxd_t *xd;
3837 3838
	struct dt_lock *dtlck = *dtlock;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
3839 3840
	int xsi, n;
	s64 bn = 0;
3841
	struct metapage *mp = 0;
Dave Kleikamp's avatar
Dave Kleikamp committed
3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853

	klen = key->namlen;
	kname = key->name;

	/* allocate a free slot */
	hsi = fsi = p->header.freelist;
	h = &p->slot[fsi];
	p->header.freelist = h->next;
	--p->header.freecnt;

	/* open new linelock */
	if (dtlck->index >= dtlck->maxcnt)
3854
		dtlck = (struct dt_lock *) txLinelock(dtlck);
Dave Kleikamp's avatar
Dave Kleikamp committed
3855

3856
	lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
3857 3858 3859 3860
	lv->offset = hsi;

	/* write head/only segment */
	if (p->header.flag & BT_LEAF) {
3861
		lh = (struct ldtentry *) h;
Dave Kleikamp's avatar
Dave Kleikamp committed
3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875
		lh->next = h->next;
		lh->inumber = data->leaf.ino;	/* little-endian */
		lh->namlen = klen;
		name = lh->name;
		if (data->leaf.ip) {
			len = min(klen, DTLHDRDATALEN);
			if (!(p->header.flag & BT_ROOT))
				bn = addressPXD(&p->header.self);
			lh->index = cpu_to_le32(add_index(data->leaf.tid,
							  data->leaf.ip,
							  bn, index));
		} else
			len = min(klen, DTLHDRDATALEN_LEGACY);
	} else {
3876
		ih = (struct idtentry *) h;
Dave Kleikamp's avatar
Dave Kleikamp committed
3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909
		ih->next = h->next;
		xd = (pxd_t *) ih;
		*xd = data->xd;
		ih->namlen = klen;
		name = ih->name;
		len = min(klen, DTIHDRDATALEN);
	}

	UniStrncpy_le(name, kname, len);

	n = 1;
	xsi = hsi;

	/* write additional segment(s) */
	t = h;
	klen -= len;
	while (klen) {
		/* get free slot */
		fsi = p->header.freelist;
		t = &p->slot[fsi];
		p->header.freelist = t->next;
		--p->header.freecnt;

		/* is next slot contiguous ? */
		if (fsi != xsi + 1) {
			/* close current linelock */
			lv->length = n;
			dtlck->index++;

			/* open new linelock */
			if (dtlck->index < dtlck->maxcnt)
				lv++;
			else {
3910 3911
				dtlck = (struct dt_lock *) txLinelock(dtlck);
				lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950
			}

			lv->offset = fsi;
			n = 0;
		}

		kname += len;
		len = min(klen, DTSLOTDATALEN);
		UniStrncpy_le(t->name, kname, len);

		n++;
		xsi = fsi;
		klen -= len;
	}

	/* close current linelock */
	lv->length = n;
	dtlck->index++;

	*dtlock = dtlck;

	/* terminate last/only segment */
	if (h == t) {
		/* single segment entry */
		if (p->header.flag & BT_LEAF)
			lh->next = -1;
		else
			ih->next = -1;
	} else
		/* multi-segment entry */
		t->next = -1;

	/* if insert into middle, shift right succeeding entries in stbl */
	stbl = DT_GETSTBL(p);
	nextindex = p->header.nextindex;
	if (index < nextindex) {
		memmove(stbl + index + 1, stbl + index, nextindex - index);

		if ((p->header.flag & BT_LEAF) && data->leaf.ip) {
3951 3952
			s64 lblock;

Dave Kleikamp's avatar
Dave Kleikamp committed
3953 3954 3955 3956 3957 3958
			/*
			 * Need to update slot number for entries that moved
			 * in the stbl
			 */
			mp = 0;
			for (n = index + 1; n <= nextindex; n++) {
3959
				lh = (struct ldtentry *) & (p->slot[stbl[n]]);
Dave Kleikamp's avatar
Dave Kleikamp committed
3960 3961
				modify_index(data->leaf.tid, data->leaf.ip,
					     le32_to_cpu(lh->index), bn, n,
3962
					     &mp, &lblock);
Dave Kleikamp's avatar
Dave Kleikamp committed
3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984
			}
			if (mp)
				release_metapage(mp);
		}
	}

	stbl[index] = hsi;

	/* advance next available entry index of stbl */
	++p->header.nextindex;
}


/*
 *	dtMoveEntry()
 *
 * function: move entries from split/left page to new/right page
 *
 *	nextindex of dst page and freelist/freecnt of both pages
 *	are updated.
 */
static void dtMoveEntry(dtpage_t * sp, int si, dtpage_t * dp,
3985
			struct dt_lock ** sdtlock, struct dt_lock ** ddtlock,
Dave Kleikamp's avatar
Dave Kleikamp committed
3986 3987 3988 3989 3990 3991 3992
			int do_index)
{
	int ssi, next;		/* src slot index */
	int di;			/* dst entry index */
	int dsi;		/* dst slot index */
	s8 *sstbl, *dstbl;	/* sorted entry table */
	int snamlen, len;
3993 3994 3995 3996 3997
	struct ldtentry *slh, *dlh = 0;
	struct idtentry *sih, *dih = 0;
	struct dtslot *h, *s, *d;
	struct dt_lock *sdtlck = *sdtlock, *ddtlck = *ddtlock;
	struct lv *slv, *dlv;
Dave Kleikamp's avatar
Dave Kleikamp committed
3998 3999 4000 4001 4002 4003 4004 4005 4006 4007
	int xssi, ns, nd;
	int sfsi;

	sstbl = (s8 *) & sp->slot[sp->header.stblindex];
	dstbl = (s8 *) & dp->slot[dp->header.stblindex];

	dsi = dp->header.freelist;	/* first (whole page) free slot */
	sfsi = sp->header.freelist;

	/* linelock destination entry slot */
4008
	dlv = & ddtlck->lv[ddtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
4009 4010 4011
	dlv->offset = dsi;

	/* linelock source entry slot */
4012
	slv = & sdtlck->lv[sdtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033
	slv->offset = sstbl[si];
	xssi = slv->offset - 1;

	/*
	 * move entries
	 */
	ns = nd = 0;
	for (di = 0; si < sp->header.nextindex; si++, di++) {
		ssi = sstbl[si];
		dstbl[di] = dsi;

		/* is next slot contiguous ? */
		if (ssi != xssi + 1) {
			/* close current linelock */
			slv->length = ns;
			sdtlck->index++;

			/* open new linelock */
			if (sdtlck->index < sdtlck->maxcnt)
				slv++;
			else {
4034 4035
				sdtlck = (struct dt_lock *) txLinelock(sdtlck);
				slv = & sdtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051
			}

			slv->offset = ssi;
			ns = 0;
		}

		/*
		 * move head/only segment of an entry
		 */
		/* get dst slot */
		h = d = &dp->slot[dsi];

		/* get src slot and move */
		s = &sp->slot[ssi];
		if (sp->header.flag & BT_LEAF) {
			/* get source entry */
4052 4053
			slh = (struct ldtentry *) s;
			dlh = (struct ldtentry *) h;
Dave Kleikamp's avatar
Dave Kleikamp committed
4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069
			snamlen = slh->namlen;

			if (do_index) {
				len = min(snamlen, DTLHDRDATALEN);
				dlh->index = slh->index; /* little-endian */
			} else
				len = min(snamlen, DTLHDRDATALEN_LEGACY);

			memcpy(dlh, slh, 6 + len * 2);

			next = slh->next;

			/* update dst head/only segment next field */
			dsi++;
			dlh->next = dsi;
		} else {
4070
			sih = (struct idtentry *) s;
Dave Kleikamp's avatar
Dave Kleikamp committed
4071 4072 4073
			snamlen = sih->namlen;

			len = min(snamlen, DTIHDRDATALEN);
4074
			dih = (struct idtentry *) h;
Dave Kleikamp's avatar
Dave Kleikamp committed
4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106
			memcpy(dih, sih, 10 + len * 2);
			next = sih->next;

			dsi++;
			dih->next = dsi;
		}

		/* free src head/only segment */
		s->next = sfsi;
		s->cnt = 1;
		sfsi = ssi;

		ns++;
		nd++;
		xssi = ssi;

		/*
		 * move additional segment(s) of the entry
		 */
		snamlen -= len;
		while ((ssi = next) >= 0) {
			/* is next slot contiguous ? */
			if (ssi != xssi + 1) {
				/* close current linelock */
				slv->length = ns;
				sdtlck->index++;

				/* open new linelock */
				if (sdtlck->index < sdtlck->maxcnt)
					slv++;
				else {
					sdtlck =
4107
					    (struct dt_lock *)
Dave Kleikamp's avatar
Dave Kleikamp committed
4108
					    txLinelock(sdtlck);
4109
					slv = & sdtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184
				}

				slv->offset = ssi;
				ns = 0;
			}

			/* get next source segment */
			s = &sp->slot[ssi];

			/* get next destination free slot */
			d++;

			len = min(snamlen, DTSLOTDATALEN);
			UniStrncpy(d->name, s->name, len);

			ns++;
			nd++;
			xssi = ssi;

			dsi++;
			d->next = dsi;

			/* free source segment */
			next = s->next;
			s->next = sfsi;
			s->cnt = 1;
			sfsi = ssi;

			snamlen -= len;
		}		/* end while */

		/* terminate dst last/only segment */
		if (h == d) {
			/* single segment entry */
			if (dp->header.flag & BT_LEAF)
				dlh->next = -1;
			else
				dih->next = -1;
		} else
			/* multi-segment entry */
			d->next = -1;
	}			/* end for */

	/* close current linelock */
	slv->length = ns;
	sdtlck->index++;
	*sdtlock = sdtlck;

	dlv->length = nd;
	ddtlck->index++;
	*ddtlock = ddtlck;

	/* update source header */
	sp->header.freelist = sfsi;
	sp->header.freecnt += nd;

	/* update destination header */
	dp->header.nextindex = di;

	dp->header.freelist = dsi;
	dp->header.freecnt -= nd;
}


/*
 *	dtDeleteEntry()
 *
 * function: free a (leaf/internal) entry
 *
 * log freelist header, stbl, and each segment slot of entry
 * (even though last/only segment next field is modified,
 * physical image logging requires all segment slots of
 * the entry logged to avoid applying previous updates
 * to the same slots)
 */
4185
static void dtDeleteEntry(dtpage_t * p, int fi, struct dt_lock ** dtlock)
Dave Kleikamp's avatar
Dave Kleikamp committed
4186 4187 4188
{
	int fsi;		/* free entry slot index */
	s8 *stbl;
4189
	struct dtslot *t;
Dave Kleikamp's avatar
Dave Kleikamp committed
4190
	int si, freecnt;
4191 4192
	struct dt_lock *dtlck = *dtlock;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
4193 4194 4195 4196 4197 4198 4199 4200
	int xsi, n;

	/* get free entry slot index */
	stbl = DT_GETSTBL(p);
	fsi = stbl[fi];

	/* open new linelock */
	if (dtlck->index >= dtlck->maxcnt)
4201 4202
		dtlck = (struct dt_lock *) txLinelock(dtlck);
	lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
4203 4204 4205 4206 4207 4208

	lv->offset = fsi;

	/* get the head/only segment */
	t = &p->slot[fsi];
	if (p->header.flag & BT_LEAF)
4209
		si = ((struct ldtentry *) t)->next;
Dave Kleikamp's avatar
Dave Kleikamp committed
4210
	else
4211
		si = ((struct idtentry *) t)->next;
Dave Kleikamp's avatar
Dave Kleikamp committed
4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229
	t->next = si;
	t->cnt = 1;

	n = freecnt = 1;
	xsi = fsi;

	/* find the last/only segment */
	while (si >= 0) {
		/* is next slot contiguous ? */
		if (si != xsi + 1) {
			/* close current linelock */
			lv->length = n;
			dtlck->index++;

			/* open new linelock */
			if (dtlck->index < dtlck->maxcnt)
				lv++;
			else {
4230 4231
				dtlck = (struct dt_lock *) txLinelock(dtlck);
				lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279
			}

			lv->offset = si;
			n = 0;
		}

		n++;
		xsi = si;
		freecnt++;

		t = &p->slot[si];
		t->cnt = 1;
		si = t->next;
	}

	/* close current linelock */
	lv->length = n;
	dtlck->index++;

	*dtlock = dtlck;

	/* update freelist */
	t->next = p->header.freelist;
	p->header.freelist = fsi;
	p->header.freecnt += freecnt;

	/* if delete from middle,
	 * shift left the succedding entries in the stbl
	 */
	si = p->header.nextindex;
	if (fi < si - 1)
		memmove(&stbl[fi], &stbl[fi + 1], si - fi - 1);

	p->header.nextindex--;
}


/*
 *	dtTruncateEntry()
 *
 * function: truncate a (leaf/internal) entry
 *
 * log freelist header, stbl, and each segment slot of entry
 * (even though last/only segment next field is modified,
 * physical image logging requires all segment slots of
 * the entry logged to avoid applying previous updates
 * to the same slots)
 */
4280
static void dtTruncateEntry(dtpage_t * p, int ti, struct dt_lock ** dtlock)
Dave Kleikamp's avatar
Dave Kleikamp committed
4281 4282 4283
{
	int tsi;		/* truncate entry slot index */
	s8 *stbl;
4284
	struct dtslot *t;
Dave Kleikamp's avatar
Dave Kleikamp committed
4285
	int si, freecnt;
4286 4287
	struct dt_lock *dtlck = *dtlock;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
4288 4289 4290 4291 4292 4293 4294 4295
	int fsi, xsi, n;

	/* get free entry slot index */
	stbl = DT_GETSTBL(p);
	tsi = stbl[ti];

	/* open new linelock */
	if (dtlck->index >= dtlck->maxcnt)
4296 4297
		dtlck = (struct dt_lock *) txLinelock(dtlck);
	lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
4298 4299 4300 4301 4302 4303

	lv->offset = tsi;

	/* get the head/only segment */
	t = &p->slot[tsi];
	ASSERT(p->header.flag & BT_INTERNAL);
4304 4305 4306
	((struct idtentry *) t)->namlen = 0;
	si = ((struct idtentry *) t)->next;
	((struct idtentry *) t)->next = -1;
Dave Kleikamp's avatar
Dave Kleikamp committed
4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324

	n = 1;
	freecnt = 0;
	fsi = si;
	xsi = tsi;

	/* find the last/only segment */
	while (si >= 0) {
		/* is next slot contiguous ? */
		if (si != xsi + 1) {
			/* close current linelock */
			lv->length = n;
			dtlck->index++;

			/* open new linelock */
			if (dtlck->index < dtlck->maxcnt)
				lv++;
			else {
4325 4326
				dtlck = (struct dt_lock *) txLinelock(dtlck);
				lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361
			}

			lv->offset = si;
			n = 0;
		}

		n++;
		xsi = si;
		freecnt++;

		t = &p->slot[si];
		t->cnt = 1;
		si = t->next;
	}

	/* close current linelock */
	lv->length = n;
	dtlck->index++;

	*dtlock = dtlck;

	/* update freelist */
	if (freecnt == 0)
		return;
	t->next = p->header.freelist;
	p->header.freelist = fsi;
	p->header.freecnt += freecnt;
}


/*
 *	dtLinelockFreelist()
 */
static void dtLinelockFreelist(dtpage_t * p,	/* directory page */
			       int m,	/* max slot index */
4362
			       struct dt_lock ** dtlock)
Dave Kleikamp's avatar
Dave Kleikamp committed
4363 4364
{
	int fsi;		/* free entry slot index */
4365
	struct dtslot *t;
Dave Kleikamp's avatar
Dave Kleikamp committed
4366
	int si;
4367 4368
	struct dt_lock *dtlck = *dtlock;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
4369 4370 4371 4372 4373 4374 4375
	int xsi, n;

	/* get free entry slot index */
	fsi = p->header.freelist;

	/* open new linelock */
	if (dtlck->index >= dtlck->maxcnt)
4376 4377
		dtlck = (struct dt_lock *) txLinelock(dtlck);
	lv = & dtlck->lv[dtlck->index];
Dave Kleikamp's avatar
Dave Kleikamp committed
4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398

	lv->offset = fsi;

	n = 1;
	xsi = fsi;

	t = &p->slot[fsi];
	si = t->next;

	/* find the last/only segment */
	while (si < m && si >= 0) {
		/* is next slot contiguous ? */
		if (si != xsi + 1) {
			/* close current linelock */
			lv->length = n;
			dtlck->index++;

			/* open new linelock */
			if (dtlck->index < dtlck->maxcnt)
				lv++;
			else {
4399 4400
				dtlck = (struct dt_lock *) txLinelock(dtlck);
				lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435
			}

			lv->offset = si;
			n = 0;
		}

		n++;
		xsi = si;

		t = &p->slot[si];
		si = t->next;
	}

	/* close current linelock */
	lv->length = n;
	dtlck->index++;

	*dtlock = dtlck;
}


/*
 * NAME: dtModify
 *
 * FUNCTION: Modify the inode number part of a directory entry
 *
 * PARAMETERS:
 *	tid	- Transaction id
 *	ip	- Inode of parent directory
 *	key	- Name of entry to be modified
 *	orig_ino	- Original inode number expected in entry
 *	new_ino	- New inode number to put into entry
 *	flag	- JFS_RENAME
 *
 * RETURNS:
4436 4437
 *	-ESTALE	- If entry found does not match orig_ino passed in
 *	-ENOENT	- If no entry can be found to match key
Dave Kleikamp's avatar
Dave Kleikamp committed
4438 4439 4440
 *	0	- If successfully modified entry
 */
int dtModify(tid_t tid, struct inode *ip,
4441
	 struct component_name * key, ino_t * orig_ino, ino_t new_ino, int flag)
Dave Kleikamp's avatar
Dave Kleikamp committed
4442 4443 4444
{
	int rc;
	s64 bn;
4445
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
4446 4447
	dtpage_t *p;
	int index;
4448 4449 4450 4451
	struct btstack btstack;
	struct tlock *tlck;
	struct dt_lock *dtlck;
	struct lv *lv;
Dave Kleikamp's avatar
Dave Kleikamp committed
4452 4453
	s8 *stbl;
	int entry_si;		/* entry slot index */
4454
	struct ldtentry *entry;
Dave Kleikamp's avatar
Dave Kleikamp committed
4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471

	/*
	 *      search for the entry to modify:
	 *
	 * dtSearch() returns (leaf page pinned, index at which to modify).
	 */
	if ((rc = dtSearch(ip, key, orig_ino, &btstack, flag)))
		return rc;

	/* retrieve search result */
	DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);

	BT_MARK_DIRTY(mp, ip);
	/*
	 * acquire a transaction lock on the leaf page of named entry
	 */
	tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
4472
	dtlck = (struct dt_lock *) & tlck->lock;
Dave Kleikamp's avatar
Dave Kleikamp committed
4473 4474 4475 4476 4477 4478 4479

	/* get slot index of the entry */
	stbl = DT_GETSTBL(p);
	entry_si = stbl[index];

	/* linelock entry */
	ASSERT(dtlck->index == 0);
4480
	lv = & dtlck->lv[0];
Dave Kleikamp's avatar
Dave Kleikamp committed
4481 4482 4483 4484 4485
	lv->offset = entry_si;
	lv->length = 1;
	dtlck->index++;

	/* get the head/only segment */
4486
	entry = (struct ldtentry *) & p->slot[entry_si];
Dave Kleikamp's avatar
Dave Kleikamp committed
4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505

	/* substitute the inode number of the entry */
	entry->inumber = cpu_to_le32(new_ino);

	/* unpin the leaf page */
	DT_PUTPAGE(mp);

	return 0;
}

#ifdef _JFS_DEBUG_DTREE
/*
 *	dtDisplayTree()
 *
 * function: traverse forward
 */
int dtDisplayTree(struct inode *ip)
{
	int rc;
4506
	struct metapage *mp;
Dave Kleikamp's avatar
Dave Kleikamp committed
4507 4508 4509 4510
	dtpage_t *p;
	s64 bn, pbn;
	int index, lastindex, v, h;
	pxd_t *xd;
4511 4512 4513
	struct btstack btstack;
	struct btframe *btsp;
	struct btframe *parent;
Dave Kleikamp's avatar
Dave Kleikamp committed
4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643
	u8 *stbl;
	int psize = 256;

	printk("display B+-tree.\n");

	/* clear stack */
	btsp = btstack.stack;

	/*
	 * start with root
	 *
	 * root resides in the inode
	 */
	bn = 0;
	v = h = 0;

	/*
	 * first access of each page:
	 */
      newPage:
	DT_GETPAGE(ip, bn, mp, psize, p, rc);
	if (rc)
		return rc;

	/* process entries forward from first index */
	index = 0;
	lastindex = p->header.nextindex - 1;

	if (p->header.flag & BT_INTERNAL) {
		/*
		 * first access of each internal page
		 */
		printf("internal page ");
		dtDisplayPage(ip, bn, p);

		goto getChild;
	} else {		/* (p->header.flag & BT_LEAF) */

		/*
		 * first access of each leaf page
		 */
		printf("leaf page ");
		dtDisplayPage(ip, bn, p);

		/*
		 * process leaf page entries
		 *
		 for ( ; index <= lastindex; index++)
		 {
		 }
		 */

		/* unpin the leaf page */
		DT_PUTPAGE(mp);
	}

	/*
	 * go back up to the parent page
	 */
      getParent:
	/* pop/restore parent entry for the current child page */
	if ((parent = (btsp == btstack.stack ? NULL : --btsp)) == NULL)
		/* current page must have been root */
		return;

	/*
	 * parent page scan completed
	 */
	if ((index = parent->index) == (lastindex = parent->lastindex)) {
		/* go back up to the parent page */
		goto getParent;
	}

	/*
	 * parent page has entries remaining
	 */
	/* get back the parent page */
	bn = parent->bn;
	/* v = parent->level; */
	DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
	if (rc)
		return rc;

	/* get next parent entry */
	index++;

	/*
	 * internal page: go down to child page of current entry
	 */
      getChild:
	/* push/save current parent entry for the child page */
	btsp->bn = pbn = bn;
	btsp->index = index;
	btsp->lastindex = lastindex;
	/* btsp->level = v; */
	/* btsp->node = h; */
	++btsp;

	/* get current entry for the child page */
	stbl = DT_GETSTBL(p);
	xd = (pxd_t *) & p->slot[stbl[index]];

	/*
	 * first access of each internal entry:
	 */

	/* get child page */
	bn = addressPXD(xd);
	psize = lengthPXD(xd) << ip->i_ipmnt->i_l2bsize;

	printk("traverse down 0x%Lx[%d]->0x%Lx\n", pbn, index, bn);
	v++;
	h = index;

	/* release parent page */
	DT_PUTPAGE(mp);

	/* process the child page */
	goto newPage;
}


/*
 *	dtDisplayPage()
 *
 * function: display page
 */
int dtDisplayPage(struct inode *ip, s64 bn, dtpage_t * p)
{
	int rc;
4644 4645 4646
	struct metapage *mp;
	struct ldtentry *lh;
	struct idtentry *ih;
Dave Kleikamp's avatar
Dave Kleikamp committed
4647 4648 4649 4650
	pxd_t *xd;
	int i, j;
	u8 *stbl;
	wchar_t name[JFS_NAME_MAX + 1];
4651
	struct component_name key = { 0, name };
Dave Kleikamp's avatar
Dave Kleikamp committed
4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670
	int freepage = 0;

	if (p == NULL) {
		freepage = 1;
		DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
		if (rc)
			return rc;
	}

	/* display page control */
	printk("bn:0x%Lx flag:0x%08x nextindex:%d\n",
	       bn, p->header.flag, p->header.nextindex);

	/* display entries */
	stbl = DT_GETSTBL(p);
	for (i = 0, j = 1; i < p->header.nextindex; i++, j++) {
		dtGetKey(p, i, &key, JFS_SBI(ip->i_sb)->mntflag);
		key.name[key.namlen] = '\0';
		if (p->header.flag & BT_LEAF) {
4671
			lh = (struct ldtentry *) & p->slot[stbl[i]];
Dave Kleikamp's avatar
Dave Kleikamp committed
4672 4673 4674
			printf("\t[%d] %s:%d", i, key.name,
			       le32_to_cpu(lh->inumber));
		} else {
4675
			ih = (struct idtentry *) & p->slot[stbl[i]];
Dave Kleikamp's avatar
Dave Kleikamp committed
4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694
			xd = (pxd_t *) ih;
			bn = addressPXD(xd);
			printf("\t[%d] %s:0x%Lx", i, key.name, bn);
		}

		if (j == 4) {
			printf("\n");
			j = 0;
		}
	}

	printf("\n");

	if (freepage)
		DT_PUTPAGE(mp);

	return 0;
}
#endif				/* _JFS_DEBUG_DTREE */