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_mount.h" 25 #include "xfs_inode.h" 26 #include "xfs_btree.h" 27 #include "xfs_ialloc.h" 28 #include "xfs_ialloc_btree.h" 29 #include "xfs_itable.h" 30 #include "xfs_error.h" 31 #include "xfs_trace.h" 32 #include "xfs_icache.h" 33 34 STATIC int 35 xfs_internal_inum( 36 xfs_mount_t *mp, 37 xfs_ino_t ino) 38 { 39 return (ino == mp->m_sb.sb_rbmino || ino == mp->m_sb.sb_rsumino || 40 (xfs_sb_version_hasquota(&mp->m_sb) && 41 xfs_is_quota_inode(&mp->m_sb, ino))); 42 } 43 44 /* 45 * Return stat information for one inode. 46 * Return 0 if ok, else errno. 47 */ 48 int 49 xfs_bulkstat_one_int( 50 struct xfs_mount *mp, /* mount point for filesystem */ 51 xfs_ino_t ino, /* inode to get data for */ 52 void __user *buffer, /* buffer to place output in */ 53 int ubsize, /* size of buffer */ 54 bulkstat_one_fmt_pf formatter, /* formatter, copy to user */ 55 int *ubused, /* bytes used by me */ 56 int *stat) /* BULKSTAT_RV_... */ 57 { 58 struct xfs_icdinode *dic; /* dinode core info pointer */ 59 struct xfs_inode *ip; /* incore inode pointer */ 60 struct inode *inode; 61 struct xfs_bstat *buf; /* return buffer */ 62 int error = 0; /* error value */ 63 64 *stat = BULKSTAT_RV_NOTHING; 65 66 if (!buffer || xfs_internal_inum(mp, ino)) 67 return -EINVAL; 68 69 buf = kmem_zalloc(sizeof(*buf), KM_SLEEP | KM_MAYFAIL); 70 if (!buf) 71 return -ENOMEM; 72 73 error = xfs_iget(mp, NULL, ino, 74 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED), 75 XFS_ILOCK_SHARED, &ip); 76 if (error) 77 goto out_free; 78 79 ASSERT(ip != NULL); 80 ASSERT(ip->i_imap.im_blkno != 0); 81 inode = VFS_I(ip); 82 83 dic = &ip->i_d; 84 85 /* xfs_iget returns the following without needing 86 * further change. 87 */ 88 buf->bs_projid_lo = dic->di_projid_lo; 89 buf->bs_projid_hi = dic->di_projid_hi; 90 buf->bs_ino = ino; 91 buf->bs_uid = dic->di_uid; 92 buf->bs_gid = dic->di_gid; 93 buf->bs_size = dic->di_size; 94 95 buf->bs_nlink = inode->i_nlink; 96 buf->bs_atime.tv_sec = inode->i_atime.tv_sec; 97 buf->bs_atime.tv_nsec = inode->i_atime.tv_nsec; 98 buf->bs_mtime.tv_sec = inode->i_mtime.tv_sec; 99 buf->bs_mtime.tv_nsec = inode->i_mtime.tv_nsec; 100 buf->bs_ctime.tv_sec = inode->i_ctime.tv_sec; 101 buf->bs_ctime.tv_nsec = inode->i_ctime.tv_nsec; 102 buf->bs_gen = inode->i_generation; 103 buf->bs_mode = inode->i_mode; 104 105 buf->bs_xflags = xfs_ip2xflags(ip); 106 buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog; 107 buf->bs_extents = dic->di_nextents; 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 if (dic->di_version == 3) { 115 if (dic->di_flags2 & XFS_DIFLAG2_COWEXTSIZE) 116 buf->bs_cowextsize = dic->di_cowextsize << 117 mp->m_sb.sb_blocklog; 118 } 119 120 switch (dic->di_format) { 121 case XFS_DINODE_FMT_DEV: 122 buf->bs_rdev = ip->i_df.if_u2.if_rdev; 123 buf->bs_blksize = BLKDEV_IOSIZE; 124 buf->bs_blocks = 0; 125 break; 126 case XFS_DINODE_FMT_LOCAL: 127 case XFS_DINODE_FMT_UUID: 128 buf->bs_rdev = 0; 129 buf->bs_blksize = mp->m_sb.sb_blocksize; 130 buf->bs_blocks = 0; 131 break; 132 case XFS_DINODE_FMT_EXTENTS: 133 case XFS_DINODE_FMT_BTREE: 134 buf->bs_rdev = 0; 135 buf->bs_blksize = mp->m_sb.sb_blocksize; 136 buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks; 137 break; 138 } 139 xfs_iunlock(ip, XFS_ILOCK_SHARED); 140 IRELE(ip); 141 142 error = formatter(buffer, ubsize, ubused, buf); 143 if (!error) 144 *stat = BULKSTAT_RV_DIDONE; 145 146 out_free: 147 kmem_free(buf); 148 return error; 149 } 150 151 /* Return 0 on success or positive error */ 152 STATIC int 153 xfs_bulkstat_one_fmt( 154 void __user *ubuffer, 155 int ubsize, 156 int *ubused, 157 const xfs_bstat_t *buffer) 158 { 159 if (ubsize < sizeof(*buffer)) 160 return -ENOMEM; 161 if (copy_to_user(ubuffer, buffer, sizeof(*buffer))) 162 return -EFAULT; 163 if (ubused) 164 *ubused = sizeof(*buffer); 165 return 0; 166 } 167 168 int 169 xfs_bulkstat_one( 170 xfs_mount_t *mp, /* mount point for filesystem */ 171 xfs_ino_t ino, /* inode number to get data for */ 172 void __user *buffer, /* buffer to place output in */ 173 int ubsize, /* size of buffer */ 174 int *ubused, /* bytes used by me */ 175 int *stat) /* BULKSTAT_RV_... */ 176 { 177 return xfs_bulkstat_one_int(mp, ino, buffer, ubsize, 178 xfs_bulkstat_one_fmt, ubused, stat); 179 } 180 181 /* 182 * Loop over all clusters in a chunk for a given incore inode allocation btree 183 * record. Do a readahead if there are any allocated inodes in that cluster. 184 */ 185 STATIC void 186 xfs_bulkstat_ichunk_ra( 187 struct xfs_mount *mp, 188 xfs_agnumber_t agno, 189 struct xfs_inobt_rec_incore *irec) 190 { 191 xfs_agblock_t agbno; 192 struct blk_plug plug; 193 int blks_per_cluster; 194 int inodes_per_cluster; 195 int i; /* inode chunk index */ 196 197 agbno = XFS_AGINO_TO_AGBNO(mp, irec->ir_startino); 198 blks_per_cluster = xfs_icluster_size_fsb(mp); 199 inodes_per_cluster = blks_per_cluster << mp->m_sb.sb_inopblog; 200 201 blk_start_plug(&plug); 202 for (i = 0; i < XFS_INODES_PER_CHUNK; 203 i += inodes_per_cluster, agbno += blks_per_cluster) { 204 if (xfs_inobt_maskn(i, inodes_per_cluster) & ~irec->ir_free) { 205 xfs_btree_reada_bufs(mp, agno, agbno, blks_per_cluster, 206 &xfs_inode_buf_ops); 207 } 208 } 209 blk_finish_plug(&plug); 210 } 211 212 /* 213 * Lookup the inode chunk that the given inode lives in and then get the record 214 * if we found the chunk. If the inode was not the last in the chunk and there 215 * are some left allocated, update the data for the pointed-to record as well as 216 * return the count of grabbed inodes. 217 */ 218 STATIC int 219 xfs_bulkstat_grab_ichunk( 220 struct xfs_btree_cur *cur, /* btree cursor */ 221 xfs_agino_t agino, /* starting inode of chunk */ 222 int *icount,/* return # of inodes grabbed */ 223 struct xfs_inobt_rec_incore *irec) /* btree record */ 224 { 225 int idx; /* index into inode chunk */ 226 int stat; 227 int error = 0; 228 229 /* Lookup the inode chunk that this inode lives in */ 230 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &stat); 231 if (error) 232 return error; 233 if (!stat) { 234 *icount = 0; 235 return error; 236 } 237 238 /* Get the record, should always work */ 239 error = xfs_inobt_get_rec(cur, irec, &stat); 240 if (error) 241 return error; 242 XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, stat == 1); 243 244 /* Check if the record contains the inode in request */ 245 if (irec->ir_startino + XFS_INODES_PER_CHUNK <= agino) { 246 *icount = 0; 247 return 0; 248 } 249 250 idx = agino - irec->ir_startino + 1; 251 if (idx < XFS_INODES_PER_CHUNK && 252 (xfs_inobt_maskn(idx, XFS_INODES_PER_CHUNK - idx) & ~irec->ir_free)) { 253 int i; 254 255 /* We got a right chunk with some left inodes allocated at it. 256 * Grab the chunk record. Mark all the uninteresting inodes 257 * free -- because they're before our start point. 258 */ 259 for (i = 0; i < idx; i++) { 260 if (XFS_INOBT_MASK(i) & ~irec->ir_free) 261 irec->ir_freecount++; 262 } 263 264 irec->ir_free |= xfs_inobt_maskn(0, idx); 265 *icount = irec->ir_count - irec->ir_freecount; 266 } 267 268 return 0; 269 } 270 271 #define XFS_BULKSTAT_UBLEFT(ubleft) ((ubleft) >= statstruct_size) 272 273 struct xfs_bulkstat_agichunk { 274 char __user **ac_ubuffer;/* pointer into user's buffer */ 275 int ac_ubleft; /* bytes left in user's buffer */ 276 int ac_ubelem; /* spaces used in user's buffer */ 277 }; 278 279 /* 280 * Process inodes in chunk with a pointer to a formatter function 281 * that will iget the inode and fill in the appropriate structure. 282 */ 283 static int 284 xfs_bulkstat_ag_ichunk( 285 struct xfs_mount *mp, 286 xfs_agnumber_t agno, 287 struct xfs_inobt_rec_incore *irbp, 288 bulkstat_one_pf formatter, 289 size_t statstruct_size, 290 struct xfs_bulkstat_agichunk *acp, 291 xfs_agino_t *last_agino) 292 { 293 char __user **ubufp = acp->ac_ubuffer; 294 int chunkidx; 295 int error = 0; 296 xfs_agino_t agino = irbp->ir_startino; 297 298 for (chunkidx = 0; chunkidx < XFS_INODES_PER_CHUNK; 299 chunkidx++, agino++) { 300 int fmterror; 301 int ubused; 302 303 /* inode won't fit in buffer, we are done */ 304 if (acp->ac_ubleft < statstruct_size) 305 break; 306 307 /* Skip if this inode is free */ 308 if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free) 309 continue; 310 311 /* Get the inode and fill in a single buffer */ 312 ubused = statstruct_size; 313 error = formatter(mp, XFS_AGINO_TO_INO(mp, agno, agino), 314 *ubufp, acp->ac_ubleft, &ubused, &fmterror); 315 316 if (fmterror == BULKSTAT_RV_GIVEUP || 317 (error && error != -ENOENT && error != -EINVAL)) { 318 acp->ac_ubleft = 0; 319 ASSERT(error); 320 break; 321 } 322 323 /* be careful not to leak error if at end of chunk */ 324 if (fmterror == BULKSTAT_RV_NOTHING || error) { 325 error = 0; 326 continue; 327 } 328 329 *ubufp += ubused; 330 acp->ac_ubleft -= ubused; 331 acp->ac_ubelem++; 332 } 333 334 /* 335 * Post-update *last_agino. At this point, agino will always point one 336 * inode past the last inode we processed successfully. Hence we 337 * substract that inode when setting the *last_agino cursor so that we 338 * return the correct cookie to userspace. On the next bulkstat call, 339 * the inode under the lastino cookie will be skipped as we have already 340 * processed it here. 341 */ 342 *last_agino = agino - 1; 343 344 return error; 345 } 346 347 /* 348 * Return stat information in bulk (by-inode) for the filesystem. 349 */ 350 int /* error status */ 351 xfs_bulkstat( 352 xfs_mount_t *mp, /* mount point for filesystem */ 353 xfs_ino_t *lastinop, /* last inode returned */ 354 int *ubcountp, /* size of buffer/count returned */ 355 bulkstat_one_pf formatter, /* func that'd fill a single buf */ 356 size_t statstruct_size, /* sizeof struct filling */ 357 char __user *ubuffer, /* buffer with inode stats */ 358 int *done) /* 1 if there are more stats to get */ 359 { 360 xfs_buf_t *agbp; /* agi header buffer */ 361 xfs_agino_t agino; /* inode # in allocation group */ 362 xfs_agnumber_t agno; /* allocation group number */ 363 xfs_btree_cur_t *cur; /* btree cursor for ialloc btree */ 364 xfs_inobt_rec_incore_t *irbuf; /* start of irec buffer */ 365 int nirbuf; /* size of irbuf */ 366 int ubcount; /* size of user's buffer */ 367 struct xfs_bulkstat_agichunk ac; 368 int error = 0; 369 370 /* 371 * Get the last inode value, see if there's nothing to do. 372 */ 373 agno = XFS_INO_TO_AGNO(mp, *lastinop); 374 agino = XFS_INO_TO_AGINO(mp, *lastinop); 375 if (agno >= mp->m_sb.sb_agcount || 376 *lastinop != XFS_AGINO_TO_INO(mp, agno, agino)) { 377 *done = 1; 378 *ubcountp = 0; 379 return 0; 380 } 381 382 ubcount = *ubcountp; /* statstruct's */ 383 ac.ac_ubuffer = &ubuffer; 384 ac.ac_ubleft = ubcount * statstruct_size; /* bytes */; 385 ac.ac_ubelem = 0; 386 387 *ubcountp = 0; 388 *done = 0; 389 390 irbuf = kmem_zalloc_large(PAGE_SIZE * 4, KM_SLEEP); 391 if (!irbuf) 392 return -ENOMEM; 393 nirbuf = (PAGE_SIZE * 4) / sizeof(*irbuf); 394 395 /* 396 * Loop over the allocation groups, starting from the last 397 * inode returned; 0 means start of the allocation group. 398 */ 399 while (agno < mp->m_sb.sb_agcount) { 400 struct xfs_inobt_rec_incore *irbp = irbuf; 401 struct xfs_inobt_rec_incore *irbufend = irbuf + nirbuf; 402 bool end_of_ag = false; 403 int icount = 0; 404 int stat; 405 406 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp); 407 if (error) 408 break; 409 /* 410 * Allocate and initialize a btree cursor for ialloc btree. 411 */ 412 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno, 413 XFS_BTNUM_INO); 414 if (agino > 0) { 415 /* 416 * In the middle of an allocation group, we need to get 417 * the remainder of the chunk we're in. 418 */ 419 struct xfs_inobt_rec_incore r; 420 421 error = xfs_bulkstat_grab_ichunk(cur, agino, &icount, &r); 422 if (error) 423 goto del_cursor; 424 if (icount) { 425 irbp->ir_startino = r.ir_startino; 426 irbp->ir_holemask = r.ir_holemask; 427 irbp->ir_count = r.ir_count; 428 irbp->ir_freecount = r.ir_freecount; 429 irbp->ir_free = r.ir_free; 430 irbp++; 431 } 432 /* Increment to the next record */ 433 error = xfs_btree_increment(cur, 0, &stat); 434 } else { 435 /* Start of ag. Lookup the first inode chunk */ 436 error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &stat); 437 } 438 if (error || stat == 0) { 439 end_of_ag = true; 440 goto del_cursor; 441 } 442 443 /* 444 * Loop through inode btree records in this ag, 445 * until we run out of inodes or space in the buffer. 446 */ 447 while (irbp < irbufend && icount < ubcount) { 448 struct xfs_inobt_rec_incore r; 449 450 error = xfs_inobt_get_rec(cur, &r, &stat); 451 if (error || stat == 0) { 452 end_of_ag = true; 453 goto del_cursor; 454 } 455 456 /* 457 * If this chunk has any allocated inodes, save it. 458 * Also start read-ahead now for this chunk. 459 */ 460 if (r.ir_freecount < r.ir_count) { 461 xfs_bulkstat_ichunk_ra(mp, agno, &r); 462 irbp->ir_startino = r.ir_startino; 463 irbp->ir_holemask = r.ir_holemask; 464 irbp->ir_count = r.ir_count; 465 irbp->ir_freecount = r.ir_freecount; 466 irbp->ir_free = r.ir_free; 467 irbp++; 468 icount += r.ir_count - r.ir_freecount; 469 } 470 error = xfs_btree_increment(cur, 0, &stat); 471 if (error || stat == 0) { 472 end_of_ag = true; 473 goto del_cursor; 474 } 475 cond_resched(); 476 } 477 478 /* 479 * Drop the btree buffers and the agi buffer as we can't hold any 480 * of the locks these represent when calling iget. If there is a 481 * pending error, then we are done. 482 */ 483 del_cursor: 484 xfs_btree_del_cursor(cur, error ? 485 XFS_BTREE_ERROR : XFS_BTREE_NOERROR); 486 xfs_buf_relse(agbp); 487 if (error) 488 break; 489 /* 490 * Now format all the good inodes into the user's buffer. The 491 * call to xfs_bulkstat_ag_ichunk() sets up the agino pointer 492 * for the next loop iteration. 493 */ 494 irbufend = irbp; 495 for (irbp = irbuf; 496 irbp < irbufend && ac.ac_ubleft >= statstruct_size; 497 irbp++) { 498 error = xfs_bulkstat_ag_ichunk(mp, agno, irbp, 499 formatter, statstruct_size, &ac, 500 &agino); 501 if (error) 502 break; 503 504 cond_resched(); 505 } 506 507 /* 508 * If we've run out of space or had a formatting error, we 509 * are now done 510 */ 511 if (ac.ac_ubleft < statstruct_size || error) 512 break; 513 514 if (end_of_ag) { 515 agno++; 516 agino = 0; 517 } 518 } 519 /* 520 * Done, we're either out of filesystem or space to put the data. 521 */ 522 kmem_free(irbuf); 523 *ubcountp = ac.ac_ubelem; 524 525 /* 526 * We found some inodes, so clear the error status and return them. 527 * The lastino pointer will point directly at the inode that triggered 528 * any error that occurred, so on the next call the error will be 529 * triggered again and propagated to userspace as there will be no 530 * formatted inodes in the buffer. 531 */ 532 if (ac.ac_ubelem) 533 error = 0; 534 535 /* 536 * If we ran out of filesystem, lastino will point off the end of 537 * the filesystem so the next call will return immediately. 538 */ 539 *lastinop = XFS_AGINO_TO_INO(mp, agno, agino); 540 if (agno >= mp->m_sb.sb_agcount) 541 *done = 1; 542 543 return error; 544 } 545 546 int 547 xfs_inumbers_fmt( 548 void __user *ubuffer, /* buffer to write to */ 549 const struct xfs_inogrp *buffer, /* buffer to read from */ 550 long count, /* # of elements to read */ 551 long *written) /* # of bytes written */ 552 { 553 if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer))) 554 return -EFAULT; 555 *written = count * sizeof(*buffer); 556 return 0; 557 } 558 559 /* 560 * Return inode number table for the filesystem. 561 */ 562 int /* error status */ 563 xfs_inumbers( 564 struct xfs_mount *mp,/* mount point for filesystem */ 565 xfs_ino_t *lastino,/* last inode returned */ 566 int *count,/* size of buffer/count returned */ 567 void __user *ubuffer,/* buffer with inode descriptions */ 568 inumbers_fmt_pf formatter) 569 { 570 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, *lastino); 571 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, *lastino); 572 struct xfs_btree_cur *cur = NULL; 573 struct xfs_buf *agbp = NULL; 574 struct xfs_inogrp *buffer; 575 int bcount; 576 int left = *count; 577 int bufidx = 0; 578 int error = 0; 579 580 *count = 0; 581 if (agno >= mp->m_sb.sb_agcount || 582 *lastino != XFS_AGINO_TO_INO(mp, agno, agino)) 583 return error; 584 585 bcount = MIN(left, (int)(PAGE_SIZE / sizeof(*buffer))); 586 buffer = kmem_zalloc(bcount * sizeof(*buffer), KM_SLEEP); 587 do { 588 struct xfs_inobt_rec_incore r; 589 int stat; 590 591 if (!agbp) { 592 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp); 593 if (error) 594 break; 595 596 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno, 597 XFS_BTNUM_INO); 598 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_GE, 599 &stat); 600 if (error) 601 break; 602 if (!stat) 603 goto next_ag; 604 } 605 606 error = xfs_inobt_get_rec(cur, &r, &stat); 607 if (error) 608 break; 609 if (!stat) 610 goto next_ag; 611 612 agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1; 613 buffer[bufidx].xi_startino = 614 XFS_AGINO_TO_INO(mp, agno, r.ir_startino); 615 buffer[bufidx].xi_alloccount = r.ir_count - r.ir_freecount; 616 buffer[bufidx].xi_allocmask = ~r.ir_free; 617 if (++bufidx == bcount) { 618 long written; 619 620 error = formatter(ubuffer, buffer, bufidx, &written); 621 if (error) 622 break; 623 ubuffer += written; 624 *count += bufidx; 625 bufidx = 0; 626 } 627 if (!--left) 628 break; 629 630 error = xfs_btree_increment(cur, 0, &stat); 631 if (error) 632 break; 633 if (stat) 634 continue; 635 636 next_ag: 637 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); 638 cur = NULL; 639 xfs_buf_relse(agbp); 640 agbp = NULL; 641 agino = 0; 642 agno++; 643 } while (agno < mp->m_sb.sb_agcount); 644 645 if (!error) { 646 if (bufidx) { 647 long written; 648 649 error = formatter(ubuffer, buffer, bufidx, &written); 650 if (!error) 651 *count += bufidx; 652 } 653 *lastino = XFS_AGINO_TO_INO(mp, agno, agino); 654 } 655 656 kmem_free(buffer); 657 if (cur) 658 xfs_btree_del_cursor(cur, (error ? XFS_BTREE_ERROR : 659 XFS_BTREE_NOERROR)); 660 if (agbp) 661 xfs_buf_relse(agbp); 662 663 return error; 664 } 665