1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc. 4 * All Rights Reserved. 5 */ 6 #include "xfs.h" 7 #include "xfs_fs.h" 8 #include "xfs_shared.h" 9 #include "xfs_format.h" 10 #include "xfs_log_format.h" 11 #include "xfs_trans_resv.h" 12 #include "xfs_sb.h" 13 #include "xfs_mount.h" 14 #include "xfs_btree.h" 15 #include "xfs_btree_staging.h" 16 #include "xfs_alloc_btree.h" 17 #include "xfs_alloc.h" 18 #include "xfs_extent_busy.h" 19 #include "xfs_error.h" 20 #include "xfs_trace.h" 21 #include "xfs_trans.h" 22 23 24 STATIC struct xfs_btree_cur * 25 xfs_allocbt_dup_cursor( 26 struct xfs_btree_cur *cur) 27 { 28 return xfs_allocbt_init_cursor(cur->bc_mp, cur->bc_tp, 29 cur->bc_ag.agbp, cur->bc_ag.agno, 30 cur->bc_btnum); 31 } 32 33 STATIC void 34 xfs_allocbt_set_root( 35 struct xfs_btree_cur *cur, 36 union xfs_btree_ptr *ptr, 37 int inc) 38 { 39 struct xfs_buf *agbp = cur->bc_ag.agbp; 40 struct xfs_agf *agf = agbp->b_addr; 41 int btnum = cur->bc_btnum; 42 struct xfs_perag *pag = agbp->b_pag; 43 44 ASSERT(ptr->s != 0); 45 46 agf->agf_roots[btnum] = ptr->s; 47 be32_add_cpu(&agf->agf_levels[btnum], inc); 48 pag->pagf_levels[btnum] += inc; 49 50 xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS); 51 } 52 53 STATIC int 54 xfs_allocbt_alloc_block( 55 struct xfs_btree_cur *cur, 56 union xfs_btree_ptr *start, 57 union xfs_btree_ptr *new, 58 int *stat) 59 { 60 int error; 61 xfs_agblock_t bno; 62 63 /* Allocate the new block from the freelist. If we can't, give up. */ 64 error = xfs_alloc_get_freelist(cur->bc_tp, cur->bc_ag.agbp, 65 &bno, 1); 66 if (error) 67 return error; 68 69 if (bno == NULLAGBLOCK) { 70 *stat = 0; 71 return 0; 72 } 73 74 atomic64_inc(&cur->bc_mp->m_allocbt_blks); 75 xfs_extent_busy_reuse(cur->bc_mp, cur->bc_ag.agno, bno, 1, false); 76 77 new->s = cpu_to_be32(bno); 78 79 *stat = 1; 80 return 0; 81 } 82 83 STATIC int 84 xfs_allocbt_free_block( 85 struct xfs_btree_cur *cur, 86 struct xfs_buf *bp) 87 { 88 struct xfs_buf *agbp = cur->bc_ag.agbp; 89 struct xfs_agf *agf = agbp->b_addr; 90 xfs_agblock_t bno; 91 int error; 92 93 bno = xfs_daddr_to_agbno(cur->bc_mp, XFS_BUF_ADDR(bp)); 94 error = xfs_alloc_put_freelist(cur->bc_tp, agbp, NULL, bno, 1); 95 if (error) 96 return error; 97 98 atomic64_dec(&cur->bc_mp->m_allocbt_blks); 99 xfs_extent_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1, 100 XFS_EXTENT_BUSY_SKIP_DISCARD); 101 return 0; 102 } 103 104 /* 105 * Update the longest extent in the AGF 106 */ 107 STATIC void 108 xfs_allocbt_update_lastrec( 109 struct xfs_btree_cur *cur, 110 struct xfs_btree_block *block, 111 union xfs_btree_rec *rec, 112 int ptr, 113 int reason) 114 { 115 struct xfs_agf *agf = cur->bc_ag.agbp->b_addr; 116 struct xfs_perag *pag; 117 __be32 len; 118 int numrecs; 119 120 ASSERT(cur->bc_btnum == XFS_BTNUM_CNT); 121 122 switch (reason) { 123 case LASTREC_UPDATE: 124 /* 125 * If this is the last leaf block and it's the last record, 126 * then update the size of the longest extent in the AG. 127 */ 128 if (ptr != xfs_btree_get_numrecs(block)) 129 return; 130 len = rec->alloc.ar_blockcount; 131 break; 132 case LASTREC_INSREC: 133 if (be32_to_cpu(rec->alloc.ar_blockcount) <= 134 be32_to_cpu(agf->agf_longest)) 135 return; 136 len = rec->alloc.ar_blockcount; 137 break; 138 case LASTREC_DELREC: 139 numrecs = xfs_btree_get_numrecs(block); 140 if (ptr <= numrecs) 141 return; 142 ASSERT(ptr == numrecs + 1); 143 144 if (numrecs) { 145 xfs_alloc_rec_t *rrp; 146 147 rrp = XFS_ALLOC_REC_ADDR(cur->bc_mp, block, numrecs); 148 len = rrp->ar_blockcount; 149 } else { 150 len = 0; 151 } 152 153 break; 154 default: 155 ASSERT(0); 156 return; 157 } 158 159 agf->agf_longest = len; 160 pag = cur->bc_ag.agbp->b_pag; 161 pag->pagf_longest = be32_to_cpu(len); 162 xfs_alloc_log_agf(cur->bc_tp, cur->bc_ag.agbp, XFS_AGF_LONGEST); 163 } 164 165 STATIC int 166 xfs_allocbt_get_minrecs( 167 struct xfs_btree_cur *cur, 168 int level) 169 { 170 return cur->bc_mp->m_alloc_mnr[level != 0]; 171 } 172 173 STATIC int 174 xfs_allocbt_get_maxrecs( 175 struct xfs_btree_cur *cur, 176 int level) 177 { 178 return cur->bc_mp->m_alloc_mxr[level != 0]; 179 } 180 181 STATIC void 182 xfs_allocbt_init_key_from_rec( 183 union xfs_btree_key *key, 184 union xfs_btree_rec *rec) 185 { 186 key->alloc.ar_startblock = rec->alloc.ar_startblock; 187 key->alloc.ar_blockcount = rec->alloc.ar_blockcount; 188 } 189 190 STATIC void 191 xfs_bnobt_init_high_key_from_rec( 192 union xfs_btree_key *key, 193 union xfs_btree_rec *rec) 194 { 195 __u32 x; 196 197 x = be32_to_cpu(rec->alloc.ar_startblock); 198 x += be32_to_cpu(rec->alloc.ar_blockcount) - 1; 199 key->alloc.ar_startblock = cpu_to_be32(x); 200 key->alloc.ar_blockcount = 0; 201 } 202 203 STATIC void 204 xfs_cntbt_init_high_key_from_rec( 205 union xfs_btree_key *key, 206 union xfs_btree_rec *rec) 207 { 208 key->alloc.ar_blockcount = rec->alloc.ar_blockcount; 209 key->alloc.ar_startblock = 0; 210 } 211 212 STATIC void 213 xfs_allocbt_init_rec_from_cur( 214 struct xfs_btree_cur *cur, 215 union xfs_btree_rec *rec) 216 { 217 rec->alloc.ar_startblock = cpu_to_be32(cur->bc_rec.a.ar_startblock); 218 rec->alloc.ar_blockcount = cpu_to_be32(cur->bc_rec.a.ar_blockcount); 219 } 220 221 STATIC void 222 xfs_allocbt_init_ptr_from_cur( 223 struct xfs_btree_cur *cur, 224 union xfs_btree_ptr *ptr) 225 { 226 struct xfs_agf *agf = cur->bc_ag.agbp->b_addr; 227 228 ASSERT(cur->bc_ag.agno == be32_to_cpu(agf->agf_seqno)); 229 230 ptr->s = agf->agf_roots[cur->bc_btnum]; 231 } 232 233 STATIC int64_t 234 xfs_bnobt_key_diff( 235 struct xfs_btree_cur *cur, 236 union xfs_btree_key *key) 237 { 238 xfs_alloc_rec_incore_t *rec = &cur->bc_rec.a; 239 xfs_alloc_key_t *kp = &key->alloc; 240 241 return (int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock; 242 } 243 244 STATIC int64_t 245 xfs_cntbt_key_diff( 246 struct xfs_btree_cur *cur, 247 union xfs_btree_key *key) 248 { 249 xfs_alloc_rec_incore_t *rec = &cur->bc_rec.a; 250 xfs_alloc_key_t *kp = &key->alloc; 251 int64_t diff; 252 253 diff = (int64_t)be32_to_cpu(kp->ar_blockcount) - rec->ar_blockcount; 254 if (diff) 255 return diff; 256 257 return (int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock; 258 } 259 260 STATIC int64_t 261 xfs_bnobt_diff_two_keys( 262 struct xfs_btree_cur *cur, 263 union xfs_btree_key *k1, 264 union xfs_btree_key *k2) 265 { 266 return (int64_t)be32_to_cpu(k1->alloc.ar_startblock) - 267 be32_to_cpu(k2->alloc.ar_startblock); 268 } 269 270 STATIC int64_t 271 xfs_cntbt_diff_two_keys( 272 struct xfs_btree_cur *cur, 273 union xfs_btree_key *k1, 274 union xfs_btree_key *k2) 275 { 276 int64_t diff; 277 278 diff = be32_to_cpu(k1->alloc.ar_blockcount) - 279 be32_to_cpu(k2->alloc.ar_blockcount); 280 if (diff) 281 return diff; 282 283 return be32_to_cpu(k1->alloc.ar_startblock) - 284 be32_to_cpu(k2->alloc.ar_startblock); 285 } 286 287 static xfs_failaddr_t 288 xfs_allocbt_verify( 289 struct xfs_buf *bp) 290 { 291 struct xfs_mount *mp = bp->b_mount; 292 struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp); 293 struct xfs_perag *pag = bp->b_pag; 294 xfs_failaddr_t fa; 295 unsigned int level; 296 xfs_btnum_t btnum = XFS_BTNUM_BNOi; 297 298 if (!xfs_verify_magic(bp, block->bb_magic)) 299 return __this_address; 300 301 if (xfs_sb_version_hascrc(&mp->m_sb)) { 302 fa = xfs_btree_sblock_v5hdr_verify(bp); 303 if (fa) 304 return fa; 305 } 306 307 /* 308 * The perag may not be attached during grow operations or fully 309 * initialized from the AGF during log recovery. Therefore we can only 310 * check against maximum tree depth from those contexts. 311 * 312 * Otherwise check against the per-tree limit. Peek at one of the 313 * verifier magic values to determine the type of tree we're verifying 314 * against. 315 */ 316 level = be16_to_cpu(block->bb_level); 317 if (bp->b_ops->magic[0] == cpu_to_be32(XFS_ABTC_MAGIC)) 318 btnum = XFS_BTNUM_CNTi; 319 if (pag && pag->pagf_init) { 320 if (level >= pag->pagf_levels[btnum]) 321 return __this_address; 322 } else if (level >= mp->m_ag_maxlevels) 323 return __this_address; 324 325 return xfs_btree_sblock_verify(bp, mp->m_alloc_mxr[level != 0]); 326 } 327 328 static void 329 xfs_allocbt_read_verify( 330 struct xfs_buf *bp) 331 { 332 xfs_failaddr_t fa; 333 334 if (!xfs_btree_sblock_verify_crc(bp)) 335 xfs_verifier_error(bp, -EFSBADCRC, __this_address); 336 else { 337 fa = xfs_allocbt_verify(bp); 338 if (fa) 339 xfs_verifier_error(bp, -EFSCORRUPTED, fa); 340 } 341 342 if (bp->b_error) 343 trace_xfs_btree_corrupt(bp, _RET_IP_); 344 } 345 346 static void 347 xfs_allocbt_write_verify( 348 struct xfs_buf *bp) 349 { 350 xfs_failaddr_t fa; 351 352 fa = xfs_allocbt_verify(bp); 353 if (fa) { 354 trace_xfs_btree_corrupt(bp, _RET_IP_); 355 xfs_verifier_error(bp, -EFSCORRUPTED, fa); 356 return; 357 } 358 xfs_btree_sblock_calc_crc(bp); 359 360 } 361 362 const struct xfs_buf_ops xfs_bnobt_buf_ops = { 363 .name = "xfs_bnobt", 364 .magic = { cpu_to_be32(XFS_ABTB_MAGIC), 365 cpu_to_be32(XFS_ABTB_CRC_MAGIC) }, 366 .verify_read = xfs_allocbt_read_verify, 367 .verify_write = xfs_allocbt_write_verify, 368 .verify_struct = xfs_allocbt_verify, 369 }; 370 371 const struct xfs_buf_ops xfs_cntbt_buf_ops = { 372 .name = "xfs_cntbt", 373 .magic = { cpu_to_be32(XFS_ABTC_MAGIC), 374 cpu_to_be32(XFS_ABTC_CRC_MAGIC) }, 375 .verify_read = xfs_allocbt_read_verify, 376 .verify_write = xfs_allocbt_write_verify, 377 .verify_struct = xfs_allocbt_verify, 378 }; 379 380 STATIC int 381 xfs_bnobt_keys_inorder( 382 struct xfs_btree_cur *cur, 383 union xfs_btree_key *k1, 384 union xfs_btree_key *k2) 385 { 386 return be32_to_cpu(k1->alloc.ar_startblock) < 387 be32_to_cpu(k2->alloc.ar_startblock); 388 } 389 390 STATIC int 391 xfs_bnobt_recs_inorder( 392 struct xfs_btree_cur *cur, 393 union xfs_btree_rec *r1, 394 union xfs_btree_rec *r2) 395 { 396 return be32_to_cpu(r1->alloc.ar_startblock) + 397 be32_to_cpu(r1->alloc.ar_blockcount) <= 398 be32_to_cpu(r2->alloc.ar_startblock); 399 } 400 401 STATIC int 402 xfs_cntbt_keys_inorder( 403 struct xfs_btree_cur *cur, 404 union xfs_btree_key *k1, 405 union xfs_btree_key *k2) 406 { 407 return be32_to_cpu(k1->alloc.ar_blockcount) < 408 be32_to_cpu(k2->alloc.ar_blockcount) || 409 (k1->alloc.ar_blockcount == k2->alloc.ar_blockcount && 410 be32_to_cpu(k1->alloc.ar_startblock) < 411 be32_to_cpu(k2->alloc.ar_startblock)); 412 } 413 414 STATIC int 415 xfs_cntbt_recs_inorder( 416 struct xfs_btree_cur *cur, 417 union xfs_btree_rec *r1, 418 union xfs_btree_rec *r2) 419 { 420 return be32_to_cpu(r1->alloc.ar_blockcount) < 421 be32_to_cpu(r2->alloc.ar_blockcount) || 422 (r1->alloc.ar_blockcount == r2->alloc.ar_blockcount && 423 be32_to_cpu(r1->alloc.ar_startblock) < 424 be32_to_cpu(r2->alloc.ar_startblock)); 425 } 426 427 static const struct xfs_btree_ops xfs_bnobt_ops = { 428 .rec_len = sizeof(xfs_alloc_rec_t), 429 .key_len = sizeof(xfs_alloc_key_t), 430 431 .dup_cursor = xfs_allocbt_dup_cursor, 432 .set_root = xfs_allocbt_set_root, 433 .alloc_block = xfs_allocbt_alloc_block, 434 .free_block = xfs_allocbt_free_block, 435 .update_lastrec = xfs_allocbt_update_lastrec, 436 .get_minrecs = xfs_allocbt_get_minrecs, 437 .get_maxrecs = xfs_allocbt_get_maxrecs, 438 .init_key_from_rec = xfs_allocbt_init_key_from_rec, 439 .init_high_key_from_rec = xfs_bnobt_init_high_key_from_rec, 440 .init_rec_from_cur = xfs_allocbt_init_rec_from_cur, 441 .init_ptr_from_cur = xfs_allocbt_init_ptr_from_cur, 442 .key_diff = xfs_bnobt_key_diff, 443 .buf_ops = &xfs_bnobt_buf_ops, 444 .diff_two_keys = xfs_bnobt_diff_two_keys, 445 .keys_inorder = xfs_bnobt_keys_inorder, 446 .recs_inorder = xfs_bnobt_recs_inorder, 447 }; 448 449 static const struct xfs_btree_ops xfs_cntbt_ops = { 450 .rec_len = sizeof(xfs_alloc_rec_t), 451 .key_len = sizeof(xfs_alloc_key_t), 452 453 .dup_cursor = xfs_allocbt_dup_cursor, 454 .set_root = xfs_allocbt_set_root, 455 .alloc_block = xfs_allocbt_alloc_block, 456 .free_block = xfs_allocbt_free_block, 457 .update_lastrec = xfs_allocbt_update_lastrec, 458 .get_minrecs = xfs_allocbt_get_minrecs, 459 .get_maxrecs = xfs_allocbt_get_maxrecs, 460 .init_key_from_rec = xfs_allocbt_init_key_from_rec, 461 .init_high_key_from_rec = xfs_cntbt_init_high_key_from_rec, 462 .init_rec_from_cur = xfs_allocbt_init_rec_from_cur, 463 .init_ptr_from_cur = xfs_allocbt_init_ptr_from_cur, 464 .key_diff = xfs_cntbt_key_diff, 465 .buf_ops = &xfs_cntbt_buf_ops, 466 .diff_two_keys = xfs_cntbt_diff_two_keys, 467 .keys_inorder = xfs_cntbt_keys_inorder, 468 .recs_inorder = xfs_cntbt_recs_inorder, 469 }; 470 471 /* Allocate most of a new allocation btree cursor. */ 472 STATIC struct xfs_btree_cur * 473 xfs_allocbt_init_common( 474 struct xfs_mount *mp, 475 struct xfs_trans *tp, 476 xfs_agnumber_t agno, 477 xfs_btnum_t btnum) 478 { 479 struct xfs_btree_cur *cur; 480 481 ASSERT(btnum == XFS_BTNUM_BNO || btnum == XFS_BTNUM_CNT); 482 483 cur = kmem_cache_zalloc(xfs_btree_cur_zone, GFP_NOFS | __GFP_NOFAIL); 484 485 cur->bc_tp = tp; 486 cur->bc_mp = mp; 487 cur->bc_btnum = btnum; 488 cur->bc_blocklog = mp->m_sb.sb_blocklog; 489 490 if (btnum == XFS_BTNUM_CNT) { 491 cur->bc_ops = &xfs_cntbt_ops; 492 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtc_2); 493 cur->bc_flags = XFS_BTREE_LASTREC_UPDATE; 494 } else { 495 cur->bc_ops = &xfs_bnobt_ops; 496 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtb_2); 497 } 498 499 cur->bc_ag.agno = agno; 500 cur->bc_ag.abt.active = false; 501 502 if (xfs_sb_version_hascrc(&mp->m_sb)) 503 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS; 504 505 return cur; 506 } 507 508 /* 509 * Allocate a new allocation btree cursor. 510 */ 511 struct xfs_btree_cur * /* new alloc btree cursor */ 512 xfs_allocbt_init_cursor( 513 struct xfs_mount *mp, /* file system mount point */ 514 struct xfs_trans *tp, /* transaction pointer */ 515 struct xfs_buf *agbp, /* buffer for agf structure */ 516 xfs_agnumber_t agno, /* allocation group number */ 517 xfs_btnum_t btnum) /* btree identifier */ 518 { 519 struct xfs_agf *agf = agbp->b_addr; 520 struct xfs_btree_cur *cur; 521 522 cur = xfs_allocbt_init_common(mp, tp, agno, btnum); 523 if (btnum == XFS_BTNUM_CNT) 524 cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]); 525 else 526 cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]); 527 528 cur->bc_ag.agbp = agbp; 529 530 return cur; 531 } 532 533 /* Create a free space btree cursor with a fake root for staging. */ 534 struct xfs_btree_cur * 535 xfs_allocbt_stage_cursor( 536 struct xfs_mount *mp, 537 struct xbtree_afakeroot *afake, 538 xfs_agnumber_t agno, 539 xfs_btnum_t btnum) 540 { 541 struct xfs_btree_cur *cur; 542 543 cur = xfs_allocbt_init_common(mp, NULL, agno, btnum); 544 xfs_btree_stage_afakeroot(cur, afake); 545 return cur; 546 } 547 548 /* 549 * Install a new free space btree root. Caller is responsible for invalidating 550 * and freeing the old btree blocks. 551 */ 552 void 553 xfs_allocbt_commit_staged_btree( 554 struct xfs_btree_cur *cur, 555 struct xfs_trans *tp, 556 struct xfs_buf *agbp) 557 { 558 struct xfs_agf *agf = agbp->b_addr; 559 struct xbtree_afakeroot *afake = cur->bc_ag.afake; 560 561 ASSERT(cur->bc_flags & XFS_BTREE_STAGING); 562 563 agf->agf_roots[cur->bc_btnum] = cpu_to_be32(afake->af_root); 564 agf->agf_levels[cur->bc_btnum] = cpu_to_be32(afake->af_levels); 565 xfs_alloc_log_agf(tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS); 566 567 if (cur->bc_btnum == XFS_BTNUM_BNO) { 568 xfs_btree_commit_afakeroot(cur, tp, agbp, &xfs_bnobt_ops); 569 } else { 570 cur->bc_flags |= XFS_BTREE_LASTREC_UPDATE; 571 xfs_btree_commit_afakeroot(cur, tp, agbp, &xfs_cntbt_ops); 572 } 573 } 574 575 /* 576 * Calculate number of records in an alloc btree block. 577 */ 578 int 579 xfs_allocbt_maxrecs( 580 struct xfs_mount *mp, 581 int blocklen, 582 int leaf) 583 { 584 blocklen -= XFS_ALLOC_BLOCK_LEN(mp); 585 586 if (leaf) 587 return blocklen / sizeof(xfs_alloc_rec_t); 588 return blocklen / (sizeof(xfs_alloc_key_t) + sizeof(xfs_alloc_ptr_t)); 589 } 590 591 /* Calculate the freespace btree size for some records. */ 592 xfs_extlen_t 593 xfs_allocbt_calc_size( 594 struct xfs_mount *mp, 595 unsigned long long len) 596 { 597 return xfs_btree_calc_size(mp->m_alloc_mnr, len); 598 } 599