b.liu | e958203 | 2025-04-17 19:18:16 +0800 | [diff] [blame^] | 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_sb.h" |
| 13 | #include "xfs_mount.h" |
| 14 | #include "xfs_btree.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 | |
| 23 | static struct xfs_btree_cur * |
| 24 | xfs_refcountbt_dup_cursor( |
| 25 | struct xfs_btree_cur *cur) |
| 26 | { |
| 27 | return xfs_refcountbt_init_cursor(cur->bc_mp, cur->bc_tp, |
| 28 | cur->bc_private.a.agbp, cur->bc_private.a.agno); |
| 29 | } |
| 30 | |
| 31 | STATIC void |
| 32 | xfs_refcountbt_set_root( |
| 33 | struct xfs_btree_cur *cur, |
| 34 | union xfs_btree_ptr *ptr, |
| 35 | int inc) |
| 36 | { |
| 37 | struct xfs_buf *agbp = cur->bc_private.a.agbp; |
| 38 | struct xfs_agf *agf = XFS_BUF_TO_AGF(agbp); |
| 39 | xfs_agnumber_t seqno = be32_to_cpu(agf->agf_seqno); |
| 40 | struct xfs_perag *pag = xfs_perag_get(cur->bc_mp, seqno); |
| 41 | |
| 42 | ASSERT(ptr->s != 0); |
| 43 | |
| 44 | agf->agf_refcount_root = ptr->s; |
| 45 | be32_add_cpu(&agf->agf_refcount_level, inc); |
| 46 | pag->pagf_refcount_level += inc; |
| 47 | xfs_perag_put(pag); |
| 48 | |
| 49 | xfs_alloc_log_agf(cur->bc_tp, agbp, |
| 50 | XFS_AGF_REFCOUNT_ROOT | XFS_AGF_REFCOUNT_LEVEL); |
| 51 | } |
| 52 | |
| 53 | STATIC int |
| 54 | xfs_refcountbt_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 | struct xfs_buf *agbp = cur->bc_private.a.agbp; |
| 61 | struct xfs_agf *agf = XFS_BUF_TO_AGF(agbp); |
| 62 | struct xfs_alloc_arg args; /* block allocation args */ |
| 63 | int error; /* error return value */ |
| 64 | |
| 65 | memset(&args, 0, sizeof(args)); |
| 66 | args.tp = cur->bc_tp; |
| 67 | args.mp = cur->bc_mp; |
| 68 | args.type = XFS_ALLOCTYPE_NEAR_BNO; |
| 69 | args.fsbno = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_private.a.agno, |
| 70 | xfs_refc_block(args.mp)); |
| 71 | args.oinfo = XFS_RMAP_OINFO_REFC; |
| 72 | args.minlen = args.maxlen = args.prod = 1; |
| 73 | args.resv = XFS_AG_RESV_METADATA; |
| 74 | |
| 75 | error = xfs_alloc_vextent(&args); |
| 76 | if (error) |
| 77 | goto out_error; |
| 78 | trace_xfs_refcountbt_alloc_block(cur->bc_mp, cur->bc_private.a.agno, |
| 79 | args.agbno, 1); |
| 80 | if (args.fsbno == NULLFSBLOCK) { |
| 81 | *stat = 0; |
| 82 | return 0; |
| 83 | } |
| 84 | ASSERT(args.agno == cur->bc_private.a.agno); |
| 85 | ASSERT(args.len == 1); |
| 86 | |
| 87 | new->s = cpu_to_be32(args.agbno); |
| 88 | be32_add_cpu(&agf->agf_refcount_blocks, 1); |
| 89 | xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_REFCOUNT_BLOCKS); |
| 90 | |
| 91 | *stat = 1; |
| 92 | return 0; |
| 93 | |
| 94 | out_error: |
| 95 | return error; |
| 96 | } |
| 97 | |
| 98 | STATIC int |
| 99 | xfs_refcountbt_free_block( |
| 100 | struct xfs_btree_cur *cur, |
| 101 | struct xfs_buf *bp) |
| 102 | { |
| 103 | struct xfs_mount *mp = cur->bc_mp; |
| 104 | struct xfs_buf *agbp = cur->bc_private.a.agbp; |
| 105 | struct xfs_agf *agf = XFS_BUF_TO_AGF(agbp); |
| 106 | xfs_fsblock_t fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp)); |
| 107 | int error; |
| 108 | |
| 109 | trace_xfs_refcountbt_free_block(cur->bc_mp, cur->bc_private.a.agno, |
| 110 | XFS_FSB_TO_AGBNO(cur->bc_mp, fsbno), 1); |
| 111 | be32_add_cpu(&agf->agf_refcount_blocks, -1); |
| 112 | xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_REFCOUNT_BLOCKS); |
| 113 | error = xfs_free_extent(cur->bc_tp, fsbno, 1, &XFS_RMAP_OINFO_REFC, |
| 114 | XFS_AG_RESV_METADATA); |
| 115 | if (error) |
| 116 | return error; |
| 117 | |
| 118 | return error; |
| 119 | } |
| 120 | |
| 121 | STATIC int |
| 122 | xfs_refcountbt_get_minrecs( |
| 123 | struct xfs_btree_cur *cur, |
| 124 | int level) |
| 125 | { |
| 126 | return cur->bc_mp->m_refc_mnr[level != 0]; |
| 127 | } |
| 128 | |
| 129 | STATIC int |
| 130 | xfs_refcountbt_get_maxrecs( |
| 131 | struct xfs_btree_cur *cur, |
| 132 | int level) |
| 133 | { |
| 134 | return cur->bc_mp->m_refc_mxr[level != 0]; |
| 135 | } |
| 136 | |
| 137 | STATIC void |
| 138 | xfs_refcountbt_init_key_from_rec( |
| 139 | union xfs_btree_key *key, |
| 140 | union xfs_btree_rec *rec) |
| 141 | { |
| 142 | key->refc.rc_startblock = rec->refc.rc_startblock; |
| 143 | } |
| 144 | |
| 145 | STATIC void |
| 146 | xfs_refcountbt_init_high_key_from_rec( |
| 147 | union xfs_btree_key *key, |
| 148 | union xfs_btree_rec *rec) |
| 149 | { |
| 150 | __u32 x; |
| 151 | |
| 152 | x = be32_to_cpu(rec->refc.rc_startblock); |
| 153 | x += be32_to_cpu(rec->refc.rc_blockcount) - 1; |
| 154 | key->refc.rc_startblock = cpu_to_be32(x); |
| 155 | } |
| 156 | |
| 157 | STATIC void |
| 158 | xfs_refcountbt_init_rec_from_cur( |
| 159 | struct xfs_btree_cur *cur, |
| 160 | union xfs_btree_rec *rec) |
| 161 | { |
| 162 | rec->refc.rc_startblock = cpu_to_be32(cur->bc_rec.rc.rc_startblock); |
| 163 | rec->refc.rc_blockcount = cpu_to_be32(cur->bc_rec.rc.rc_blockcount); |
| 164 | rec->refc.rc_refcount = cpu_to_be32(cur->bc_rec.rc.rc_refcount); |
| 165 | } |
| 166 | |
| 167 | STATIC void |
| 168 | xfs_refcountbt_init_ptr_from_cur( |
| 169 | struct xfs_btree_cur *cur, |
| 170 | union xfs_btree_ptr *ptr) |
| 171 | { |
| 172 | struct xfs_agf *agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp); |
| 173 | |
| 174 | ASSERT(cur->bc_private.a.agno == be32_to_cpu(agf->agf_seqno)); |
| 175 | |
| 176 | ptr->s = agf->agf_refcount_root; |
| 177 | } |
| 178 | |
| 179 | STATIC int64_t |
| 180 | xfs_refcountbt_key_diff( |
| 181 | struct xfs_btree_cur *cur, |
| 182 | union xfs_btree_key *key) |
| 183 | { |
| 184 | struct xfs_refcount_irec *rec = &cur->bc_rec.rc; |
| 185 | struct xfs_refcount_key *kp = &key->refc; |
| 186 | |
| 187 | return (int64_t)be32_to_cpu(kp->rc_startblock) - rec->rc_startblock; |
| 188 | } |
| 189 | |
| 190 | STATIC int64_t |
| 191 | xfs_refcountbt_diff_two_keys( |
| 192 | struct xfs_btree_cur *cur, |
| 193 | union xfs_btree_key *k1, |
| 194 | union xfs_btree_key *k2) |
| 195 | { |
| 196 | return (int64_t)be32_to_cpu(k1->refc.rc_startblock) - |
| 197 | be32_to_cpu(k2->refc.rc_startblock); |
| 198 | } |
| 199 | |
| 200 | STATIC xfs_failaddr_t |
| 201 | xfs_refcountbt_verify( |
| 202 | struct xfs_buf *bp) |
| 203 | { |
| 204 | struct xfs_mount *mp = bp->b_mount; |
| 205 | struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp); |
| 206 | struct xfs_perag *pag = bp->b_pag; |
| 207 | xfs_failaddr_t fa; |
| 208 | unsigned int level; |
| 209 | |
| 210 | if (!xfs_verify_magic(bp, block->bb_magic)) |
| 211 | return __this_address; |
| 212 | |
| 213 | if (!xfs_sb_version_hasreflink(&mp->m_sb)) |
| 214 | return __this_address; |
| 215 | fa = xfs_btree_sblock_v5hdr_verify(bp); |
| 216 | if (fa) |
| 217 | return fa; |
| 218 | |
| 219 | level = be16_to_cpu(block->bb_level); |
| 220 | if (pag && pag->pagf_init) { |
| 221 | if (level >= pag->pagf_refcount_level) |
| 222 | return __this_address; |
| 223 | } else if (level >= mp->m_refc_maxlevels) |
| 224 | return __this_address; |
| 225 | |
| 226 | return xfs_btree_sblock_verify(bp, mp->m_refc_mxr[level != 0]); |
| 227 | } |
| 228 | |
| 229 | STATIC void |
| 230 | xfs_refcountbt_read_verify( |
| 231 | struct xfs_buf *bp) |
| 232 | { |
| 233 | xfs_failaddr_t fa; |
| 234 | |
| 235 | if (!xfs_btree_sblock_verify_crc(bp)) |
| 236 | xfs_verifier_error(bp, -EFSBADCRC, __this_address); |
| 237 | else { |
| 238 | fa = xfs_refcountbt_verify(bp); |
| 239 | if (fa) |
| 240 | xfs_verifier_error(bp, -EFSCORRUPTED, fa); |
| 241 | } |
| 242 | |
| 243 | if (bp->b_error) |
| 244 | trace_xfs_btree_corrupt(bp, _RET_IP_); |
| 245 | } |
| 246 | |
| 247 | STATIC void |
| 248 | xfs_refcountbt_write_verify( |
| 249 | struct xfs_buf *bp) |
| 250 | { |
| 251 | xfs_failaddr_t fa; |
| 252 | |
| 253 | fa = xfs_refcountbt_verify(bp); |
| 254 | if (fa) { |
| 255 | trace_xfs_btree_corrupt(bp, _RET_IP_); |
| 256 | xfs_verifier_error(bp, -EFSCORRUPTED, fa); |
| 257 | return; |
| 258 | } |
| 259 | xfs_btree_sblock_calc_crc(bp); |
| 260 | |
| 261 | } |
| 262 | |
| 263 | const struct xfs_buf_ops xfs_refcountbt_buf_ops = { |
| 264 | .name = "xfs_refcountbt", |
| 265 | .magic = { 0, cpu_to_be32(XFS_REFC_CRC_MAGIC) }, |
| 266 | .verify_read = xfs_refcountbt_read_verify, |
| 267 | .verify_write = xfs_refcountbt_write_verify, |
| 268 | .verify_struct = xfs_refcountbt_verify, |
| 269 | }; |
| 270 | |
| 271 | STATIC int |
| 272 | xfs_refcountbt_keys_inorder( |
| 273 | struct xfs_btree_cur *cur, |
| 274 | union xfs_btree_key *k1, |
| 275 | union xfs_btree_key *k2) |
| 276 | { |
| 277 | return be32_to_cpu(k1->refc.rc_startblock) < |
| 278 | be32_to_cpu(k2->refc.rc_startblock); |
| 279 | } |
| 280 | |
| 281 | STATIC int |
| 282 | xfs_refcountbt_recs_inorder( |
| 283 | struct xfs_btree_cur *cur, |
| 284 | union xfs_btree_rec *r1, |
| 285 | union xfs_btree_rec *r2) |
| 286 | { |
| 287 | return be32_to_cpu(r1->refc.rc_startblock) + |
| 288 | be32_to_cpu(r1->refc.rc_blockcount) <= |
| 289 | be32_to_cpu(r2->refc.rc_startblock); |
| 290 | } |
| 291 | |
| 292 | static const struct xfs_btree_ops xfs_refcountbt_ops = { |
| 293 | .rec_len = sizeof(struct xfs_refcount_rec), |
| 294 | .key_len = sizeof(struct xfs_refcount_key), |
| 295 | |
| 296 | .dup_cursor = xfs_refcountbt_dup_cursor, |
| 297 | .set_root = xfs_refcountbt_set_root, |
| 298 | .alloc_block = xfs_refcountbt_alloc_block, |
| 299 | .free_block = xfs_refcountbt_free_block, |
| 300 | .get_minrecs = xfs_refcountbt_get_minrecs, |
| 301 | .get_maxrecs = xfs_refcountbt_get_maxrecs, |
| 302 | .init_key_from_rec = xfs_refcountbt_init_key_from_rec, |
| 303 | .init_high_key_from_rec = xfs_refcountbt_init_high_key_from_rec, |
| 304 | .init_rec_from_cur = xfs_refcountbt_init_rec_from_cur, |
| 305 | .init_ptr_from_cur = xfs_refcountbt_init_ptr_from_cur, |
| 306 | .key_diff = xfs_refcountbt_key_diff, |
| 307 | .buf_ops = &xfs_refcountbt_buf_ops, |
| 308 | .diff_two_keys = xfs_refcountbt_diff_two_keys, |
| 309 | .keys_inorder = xfs_refcountbt_keys_inorder, |
| 310 | .recs_inorder = xfs_refcountbt_recs_inorder, |
| 311 | }; |
| 312 | |
| 313 | /* |
| 314 | * Allocate a new refcount btree cursor. |
| 315 | */ |
| 316 | struct xfs_btree_cur * |
| 317 | xfs_refcountbt_init_cursor( |
| 318 | struct xfs_mount *mp, |
| 319 | struct xfs_trans *tp, |
| 320 | struct xfs_buf *agbp, |
| 321 | xfs_agnumber_t agno) |
| 322 | { |
| 323 | struct xfs_agf *agf = XFS_BUF_TO_AGF(agbp); |
| 324 | struct xfs_btree_cur *cur; |
| 325 | |
| 326 | ASSERT(agno != NULLAGNUMBER); |
| 327 | ASSERT(agno < mp->m_sb.sb_agcount); |
| 328 | cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS); |
| 329 | |
| 330 | cur->bc_tp = tp; |
| 331 | cur->bc_mp = mp; |
| 332 | cur->bc_btnum = XFS_BTNUM_REFC; |
| 333 | cur->bc_blocklog = mp->m_sb.sb_blocklog; |
| 334 | cur->bc_ops = &xfs_refcountbt_ops; |
| 335 | cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_refcbt_2); |
| 336 | |
| 337 | cur->bc_nlevels = be32_to_cpu(agf->agf_refcount_level); |
| 338 | |
| 339 | cur->bc_private.a.agbp = agbp; |
| 340 | cur->bc_private.a.agno = agno; |
| 341 | cur->bc_flags |= XFS_BTREE_CRC_BLOCKS; |
| 342 | |
| 343 | cur->bc_private.a.priv.refc.nr_ops = 0; |
| 344 | cur->bc_private.a.priv.refc.shape_changes = 0; |
| 345 | |
| 346 | return cur; |
| 347 | } |
| 348 | |
| 349 | /* |
| 350 | * Calculate the number of records in a refcount btree block. |
| 351 | */ |
| 352 | int |
| 353 | xfs_refcountbt_maxrecs( |
| 354 | int blocklen, |
| 355 | bool leaf) |
| 356 | { |
| 357 | blocklen -= XFS_REFCOUNT_BLOCK_LEN; |
| 358 | |
| 359 | if (leaf) |
| 360 | return blocklen / sizeof(struct xfs_refcount_rec); |
| 361 | return blocklen / (sizeof(struct xfs_refcount_key) + |
| 362 | sizeof(xfs_refcount_ptr_t)); |
| 363 | } |
| 364 | |
| 365 | /* Compute the maximum height of a refcount btree. */ |
| 366 | void |
| 367 | xfs_refcountbt_compute_maxlevels( |
| 368 | struct xfs_mount *mp) |
| 369 | { |
| 370 | mp->m_refc_maxlevels = xfs_btree_compute_maxlevels( |
| 371 | mp->m_refc_mnr, mp->m_sb.sb_agblocks); |
| 372 | } |
| 373 | |
| 374 | /* Calculate the refcount btree size for some records. */ |
| 375 | xfs_extlen_t |
| 376 | xfs_refcountbt_calc_size( |
| 377 | struct xfs_mount *mp, |
| 378 | unsigned long long len) |
| 379 | { |
| 380 | return xfs_btree_calc_size(mp->m_refc_mnr, len); |
| 381 | } |
| 382 | |
| 383 | /* |
| 384 | * Calculate the maximum refcount btree size. |
| 385 | */ |
| 386 | xfs_extlen_t |
| 387 | xfs_refcountbt_max_size( |
| 388 | struct xfs_mount *mp, |
| 389 | xfs_agblock_t agblocks) |
| 390 | { |
| 391 | /* Bail out if we're uninitialized, which can happen in mkfs. */ |
| 392 | if (mp->m_refc_mxr[0] == 0) |
| 393 | return 0; |
| 394 | |
| 395 | return xfs_refcountbt_calc_size(mp, agblocks); |
| 396 | } |
| 397 | |
| 398 | /* |
| 399 | * Figure out how many blocks to reserve and how many are used by this btree. |
| 400 | */ |
| 401 | int |
| 402 | xfs_refcountbt_calc_reserves( |
| 403 | struct xfs_mount *mp, |
| 404 | struct xfs_trans *tp, |
| 405 | xfs_agnumber_t agno, |
| 406 | xfs_extlen_t *ask, |
| 407 | xfs_extlen_t *used) |
| 408 | { |
| 409 | struct xfs_buf *agbp; |
| 410 | struct xfs_agf *agf; |
| 411 | xfs_agblock_t agblocks; |
| 412 | xfs_extlen_t tree_len; |
| 413 | int error; |
| 414 | |
| 415 | if (!xfs_sb_version_hasreflink(&mp->m_sb)) |
| 416 | return 0; |
| 417 | |
| 418 | |
| 419 | error = xfs_alloc_read_agf(mp, tp, agno, 0, &agbp); |
| 420 | if (error) |
| 421 | return error; |
| 422 | |
| 423 | agf = XFS_BUF_TO_AGF(agbp); |
| 424 | agblocks = be32_to_cpu(agf->agf_length); |
| 425 | tree_len = be32_to_cpu(agf->agf_refcount_blocks); |
| 426 | xfs_trans_brelse(tp, agbp); |
| 427 | |
| 428 | /* |
| 429 | * The log is permanently allocated, so the space it occupies will |
| 430 | * never be available for the kinds of things that would require btree |
| 431 | * expansion. We therefore can pretend the space isn't there. |
| 432 | */ |
| 433 | if (mp->m_sb.sb_logstart && |
| 434 | XFS_FSB_TO_AGNO(mp, mp->m_sb.sb_logstart) == agno) |
| 435 | agblocks -= mp->m_sb.sb_logblocks; |
| 436 | |
| 437 | *ask += xfs_refcountbt_max_size(mp, agblocks); |
| 438 | *used += tree_len; |
| 439 | |
| 440 | return error; |
| 441 | } |