1 // SPDX-License-Identifier: GPL-2.0+ 2 /* 3 * Copyright (C) 2016 Oracle. All Rights Reserved. 4 * Author: Darrick J. Wong <darrick.wong@oracle.com> 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_mount.h" 13 #include "xfs_btree.h" 14 #include "xfs_btree_staging.h" 15 #include "xfs_refcount_btree.h" 16 #include "xfs_alloc.h" 17 #include "xfs_error.h" 18 #include "xfs_trace.h" 19 #include "xfs_trans.h" 20 #include "xfs_bit.h" 21 #include "xfs_rmap.h" 22 #include "xfs_ag.h" 23 24 static struct kmem_cache *xfs_refcountbt_cur_cache; 25 26 static struct xfs_btree_cur * 27 xfs_refcountbt_dup_cursor( 28 struct xfs_btree_cur *cur) 29 { 30 return xfs_refcountbt_init_cursor(cur->bc_mp, cur->bc_tp, 31 cur->bc_ag.agbp, cur->bc_ag.pag); 32 } 33 34 STATIC void 35 xfs_refcountbt_set_root( 36 struct xfs_btree_cur *cur, 37 const union xfs_btree_ptr *ptr, 38 int inc) 39 { 40 struct xfs_buf *agbp = cur->bc_ag.agbp; 41 struct xfs_agf *agf = agbp->b_addr; 42 struct xfs_perag *pag = agbp->b_pag; 43 44 ASSERT(ptr->s != 0); 45 46 agf->agf_refcount_root = ptr->s; 47 be32_add_cpu(&agf->agf_refcount_level, inc); 48 pag->pagf_refcount_level += inc; 49 50 xfs_alloc_log_agf(cur->bc_tp, agbp, 51 XFS_AGF_REFCOUNT_ROOT | XFS_AGF_REFCOUNT_LEVEL); 52 } 53 54 STATIC int 55 xfs_refcountbt_alloc_block( 56 struct xfs_btree_cur *cur, 57 const union xfs_btree_ptr *start, 58 union xfs_btree_ptr *new, 59 int *stat) 60 { 61 struct xfs_buf *agbp = cur->bc_ag.agbp; 62 struct xfs_agf *agf = agbp->b_addr; 63 struct xfs_alloc_arg args; /* block allocation args */ 64 int error; /* error return value */ 65 66 memset(&args, 0, sizeof(args)); 67 args.tp = cur->bc_tp; 68 args.mp = cur->bc_mp; 69 args.type = XFS_ALLOCTYPE_NEAR_BNO; 70 args.fsbno = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_ag.pag->pag_agno, 71 xfs_refc_block(args.mp)); 72 args.oinfo = XFS_RMAP_OINFO_REFC; 73 args.minlen = args.maxlen = args.prod = 1; 74 args.resv = XFS_AG_RESV_METADATA; 75 76 error = xfs_alloc_vextent(&args); 77 if (error) 78 goto out_error; 79 trace_xfs_refcountbt_alloc_block(cur->bc_mp, cur->bc_ag.pag->pag_agno, 80 args.agbno, 1); 81 if (args.fsbno == NULLFSBLOCK) { 82 *stat = 0; 83 return 0; 84 } 85 ASSERT(args.agno == cur->bc_ag.pag->pag_agno); 86 ASSERT(args.len == 1); 87 88 new->s = cpu_to_be32(args.agbno); 89 be32_add_cpu(&agf->agf_refcount_blocks, 1); 90 xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_REFCOUNT_BLOCKS); 91 92 *stat = 1; 93 return 0; 94 95 out_error: 96 return error; 97 } 98 99 STATIC int 100 xfs_refcountbt_free_block( 101 struct xfs_btree_cur *cur, 102 struct xfs_buf *bp) 103 { 104 struct xfs_mount *mp = cur->bc_mp; 105 struct xfs_buf *agbp = cur->bc_ag.agbp; 106 struct xfs_agf *agf = agbp->b_addr; 107 xfs_fsblock_t fsbno = XFS_DADDR_TO_FSB(mp, xfs_buf_daddr(bp)); 108 int error; 109 110 trace_xfs_refcountbt_free_block(cur->bc_mp, cur->bc_ag.pag->pag_agno, 111 XFS_FSB_TO_AGBNO(cur->bc_mp, fsbno), 1); 112 be32_add_cpu(&agf->agf_refcount_blocks, -1); 113 xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_REFCOUNT_BLOCKS); 114 error = xfs_free_extent(cur->bc_tp, fsbno, 1, &XFS_RMAP_OINFO_REFC, 115 XFS_AG_RESV_METADATA); 116 if (error) 117 return error; 118 119 return error; 120 } 121 122 STATIC int 123 xfs_refcountbt_get_minrecs( 124 struct xfs_btree_cur *cur, 125 int level) 126 { 127 return cur->bc_mp->m_refc_mnr[level != 0]; 128 } 129 130 STATIC int 131 xfs_refcountbt_get_maxrecs( 132 struct xfs_btree_cur *cur, 133 int level) 134 { 135 return cur->bc_mp->m_refc_mxr[level != 0]; 136 } 137 138 STATIC void 139 xfs_refcountbt_init_key_from_rec( 140 union xfs_btree_key *key, 141 const union xfs_btree_rec *rec) 142 { 143 key->refc.rc_startblock = rec->refc.rc_startblock; 144 } 145 146 STATIC void 147 xfs_refcountbt_init_high_key_from_rec( 148 union xfs_btree_key *key, 149 const union xfs_btree_rec *rec) 150 { 151 __u32 x; 152 153 x = be32_to_cpu(rec->refc.rc_startblock); 154 x += be32_to_cpu(rec->refc.rc_blockcount) - 1; 155 key->refc.rc_startblock = cpu_to_be32(x); 156 } 157 158 STATIC void 159 xfs_refcountbt_init_rec_from_cur( 160 struct xfs_btree_cur *cur, 161 union xfs_btree_rec *rec) 162 { 163 rec->refc.rc_startblock = cpu_to_be32(cur->bc_rec.rc.rc_startblock); 164 rec->refc.rc_blockcount = cpu_to_be32(cur->bc_rec.rc.rc_blockcount); 165 rec->refc.rc_refcount = cpu_to_be32(cur->bc_rec.rc.rc_refcount); 166 } 167 168 STATIC void 169 xfs_refcountbt_init_ptr_from_cur( 170 struct xfs_btree_cur *cur, 171 union xfs_btree_ptr *ptr) 172 { 173 struct xfs_agf *agf = cur->bc_ag.agbp->b_addr; 174 175 ASSERT(cur->bc_ag.pag->pag_agno == be32_to_cpu(agf->agf_seqno)); 176 177 ptr->s = agf->agf_refcount_root; 178 } 179 180 STATIC int64_t 181 xfs_refcountbt_key_diff( 182 struct xfs_btree_cur *cur, 183 const union xfs_btree_key *key) 184 { 185 struct xfs_refcount_irec *rec = &cur->bc_rec.rc; 186 const struct xfs_refcount_key *kp = &key->refc; 187 188 return (int64_t)be32_to_cpu(kp->rc_startblock) - rec->rc_startblock; 189 } 190 191 STATIC int64_t 192 xfs_refcountbt_diff_two_keys( 193 struct xfs_btree_cur *cur, 194 const union xfs_btree_key *k1, 195 const union xfs_btree_key *k2) 196 { 197 return (int64_t)be32_to_cpu(k1->refc.rc_startblock) - 198 be32_to_cpu(k2->refc.rc_startblock); 199 } 200 201 STATIC xfs_failaddr_t 202 xfs_refcountbt_verify( 203 struct xfs_buf *bp) 204 { 205 struct xfs_mount *mp = bp->b_mount; 206 struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp); 207 struct xfs_perag *pag = bp->b_pag; 208 xfs_failaddr_t fa; 209 unsigned int level; 210 211 if (!xfs_verify_magic(bp, block->bb_magic)) 212 return __this_address; 213 214 if (!xfs_has_reflink(mp)) 215 return __this_address; 216 fa = xfs_btree_sblock_v5hdr_verify(bp); 217 if (fa) 218 return fa; 219 220 level = be16_to_cpu(block->bb_level); 221 if (pag && pag->pagf_init) { 222 if (level >= pag->pagf_refcount_level) 223 return __this_address; 224 } else if (level >= mp->m_refc_maxlevels) 225 return __this_address; 226 227 return xfs_btree_sblock_verify(bp, mp->m_refc_mxr[level != 0]); 228 } 229 230 STATIC void 231 xfs_refcountbt_read_verify( 232 struct xfs_buf *bp) 233 { 234 xfs_failaddr_t fa; 235 236 if (!xfs_btree_sblock_verify_crc(bp)) 237 xfs_verifier_error(bp, -EFSBADCRC, __this_address); 238 else { 239 fa = xfs_refcountbt_verify(bp); 240 if (fa) 241 xfs_verifier_error(bp, -EFSCORRUPTED, fa); 242 } 243 244 if (bp->b_error) 245 trace_xfs_btree_corrupt(bp, _RET_IP_); 246 } 247 248 STATIC void 249 xfs_refcountbt_write_verify( 250 struct xfs_buf *bp) 251 { 252 xfs_failaddr_t fa; 253 254 fa = xfs_refcountbt_verify(bp); 255 if (fa) { 256 trace_xfs_btree_corrupt(bp, _RET_IP_); 257 xfs_verifier_error(bp, -EFSCORRUPTED, fa); 258 return; 259 } 260 xfs_btree_sblock_calc_crc(bp); 261 262 } 263 264 const struct xfs_buf_ops xfs_refcountbt_buf_ops = { 265 .name = "xfs_refcountbt", 266 .magic = { 0, cpu_to_be32(XFS_REFC_CRC_MAGIC) }, 267 .verify_read = xfs_refcountbt_read_verify, 268 .verify_write = xfs_refcountbt_write_verify, 269 .verify_struct = xfs_refcountbt_verify, 270 }; 271 272 STATIC int 273 xfs_refcountbt_keys_inorder( 274 struct xfs_btree_cur *cur, 275 const union xfs_btree_key *k1, 276 const union xfs_btree_key *k2) 277 { 278 return be32_to_cpu(k1->refc.rc_startblock) < 279 be32_to_cpu(k2->refc.rc_startblock); 280 } 281 282 STATIC int 283 xfs_refcountbt_recs_inorder( 284 struct xfs_btree_cur *cur, 285 const union xfs_btree_rec *r1, 286 const union xfs_btree_rec *r2) 287 { 288 return be32_to_cpu(r1->refc.rc_startblock) + 289 be32_to_cpu(r1->refc.rc_blockcount) <= 290 be32_to_cpu(r2->refc.rc_startblock); 291 } 292 293 static const struct xfs_btree_ops xfs_refcountbt_ops = { 294 .rec_len = sizeof(struct xfs_refcount_rec), 295 .key_len = sizeof(struct xfs_refcount_key), 296 297 .dup_cursor = xfs_refcountbt_dup_cursor, 298 .set_root = xfs_refcountbt_set_root, 299 .alloc_block = xfs_refcountbt_alloc_block, 300 .free_block = xfs_refcountbt_free_block, 301 .get_minrecs = xfs_refcountbt_get_minrecs, 302 .get_maxrecs = xfs_refcountbt_get_maxrecs, 303 .init_key_from_rec = xfs_refcountbt_init_key_from_rec, 304 .init_high_key_from_rec = xfs_refcountbt_init_high_key_from_rec, 305 .init_rec_from_cur = xfs_refcountbt_init_rec_from_cur, 306 .init_ptr_from_cur = xfs_refcountbt_init_ptr_from_cur, 307 .key_diff = xfs_refcountbt_key_diff, 308 .buf_ops = &xfs_refcountbt_buf_ops, 309 .diff_two_keys = xfs_refcountbt_diff_two_keys, 310 .keys_inorder = xfs_refcountbt_keys_inorder, 311 .recs_inorder = xfs_refcountbt_recs_inorder, 312 }; 313 314 /* 315 * Initialize a new refcount btree cursor. 316 */ 317 static struct xfs_btree_cur * 318 xfs_refcountbt_init_common( 319 struct xfs_mount *mp, 320 struct xfs_trans *tp, 321 struct xfs_perag *pag) 322 { 323 struct xfs_btree_cur *cur; 324 325 ASSERT(pag->pag_agno < mp->m_sb.sb_agcount); 326 327 cur = xfs_btree_alloc_cursor(mp, tp, XFS_BTNUM_REFC, 328 mp->m_refc_maxlevels, xfs_refcountbt_cur_cache); 329 cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_refcbt_2); 330 331 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS; 332 333 /* take a reference for the cursor */ 334 atomic_inc(&pag->pag_ref); 335 cur->bc_ag.pag = pag; 336 337 cur->bc_ag.refc.nr_ops = 0; 338 cur->bc_ag.refc.shape_changes = 0; 339 cur->bc_ops = &xfs_refcountbt_ops; 340 return cur; 341 } 342 343 /* Create a btree cursor. */ 344 struct xfs_btree_cur * 345 xfs_refcountbt_init_cursor( 346 struct xfs_mount *mp, 347 struct xfs_trans *tp, 348 struct xfs_buf *agbp, 349 struct xfs_perag *pag) 350 { 351 struct xfs_agf *agf = agbp->b_addr; 352 struct xfs_btree_cur *cur; 353 354 cur = xfs_refcountbt_init_common(mp, tp, pag); 355 cur->bc_nlevels = be32_to_cpu(agf->agf_refcount_level); 356 cur->bc_ag.agbp = agbp; 357 return cur; 358 } 359 360 /* Create a btree cursor with a fake root for staging. */ 361 struct xfs_btree_cur * 362 xfs_refcountbt_stage_cursor( 363 struct xfs_mount *mp, 364 struct xbtree_afakeroot *afake, 365 struct xfs_perag *pag) 366 { 367 struct xfs_btree_cur *cur; 368 369 cur = xfs_refcountbt_init_common(mp, NULL, pag); 370 xfs_btree_stage_afakeroot(cur, afake); 371 return cur; 372 } 373 374 /* 375 * Swap in the new btree root. Once we pass this point the newly rebuilt btree 376 * is in place and we have to kill off all the old btree blocks. 377 */ 378 void 379 xfs_refcountbt_commit_staged_btree( 380 struct xfs_btree_cur *cur, 381 struct xfs_trans *tp, 382 struct xfs_buf *agbp) 383 { 384 struct xfs_agf *agf = agbp->b_addr; 385 struct xbtree_afakeroot *afake = cur->bc_ag.afake; 386 387 ASSERT(cur->bc_flags & XFS_BTREE_STAGING); 388 389 agf->agf_refcount_root = cpu_to_be32(afake->af_root); 390 agf->agf_refcount_level = cpu_to_be32(afake->af_levels); 391 agf->agf_refcount_blocks = cpu_to_be32(afake->af_blocks); 392 xfs_alloc_log_agf(tp, agbp, XFS_AGF_REFCOUNT_BLOCKS | 393 XFS_AGF_REFCOUNT_ROOT | 394 XFS_AGF_REFCOUNT_LEVEL); 395 xfs_btree_commit_afakeroot(cur, tp, agbp, &xfs_refcountbt_ops); 396 } 397 398 /* Calculate number of records in a refcount btree block. */ 399 static inline unsigned int 400 xfs_refcountbt_block_maxrecs( 401 unsigned int blocklen, 402 bool leaf) 403 { 404 if (leaf) 405 return blocklen / sizeof(struct xfs_refcount_rec); 406 return blocklen / (sizeof(struct xfs_refcount_key) + 407 sizeof(xfs_refcount_ptr_t)); 408 } 409 410 /* 411 * Calculate the number of records in a refcount btree block. 412 */ 413 int 414 xfs_refcountbt_maxrecs( 415 int blocklen, 416 bool leaf) 417 { 418 blocklen -= XFS_REFCOUNT_BLOCK_LEN; 419 return xfs_refcountbt_block_maxrecs(blocklen, leaf); 420 } 421 422 /* Compute the max possible height of the maximally sized refcount btree. */ 423 unsigned int 424 xfs_refcountbt_maxlevels_ondisk(void) 425 { 426 unsigned int minrecs[2]; 427 unsigned int blocklen; 428 429 blocklen = XFS_MIN_CRC_BLOCKSIZE - XFS_BTREE_SBLOCK_CRC_LEN; 430 431 minrecs[0] = xfs_refcountbt_block_maxrecs(blocklen, true) / 2; 432 minrecs[1] = xfs_refcountbt_block_maxrecs(blocklen, false) / 2; 433 434 return xfs_btree_compute_maxlevels(minrecs, XFS_MAX_CRC_AG_BLOCKS); 435 } 436 437 /* Compute the maximum height of a refcount btree. */ 438 void 439 xfs_refcountbt_compute_maxlevels( 440 struct xfs_mount *mp) 441 { 442 if (!xfs_has_reflink(mp)) { 443 mp->m_refc_maxlevels = 0; 444 return; 445 } 446 447 mp->m_refc_maxlevels = xfs_btree_compute_maxlevels( 448 mp->m_refc_mnr, mp->m_sb.sb_agblocks); 449 ASSERT(mp->m_refc_maxlevels <= xfs_refcountbt_maxlevels_ondisk()); 450 } 451 452 /* Calculate the refcount btree size for some records. */ 453 xfs_extlen_t 454 xfs_refcountbt_calc_size( 455 struct xfs_mount *mp, 456 unsigned long long len) 457 { 458 return xfs_btree_calc_size(mp->m_refc_mnr, len); 459 } 460 461 /* 462 * Calculate the maximum refcount btree size. 463 */ 464 xfs_extlen_t 465 xfs_refcountbt_max_size( 466 struct xfs_mount *mp, 467 xfs_agblock_t agblocks) 468 { 469 /* Bail out if we're uninitialized, which can happen in mkfs. */ 470 if (mp->m_refc_mxr[0] == 0) 471 return 0; 472 473 return xfs_refcountbt_calc_size(mp, agblocks); 474 } 475 476 /* 477 * Figure out how many blocks to reserve and how many are used by this btree. 478 */ 479 int 480 xfs_refcountbt_calc_reserves( 481 struct xfs_mount *mp, 482 struct xfs_trans *tp, 483 struct xfs_perag *pag, 484 xfs_extlen_t *ask, 485 xfs_extlen_t *used) 486 { 487 struct xfs_buf *agbp; 488 struct xfs_agf *agf; 489 xfs_agblock_t agblocks; 490 xfs_extlen_t tree_len; 491 int error; 492 493 if (!xfs_has_reflink(mp)) 494 return 0; 495 496 error = xfs_alloc_read_agf(pag, tp, 0, &agbp); 497 if (error) 498 return error; 499 500 agf = agbp->b_addr; 501 agblocks = be32_to_cpu(agf->agf_length); 502 tree_len = be32_to_cpu(agf->agf_refcount_blocks); 503 xfs_trans_brelse(tp, agbp); 504 505 /* 506 * The log is permanently allocated, so the space it occupies will 507 * never be available for the kinds of things that would require btree 508 * expansion. We therefore can pretend the space isn't there. 509 */ 510 if (xfs_ag_contains_log(mp, pag->pag_agno)) 511 agblocks -= mp->m_sb.sb_logblocks; 512 513 *ask += xfs_refcountbt_max_size(mp, agblocks); 514 *used += tree_len; 515 516 return error; 517 } 518 519 int __init 520 xfs_refcountbt_init_cur_cache(void) 521 { 522 xfs_refcountbt_cur_cache = kmem_cache_create("xfs_refcbt_cur", 523 xfs_btree_cur_sizeof(xfs_refcountbt_maxlevels_ondisk()), 524 0, 0, NULL); 525 526 if (!xfs_refcountbt_cur_cache) 527 return -ENOMEM; 528 return 0; 529 } 530 531 void 532 xfs_refcountbt_destroy_cur_cache(void) 533 { 534 kmem_cache_destroy(xfs_refcountbt_cur_cache); 535 xfs_refcountbt_cur_cache = NULL; 536 } 537