1 /* 2 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc. 3 * All Rights Reserved. 4 * 5 * This program is free software; you can redistribute it and/or 6 * modify it under the terms of the GNU General Public License as 7 * published by the Free Software Foundation. 8 * 9 * This program is distributed in the hope that it would be useful, 10 * but WITHOUT ANY WARRANTY; without even the implied warranty of 11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 12 * GNU General Public License for more details. 13 * 14 * You should have received a copy of the GNU General Public License 15 * along with this program; if not, write the Free Software Foundation, 16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA 17 */ 18 #include "xfs.h" 19 #include "xfs_fs.h" 20 #include "xfs_shared.h" 21 #include "xfs_format.h" 22 #include "xfs_log_format.h" 23 #include "xfs_trans_resv.h" 24 #include "xfs_inum.h" 25 #include "xfs_sb.h" 26 #include "xfs_ag.h" 27 #include "xfs_mount.h" 28 #include "xfs_inode.h" 29 #include "xfs_btree.h" 30 #include "xfs_ialloc.h" 31 #include "xfs_ialloc_btree.h" 32 #include "xfs_itable.h" 33 #include "xfs_error.h" 34 #include "xfs_trace.h" 35 #include "xfs_icache.h" 36 #include "xfs_dinode.h" 37 38 STATIC int 39 xfs_internal_inum( 40 xfs_mount_t *mp, 41 xfs_ino_t ino) 42 { 43 return (ino == mp->m_sb.sb_rbmino || ino == mp->m_sb.sb_rsumino || 44 (xfs_sb_version_hasquota(&mp->m_sb) && 45 xfs_is_quota_inode(&mp->m_sb, ino))); 46 } 47 48 /* 49 * Return stat information for one inode. 50 * Return 0 if ok, else errno. 51 */ 52 int 53 xfs_bulkstat_one_int( 54 struct xfs_mount *mp, /* mount point for filesystem */ 55 xfs_ino_t ino, /* inode to get data for */ 56 void __user *buffer, /* buffer to place output in */ 57 int ubsize, /* size of buffer */ 58 bulkstat_one_fmt_pf formatter, /* formatter, copy to user */ 59 int *ubused, /* bytes used by me */ 60 int *stat) /* BULKSTAT_RV_... */ 61 { 62 struct xfs_icdinode *dic; /* dinode core info pointer */ 63 struct xfs_inode *ip; /* incore inode pointer */ 64 struct xfs_bstat *buf; /* return buffer */ 65 int error = 0; /* error value */ 66 67 *stat = BULKSTAT_RV_NOTHING; 68 69 if (!buffer || xfs_internal_inum(mp, ino)) 70 return -EINVAL; 71 72 buf = kmem_alloc(sizeof(*buf), KM_SLEEP | KM_MAYFAIL); 73 if (!buf) 74 return -ENOMEM; 75 76 error = xfs_iget(mp, NULL, ino, 77 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED), 78 XFS_ILOCK_SHARED, &ip); 79 if (error) 80 goto out_free; 81 82 ASSERT(ip != NULL); 83 ASSERT(ip->i_imap.im_blkno != 0); 84 85 dic = &ip->i_d; 86 87 /* xfs_iget returns the following without needing 88 * further change. 89 */ 90 buf->bs_nlink = dic->di_nlink; 91 buf->bs_projid_lo = dic->di_projid_lo; 92 buf->bs_projid_hi = dic->di_projid_hi; 93 buf->bs_ino = ino; 94 buf->bs_mode = dic->di_mode; 95 buf->bs_uid = dic->di_uid; 96 buf->bs_gid = dic->di_gid; 97 buf->bs_size = dic->di_size; 98 buf->bs_atime.tv_sec = dic->di_atime.t_sec; 99 buf->bs_atime.tv_nsec = dic->di_atime.t_nsec; 100 buf->bs_mtime.tv_sec = dic->di_mtime.t_sec; 101 buf->bs_mtime.tv_nsec = dic->di_mtime.t_nsec; 102 buf->bs_ctime.tv_sec = dic->di_ctime.t_sec; 103 buf->bs_ctime.tv_nsec = dic->di_ctime.t_nsec; 104 buf->bs_xflags = xfs_ip2xflags(ip); 105 buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog; 106 buf->bs_extents = dic->di_nextents; 107 buf->bs_gen = dic->di_gen; 108 memset(buf->bs_pad, 0, sizeof(buf->bs_pad)); 109 buf->bs_dmevmask = dic->di_dmevmask; 110 buf->bs_dmstate = dic->di_dmstate; 111 buf->bs_aextents = dic->di_anextents; 112 buf->bs_forkoff = XFS_IFORK_BOFF(ip); 113 114 switch (dic->di_format) { 115 case XFS_DINODE_FMT_DEV: 116 buf->bs_rdev = ip->i_df.if_u2.if_rdev; 117 buf->bs_blksize = BLKDEV_IOSIZE; 118 buf->bs_blocks = 0; 119 break; 120 case XFS_DINODE_FMT_LOCAL: 121 case XFS_DINODE_FMT_UUID: 122 buf->bs_rdev = 0; 123 buf->bs_blksize = mp->m_sb.sb_blocksize; 124 buf->bs_blocks = 0; 125 break; 126 case XFS_DINODE_FMT_EXTENTS: 127 case XFS_DINODE_FMT_BTREE: 128 buf->bs_rdev = 0; 129 buf->bs_blksize = mp->m_sb.sb_blocksize; 130 buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks; 131 break; 132 } 133 xfs_iunlock(ip, XFS_ILOCK_SHARED); 134 IRELE(ip); 135 136 error = formatter(buffer, ubsize, ubused, buf); 137 if (!error) 138 *stat = BULKSTAT_RV_DIDONE; 139 140 out_free: 141 kmem_free(buf); 142 return error; 143 } 144 145 /* Return 0 on success or positive error */ 146 STATIC int 147 xfs_bulkstat_one_fmt( 148 void __user *ubuffer, 149 int ubsize, 150 int *ubused, 151 const xfs_bstat_t *buffer) 152 { 153 if (ubsize < sizeof(*buffer)) 154 return -ENOMEM; 155 if (copy_to_user(ubuffer, buffer, sizeof(*buffer))) 156 return -EFAULT; 157 if (ubused) 158 *ubused = sizeof(*buffer); 159 return 0; 160 } 161 162 int 163 xfs_bulkstat_one( 164 xfs_mount_t *mp, /* mount point for filesystem */ 165 xfs_ino_t ino, /* inode number to get data for */ 166 void __user *buffer, /* buffer to place output in */ 167 int ubsize, /* size of buffer */ 168 int *ubused, /* bytes used by me */ 169 int *stat) /* BULKSTAT_RV_... */ 170 { 171 return xfs_bulkstat_one_int(mp, ino, buffer, ubsize, 172 xfs_bulkstat_one_fmt, ubused, stat); 173 } 174 175 /* 176 * Loop over all clusters in a chunk for a given incore inode allocation btree 177 * record. Do a readahead if there are any allocated inodes in that cluster. 178 */ 179 STATIC void 180 xfs_bulkstat_ichunk_ra( 181 struct xfs_mount *mp, 182 xfs_agnumber_t agno, 183 struct xfs_inobt_rec_incore *irec) 184 { 185 xfs_agblock_t agbno; 186 struct blk_plug plug; 187 int blks_per_cluster; 188 int inodes_per_cluster; 189 int i; /* inode chunk index */ 190 191 agbno = XFS_AGINO_TO_AGBNO(mp, irec->ir_startino); 192 blks_per_cluster = xfs_icluster_size_fsb(mp); 193 inodes_per_cluster = blks_per_cluster << mp->m_sb.sb_inopblog; 194 195 blk_start_plug(&plug); 196 for (i = 0; i < XFS_INODES_PER_CHUNK; 197 i += inodes_per_cluster, agbno += blks_per_cluster) { 198 if (xfs_inobt_maskn(i, inodes_per_cluster) & ~irec->ir_free) { 199 xfs_btree_reada_bufs(mp, agno, agbno, blks_per_cluster, 200 &xfs_inode_buf_ops); 201 } 202 } 203 blk_finish_plug(&plug); 204 } 205 206 /* 207 * Lookup the inode chunk that the given inode lives in and then get the record 208 * if we found the chunk. If the inode was not the last in the chunk and there 209 * are some left allocated, update the data for the pointed-to record as well as 210 * return the count of grabbed inodes. 211 */ 212 STATIC int 213 xfs_bulkstat_grab_ichunk( 214 struct xfs_btree_cur *cur, /* btree cursor */ 215 xfs_agino_t agino, /* starting inode of chunk */ 216 int *icount,/* return # of inodes grabbed */ 217 struct xfs_inobt_rec_incore *irec) /* btree record */ 218 { 219 int idx; /* index into inode chunk */ 220 int stat; 221 int error = 0; 222 223 /* Lookup the inode chunk that this inode lives in */ 224 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &stat); 225 if (error) 226 return error; 227 if (!stat) { 228 *icount = 0; 229 return error; 230 } 231 232 /* Get the record, should always work */ 233 error = xfs_inobt_get_rec(cur, irec, &stat); 234 if (error) 235 return error; 236 XFS_WANT_CORRUPTED_RETURN(stat == 1); 237 238 /* Check if the record contains the inode in request */ 239 if (irec->ir_startino + XFS_INODES_PER_CHUNK <= agino) 240 return -EINVAL; 241 242 idx = agino - irec->ir_startino + 1; 243 if (idx < XFS_INODES_PER_CHUNK && 244 (xfs_inobt_maskn(idx, XFS_INODES_PER_CHUNK - idx) & ~irec->ir_free)) { 245 int i; 246 247 /* We got a right chunk with some left inodes allocated at it. 248 * Grab the chunk record. Mark all the uninteresting inodes 249 * free -- because they're before our start point. 250 */ 251 for (i = 0; i < idx; i++) { 252 if (XFS_INOBT_MASK(i) & ~irec->ir_free) 253 irec->ir_freecount++; 254 } 255 256 irec->ir_free |= xfs_inobt_maskn(0, idx); 257 *icount = XFS_INODES_PER_CHUNK - irec->ir_freecount; 258 } 259 260 return 0; 261 } 262 263 #define XFS_BULKSTAT_UBLEFT(ubleft) ((ubleft) >= statstruct_size) 264 265 struct xfs_bulkstat_agichunk { 266 char __user **ac_ubuffer;/* pointer into user's buffer */ 267 int ac_ubleft; /* bytes left in user's buffer */ 268 int ac_ubelem; /* spaces used in user's buffer */ 269 }; 270 271 /* 272 * Process inodes in chunk with a pointer to a formatter function 273 * that will iget the inode and fill in the appropriate structure. 274 */ 275 static int 276 xfs_bulkstat_ag_ichunk( 277 struct xfs_mount *mp, 278 xfs_agnumber_t agno, 279 struct xfs_inobt_rec_incore *irbp, 280 bulkstat_one_pf formatter, 281 size_t statstruct_size, 282 struct xfs_bulkstat_agichunk *acp, 283 xfs_ino_t *lastino) 284 { 285 char __user **ubufp = acp->ac_ubuffer; 286 int chunkidx; 287 int error = 0; 288 xfs_agino_t agino; 289 290 agino = irbp->ir_startino; 291 for (chunkidx = 0; chunkidx < XFS_INODES_PER_CHUNK; 292 chunkidx++, agino++) { 293 int fmterror; 294 int ubused; 295 xfs_ino_t ino = XFS_AGINO_TO_INO(mp, agno, agino); 296 297 /* Skip if this inode is free */ 298 if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free) { 299 *lastino = ino; 300 continue; 301 } 302 303 /* Get the inode and fill in a single buffer */ 304 ubused = statstruct_size; 305 error = formatter(mp, ino, *ubufp, acp->ac_ubleft, 306 &ubused, &fmterror); 307 if (fmterror == BULKSTAT_RV_GIVEUP || 308 (error && error != -ENOENT && error != -EINVAL)) { 309 acp->ac_ubleft = 0; 310 ASSERT(error); 311 break; 312 } 313 314 /* be careful not to leak error if at end of chunk */ 315 if (fmterror == BULKSTAT_RV_NOTHING || error) { 316 *lastino = ino; 317 error = 0; 318 continue; 319 } 320 321 *ubufp += ubused; 322 acp->ac_ubleft -= ubused; 323 acp->ac_ubelem++; 324 *lastino = ino; 325 326 if (acp->ac_ubleft < statstruct_size) 327 break; 328 } 329 330 return error; 331 } 332 333 /* 334 * Return stat information in bulk (by-inode) for the filesystem. 335 */ 336 int /* error status */ 337 xfs_bulkstat( 338 xfs_mount_t *mp, /* mount point for filesystem */ 339 xfs_ino_t *lastinop, /* last inode returned */ 340 int *ubcountp, /* size of buffer/count returned */ 341 bulkstat_one_pf formatter, /* func that'd fill a single buf */ 342 size_t statstruct_size, /* sizeof struct filling */ 343 char __user *ubuffer, /* buffer with inode stats */ 344 int *done) /* 1 if there are more stats to get */ 345 { 346 xfs_buf_t *agbp; /* agi header buffer */ 347 xfs_agi_t *agi; /* agi header data */ 348 xfs_agino_t agino; /* inode # in allocation group */ 349 xfs_agnumber_t agno; /* allocation group number */ 350 xfs_btree_cur_t *cur; /* btree cursor for ialloc btree */ 351 size_t irbsize; /* size of irec buffer in bytes */ 352 xfs_inobt_rec_incore_t *irbuf; /* start of irec buffer */ 353 xfs_ino_t lastino; /* last inode number returned */ 354 int nirbuf; /* size of irbuf */ 355 int rval; /* return value error code */ 356 int ubcount; /* size of user's buffer */ 357 struct xfs_bulkstat_agichunk ac; 358 int error = 0; 359 360 /* 361 * Get the last inode value, see if there's nothing to do. 362 */ 363 lastino = *lastinop; 364 agno = XFS_INO_TO_AGNO(mp, lastino); 365 agino = XFS_INO_TO_AGINO(mp, lastino); 366 if (agno >= mp->m_sb.sb_agcount || 367 lastino != XFS_AGINO_TO_INO(mp, agno, agino)) { 368 *done = 1; 369 *ubcountp = 0; 370 return 0; 371 } 372 373 ubcount = *ubcountp; /* statstruct's */ 374 ac.ac_ubuffer = &ubuffer; 375 ac.ac_ubleft = ubcount * statstruct_size; /* bytes */; 376 ac.ac_ubelem = 0; 377 378 *ubcountp = 0; 379 *done = 0; 380 381 irbuf = kmem_zalloc_greedy(&irbsize, PAGE_SIZE, PAGE_SIZE * 4); 382 if (!irbuf) 383 return -ENOMEM; 384 385 nirbuf = irbsize / sizeof(*irbuf); 386 387 /* 388 * Loop over the allocation groups, starting from the last 389 * inode returned; 0 means start of the allocation group. 390 */ 391 rval = 0; 392 while (agno < mp->m_sb.sb_agcount) { 393 struct xfs_inobt_rec_incore *irbp = irbuf; 394 struct xfs_inobt_rec_incore *irbufend = irbuf + nirbuf; 395 bool end_of_ag = false; 396 int icount = 0; 397 int stat; 398 399 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp); 400 if (error) 401 break; 402 agi = XFS_BUF_TO_AGI(agbp); 403 /* 404 * Allocate and initialize a btree cursor for ialloc btree. 405 */ 406 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno, 407 XFS_BTNUM_INO); 408 if (agino > 0) { 409 /* 410 * In the middle of an allocation group, we need to get 411 * the remainder of the chunk we're in. 412 */ 413 struct xfs_inobt_rec_incore r; 414 415 error = xfs_bulkstat_grab_ichunk(cur, agino, &icount, &r); 416 if (error) 417 goto del_cursor; 418 if (icount) { 419 irbp->ir_startino = r.ir_startino; 420 irbp->ir_freecount = r.ir_freecount; 421 irbp->ir_free = r.ir_free; 422 irbp++; 423 agino = r.ir_startino + XFS_INODES_PER_CHUNK; 424 } 425 /* Increment to the next record */ 426 error = xfs_btree_increment(cur, 0, &stat); 427 } else { 428 /* Start of ag. Lookup the first inode chunk */ 429 error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &stat); 430 } 431 if (error || stat == 0) { 432 end_of_ag = true; 433 goto del_cursor; 434 } 435 436 /* 437 * Loop through inode btree records in this ag, 438 * until we run out of inodes or space in the buffer. 439 */ 440 while (irbp < irbufend && icount < ubcount) { 441 struct xfs_inobt_rec_incore r; 442 443 error = xfs_inobt_get_rec(cur, &r, &stat); 444 if (error || stat == 0) { 445 end_of_ag = true; 446 goto del_cursor; 447 } 448 449 /* 450 * If this chunk has any allocated inodes, save it. 451 * Also start read-ahead now for this chunk. 452 */ 453 if (r.ir_freecount < XFS_INODES_PER_CHUNK) { 454 xfs_bulkstat_ichunk_ra(mp, agno, &r); 455 irbp->ir_startino = r.ir_startino; 456 irbp->ir_freecount = r.ir_freecount; 457 irbp->ir_free = r.ir_free; 458 irbp++; 459 icount += XFS_INODES_PER_CHUNK - r.ir_freecount; 460 } 461 /* 462 * Set agino to after this chunk and bump the cursor. 463 */ 464 agino = r.ir_startino + XFS_INODES_PER_CHUNK; 465 error = xfs_btree_increment(cur, 0, &stat); 466 if (error || stat == 0) { 467 end_of_ag = true; 468 goto del_cursor; 469 } 470 cond_resched(); 471 } 472 473 /* 474 * Drop the btree buffers and the agi buffer as we can't hold any 475 * of the locks these represent when calling iget. If there is a 476 * pending error, then we are done. 477 */ 478 del_cursor: 479 xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR); 480 xfs_buf_relse(agbp); 481 if (error) 482 break; 483 /* 484 * Now format all the good inodes into the user's buffer. 485 */ 486 irbufend = irbp; 487 for (irbp = irbuf; 488 irbp < irbufend && ac.ac_ubleft >= statstruct_size; 489 irbp++) { 490 error = xfs_bulkstat_ag_ichunk(mp, agno, irbp, 491 formatter, statstruct_size, &ac, 492 &lastino); 493 if (error) 494 rval = error; 495 496 cond_resched(); 497 } 498 499 /* If we've run out of space, we are done */ 500 if (ac.ac_ubleft < statstruct_size) 501 break; 502 503 if (end_of_ag) { 504 agno++; 505 agino = 0; 506 } else 507 agino = XFS_INO_TO_AGINO(mp, lastino); 508 } 509 /* 510 * Done, we're either out of filesystem or space to put the data. 511 */ 512 kmem_free(irbuf); 513 *ubcountp = ac.ac_ubelem; 514 /* 515 * Found some inodes, return them now and return the error next time. 516 */ 517 if (ac.ac_ubelem) 518 rval = 0; 519 if (agno >= mp->m_sb.sb_agcount) { 520 /* 521 * If we ran out of filesystem, mark lastino as off 522 * the end of the filesystem, so the next call 523 * will return immediately. 524 */ 525 *lastinop = (xfs_ino_t)XFS_AGINO_TO_INO(mp, agno, 0); 526 *done = 1; 527 } else 528 *lastinop = (xfs_ino_t)lastino; 529 530 return rval; 531 } 532 533 int 534 xfs_inumbers_fmt( 535 void __user *ubuffer, /* buffer to write to */ 536 const struct xfs_inogrp *buffer, /* buffer to read from */ 537 long count, /* # of elements to read */ 538 long *written) /* # of bytes written */ 539 { 540 if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer))) 541 return -EFAULT; 542 *written = count * sizeof(*buffer); 543 return 0; 544 } 545 546 /* 547 * Return inode number table for the filesystem. 548 */ 549 int /* error status */ 550 xfs_inumbers( 551 struct xfs_mount *mp,/* mount point for filesystem */ 552 xfs_ino_t *lastino,/* last inode returned */ 553 int *count,/* size of buffer/count returned */ 554 void __user *ubuffer,/* buffer with inode descriptions */ 555 inumbers_fmt_pf formatter) 556 { 557 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, *lastino); 558 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, *lastino); 559 struct xfs_btree_cur *cur = NULL; 560 struct xfs_buf *agbp = NULL; 561 struct xfs_inogrp *buffer; 562 int bcount; 563 int left = *count; 564 int bufidx = 0; 565 int error = 0; 566 567 *count = 0; 568 if (agno >= mp->m_sb.sb_agcount || 569 *lastino != XFS_AGINO_TO_INO(mp, agno, agino)) 570 return error; 571 572 bcount = MIN(left, (int)(PAGE_SIZE / sizeof(*buffer))); 573 buffer = kmem_alloc(bcount * sizeof(*buffer), KM_SLEEP); 574 do { 575 struct xfs_inobt_rec_incore r; 576 int stat; 577 578 if (!agbp) { 579 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp); 580 if (error) 581 break; 582 583 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno, 584 XFS_BTNUM_INO); 585 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_GE, 586 &stat); 587 if (error) 588 break; 589 if (!stat) 590 goto next_ag; 591 } 592 593 error = xfs_inobt_get_rec(cur, &r, &stat); 594 if (error) 595 break; 596 if (!stat) 597 goto next_ag; 598 599 agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1; 600 buffer[bufidx].xi_startino = 601 XFS_AGINO_TO_INO(mp, agno, r.ir_startino); 602 buffer[bufidx].xi_alloccount = 603 XFS_INODES_PER_CHUNK - r.ir_freecount; 604 buffer[bufidx].xi_allocmask = ~r.ir_free; 605 if (++bufidx == bcount) { 606 long written; 607 608 error = formatter(ubuffer, buffer, bufidx, &written); 609 if (error) 610 break; 611 ubuffer += written; 612 *count += bufidx; 613 bufidx = 0; 614 } 615 if (!--left) 616 break; 617 618 error = xfs_btree_increment(cur, 0, &stat); 619 if (error) 620 break; 621 if (stat) 622 continue; 623 624 next_ag: 625 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); 626 cur = NULL; 627 xfs_buf_relse(agbp); 628 agbp = NULL; 629 agino = 0; 630 agno++; 631 } while (agno < mp->m_sb.sb_agcount); 632 633 if (!error) { 634 if (bufidx) { 635 long written; 636 637 error = formatter(ubuffer, buffer, bufidx, &written); 638 if (!error) 639 *count += bufidx; 640 } 641 *lastino = XFS_AGINO_TO_INO(mp, agno, agino); 642 } 643 644 kmem_free(buffer); 645 if (cur) 646 xfs_btree_del_cursor(cur, (error ? XFS_BTREE_ERROR : 647 XFS_BTREE_NOERROR)); 648 if (agbp) 649 xfs_buf_relse(agbp); 650 651 return error; 652 } 653