xfs_ialloc_btree.c 41.2 KB
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 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 1180 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 1229 1230 1231 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 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 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 1401 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 1437 1438 1439
/*
 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
 * All Rights Reserved.
 *
 * 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.
 *
 * This program is distributed in the hope that it would 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 the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_types.h"
#include "xfs_bit.h"
#include "xfs_log.h"
#include "xfs_inum.h"
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_dir2.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
#include "xfs_bmap_btree.h"
#include "xfs_alloc_btree.h"
#include "xfs_ialloc_btree.h"
#include "xfs_dir2_sf.h"
#include "xfs_attr_sf.h"
#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_btree.h"
#include "xfs_ialloc.h"
#include "xfs_alloc.h"
#include "xfs_error.h"

STATIC void xfs_inobt_log_block(xfs_trans_t *, xfs_buf_t *, int);
STATIC void xfs_inobt_log_keys(xfs_btree_cur_t *, xfs_buf_t *, int, int);
STATIC void xfs_inobt_log_ptrs(xfs_btree_cur_t *, xfs_buf_t *, int, int);
STATIC void xfs_inobt_log_recs(xfs_btree_cur_t *, xfs_buf_t *, int, int);
STATIC int xfs_inobt_newroot(xfs_btree_cur_t *, int *);
STATIC int xfs_inobt_split(xfs_btree_cur_t *, int, xfs_agblock_t *,
		xfs_inobt_key_t *, xfs_btree_cur_t **, int *);

/*
 * Single level of the xfs_inobt_delete record deletion routine.
 * Delete record pointed to by cur/level.
 * Remove the record from its block then rebalance the tree.
 * Return 0 for error, 1 for done, 2 to go on to the next level.
 */
STATIC int				/* error */
xfs_inobt_delrec(
	xfs_btree_cur_t		*cur,	/* btree cursor */
	int			level,	/* level removing record from */
	int			*stat)	/* fail/done/go-on */
{
	xfs_buf_t		*agbp;	/* buffer for a.g. inode header */
	xfs_mount_t		*mp;	/* mount structure */
	xfs_agi_t		*agi;	/* allocation group inode header */
	xfs_inobt_block_t	*block;	/* btree block record/key lives in */
	xfs_agblock_t		bno;	/* btree block number */
	xfs_buf_t		*bp;	/* buffer for block */
	int			error;	/* error return value */
	int			i;	/* loop index */
	xfs_inobt_key_t		key;	/* kp points here if block is level 0 */
	xfs_inobt_key_t		*kp = NULL;	/* pointer to btree keys */
	xfs_agblock_t		lbno;	/* left block's block number */
	xfs_buf_t		*lbp;	/* left block's buffer pointer */
	xfs_inobt_block_t	*left;	/* left btree block */
	xfs_inobt_key_t		*lkp;	/* left block key pointer */
	xfs_inobt_ptr_t		*lpp;	/* left block address pointer */
	int			lrecs = 0;	/* number of records in left block */
	xfs_inobt_rec_t		*lrp;	/* left block record pointer */
	xfs_inobt_ptr_t		*pp = NULL;	/* pointer to btree addresses */
	int			ptr;	/* index in btree block for this rec */
	xfs_agblock_t		rbno;	/* right block's block number */
	xfs_buf_t		*rbp;	/* right block's buffer pointer */
	xfs_inobt_block_t	*right;	/* right btree block */
	xfs_inobt_key_t		*rkp;	/* right block key pointer */
	xfs_inobt_rec_t		*rp;	/* pointer to btree records */
	xfs_inobt_ptr_t		*rpp;	/* right block address pointer */
	int			rrecs = 0;	/* number of records in right block */
	int			numrecs;
	xfs_inobt_rec_t		*rrp;	/* right block record pointer */
	xfs_btree_cur_t		*tcur;	/* temporary btree cursor */

	mp = cur->bc_mp;

	/*
	 * Get the index of the entry being deleted, check for nothing there.
	 */
	ptr = cur->bc_ptrs[level];
	if (ptr == 0) {
		*stat = 0;
		return 0;
	}

	/*
	 * Get the buffer & block containing the record or key/ptr.
	 */
	bp = cur->bc_bufs[level];
	block = XFS_BUF_TO_INOBT_BLOCK(bp);
#ifdef DEBUG
	if ((error = xfs_btree_check_sblock(cur, block, level, bp)))
		return error;
#endif
	/*
	 * Fail if we're off the end of the block.
	 */

	numrecs = be16_to_cpu(block->bb_numrecs);
	if (ptr > numrecs) {
		*stat = 0;
		return 0;
	}
	/*
	 * It's a nonleaf.  Excise the key and ptr being deleted, by
	 * sliding the entries past them down one.
	 * Log the changed areas of the block.
	 */
	if (level > 0) {
		kp = XFS_INOBT_KEY_ADDR(block, 1, cur);
		pp = XFS_INOBT_PTR_ADDR(block, 1, cur);
#ifdef DEBUG
		for (i = ptr; i < numrecs; i++) {
			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(pp[i]), level)))
				return error;
		}
#endif
		if (ptr < numrecs) {
			memmove(&kp[ptr - 1], &kp[ptr],
				(numrecs - ptr) * sizeof(*kp));
			memmove(&pp[ptr - 1], &pp[ptr],
				(numrecs - ptr) * sizeof(*kp));
			xfs_inobt_log_keys(cur, bp, ptr, numrecs - 1);
			xfs_inobt_log_ptrs(cur, bp, ptr, numrecs - 1);
		}
	}
	/*
	 * It's a leaf.  Excise the record being deleted, by sliding the
	 * entries past it down one.  Log the changed areas of the block.
	 */
	else {
		rp = XFS_INOBT_REC_ADDR(block, 1, cur);
		if (ptr < numrecs) {
			memmove(&rp[ptr - 1], &rp[ptr],
				(numrecs - ptr) * sizeof(*rp));
			xfs_inobt_log_recs(cur, bp, ptr, numrecs - 1);
		}
		/*
		 * If it's the first record in the block, we'll need a key
		 * structure to pass up to the next level (updkey).
		 */
		if (ptr == 1) {
			key.ir_startino = rp->ir_startino;
			kp = &key;
		}
	}
	/*
	 * Decrement and log the number of entries in the block.
	 */
	numrecs--;
	block->bb_numrecs = cpu_to_be16(numrecs);
	xfs_inobt_log_block(cur->bc_tp, bp, XFS_BB_NUMRECS);
	/*
	 * Is this the root level?  If so, we're almost done.
	 */
	if (level == cur->bc_nlevels - 1) {
		/*
		 * If this is the root level,
		 * and there's only one entry left,
		 * and it's NOT the leaf level,
		 * then we can get rid of this level.
		 */
		if (numrecs == 1 && level > 0) {
			agbp = cur->bc_private.a.agbp;
			agi = XFS_BUF_TO_AGI(agbp);
			/*
			 * pp is still set to the first pointer in the block.
			 * Make it the new root of the btree.
			 */
			bno = be32_to_cpu(agi->agi_root);
			agi->agi_root = *pp;
			be32_add_cpu(&agi->agi_level, -1);
			/*
			 * Free the block.
			 */
			if ((error = xfs_free_extent(cur->bc_tp,
				XFS_AGB_TO_FSB(mp, cur->bc_private.a.agno, bno), 1)))
				return error;
			xfs_trans_binval(cur->bc_tp, bp);
			xfs_ialloc_log_agi(cur->bc_tp, agbp,
				XFS_AGI_ROOT | XFS_AGI_LEVEL);
			/*
			 * Update the cursor so there's one fewer level.
			 */
			cur->bc_bufs[level] = NULL;
			cur->bc_nlevels--;
		} else if (level > 0 &&
			   (error = xfs_btree_decrement(cur, level, &i)))
			return error;
		*stat = 1;
		return 0;
	}
	/*
	 * If we deleted the leftmost entry in the block, update the
	 * key values above us in the tree.
	 */
	if (ptr == 1 && (error = xfs_btree_updkey(cur, (union xfs_btree_key *)kp, level + 1)))
		return error;
	/*
	 * If the number of records remaining in the block is at least
	 * the minimum, we're done.
	 */
	if (numrecs >= XFS_INOBT_BLOCK_MINRECS(level, cur)) {
		if (level > 0 &&
		    (error = xfs_btree_decrement(cur, level, &i)))
			return error;
		*stat = 1;
		return 0;
	}
	/*
	 * Otherwise, we have to move some records around to keep the
	 * tree balanced.  Look at the left and right sibling blocks to
	 * see if we can re-balance by moving only one record.
	 */
	rbno = be32_to_cpu(block->bb_rightsib);
	lbno = be32_to_cpu(block->bb_leftsib);
	bno = NULLAGBLOCK;
	ASSERT(rbno != NULLAGBLOCK || lbno != NULLAGBLOCK);
	/*
	 * Duplicate the cursor so our btree manipulations here won't
	 * disrupt the next level up.
	 */
	if ((error = xfs_btree_dup_cursor(cur, &tcur)))
		return error;
	/*
	 * If there's a right sibling, see if it's ok to shift an entry
	 * out of it.
	 */
	if (rbno != NULLAGBLOCK) {
		/*
		 * Move the temp cursor to the last entry in the next block.
		 * Actually any entry but the first would suffice.
		 */
		i = xfs_btree_lastrec(tcur, level);
		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
		if ((error = xfs_btree_increment(tcur, level, &i)))
			goto error0;
		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
		i = xfs_btree_lastrec(tcur, level);
		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
		/*
		 * Grab a pointer to the block.
		 */
		rbp = tcur->bc_bufs[level];
		right = XFS_BUF_TO_INOBT_BLOCK(rbp);
#ifdef DEBUG
		if ((error = xfs_btree_check_sblock(cur, right, level, rbp)))
			goto error0;
#endif
		/*
		 * Grab the current block number, for future use.
		 */
		bno = be32_to_cpu(right->bb_leftsib);
		/*
		 * If right block is full enough so that removing one entry
		 * won't make it too empty, and left-shifting an entry out
		 * of right to us works, we're done.
		 */
		if (be16_to_cpu(right->bb_numrecs) - 1 >=
		     XFS_INOBT_BLOCK_MINRECS(level, cur)) {
			if ((error = xfs_btree_lshift(tcur, level, &i)))
				goto error0;
			if (i) {
				ASSERT(be16_to_cpu(block->bb_numrecs) >=
				       XFS_INOBT_BLOCK_MINRECS(level, cur));
				xfs_btree_del_cursor(tcur,
						     XFS_BTREE_NOERROR);
				if (level > 0 &&
				    (error = xfs_btree_decrement(cur, level,
						&i)))
					return error;
				*stat = 1;
				return 0;
			}
		}
		/*
		 * Otherwise, grab the number of records in right for
		 * future reference, and fix up the temp cursor to point
		 * to our block again (last record).
		 */
		rrecs = be16_to_cpu(right->bb_numrecs);
		if (lbno != NULLAGBLOCK) {
			xfs_btree_firstrec(tcur, level);
			if ((error = xfs_btree_decrement(tcur, level, &i)))
				goto error0;
		}
	}
	/*
	 * If there's a left sibling, see if it's ok to shift an entry
	 * out of it.
	 */
	if (lbno != NULLAGBLOCK) {
		/*
		 * Move the temp cursor to the first entry in the
		 * previous block.
		 */
		xfs_btree_firstrec(tcur, level);
		if ((error = xfs_btree_decrement(tcur, level, &i)))
			goto error0;
		xfs_btree_firstrec(tcur, level);
		/*
		 * Grab a pointer to the block.
		 */
		lbp = tcur->bc_bufs[level];
		left = XFS_BUF_TO_INOBT_BLOCK(lbp);
#ifdef DEBUG
		if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
			goto error0;
#endif
		/*
		 * Grab the current block number, for future use.
		 */
		bno = be32_to_cpu(left->bb_rightsib);
		/*
		 * If left block is full enough so that removing one entry
		 * won't make it too empty, and right-shifting an entry out
		 * of left to us works, we're done.
		 */
		if (be16_to_cpu(left->bb_numrecs) - 1 >=
		     XFS_INOBT_BLOCK_MINRECS(level, cur)) {
			if ((error = xfs_btree_rshift(tcur, level, &i)))
				goto error0;
			if (i) {
				ASSERT(be16_to_cpu(block->bb_numrecs) >=
				       XFS_INOBT_BLOCK_MINRECS(level, cur));
				xfs_btree_del_cursor(tcur,
						     XFS_BTREE_NOERROR);
				if (level == 0)
					cur->bc_ptrs[0]++;
				*stat = 1;
				return 0;
			}
		}
		/*
		 * Otherwise, grab the number of records in right for
		 * future reference.
		 */
		lrecs = be16_to_cpu(left->bb_numrecs);
	}
	/*
	 * Delete the temp cursor, we're done with it.
	 */
	xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
	/*
	 * If here, we need to do a join to keep the tree balanced.
	 */
	ASSERT(bno != NULLAGBLOCK);
	/*
	 * See if we can join with the left neighbor block.
	 */
	if (lbno != NULLAGBLOCK &&
	    lrecs + numrecs <= XFS_INOBT_BLOCK_MAXRECS(level, cur)) {
		/*
		 * Set "right" to be the starting block,
		 * "left" to be the left neighbor.
		 */
		rbno = bno;
		right = block;
		rrecs = be16_to_cpu(right->bb_numrecs);
		rbp = bp;
		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
				cur->bc_private.a.agno, lbno, 0, &lbp,
				XFS_INO_BTREE_REF)))
			return error;
		left = XFS_BUF_TO_INOBT_BLOCK(lbp);
		lrecs = be16_to_cpu(left->bb_numrecs);
		if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
			return error;
	}
	/*
	 * If that won't work, see if we can join with the right neighbor block.
	 */
	else if (rbno != NULLAGBLOCK &&
		 rrecs + numrecs <= XFS_INOBT_BLOCK_MAXRECS(level, cur)) {
		/*
		 * Set "left" to be the starting block,
		 * "right" to be the right neighbor.
		 */
		lbno = bno;
		left = block;
		lrecs = be16_to_cpu(left->bb_numrecs);
		lbp = bp;
		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
				cur->bc_private.a.agno, rbno, 0, &rbp,
				XFS_INO_BTREE_REF)))
			return error;
		right = XFS_BUF_TO_INOBT_BLOCK(rbp);
		rrecs = be16_to_cpu(right->bb_numrecs);
		if ((error = xfs_btree_check_sblock(cur, right, level, rbp)))
			return error;
	}
	/*
	 * Otherwise, we can't fix the imbalance.
	 * Just return.  This is probably a logic error, but it's not fatal.
	 */
	else {
		if (level > 0 && (error = xfs_btree_decrement(cur, level, &i)))
			return error;
		*stat = 1;
		return 0;
	}
	/*
	 * We're now going to join "left" and "right" by moving all the stuff
	 * in "right" to "left" and deleting "right".
	 */
	if (level > 0) {
		/*
		 * It's a non-leaf.  Move keys and pointers.
		 */
		lkp = XFS_INOBT_KEY_ADDR(left, lrecs + 1, cur);
		lpp = XFS_INOBT_PTR_ADDR(left, lrecs + 1, cur);
		rkp = XFS_INOBT_KEY_ADDR(right, 1, cur);
		rpp = XFS_INOBT_PTR_ADDR(right, 1, cur);
#ifdef DEBUG
		for (i = 0; i < rrecs; i++) {
			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(rpp[i]), level)))
				return error;
		}
#endif
		memcpy(lkp, rkp, rrecs * sizeof(*lkp));
		memcpy(lpp, rpp, rrecs * sizeof(*lpp));
		xfs_inobt_log_keys(cur, lbp, lrecs + 1, lrecs + rrecs);
		xfs_inobt_log_ptrs(cur, lbp, lrecs + 1, lrecs + rrecs);
	} else {
		/*
		 * It's a leaf.  Move records.
		 */
		lrp = XFS_INOBT_REC_ADDR(left, lrecs + 1, cur);
		rrp = XFS_INOBT_REC_ADDR(right, 1, cur);
		memcpy(lrp, rrp, rrecs * sizeof(*lrp));
		xfs_inobt_log_recs(cur, lbp, lrecs + 1, lrecs + rrecs);
	}
	/*
	 * If we joined with the left neighbor, set the buffer in the
	 * cursor to the left block, and fix up the index.
	 */
	if (bp != lbp) {
		xfs_btree_setbuf(cur, level, lbp);
		cur->bc_ptrs[level] += lrecs;
	}
	/*
	 * If we joined with the right neighbor and there's a level above
	 * us, increment the cursor at that level.
	 */
	else if (level + 1 < cur->bc_nlevels &&
		 (error = xfs_btree_increment(cur, level + 1, &i)))
		return error;
	/*
	 * Fix up the number of records in the surviving block.
	 */
	lrecs += rrecs;
	left->bb_numrecs = cpu_to_be16(lrecs);
	/*
	 * Fix up the right block pointer in the surviving block, and log it.
	 */
	left->bb_rightsib = right->bb_rightsib;
	xfs_inobt_log_block(cur->bc_tp, lbp, XFS_BB_NUMRECS | XFS_BB_RIGHTSIB);
	/*
	 * If there is a right sibling now, make it point to the
	 * remaining block.
	 */
	if (be32_to_cpu(left->bb_rightsib) != NULLAGBLOCK) {
		xfs_inobt_block_t	*rrblock;
		xfs_buf_t		*rrbp;

		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
				cur->bc_private.a.agno, be32_to_cpu(left->bb_rightsib), 0,
				&rrbp, XFS_INO_BTREE_REF)))
			return error;
		rrblock = XFS_BUF_TO_INOBT_BLOCK(rrbp);
		if ((error = xfs_btree_check_sblock(cur, rrblock, level, rrbp)))
			return error;
		rrblock->bb_leftsib = cpu_to_be32(lbno);
		xfs_inobt_log_block(cur->bc_tp, rrbp, XFS_BB_LEFTSIB);
	}
	/*
	 * Free the deleting block.
	 */
	if ((error = xfs_free_extent(cur->bc_tp, XFS_AGB_TO_FSB(mp,
				     cur->bc_private.a.agno, rbno), 1)))
		return error;
	xfs_trans_binval(cur->bc_tp, rbp);
	/*
	 * Readjust the ptr at this level if it's not a leaf, since it's
	 * still pointing at the deletion point, which makes the cursor
	 * inconsistent.  If this makes the ptr 0, the caller fixes it up.
	 * We can't use decrement because it would change the next level up.
	 */
	if (level > 0)
		cur->bc_ptrs[level]--;
	/*
	 * Return value means the next level up has something to do.
	 */
	*stat = 2;
	return 0;

error0:
	xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR);
	return error;
}

/*
 * Insert one record/level.  Return information to the caller
 * allowing the next level up to proceed if necessary.
 */
STATIC int				/* error */
xfs_inobt_insrec(
	xfs_btree_cur_t		*cur,	/* btree cursor */
	int			level,	/* level to insert record at */
	xfs_agblock_t		*bnop,	/* i/o: block number inserted */
	xfs_inobt_rec_t		*recp,	/* i/o: record data inserted */
	xfs_btree_cur_t		**curp,	/* output: new cursor replacing cur */
	int			*stat)	/* success/failure */
{
	xfs_inobt_block_t	*block;	/* btree block record/key lives in */
	xfs_buf_t		*bp;	/* buffer for block */
	int			error;	/* error return value */
	int			i;	/* loop index */
	xfs_inobt_key_t		key;	/* key value being inserted */
	xfs_inobt_key_t		*kp=NULL;	/* pointer to btree keys */
	xfs_agblock_t		nbno;	/* block number of allocated block */
	xfs_btree_cur_t		*ncur;	/* new cursor to be used at next lvl */
	xfs_inobt_key_t		nkey;	/* new key value, from split */
	xfs_inobt_rec_t		nrec;	/* new record value, for caller */
	int			numrecs;
	int			optr;	/* old ptr value */
	xfs_inobt_ptr_t		*pp;	/* pointer to btree addresses */
	int			ptr;	/* index in btree block for this rec */
	xfs_inobt_rec_t		*rp=NULL;	/* pointer to btree records */

	/*
	 * GCC doesn't understand the (arguably complex) control flow in
	 * this function and complains about uninitialized structure fields
	 * without this.
	 */
	memset(&nrec, 0, sizeof(nrec));

	/*
	 * If we made it to the root level, allocate a new root block
	 * and we're done.
	 */
	if (level >= cur->bc_nlevels) {
		error = xfs_inobt_newroot(cur, &i);
		*bnop = NULLAGBLOCK;
		*stat = i;
		return error;
	}
	/*
	 * Make a key out of the record data to be inserted, and save it.
	 */
	key.ir_startino = recp->ir_startino;
	optr = ptr = cur->bc_ptrs[level];
	/*
	 * If we're off the left edge, return failure.
	 */
	if (ptr == 0) {
		*stat = 0;
		return 0;
	}
	/*
	 * Get pointers to the btree buffer and block.
	 */
	bp = cur->bc_bufs[level];
	block = XFS_BUF_TO_INOBT_BLOCK(bp);
	numrecs = be16_to_cpu(block->bb_numrecs);
#ifdef DEBUG
	if ((error = xfs_btree_check_sblock(cur, block, level, bp)))
		return error;
	/*
	 * Check that the new entry is being inserted in the right place.
	 */
	if (ptr <= numrecs) {
		if (level == 0) {
			rp = XFS_INOBT_REC_ADDR(block, ptr, cur);
			xfs_btree_check_rec(cur->bc_btnum, recp, rp);
		} else {
			kp = XFS_INOBT_KEY_ADDR(block, ptr, cur);
			xfs_btree_check_key(cur->bc_btnum, &key, kp);
		}
	}
#endif
	nbno = NULLAGBLOCK;
	ncur = NULL;
	/*
	 * If the block is full, we can't insert the new entry until we
	 * make the block un-full.
	 */
	if (numrecs == XFS_INOBT_BLOCK_MAXRECS(level, cur)) {
		/*
		 * First, try shifting an entry to the right neighbor.
		 */
		if ((error = xfs_btree_rshift(cur, level, &i)))
			return error;
		if (i) {
			/* nothing */
		}
		/*
		 * Next, try shifting an entry to the left neighbor.
		 */
		else {
			if ((error = xfs_btree_lshift(cur, level, &i)))
				return error;
			if (i) {
				optr = ptr = cur->bc_ptrs[level];
			} else {
				/*
				 * Next, try splitting the current block
				 * in half. If this works we have to
				 * re-set our variables because
				 * we could be in a different block now.
				 */
				if ((error = xfs_inobt_split(cur, level, &nbno,
						&nkey, &ncur, &i)))
					return error;
				if (i) {
					bp = cur->bc_bufs[level];
					block = XFS_BUF_TO_INOBT_BLOCK(bp);
#ifdef DEBUG
					if ((error = xfs_btree_check_sblock(cur,
							block, level, bp)))
						return error;
#endif
					ptr = cur->bc_ptrs[level];
					nrec.ir_startino = nkey.ir_startino;
				} else {
					/*
					 * Otherwise the insert fails.
					 */
					*stat = 0;
					return 0;
				}
			}
		}
	}
	/*
	 * At this point we know there's room for our new entry in the block
	 * we're pointing at.
	 */
	numrecs = be16_to_cpu(block->bb_numrecs);
	if (level > 0) {
		/*
		 * It's a non-leaf entry.  Make a hole for the new data
		 * in the key and ptr regions of the block.
		 */
		kp = XFS_INOBT_KEY_ADDR(block, 1, cur);
		pp = XFS_INOBT_PTR_ADDR(block, 1, cur);
#ifdef DEBUG
		for (i = numrecs; i >= ptr; i--) {
			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(pp[i - 1]), level)))
				return error;
		}
#endif
		memmove(&kp[ptr], &kp[ptr - 1],
			(numrecs - ptr + 1) * sizeof(*kp));
		memmove(&pp[ptr], &pp[ptr - 1],
			(numrecs - ptr + 1) * sizeof(*pp));
		/*
		 * Now stuff the new data in, bump numrecs and log the new data.
		 */
#ifdef DEBUG
		if ((error = xfs_btree_check_sptr(cur, *bnop, level)))
			return error;
#endif
		kp[ptr - 1] = key;
		pp[ptr - 1] = cpu_to_be32(*bnop);
		numrecs++;
		block->bb_numrecs = cpu_to_be16(numrecs);
		xfs_inobt_log_keys(cur, bp, ptr, numrecs);
		xfs_inobt_log_ptrs(cur, bp, ptr, numrecs);
	} else {
		/*
		 * It's a leaf entry.  Make a hole for the new record.
		 */
		rp = XFS_INOBT_REC_ADDR(block, 1, cur);
		memmove(&rp[ptr], &rp[ptr - 1],
			(numrecs - ptr + 1) * sizeof(*rp));
		/*
		 * Now stuff the new record in, bump numrecs
		 * and log the new data.
		 */
		rp[ptr - 1] = *recp;
		numrecs++;
		block->bb_numrecs = cpu_to_be16(numrecs);
		xfs_inobt_log_recs(cur, bp, ptr, numrecs);
	}
	/*
	 * Log the new number of records in the btree header.
	 */
	xfs_inobt_log_block(cur->bc_tp, bp, XFS_BB_NUMRECS);
#ifdef DEBUG
	/*
	 * Check that the key/record is in the right place, now.
	 */
	if (ptr < numrecs) {
		if (level == 0)
			xfs_btree_check_rec(cur->bc_btnum, rp + ptr - 1,
				rp + ptr);
		else
			xfs_btree_check_key(cur->bc_btnum, kp + ptr - 1,
				kp + ptr);
	}
#endif
	/*
	 * If we inserted at the start of a block, update the parents' keys.
	 */
	if (optr == 1 && (error = xfs_btree_updkey(cur, (union xfs_btree_key *)&key, level + 1)))
		return error;
	/*
	 * Return the new block number, if any.
	 * If there is one, give back a record value and a cursor too.
	 */
	*bnop = nbno;
	if (nbno != NULLAGBLOCK) {
		*recp = nrec;
		*curp = ncur;
	}
	*stat = 1;
	return 0;
}

/*
 * Log header fields from a btree block.
 */
STATIC void
xfs_inobt_log_block(
	xfs_trans_t		*tp,	/* transaction pointer */
	xfs_buf_t		*bp,	/* buffer containing btree block */
	int			fields)	/* mask of fields: XFS_BB_... */
{
	int			first;	/* first byte offset logged */
	int			last;	/* last byte offset logged */
	static const short	offsets[] = {	/* table of offsets */
		offsetof(xfs_inobt_block_t, bb_magic),
		offsetof(xfs_inobt_block_t, bb_level),
		offsetof(xfs_inobt_block_t, bb_numrecs),
		offsetof(xfs_inobt_block_t, bb_leftsib),
		offsetof(xfs_inobt_block_t, bb_rightsib),
		sizeof(xfs_inobt_block_t)
	};

	xfs_btree_offsets(fields, offsets, XFS_BB_NUM_BITS, &first, &last);
	xfs_trans_log_buf(tp, bp, first, last);
}

/*
 * Log keys from a btree block (nonleaf).
 */
STATIC void
xfs_inobt_log_keys(
	xfs_btree_cur_t		*cur,	/* btree cursor */
	xfs_buf_t		*bp,	/* buffer containing btree block */
	int			kfirst,	/* index of first key to log */
	int			klast)	/* index of last key to log */
{
	xfs_inobt_block_t	*block;	/* btree block to log from */
	int			first;	/* first byte offset logged */
	xfs_inobt_key_t		*kp;	/* key pointer in btree block */
	int			last;	/* last byte offset logged */

	block = XFS_BUF_TO_INOBT_BLOCK(bp);
	kp = XFS_INOBT_KEY_ADDR(block, 1, cur);
	first = (int)((xfs_caddr_t)&kp[kfirst - 1] - (xfs_caddr_t)block);
	last = (int)(((xfs_caddr_t)&kp[klast] - 1) - (xfs_caddr_t)block);
	xfs_trans_log_buf(cur->bc_tp, bp, first, last);
}

/*
 * Log block pointer fields from a btree block (nonleaf).
 */
STATIC void
xfs_inobt_log_ptrs(
	xfs_btree_cur_t		*cur,	/* btree cursor */
	xfs_buf_t		*bp,	/* buffer containing btree block */
	int			pfirst,	/* index of first pointer to log */
	int			plast)	/* index of last pointer to log */
{
	xfs_inobt_block_t	*block;	/* btree block to log from */
	int			first;	/* first byte offset logged */
	int			last;	/* last byte offset logged */
	xfs_inobt_ptr_t		*pp;	/* block-pointer pointer in btree blk */

	block = XFS_BUF_TO_INOBT_BLOCK(bp);
	pp = XFS_INOBT_PTR_ADDR(block, 1, cur);
	first = (int)((xfs_caddr_t)&pp[pfirst - 1] - (xfs_caddr_t)block);
	last = (int)(((xfs_caddr_t)&pp[plast] - 1) - (xfs_caddr_t)block);
	xfs_trans_log_buf(cur->bc_tp, bp, first, last);
}

/*
 * Log records from a btree block (leaf).
 */
STATIC void
xfs_inobt_log_recs(
	xfs_btree_cur_t		*cur,	/* btree cursor */
	xfs_buf_t		*bp,	/* buffer containing btree block */
	int			rfirst,	/* index of first record to log */
	int			rlast)	/* index of last record to log */
{
	xfs_inobt_block_t	*block;	/* btree block to log from */
	int			first;	/* first byte offset logged */
	int			last;	/* last byte offset logged */
	xfs_inobt_rec_t		*rp;	/* record pointer for btree block */

	block = XFS_BUF_TO_INOBT_BLOCK(bp);
	rp = XFS_INOBT_REC_ADDR(block, 1, cur);
	first = (int)((xfs_caddr_t)&rp[rfirst - 1] - (xfs_caddr_t)block);
	last = (int)(((xfs_caddr_t)&rp[rlast] - 1) - (xfs_caddr_t)block);
	xfs_trans_log_buf(cur->bc_tp, bp, first, last);
}

/*
 * Allocate a new root block, fill it in.
 */
STATIC int				/* error */
xfs_inobt_newroot(
	xfs_btree_cur_t		*cur,	/* btree cursor */
	int			*stat)	/* success/failure */
{
	xfs_agi_t		*agi;	/* a.g. inode header */
	xfs_alloc_arg_t		args;	/* allocation argument structure */
	xfs_inobt_block_t	*block;	/* one half of the old root block */
	xfs_buf_t		*bp;	/* buffer containing block */
	int			error;	/* error return value */
	xfs_inobt_key_t		*kp;	/* btree key pointer */
	xfs_agblock_t		lbno;	/* left block number */
	xfs_buf_t		*lbp;	/* left buffer pointer */
	xfs_inobt_block_t	*left;	/* left btree block */
	xfs_buf_t		*nbp;	/* new (root) buffer */
	xfs_inobt_block_t	*new;	/* new (root) btree block */
	int			nptr;	/* new value for key index, 1 or 2 */
	xfs_inobt_ptr_t		*pp;	/* btree address pointer */
	xfs_agblock_t		rbno;	/* right block number */
	xfs_buf_t		*rbp;	/* right buffer pointer */
	xfs_inobt_block_t	*right;	/* right btree block */
	xfs_inobt_rec_t		*rp;	/* btree record pointer */

	ASSERT(cur->bc_nlevels < XFS_IN_MAXLEVELS(cur->bc_mp));

	/*
	 * Get a block & a buffer.
	 */
	agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
	args.tp = cur->bc_tp;
	args.mp = cur->bc_mp;
	args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_private.a.agno,
		be32_to_cpu(agi->agi_root));
	args.mod = args.minleft = args.alignment = args.total = args.wasdel =
		args.isfl = args.userdata = args.minalignslop = 0;
	args.minlen = args.maxlen = args.prod = 1;
	args.type = XFS_ALLOCTYPE_NEAR_BNO;
	if ((error = xfs_alloc_vextent(&args)))
		return error;
	/*
	 * None available, we fail.
	 */
	if (args.fsbno == NULLFSBLOCK) {
		*stat = 0;
		return 0;
	}
	ASSERT(args.len == 1);
	nbp = xfs_btree_get_bufs(args.mp, args.tp, args.agno, args.agbno, 0);
	new = XFS_BUF_TO_INOBT_BLOCK(nbp);
	/*
	 * Set the root data in the a.g. inode structure.
	 */
	agi->agi_root = cpu_to_be32(args.agbno);
	be32_add_cpu(&agi->agi_level, 1);
	xfs_ialloc_log_agi(args.tp, cur->bc_private.a.agbp,
		XFS_AGI_ROOT | XFS_AGI_LEVEL);
	/*
	 * At the previous root level there are now two blocks: the old
	 * root, and the new block generated when it was split.
	 * We don't know which one the cursor is pointing at, so we
	 * set up variables "left" and "right" for each case.
	 */
	bp = cur->bc_bufs[cur->bc_nlevels - 1];
	block = XFS_BUF_TO_INOBT_BLOCK(bp);
#ifdef DEBUG
	if ((error = xfs_btree_check_sblock(cur, block, cur->bc_nlevels - 1, bp)))
		return error;
#endif
	if (be32_to_cpu(block->bb_rightsib) != NULLAGBLOCK) {
		/*
		 * Our block is left, pick up the right block.
		 */
		lbp = bp;
		lbno = XFS_DADDR_TO_AGBNO(args.mp, XFS_BUF_ADDR(lbp));
		left = block;
		rbno = be32_to_cpu(left->bb_rightsib);
		if ((error = xfs_btree_read_bufs(args.mp, args.tp, args.agno,
				rbno, 0, &rbp, XFS_INO_BTREE_REF)))
			return error;
		bp = rbp;
		right = XFS_BUF_TO_INOBT_BLOCK(rbp);
		if ((error = xfs_btree_check_sblock(cur, right,
				cur->bc_nlevels - 1, rbp)))
			return error;
		nptr = 1;
	} else {
		/*
		 * Our block is right, pick up the left block.
		 */
		rbp = bp;
		rbno = XFS_DADDR_TO_AGBNO(args.mp, XFS_BUF_ADDR(rbp));
		right = block;
		lbno = be32_to_cpu(right->bb_leftsib);
		if ((error = xfs_btree_read_bufs(args.mp, args.tp, args.agno,
				lbno, 0, &lbp, XFS_INO_BTREE_REF)))
			return error;
		bp = lbp;
		left = XFS_BUF_TO_INOBT_BLOCK(lbp);
		if ((error = xfs_btree_check_sblock(cur, left,
				cur->bc_nlevels - 1, lbp)))
			return error;
		nptr = 2;
	}
	/*
	 * Fill in the new block's btree header and log it.
	 */
	new->bb_magic = cpu_to_be32(xfs_magics[cur->bc_btnum]);
	new->bb_level = cpu_to_be16(cur->bc_nlevels);
	new->bb_numrecs = cpu_to_be16(2);
	new->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
	new->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
	xfs_inobt_log_block(args.tp, nbp, XFS_BB_ALL_BITS);
	ASSERT(lbno != NULLAGBLOCK && rbno != NULLAGBLOCK);
	/*
	 * Fill in the key data in the new root.
	 */
	kp = XFS_INOBT_KEY_ADDR(new, 1, cur);
	if (be16_to_cpu(left->bb_level) > 0) {
		kp[0] = *XFS_INOBT_KEY_ADDR(left, 1, cur);
		kp[1] = *XFS_INOBT_KEY_ADDR(right, 1, cur);
	} else {
		rp = XFS_INOBT_REC_ADDR(left, 1, cur);
		kp[0].ir_startino = rp->ir_startino;
		rp = XFS_INOBT_REC_ADDR(right, 1, cur);
		kp[1].ir_startino = rp->ir_startino;
	}
	xfs_inobt_log_keys(cur, nbp, 1, 2);
	/*
	 * Fill in the pointer data in the new root.
	 */
	pp = XFS_INOBT_PTR_ADDR(new, 1, cur);
	pp[0] = cpu_to_be32(lbno);
	pp[1] = cpu_to_be32(rbno);
	xfs_inobt_log_ptrs(cur, nbp, 1, 2);
	/*
	 * Fix up the cursor.
	 */
	xfs_btree_setbuf(cur, cur->bc_nlevels, nbp);
	cur->bc_ptrs[cur->bc_nlevels] = nptr;
	cur->bc_nlevels++;
	*stat = 1;
	return 0;
}

/*
 * Split cur/level block in half.
 * Return new block number and its first record (to be inserted into parent).
 */
STATIC int				/* error */
xfs_inobt_split(
	xfs_btree_cur_t		*cur,	/* btree cursor */
	int			level,	/* level to split */
	xfs_agblock_t		*bnop,	/* output: block number allocated */
	xfs_inobt_key_t		*keyp,	/* output: first key of new block */
	xfs_btree_cur_t		**curp,	/* output: new cursor */
	int			*stat)	/* success/failure */
{
	xfs_alloc_arg_t		args;	/* allocation argument structure */
	int			error;	/* error return value */
	int			i;	/* loop index/record number */
	xfs_agblock_t		lbno;	/* left (current) block number */
	xfs_buf_t		*lbp;	/* buffer for left block */
	xfs_inobt_block_t	*left;	/* left (current) btree block */
	xfs_inobt_key_t		*lkp;	/* left btree key pointer */
	xfs_inobt_ptr_t		*lpp;	/* left btree address pointer */
	xfs_inobt_rec_t		*lrp;	/* left btree record pointer */
	xfs_buf_t		*rbp;	/* buffer for right block */
	xfs_inobt_block_t	*right;	/* right (new) btree block */
	xfs_inobt_key_t		*rkp;	/* right btree key pointer */
	xfs_inobt_ptr_t		*rpp;	/* right btree address pointer */
	xfs_inobt_rec_t		*rrp;	/* right btree record pointer */

	/*
	 * Set up left block (current one).
	 */
	lbp = cur->bc_bufs[level];
	args.tp = cur->bc_tp;
	args.mp = cur->bc_mp;
	lbno = XFS_DADDR_TO_AGBNO(args.mp, XFS_BUF_ADDR(lbp));
	/*
	 * Allocate the new block.
	 * If we can't do it, we're toast.  Give up.
	 */
	args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_private.a.agno, lbno);
	args.mod = args.minleft = args.alignment = args.total = args.wasdel =
		args.isfl = args.userdata = args.minalignslop = 0;
	args.minlen = args.maxlen = args.prod = 1;
	args.type = XFS_ALLOCTYPE_NEAR_BNO;
	if ((error = xfs_alloc_vextent(&args)))
		return error;
	if (args.fsbno == NULLFSBLOCK) {
		*stat = 0;
		return 0;
	}
	ASSERT(args.len == 1);
	rbp = xfs_btree_get_bufs(args.mp, args.tp, args.agno, args.agbno, 0);
	/*
	 * Set up the new block as "right".
	 */
	right = XFS_BUF_TO_INOBT_BLOCK(rbp);
	/*
	 * "Left" is the current (according to the cursor) block.
	 */
	left = XFS_BUF_TO_INOBT_BLOCK(lbp);
#ifdef DEBUG
	if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
		return error;
#endif
	/*
	 * Fill in the btree header for the new block.
	 */
	right->bb_magic = cpu_to_be32(xfs_magics[cur->bc_btnum]);
	right->bb_level = left->bb_level;
	right->bb_numrecs = cpu_to_be16(be16_to_cpu(left->bb_numrecs) / 2);
	/*
	 * Make sure that if there's an odd number of entries now, that
	 * each new block will have the same number of entries.
	 */
	if ((be16_to_cpu(left->bb_numrecs) & 1) &&
	    cur->bc_ptrs[level] <= be16_to_cpu(right->bb_numrecs) + 1)
		be16_add_cpu(&right->bb_numrecs, 1);
	i = be16_to_cpu(left->bb_numrecs) - be16_to_cpu(right->bb_numrecs) + 1;
	/*
	 * For non-leaf blocks, copy keys and addresses over to the new block.
	 */
	if (level > 0) {
		lkp = XFS_INOBT_KEY_ADDR(left, i, cur);
		lpp = XFS_INOBT_PTR_ADDR(left, i, cur);
		rkp = XFS_INOBT_KEY_ADDR(right, 1, cur);
		rpp = XFS_INOBT_PTR_ADDR(right, 1, cur);
#ifdef DEBUG
		for (i = 0; i < be16_to_cpu(right->bb_numrecs); i++) {
			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(lpp[i]), level)))
				return error;
		}
#endif
		memcpy(rkp, lkp, be16_to_cpu(right->bb_numrecs) * sizeof(*rkp));
		memcpy(rpp, lpp, be16_to_cpu(right->bb_numrecs) * sizeof(*rpp));
		xfs_inobt_log_keys(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
		xfs_inobt_log_ptrs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
		*keyp = *rkp;
	}
	/*
	 * For leaf blocks, copy records over to the new block.
	 */
	else {
		lrp = XFS_INOBT_REC_ADDR(left, i, cur);
		rrp = XFS_INOBT_REC_ADDR(right, 1, cur);
		memcpy(rrp, lrp, be16_to_cpu(right->bb_numrecs) * sizeof(*rrp));
		xfs_inobt_log_recs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
		keyp->ir_startino = rrp->ir_startino;
	}
	/*
	 * Find the left block number by looking in the buffer.
	 * Adjust numrecs, sibling pointers.
	 */
	be16_add_cpu(&left->bb_numrecs, -(be16_to_cpu(right->bb_numrecs)));
	right->bb_rightsib = left->bb_rightsib;
	left->bb_rightsib = cpu_to_be32(args.agbno);
	right->bb_leftsib = cpu_to_be32(lbno);
	xfs_inobt_log_block(args.tp, rbp, XFS_BB_ALL_BITS);
	xfs_inobt_log_block(args.tp, lbp, XFS_BB_NUMRECS | XFS_BB_RIGHTSIB);
	/*
	 * If there's a block to the new block's right, make that block
	 * point back to right instead of to left.
	 */
	if (be32_to_cpu(right->bb_rightsib) != NULLAGBLOCK) {
		xfs_inobt_block_t	*rrblock;	/* rr btree block */
		xfs_buf_t		*rrbp;		/* buffer for rrblock */

		if ((error = xfs_btree_read_bufs(args.mp, args.tp, args.agno,
				be32_to_cpu(right->bb_rightsib), 0, &rrbp,
				XFS_INO_BTREE_REF)))
			return error;
		rrblock = XFS_BUF_TO_INOBT_BLOCK(rrbp);
		if ((error = xfs_btree_check_sblock(cur, rrblock, level, rrbp)))
			return error;
		rrblock->bb_leftsib = cpu_to_be32(args.agbno);
		xfs_inobt_log_block(args.tp, rrbp, XFS_BB_LEFTSIB);
	}
	/*
	 * If the cursor is really in the right block, move it there.
	 * If it's just pointing past the last entry in left, then we'll
	 * insert there, so don't change anything in that case.
	 */
	if (cur->bc_ptrs[level] > be16_to_cpu(left->bb_numrecs) + 1) {
		xfs_btree_setbuf(cur, level, rbp);
		cur->bc_ptrs[level] -= be16_to_cpu(left->bb_numrecs);
	}
	/*
	 * If there are more levels, we'll need another cursor which refers
	 * the right block, no matter where this cursor was.
	 */
	if (level + 1 < cur->bc_nlevels) {
		if ((error = xfs_btree_dup_cursor(cur, curp)))
			return error;
		(*curp)->bc_ptrs[level + 1]++;
	}
	*bnop = args.agbno;
	*stat = 1;
	return 0;
}

/*
 * Externally visible routines.
 */

/*
 * Delete the record pointed to by cur.
 * The cursor refers to the place where the record was (could be inserted)
 * when the operation returns.
 */
int					/* error */
xfs_inobt_delete(
	xfs_btree_cur_t	*cur,		/* btree cursor */
	int		*stat)		/* success/failure */
{
	int		error;
	int		i;		/* result code */
	int		level;		/* btree level */

	/*
	 * Go up the tree, starting at leaf level.
	 * If 2 is returned then a join was done; go to the next level.
	 * Otherwise we are done.
	 */
	for (level = 0, i = 2; i == 2; level++) {
		if ((error = xfs_inobt_delrec(cur, level, &i)))
			return error;
	}
	if (i == 0) {
		for (level = 1; level < cur->bc_nlevels; level++) {
			if (cur->bc_ptrs[level] == 0) {
				if ((error = xfs_btree_decrement(cur, level, &i)))
					return error;
				break;
			}
		}
	}
	*stat = i;
	return 0;
}


/*
 * Get the data from the pointed-to record.
 */
int					/* error */
xfs_inobt_get_rec(
	xfs_btree_cur_t		*cur,	/* btree cursor */
	xfs_agino_t		*ino,	/* output: starting inode of chunk */
	__int32_t		*fcnt,	/* output: number of free inodes */
	xfs_inofree_t		*free,	/* output: free inode mask */
	int			*stat)	/* output: success/failure */
{
	xfs_inobt_block_t	*block;	/* btree block */
	xfs_buf_t		*bp;	/* buffer containing btree block */
#ifdef DEBUG
	int			error;	/* error return value */
#endif
	int			ptr;	/* record number */
	xfs_inobt_rec_t		*rec;	/* record data */

	bp = cur->bc_bufs[0];
	ptr = cur->bc_ptrs[0];
	block = XFS_BUF_TO_INOBT_BLOCK(bp);
#ifdef DEBUG
	if ((error = xfs_btree_check_sblock(cur, block, 0, bp)))
		return error;
#endif
	/*
	 * Off the right end or left end, return failure.
	 */
	if (ptr > be16_to_cpu(block->bb_numrecs) || ptr <= 0) {
		*stat = 0;
		return 0;
	}
	/*
	 * Point to the record and extract its data.
	 */
	rec = XFS_INOBT_REC_ADDR(block, ptr, cur);
	*ino = be32_to_cpu(rec->ir_startino);
	*fcnt = be32_to_cpu(rec->ir_freecount);
	*free = be64_to_cpu(rec->ir_free);
	*stat = 1;
	return 0;
}

/*
 * Insert the current record at the point referenced by cur.
 * The cursor may be inconsistent on return if splits have been done.
 */
int					/* error */
xfs_inobt_insert(
	xfs_btree_cur_t	*cur,		/* btree cursor */
	int		*stat)		/* success/failure */
{
	int		error;		/* error return value */
	int		i;		/* result value, 0 for failure */
	int		level;		/* current level number in btree */
	xfs_agblock_t	nbno;		/* new block number (split result) */
	xfs_btree_cur_t	*ncur;		/* new cursor (split result) */
	xfs_inobt_rec_t	nrec;		/* record being inserted this level */
	xfs_btree_cur_t	*pcur;		/* previous level's cursor */

	level = 0;
	nbno = NULLAGBLOCK;
	nrec.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
	nrec.ir_freecount = cpu_to_be32(cur->bc_rec.i.ir_freecount);
	nrec.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
	ncur = NULL;
	pcur = cur;
	/*
	 * Loop going up the tree, starting at the leaf level.
	 * Stop when we don't get a split block, that must mean that
	 * the insert is finished with this level.
	 */
	do {
		/*
		 * Insert nrec/nbno into this level of the tree.
		 * Note if we fail, nbno will be null.
		 */
		if ((error = xfs_inobt_insrec(pcur, level++, &nbno, &nrec, &ncur,
				&i))) {
			if (pcur != cur)
				xfs_btree_del_cursor(pcur, XFS_BTREE_ERROR);
			return error;
		}
		/*
		 * See if the cursor we just used is trash.
		 * Can't trash the caller's cursor, but otherwise we should
		 * if ncur is a new cursor or we're about to be done.
		 */
		if (pcur != cur && (ncur || nbno == NULLAGBLOCK)) {
			cur->bc_nlevels = pcur->bc_nlevels;
			xfs_btree_del_cursor(pcur, XFS_BTREE_NOERROR);
		}
		/*
		 * If we got a new cursor, switch to it.
		 */
		if (ncur) {
			pcur = ncur;
			ncur = NULL;
		}
	} while (nbno != NULLAGBLOCK);
	*stat = i;
	return 0;
}

STATIC struct xfs_btree_cur *
xfs_inobt_dup_cursor(
	struct xfs_btree_cur	*cur)
{
	return xfs_inobt_init_cursor(cur->bc_mp, cur->bc_tp,
			cur->bc_private.a.agbp, cur->bc_private.a.agno);
}

STATIC int
xfs_inobt_get_maxrecs(
	struct xfs_btree_cur	*cur,
	int			level)
{
	return cur->bc_mp->m_inobt_mxr[level != 0];
}

STATIC void
xfs_inobt_init_key_from_rec(
	union xfs_btree_key	*key,
	union xfs_btree_rec	*rec)
{
	key->inobt.ir_startino = rec->inobt.ir_startino;
}

/*
 * intial value of ptr for lookup
 */
STATIC void
xfs_inobt_init_ptr_from_cur(
	struct xfs_btree_cur	*cur,
	union xfs_btree_ptr	*ptr)
{
	struct xfs_agi		*agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);

	ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));

	ptr->s = agi->agi_root;
}

STATIC __int64_t
xfs_inobt_key_diff(
	struct xfs_btree_cur	*cur,
	union xfs_btree_key	*key)
{
	return (__int64_t)be32_to_cpu(key->inobt.ir_startino) -
			  cur->bc_rec.i.ir_startino;
}

#ifdef XFS_BTREE_TRACE
ktrace_t	*xfs_inobt_trace_buf;

STATIC void
xfs_inobt_trace_enter(
	struct xfs_btree_cur	*cur,
	const char		*func,
	char			*s,
	int			type,
	int			line,
	__psunsigned_t		a0,
	__psunsigned_t		a1,
	__psunsigned_t		a2,
	__psunsigned_t		a3,
	__psunsigned_t		a4,
	__psunsigned_t		a5,
	__psunsigned_t		a6,
	__psunsigned_t		a7,
	__psunsigned_t		a8,
	__psunsigned_t		a9,
	__psunsigned_t		a10)
{
	ktrace_enter(xfs_inobt_trace_buf, (void *)(__psint_t)type,
		(void *)func, (void *)s, NULL, (void *)cur,
		(void *)a0, (void *)a1, (void *)a2, (void *)a3,
		(void *)a4, (void *)a5, (void *)a6, (void *)a7,
		(void *)a8, (void *)a9, (void *)a10);
}

STATIC void
xfs_inobt_trace_cursor(
	struct xfs_btree_cur	*cur,
	__uint32_t		*s0,
	__uint64_t		*l0,
	__uint64_t		*l1)
{
	*s0 = cur->bc_private.a.agno;
	*l0 = cur->bc_rec.i.ir_startino;
	*l1 = cur->bc_rec.i.ir_free;
}

STATIC void
xfs_inobt_trace_key(
	struct xfs_btree_cur	*cur,
	union xfs_btree_key	*key,
	__uint64_t		*l0,
	__uint64_t		*l1)
{
	*l0 = be32_to_cpu(key->inobt.ir_startino);
	*l1 = 0;
}

STATIC void
xfs_inobt_trace_record(
	struct xfs_btree_cur	*cur,
	union xfs_btree_rec	*rec,
	__uint64_t		*l0,
	__uint64_t		*l1,
	__uint64_t		*l2)
{
	*l0 = be32_to_cpu(rec->inobt.ir_startino);
	*l1 = be32_to_cpu(rec->inobt.ir_freecount);
	*l2 = be64_to_cpu(rec->inobt.ir_free);
}
#endif /* XFS_BTREE_TRACE */

static const struct xfs_btree_ops xfs_inobt_ops = {
	.rec_len		= sizeof(xfs_inobt_rec_t),
	.key_len		= sizeof(xfs_inobt_key_t),

	.dup_cursor		= xfs_inobt_dup_cursor,
	.get_maxrecs		= xfs_inobt_get_maxrecs,
	.init_key_from_rec	= xfs_inobt_init_key_from_rec,
	.init_ptr_from_cur	= xfs_inobt_init_ptr_from_cur,
	.key_diff		= xfs_inobt_key_diff,

#ifdef XFS_BTREE_TRACE
	.trace_enter		= xfs_inobt_trace_enter,
	.trace_cursor		= xfs_inobt_trace_cursor,
	.trace_key		= xfs_inobt_trace_key,
	.trace_record		= xfs_inobt_trace_record,
#endif
};

/*
 * Allocate a new inode btree cursor.
 */
struct xfs_btree_cur *				/* new inode btree cursor */
xfs_inobt_init_cursor(
	struct xfs_mount	*mp,		/* file system mount point */
	struct xfs_trans	*tp,		/* transaction pointer */
	struct xfs_buf		*agbp,		/* buffer for agi structure */
	xfs_agnumber_t		agno)		/* allocation group number */
{
	struct xfs_agi		*agi = XFS_BUF_TO_AGI(agbp);
	struct xfs_btree_cur	*cur;

	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);

	cur->bc_tp = tp;
	cur->bc_mp = mp;
	cur->bc_nlevels = be32_to_cpu(agi->agi_level);
	cur->bc_btnum = XFS_BTNUM_INO;
	cur->bc_blocklog = mp->m_sb.sb_blocklog;

	cur->bc_ops = &xfs_inobt_ops;

	cur->bc_private.a.agbp = agbp;
	cur->bc_private.a.agno = agno;

	return cur;
}