1// SPDX-License-Identifier: GPL-2.0+ 2/* 3 * Copyright (C) 2017 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_inode.h" 14#include "xfs_trans.h" 15#include "xfs_btree.h" 16#include "xfs_rmap_btree.h" 17#include "xfs_trace.h" 18#include "xfs_rmap.h" 19#include "xfs_alloc.h" 20#include "xfs_bit.h" 21#include <linux/fsmap.h> 22#include "xfs_fsmap.h" 23#include "xfs_refcount.h" 24#include "xfs_refcount_btree.h" 25#include "xfs_alloc_btree.h" 26#include "xfs_rtalloc.h" 27#include "xfs_ag.h" 28 29/* Convert an xfs_fsmap to an fsmap. */ 30static void 31xfs_fsmap_from_internal( 32 struct fsmap *dest, 33 struct xfs_fsmap *src) 34{ 35 dest->fmr_device = src->fmr_device; 36 dest->fmr_flags = src->fmr_flags; 37 dest->fmr_physical = BBTOB(src->fmr_physical); 38 dest->fmr_owner = src->fmr_owner; 39 dest->fmr_offset = BBTOB(src->fmr_offset); 40 dest->fmr_length = BBTOB(src->fmr_length); 41 dest->fmr_reserved[0] = 0; 42 dest->fmr_reserved[1] = 0; 43 dest->fmr_reserved[2] = 0; 44} 45 46/* Convert an fsmap to an xfs_fsmap. */ 47void 48xfs_fsmap_to_internal( 49 struct xfs_fsmap *dest, 50 struct fsmap *src) 51{ 52 dest->fmr_device = src->fmr_device; 53 dest->fmr_flags = src->fmr_flags; 54 dest->fmr_physical = BTOBBT(src->fmr_physical); 55 dest->fmr_owner = src->fmr_owner; 56 dest->fmr_offset = BTOBBT(src->fmr_offset); 57 dest->fmr_length = BTOBBT(src->fmr_length); 58} 59 60/* Convert an fsmap owner into an rmapbt owner. */ 61static int 62xfs_fsmap_owner_to_rmap( 63 struct xfs_rmap_irec *dest, 64 const struct xfs_fsmap *src) 65{ 66 if (!(src->fmr_flags & FMR_OF_SPECIAL_OWNER)) { 67 dest->rm_owner = src->fmr_owner; 68 return 0; 69 } 70 71 switch (src->fmr_owner) { 72 case 0: /* "lowest owner id possible" */ 73 case -1ULL: /* "highest owner id possible" */ 74 dest->rm_owner = 0; 75 break; 76 case XFS_FMR_OWN_FREE: 77 dest->rm_owner = XFS_RMAP_OWN_NULL; 78 break; 79 case XFS_FMR_OWN_UNKNOWN: 80 dest->rm_owner = XFS_RMAP_OWN_UNKNOWN; 81 break; 82 case XFS_FMR_OWN_FS: 83 dest->rm_owner = XFS_RMAP_OWN_FS; 84 break; 85 case XFS_FMR_OWN_LOG: 86 dest->rm_owner = XFS_RMAP_OWN_LOG; 87 break; 88 case XFS_FMR_OWN_AG: 89 dest->rm_owner = XFS_RMAP_OWN_AG; 90 break; 91 case XFS_FMR_OWN_INOBT: 92 dest->rm_owner = XFS_RMAP_OWN_INOBT; 93 break; 94 case XFS_FMR_OWN_INODES: 95 dest->rm_owner = XFS_RMAP_OWN_INODES; 96 break; 97 case XFS_FMR_OWN_REFC: 98 dest->rm_owner = XFS_RMAP_OWN_REFC; 99 break; 100 case XFS_FMR_OWN_COW: 101 dest->rm_owner = XFS_RMAP_OWN_COW; 102 break; 103 case XFS_FMR_OWN_DEFECTIVE: /* not implemented */ 104 /* fall through */ 105 default: 106 return -EINVAL; 107 } 108 return 0; 109} 110 111/* Convert an rmapbt owner into an fsmap owner. */ 112static int 113xfs_fsmap_owner_from_rmap( 114 struct xfs_fsmap *dest, 115 const struct xfs_rmap_irec *src) 116{ 117 dest->fmr_flags = 0; 118 if (!XFS_RMAP_NON_INODE_OWNER(src->rm_owner)) { 119 dest->fmr_owner = src->rm_owner; 120 return 0; 121 } 122 dest->fmr_flags |= FMR_OF_SPECIAL_OWNER; 123 124 switch (src->rm_owner) { 125 case XFS_RMAP_OWN_FS: 126 dest->fmr_owner = XFS_FMR_OWN_FS; 127 break; 128 case XFS_RMAP_OWN_LOG: 129 dest->fmr_owner = XFS_FMR_OWN_LOG; 130 break; 131 case XFS_RMAP_OWN_AG: 132 dest->fmr_owner = XFS_FMR_OWN_AG; 133 break; 134 case XFS_RMAP_OWN_INOBT: 135 dest->fmr_owner = XFS_FMR_OWN_INOBT; 136 break; 137 case XFS_RMAP_OWN_INODES: 138 dest->fmr_owner = XFS_FMR_OWN_INODES; 139 break; 140 case XFS_RMAP_OWN_REFC: 141 dest->fmr_owner = XFS_FMR_OWN_REFC; 142 break; 143 case XFS_RMAP_OWN_COW: 144 dest->fmr_owner = XFS_FMR_OWN_COW; 145 break; 146 case XFS_RMAP_OWN_NULL: /* "free" */ 147 dest->fmr_owner = XFS_FMR_OWN_FREE; 148 break; 149 default: 150 ASSERT(0); 151 return -EFSCORRUPTED; 152 } 153 return 0; 154} 155 156/* getfsmap query state */ 157struct xfs_getfsmap_info { 158 struct xfs_fsmap_head *head; 159 struct fsmap *fsmap_recs; /* mapping records */ 160 struct xfs_buf *agf_bp; /* AGF, for refcount queries */ 161 struct xfs_perag *pag; /* AG info, if applicable */ 162 xfs_daddr_t next_daddr; /* next daddr we expect */ 163 /* daddr of low fsmap key when we're using the rtbitmap */ 164 xfs_daddr_t low_daddr; 165 u64 missing_owner; /* owner of holes */ 166 u32 dev; /* device id */ 167 /* 168 * Low rmap key for the query. If low.rm_blockcount is nonzero, this 169 * is the second (or later) call to retrieve the recordset in pieces. 170 * xfs_getfsmap_rec_before_start will compare all records retrieved 171 * by the rmapbt query to filter out any records that start before 172 * the last record. 173 */ 174 struct xfs_rmap_irec low; 175 struct xfs_rmap_irec high; /* high rmap key */ 176 bool last; /* last extent? */ 177}; 178 179/* Associate a device with a getfsmap handler. */ 180struct xfs_getfsmap_dev { 181 u32 dev; 182 int (*fn)(struct xfs_trans *tp, 183 const struct xfs_fsmap *keys, 184 struct xfs_getfsmap_info *info); 185}; 186 187/* Compare two getfsmap device handlers. */ 188static int 189xfs_getfsmap_dev_compare( 190 const void *p1, 191 const void *p2) 192{ 193 const struct xfs_getfsmap_dev *d1 = p1; 194 const struct xfs_getfsmap_dev *d2 = p2; 195 196 return d1->dev - d2->dev; 197} 198 199/* Decide if this mapping is shared. */ 200STATIC int 201xfs_getfsmap_is_shared( 202 struct xfs_trans *tp, 203 struct xfs_getfsmap_info *info, 204 const struct xfs_rmap_irec *rec, 205 bool *stat) 206{ 207 struct xfs_mount *mp = tp->t_mountp; 208 struct xfs_btree_cur *cur; 209 xfs_agblock_t fbno; 210 xfs_extlen_t flen; 211 int error; 212 213 *stat = false; 214 if (!xfs_has_reflink(mp)) 215 return 0; 216 /* rt files will have no perag structure */ 217 if (!info->pag) 218 return 0; 219 220 /* Are there any shared blocks here? */ 221 flen = 0; 222 cur = xfs_refcountbt_init_cursor(mp, tp, info->agf_bp, info->pag); 223 224 error = xfs_refcount_find_shared(cur, rec->rm_startblock, 225 rec->rm_blockcount, &fbno, &flen, false); 226 227 xfs_btree_del_cursor(cur, error); 228 if (error) 229 return error; 230 231 *stat = flen > 0; 232 return 0; 233} 234 235static inline void 236xfs_getfsmap_format( 237 struct xfs_mount *mp, 238 struct xfs_fsmap *xfm, 239 struct xfs_getfsmap_info *info) 240{ 241 struct fsmap *rec; 242 243 trace_xfs_getfsmap_mapping(mp, xfm); 244 245 rec = &info->fsmap_recs[info->head->fmh_entries++]; 246 xfs_fsmap_from_internal(rec, xfm); 247} 248 249static inline bool 250xfs_getfsmap_rec_before_start( 251 struct xfs_getfsmap_info *info, 252 const struct xfs_rmap_irec *rec, 253 xfs_daddr_t rec_daddr) 254{ 255 if (info->low_daddr != -1ULL) 256 return rec_daddr < info->low_daddr; 257 if (info->low.rm_blockcount) 258 return xfs_rmap_compare(rec, &info->low) < 0; 259 return false; 260} 261 262/* 263 * Format a reverse mapping for getfsmap, having translated rm_startblock 264 * into the appropriate daddr units. Pass in a nonzero @len_daddr if the 265 * length could be larger than rm_blockcount in struct xfs_rmap_irec. 266 */ 267STATIC int 268xfs_getfsmap_helper( 269 struct xfs_trans *tp, 270 struct xfs_getfsmap_info *info, 271 const struct xfs_rmap_irec *rec, 272 xfs_daddr_t rec_daddr, 273 xfs_daddr_t len_daddr) 274{ 275 struct xfs_fsmap fmr; 276 struct xfs_mount *mp = tp->t_mountp; 277 bool shared; 278 int error; 279 280 if (fatal_signal_pending(current)) 281 return -EINTR; 282 283 if (len_daddr == 0) 284 len_daddr = XFS_FSB_TO_BB(mp, rec->rm_blockcount); 285 286 /* 287 * Filter out records that start before our startpoint, if the 288 * caller requested that. 289 */ 290 if (xfs_getfsmap_rec_before_start(info, rec, rec_daddr)) { 291 rec_daddr += len_daddr; 292 if (info->next_daddr < rec_daddr) 293 info->next_daddr = rec_daddr; 294 return 0; 295 } 296 297 /* Are we just counting mappings? */ 298 if (info->head->fmh_count == 0) { 299 if (info->head->fmh_entries == UINT_MAX) 300 return -ECANCELED; 301 302 if (rec_daddr > info->next_daddr) 303 info->head->fmh_entries++; 304 305 if (info->last) 306 return 0; 307 308 info->head->fmh_entries++; 309 310 rec_daddr += len_daddr; 311 if (info->next_daddr < rec_daddr) 312 info->next_daddr = rec_daddr; 313 return 0; 314 } 315 316 /* 317 * If the record starts past the last physical block we saw, 318 * then we've found a gap. Report the gap as being owned by 319 * whatever the caller specified is the missing owner. 320 */ 321 if (rec_daddr > info->next_daddr) { 322 if (info->head->fmh_entries >= info->head->fmh_count) 323 return -ECANCELED; 324 325 fmr.fmr_device = info->dev; 326 fmr.fmr_physical = info->next_daddr; 327 fmr.fmr_owner = info->missing_owner; 328 fmr.fmr_offset = 0; 329 fmr.fmr_length = rec_daddr - info->next_daddr; 330 fmr.fmr_flags = FMR_OF_SPECIAL_OWNER; 331 xfs_getfsmap_format(mp, &fmr, info); 332 } 333 334 if (info->last) 335 goto out; 336 337 /* Fill out the extent we found */ 338 if (info->head->fmh_entries >= info->head->fmh_count) 339 return -ECANCELED; 340 341 trace_xfs_fsmap_mapping(mp, info->dev, 342 info->pag ? info->pag->pag_agno : NULLAGNUMBER, rec); 343 344 fmr.fmr_device = info->dev; 345 fmr.fmr_physical = rec_daddr; 346 error = xfs_fsmap_owner_from_rmap(&fmr, rec); 347 if (error) 348 return error; 349 fmr.fmr_offset = XFS_FSB_TO_BB(mp, rec->rm_offset); 350 fmr.fmr_length = len_daddr; 351 if (rec->rm_flags & XFS_RMAP_UNWRITTEN) 352 fmr.fmr_flags |= FMR_OF_PREALLOC; 353 if (rec->rm_flags & XFS_RMAP_ATTR_FORK) 354 fmr.fmr_flags |= FMR_OF_ATTR_FORK; 355 if (rec->rm_flags & XFS_RMAP_BMBT_BLOCK) 356 fmr.fmr_flags |= FMR_OF_EXTENT_MAP; 357 if (fmr.fmr_flags == 0) { 358 error = xfs_getfsmap_is_shared(tp, info, rec, &shared); 359 if (error) 360 return error; 361 if (shared) 362 fmr.fmr_flags |= FMR_OF_SHARED; 363 } 364 365 xfs_getfsmap_format(mp, &fmr, info); 366out: 367 rec_daddr += len_daddr; 368 if (info->next_daddr < rec_daddr) 369 info->next_daddr = rec_daddr; 370 return 0; 371} 372 373/* Transform a rmapbt irec into a fsmap */ 374STATIC int 375xfs_getfsmap_datadev_helper( 376 struct xfs_btree_cur *cur, 377 const struct xfs_rmap_irec *rec, 378 void *priv) 379{ 380 struct xfs_mount *mp = cur->bc_mp; 381 struct xfs_getfsmap_info *info = priv; 382 xfs_fsblock_t fsb; 383 xfs_daddr_t rec_daddr; 384 385 fsb = XFS_AGB_TO_FSB(mp, cur->bc_ag.pag->pag_agno, rec->rm_startblock); 386 rec_daddr = XFS_FSB_TO_DADDR(mp, fsb); 387 388 return xfs_getfsmap_helper(cur->bc_tp, info, rec, rec_daddr, 0); 389} 390 391/* Transform a bnobt irec into a fsmap */ 392STATIC int 393xfs_getfsmap_datadev_bnobt_helper( 394 struct xfs_btree_cur *cur, 395 const struct xfs_alloc_rec_incore *rec, 396 void *priv) 397{ 398 struct xfs_mount *mp = cur->bc_mp; 399 struct xfs_getfsmap_info *info = priv; 400 struct xfs_rmap_irec irec; 401 xfs_daddr_t rec_daddr; 402 403 rec_daddr = XFS_AGB_TO_DADDR(mp, cur->bc_ag.pag->pag_agno, 404 rec->ar_startblock); 405 406 irec.rm_startblock = rec->ar_startblock; 407 irec.rm_blockcount = rec->ar_blockcount; 408 irec.rm_owner = XFS_RMAP_OWN_NULL; /* "free" */ 409 irec.rm_offset = 0; 410 irec.rm_flags = 0; 411 412 return xfs_getfsmap_helper(cur->bc_tp, info, &irec, rec_daddr, 0); 413} 414 415/* Set rmap flags based on the getfsmap flags */ 416static void 417xfs_getfsmap_set_irec_flags( 418 struct xfs_rmap_irec *irec, 419 const struct xfs_fsmap *fmr) 420{ 421 irec->rm_flags = 0; 422 if (fmr->fmr_flags & FMR_OF_ATTR_FORK) 423 irec->rm_flags |= XFS_RMAP_ATTR_FORK; 424 if (fmr->fmr_flags & FMR_OF_EXTENT_MAP) 425 irec->rm_flags |= XFS_RMAP_BMBT_BLOCK; 426 if (fmr->fmr_flags & FMR_OF_PREALLOC) 427 irec->rm_flags |= XFS_RMAP_UNWRITTEN; 428} 429 430/* Execute a getfsmap query against the log device. */ 431STATIC int 432xfs_getfsmap_logdev( 433 struct xfs_trans *tp, 434 const struct xfs_fsmap *keys, 435 struct xfs_getfsmap_info *info) 436{ 437 struct xfs_mount *mp = tp->t_mountp; 438 struct xfs_rmap_irec rmap; 439 xfs_daddr_t rec_daddr, len_daddr; 440 xfs_fsblock_t start_fsb, end_fsb; 441 uint64_t eofs; 442 443 eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_logblocks); 444 if (keys[0].fmr_physical >= eofs) 445 return 0; 446 start_fsb = XFS_BB_TO_FSBT(mp, 447 keys[0].fmr_physical + keys[0].fmr_length); 448 end_fsb = XFS_BB_TO_FSB(mp, min(eofs - 1, keys[1].fmr_physical)); 449 450 /* Adjust the low key if we are continuing from where we left off. */ 451 if (keys[0].fmr_length > 0) 452 info->low_daddr = XFS_FSB_TO_BB(mp, start_fsb); 453 454 trace_xfs_fsmap_low_key_linear(mp, info->dev, start_fsb); 455 trace_xfs_fsmap_high_key_linear(mp, info->dev, end_fsb); 456 457 if (start_fsb > 0) 458 return 0; 459 460 /* Fabricate an rmap entry for the external log device. */ 461 rmap.rm_startblock = 0; 462 rmap.rm_blockcount = mp->m_sb.sb_logblocks; 463 rmap.rm_owner = XFS_RMAP_OWN_LOG; 464 rmap.rm_offset = 0; 465 rmap.rm_flags = 0; 466 467 rec_daddr = XFS_FSB_TO_BB(mp, rmap.rm_startblock); 468 len_daddr = XFS_FSB_TO_BB(mp, rmap.rm_blockcount); 469 return xfs_getfsmap_helper(tp, info, &rmap, rec_daddr, len_daddr); 470} 471 472#ifdef CONFIG_XFS_RT 473/* Transform a rtbitmap "record" into a fsmap */ 474STATIC int 475xfs_getfsmap_rtdev_rtbitmap_helper( 476 struct xfs_mount *mp, 477 struct xfs_trans *tp, 478 const struct xfs_rtalloc_rec *rec, 479 void *priv) 480{ 481 struct xfs_getfsmap_info *info = priv; 482 struct xfs_rmap_irec irec; 483 xfs_rtblock_t rtbno; 484 xfs_daddr_t rec_daddr, len_daddr; 485 486 rtbno = rec->ar_startext * mp->m_sb.sb_rextsize; 487 rec_daddr = XFS_FSB_TO_BB(mp, rtbno); 488 irec.rm_startblock = rtbno; 489 490 rtbno = rec->ar_extcount * mp->m_sb.sb_rextsize; 491 len_daddr = XFS_FSB_TO_BB(mp, rtbno); 492 irec.rm_blockcount = rtbno; 493 494 irec.rm_owner = XFS_RMAP_OWN_NULL; /* "free" */ 495 irec.rm_offset = 0; 496 irec.rm_flags = 0; 497 498 return xfs_getfsmap_helper(tp, info, &irec, rec_daddr, len_daddr); 499} 500 501/* Execute a getfsmap query against the realtime device rtbitmap. */ 502STATIC int 503xfs_getfsmap_rtdev_rtbitmap( 504 struct xfs_trans *tp, 505 const struct xfs_fsmap *keys, 506 struct xfs_getfsmap_info *info) 507{ 508 509 struct xfs_rtalloc_rec alow = { 0 }; 510 struct xfs_rtalloc_rec ahigh = { 0 }; 511 struct xfs_mount *mp = tp->t_mountp; 512 xfs_rtblock_t start_rtb; 513 xfs_rtblock_t end_rtb; 514 uint64_t eofs; 515 int error; 516 517 eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_rextents * mp->m_sb.sb_rextsize); 518 if (keys[0].fmr_physical >= eofs) 519 return 0; 520 start_rtb = XFS_BB_TO_FSBT(mp, 521 keys[0].fmr_physical + keys[0].fmr_length); 522 end_rtb = XFS_BB_TO_FSB(mp, min(eofs - 1, keys[1].fmr_physical)); 523 524 info->missing_owner = XFS_FMR_OWN_UNKNOWN; 525 526 /* Adjust the low key if we are continuing from where we left off. */ 527 if (keys[0].fmr_length > 0) { 528 info->low_daddr = XFS_FSB_TO_BB(mp, start_rtb); 529 if (info->low_daddr >= eofs) 530 return 0; 531 } 532 533 trace_xfs_fsmap_low_key_linear(mp, info->dev, start_rtb); 534 trace_xfs_fsmap_high_key_linear(mp, info->dev, end_rtb); 535 536 xfs_ilock(mp->m_rbmip, XFS_ILOCK_SHARED | XFS_ILOCK_RTBITMAP); 537 538 /* 539 * Set up query parameters to return free rtextents covering the range 540 * we want. 541 */ 542 alow.ar_startext = start_rtb; 543 ahigh.ar_startext = end_rtb; 544 do_div(alow.ar_startext, mp->m_sb.sb_rextsize); 545 if (do_div(ahigh.ar_startext, mp->m_sb.sb_rextsize)) 546 ahigh.ar_startext++; 547 error = xfs_rtalloc_query_range(mp, tp, &alow, &ahigh, 548 xfs_getfsmap_rtdev_rtbitmap_helper, info); 549 if (error) 550 goto err; 551 552 /* 553 * Report any gaps at the end of the rtbitmap by simulating a null 554 * rmap starting at the block after the end of the query range. 555 */ 556 info->last = true; 557 ahigh.ar_startext = min(mp->m_sb.sb_rextents, ahigh.ar_startext); 558 559 error = xfs_getfsmap_rtdev_rtbitmap_helper(mp, tp, &ahigh, info); 560 if (error) 561 goto err; 562err: 563 xfs_iunlock(mp->m_rbmip, XFS_ILOCK_SHARED | XFS_ILOCK_RTBITMAP); 564 return error; 565} 566#endif /* CONFIG_XFS_RT */ 567 568static inline bool 569rmap_not_shareable(struct xfs_mount *mp, const struct xfs_rmap_irec *r) 570{ 571 if (!xfs_has_reflink(mp)) 572 return true; 573 if (XFS_RMAP_NON_INODE_OWNER(r->rm_owner)) 574 return true; 575 if (r->rm_flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK | 576 XFS_RMAP_UNWRITTEN)) 577 return true; 578 return false; 579} 580 581/* Execute a getfsmap query against the regular data device. */ 582STATIC int 583__xfs_getfsmap_datadev( 584 struct xfs_trans *tp, 585 const struct xfs_fsmap *keys, 586 struct xfs_getfsmap_info *info, 587 int (*query_fn)(struct xfs_trans *, 588 struct xfs_getfsmap_info *, 589 struct xfs_btree_cur **, 590 void *), 591 void *priv) 592{ 593 struct xfs_mount *mp = tp->t_mountp; 594 struct xfs_perag *pag; 595 struct xfs_btree_cur *bt_cur = NULL; 596 xfs_fsblock_t start_fsb; 597 xfs_fsblock_t end_fsb; 598 xfs_agnumber_t start_ag; 599 xfs_agnumber_t end_ag; 600 uint64_t eofs; 601 int error = 0; 602 603 eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks); 604 if (keys[0].fmr_physical >= eofs) 605 return 0; 606 start_fsb = XFS_DADDR_TO_FSB(mp, keys[0].fmr_physical); 607 end_fsb = XFS_DADDR_TO_FSB(mp, min(eofs - 1, keys[1].fmr_physical)); 608 609 /* 610 * Convert the fsmap low/high keys to AG based keys. Initialize 611 * low to the fsmap low key and max out the high key to the end 612 * of the AG. 613 */ 614 info->low.rm_offset = XFS_BB_TO_FSBT(mp, keys[0].fmr_offset); 615 error = xfs_fsmap_owner_to_rmap(&info->low, &keys[0]); 616 if (error) 617 return error; 618 info->low.rm_blockcount = XFS_BB_TO_FSBT(mp, keys[0].fmr_length); 619 xfs_getfsmap_set_irec_flags(&info->low, &keys[0]); 620 621 /* Adjust the low key if we are continuing from where we left off. */ 622 if (info->low.rm_blockcount == 0) { 623 /* No previous record from which to continue */ 624 } else if (rmap_not_shareable(mp, &info->low)) { 625 /* Last record seen was an unshareable extent */ 626 info->low.rm_owner = 0; 627 info->low.rm_offset = 0; 628 629 start_fsb += info->low.rm_blockcount; 630 if (XFS_FSB_TO_DADDR(mp, start_fsb) >= eofs) 631 return 0; 632 } else { 633 /* Last record seen was a shareable file data extent */ 634 info->low.rm_offset += info->low.rm_blockcount; 635 } 636 info->low.rm_startblock = XFS_FSB_TO_AGBNO(mp, start_fsb); 637 638 info->high.rm_startblock = -1U; 639 info->high.rm_owner = ULLONG_MAX; 640 info->high.rm_offset = ULLONG_MAX; 641 info->high.rm_blockcount = 0; 642 info->high.rm_flags = XFS_RMAP_KEY_FLAGS | XFS_RMAP_REC_FLAGS; 643 644 start_ag = XFS_FSB_TO_AGNO(mp, start_fsb); 645 end_ag = XFS_FSB_TO_AGNO(mp, end_fsb); 646 647 for_each_perag_range(mp, start_ag, end_ag, pag) { 648 /* 649 * Set the AG high key from the fsmap high key if this 650 * is the last AG that we're querying. 651 */ 652 info->pag = pag; 653 if (pag->pag_agno == end_ag) { 654 info->high.rm_startblock = XFS_FSB_TO_AGBNO(mp, 655 end_fsb); 656 info->high.rm_offset = XFS_BB_TO_FSBT(mp, 657 keys[1].fmr_offset); 658 error = xfs_fsmap_owner_to_rmap(&info->high, &keys[1]); 659 if (error) 660 break; 661 xfs_getfsmap_set_irec_flags(&info->high, &keys[1]); 662 } 663 664 if (bt_cur) { 665 xfs_btree_del_cursor(bt_cur, XFS_BTREE_NOERROR); 666 bt_cur = NULL; 667 xfs_trans_brelse(tp, info->agf_bp); 668 info->agf_bp = NULL; 669 } 670 671 error = xfs_alloc_read_agf(pag, tp, 0, &info->agf_bp); 672 if (error) 673 break; 674 675 trace_xfs_fsmap_low_key(mp, info->dev, pag->pag_agno, 676 &info->low); 677 trace_xfs_fsmap_high_key(mp, info->dev, pag->pag_agno, 678 &info->high); 679 680 error = query_fn(tp, info, &bt_cur, priv); 681 if (error) 682 break; 683 684 /* 685 * Set the AG low key to the start of the AG prior to 686 * moving on to the next AG. 687 */ 688 if (pag->pag_agno == start_ag) 689 memset(&info->low, 0, sizeof(info->low)); 690 691 /* 692 * If this is the last AG, report any gap at the end of it 693 * before we drop the reference to the perag when the loop 694 * terminates. 695 */ 696 if (pag->pag_agno == end_ag) { 697 info->last = true; 698 error = query_fn(tp, info, &bt_cur, priv); 699 if (error) 700 break; 701 } 702 info->pag = NULL; 703 } 704 705 if (bt_cur) 706 xfs_btree_del_cursor(bt_cur, error < 0 ? XFS_BTREE_ERROR : 707 XFS_BTREE_NOERROR); 708 if (info->agf_bp) { 709 xfs_trans_brelse(tp, info->agf_bp); 710 info->agf_bp = NULL; 711 } 712 if (info->pag) { 713 xfs_perag_rele(info->pag); 714 info->pag = NULL; 715 } else if (pag) { 716 /* loop termination case */ 717 xfs_perag_rele(pag); 718 } 719 720 return error; 721} 722 723/* Actually query the rmap btree. */ 724STATIC int 725xfs_getfsmap_datadev_rmapbt_query( 726 struct xfs_trans *tp, 727 struct xfs_getfsmap_info *info, 728 struct xfs_btree_cur **curpp, 729 void *priv) 730{ 731 /* Report any gap at the end of the last AG. */ 732 if (info->last) 733 return xfs_getfsmap_datadev_helper(*curpp, &info->high, info); 734 735 /* Allocate cursor for this AG and query_range it. */ 736 *curpp = xfs_rmapbt_init_cursor(tp->t_mountp, tp, info->agf_bp, 737 info->pag); 738 return xfs_rmap_query_range(*curpp, &info->low, &info->high, 739 xfs_getfsmap_datadev_helper, info); 740} 741 742/* Execute a getfsmap query against the regular data device rmapbt. */ 743STATIC int 744xfs_getfsmap_datadev_rmapbt( 745 struct xfs_trans *tp, 746 const struct xfs_fsmap *keys, 747 struct xfs_getfsmap_info *info) 748{ 749 info->missing_owner = XFS_FMR_OWN_FREE; 750 return __xfs_getfsmap_datadev(tp, keys, info, 751 xfs_getfsmap_datadev_rmapbt_query, NULL); 752} 753 754/* Actually query the bno btree. */ 755STATIC int 756xfs_getfsmap_datadev_bnobt_query( 757 struct xfs_trans *tp, 758 struct xfs_getfsmap_info *info, 759 struct xfs_btree_cur **curpp, 760 void *priv) 761{ 762 struct xfs_alloc_rec_incore *key = priv; 763 764 /* Report any gap at the end of the last AG. */ 765 if (info->last) 766 return xfs_getfsmap_datadev_bnobt_helper(*curpp, &key[1], info); 767 768 /* Allocate cursor for this AG and query_range it. */ 769 *curpp = xfs_allocbt_init_cursor(tp->t_mountp, tp, info->agf_bp, 770 info->pag, XFS_BTNUM_BNO); 771 key->ar_startblock = info->low.rm_startblock; 772 key[1].ar_startblock = info->high.rm_startblock; 773 return xfs_alloc_query_range(*curpp, key, &key[1], 774 xfs_getfsmap_datadev_bnobt_helper, info); 775} 776 777/* Execute a getfsmap query against the regular data device's bnobt. */ 778STATIC int 779xfs_getfsmap_datadev_bnobt( 780 struct xfs_trans *tp, 781 const struct xfs_fsmap *keys, 782 struct xfs_getfsmap_info *info) 783{ 784 struct xfs_alloc_rec_incore akeys[2]; 785 786 memset(akeys, 0, sizeof(akeys)); 787 info->missing_owner = XFS_FMR_OWN_UNKNOWN; 788 return __xfs_getfsmap_datadev(tp, keys, info, 789 xfs_getfsmap_datadev_bnobt_query, &akeys[0]); 790} 791 792/* Do we recognize the device? */ 793STATIC bool 794xfs_getfsmap_is_valid_device( 795 struct xfs_mount *mp, 796 struct xfs_fsmap *fm) 797{ 798 if (fm->fmr_device == 0 || fm->fmr_device == UINT_MAX || 799 fm->fmr_device == new_encode_dev(mp->m_ddev_targp->bt_dev)) 800 return true; 801 if (mp->m_logdev_targp && 802 fm->fmr_device == new_encode_dev(mp->m_logdev_targp->bt_dev)) 803 return true; 804 if (mp->m_rtdev_targp && 805 fm->fmr_device == new_encode_dev(mp->m_rtdev_targp->bt_dev)) 806 return true; 807 return false; 808} 809 810/* Ensure that the low key is less than the high key. */ 811STATIC bool 812xfs_getfsmap_check_keys( 813 struct xfs_fsmap *low_key, 814 struct xfs_fsmap *high_key) 815{ 816 if (low_key->fmr_flags & (FMR_OF_SPECIAL_OWNER | FMR_OF_EXTENT_MAP)) { 817 if (low_key->fmr_offset) 818 return false; 819 } 820 if (high_key->fmr_flags != -1U && 821 (high_key->fmr_flags & (FMR_OF_SPECIAL_OWNER | 822 FMR_OF_EXTENT_MAP))) { 823 if (high_key->fmr_offset && high_key->fmr_offset != -1ULL) 824 return false; 825 } 826 if (high_key->fmr_length && high_key->fmr_length != -1ULL) 827 return false; 828 829 if (low_key->fmr_device > high_key->fmr_device) 830 return false; 831 if (low_key->fmr_device < high_key->fmr_device) 832 return true; 833 834 if (low_key->fmr_physical > high_key->fmr_physical) 835 return false; 836 if (low_key->fmr_physical < high_key->fmr_physical) 837 return true; 838 839 if (low_key->fmr_owner > high_key->fmr_owner) 840 return false; 841 if (low_key->fmr_owner < high_key->fmr_owner) 842 return true; 843 844 if (low_key->fmr_offset > high_key->fmr_offset) 845 return false; 846 if (low_key->fmr_offset < high_key->fmr_offset) 847 return true; 848 849 return false; 850} 851 852/* 853 * There are only two devices if we didn't configure RT devices at build time. 854 */ 855#ifdef CONFIG_XFS_RT 856#define XFS_GETFSMAP_DEVS 3 857#else 858#define XFS_GETFSMAP_DEVS 2 859#endif /* CONFIG_XFS_RT */ 860 861/* 862 * Get filesystem's extents as described in head, and format for output. Fills 863 * in the supplied records array until there are no more reverse mappings to 864 * return or head.fmh_entries == head.fmh_count. In the second case, this 865 * function returns -ECANCELED to indicate that more records would have been 866 * returned. 867 * 868 * Key to Confusion 869 * ---------------- 870 * There are multiple levels of keys and counters at work here: 871 * xfs_fsmap_head.fmh_keys -- low and high fsmap keys passed in; 872 * these reflect fs-wide sector addrs. 873 * dkeys -- fmh_keys used to query each device; 874 * these are fmh_keys but w/ the low key 875 * bumped up by fmr_length. 876 * xfs_getfsmap_info.next_daddr -- next disk addr we expect to see; this 877 * is how we detect gaps in the fsmap 878 records and report them. 879 * xfs_getfsmap_info.low/high -- per-AG low/high keys computed from 880 * dkeys; used to query the metadata. 881 */ 882int 883xfs_getfsmap( 884 struct xfs_mount *mp, 885 struct xfs_fsmap_head *head, 886 struct fsmap *fsmap_recs) 887{ 888 struct xfs_trans *tp = NULL; 889 struct xfs_fsmap dkeys[2]; /* per-dev keys */ 890 struct xfs_getfsmap_dev handlers[XFS_GETFSMAP_DEVS]; 891 struct xfs_getfsmap_info info = { NULL }; 892 bool use_rmap; 893 int i; 894 int error = 0; 895 896 if (head->fmh_iflags & ~FMH_IF_VALID) 897 return -EINVAL; 898 if (!xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[0]) || 899 !xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[1])) 900 return -EINVAL; 901 if (!xfs_getfsmap_check_keys(&head->fmh_keys[0], &head->fmh_keys[1])) 902 return -EINVAL; 903 904 use_rmap = xfs_has_rmapbt(mp) && 905 has_capability_noaudit(current, CAP_SYS_ADMIN); 906 head->fmh_entries = 0; 907 908 /* Set up our device handlers. */ 909 memset(handlers, 0, sizeof(handlers)); 910 handlers[0].dev = new_encode_dev(mp->m_ddev_targp->bt_dev); 911 if (use_rmap) 912 handlers[0].fn = xfs_getfsmap_datadev_rmapbt; 913 else 914 handlers[0].fn = xfs_getfsmap_datadev_bnobt; 915 if (mp->m_logdev_targp != mp->m_ddev_targp) { 916 handlers[1].dev = new_encode_dev(mp->m_logdev_targp->bt_dev); 917 handlers[1].fn = xfs_getfsmap_logdev; 918 } 919#ifdef CONFIG_XFS_RT 920 if (mp->m_rtdev_targp) { 921 handlers[2].dev = new_encode_dev(mp->m_rtdev_targp->bt_dev); 922 handlers[2].fn = xfs_getfsmap_rtdev_rtbitmap; 923 } 924#endif /* CONFIG_XFS_RT */ 925 926 xfs_sort(handlers, XFS_GETFSMAP_DEVS, sizeof(struct xfs_getfsmap_dev), 927 xfs_getfsmap_dev_compare); 928 929 /* 930 * To continue where we left off, we allow userspace to use the 931 * last mapping from a previous call as the low key of the next. 932 * This is identified by a non-zero length in the low key. We 933 * have to increment the low key in this scenario to ensure we 934 * don't return the same mapping again, and instead return the 935 * very next mapping. 936 * 937 * If the low key mapping refers to file data, the same physical 938 * blocks could be mapped to several other files/offsets. 939 * According to rmapbt record ordering, the minimal next 940 * possible record for the block range is the next starting 941 * offset in the same inode. Therefore, each fsmap backend bumps 942 * the file offset to continue the search appropriately. For 943 * all other low key mapping types (attr blocks, metadata), each 944 * fsmap backend bumps the physical offset as there can be no 945 * other mapping for the same physical block range. 946 */ 947 dkeys[0] = head->fmh_keys[0]; 948 memset(&dkeys[1], 0xFF, sizeof(struct xfs_fsmap)); 949 950 info.next_daddr = head->fmh_keys[0].fmr_physical + 951 head->fmh_keys[0].fmr_length; 952 info.fsmap_recs = fsmap_recs; 953 info.head = head; 954 955 /* For each device we support... */ 956 for (i = 0; i < XFS_GETFSMAP_DEVS; i++) { 957 /* Is this device within the range the user asked for? */ 958 if (!handlers[i].fn) 959 continue; 960 if (head->fmh_keys[0].fmr_device > handlers[i].dev) 961 continue; 962 if (head->fmh_keys[1].fmr_device < handlers[i].dev) 963 break; 964 965 /* 966 * If this device number matches the high key, we have 967 * to pass the high key to the handler to limit the 968 * query results. If the device number exceeds the 969 * low key, zero out the low key so that we get 970 * everything from the beginning. 971 */ 972 if (handlers[i].dev == head->fmh_keys[1].fmr_device) 973 dkeys[1] = head->fmh_keys[1]; 974 if (handlers[i].dev > head->fmh_keys[0].fmr_device) 975 memset(&dkeys[0], 0, sizeof(struct xfs_fsmap)); 976 977 /* 978 * Grab an empty transaction so that we can use its recursive 979 * buffer locking abilities to detect cycles in the rmapbt 980 * without deadlocking. 981 */ 982 error = xfs_trans_alloc_empty(mp, &tp); 983 if (error) 984 break; 985 986 info.dev = handlers[i].dev; 987 info.last = false; 988 info.pag = NULL; 989 info.low_daddr = -1ULL; 990 info.low.rm_blockcount = 0; 991 error = handlers[i].fn(tp, dkeys, &info); 992 if (error) 993 break; 994 xfs_trans_cancel(tp); 995 tp = NULL; 996 info.next_daddr = 0; 997 } 998 999 if (tp) 1000 xfs_trans_cancel(tp); 1001 head->fmh_oflags = FMH_OF_DEV_T; 1002 return error; 1003} 1004