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 /* 35 * Return stat information for one inode. 36 * Return 0 if ok, else errno. 37 */ 38 int 39 xfs_bulkstat_one_int( 40 struct xfs_mount *mp, /* mount point for filesystem */ 41 xfs_ino_t ino, /* inode to get data for */ 42 void __user *buffer, /* buffer to place output in */ 43 int ubsize, /* size of buffer */ 44 bulkstat_one_fmt_pf formatter, /* formatter, copy to user */ 45 int *ubused, /* bytes used by me */ 46 int *stat) /* BULKSTAT_RV_... */ 47 { 48 struct xfs_icdinode *dic; /* dinode core info pointer */ 49 struct xfs_inode *ip; /* incore inode pointer */ 50 struct inode *inode; 51 struct xfs_bstat *buf; /* return buffer */ 52 int error = 0; /* error value */ 53 54 *stat = BULKSTAT_RV_NOTHING; 55 56 if (!buffer || xfs_internal_inum(mp, ino)) 57 return -EINVAL; 58 59 buf = kmem_zalloc(sizeof(*buf), KM_SLEEP | KM_MAYFAIL); 60 if (!buf) 61 return -ENOMEM; 62 63 error = xfs_iget(mp, NULL, ino, 64 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED), 65 XFS_ILOCK_SHARED, &ip); 66 if (error) 67 goto out_free; 68 69 ASSERT(ip != NULL); 70 ASSERT(ip->i_imap.im_blkno != 0); 71 inode = VFS_I(ip); 72 73 dic = &ip->i_d; 74 75 /* xfs_iget returns the following without needing 76 * further change. 77 */ 78 buf->bs_projid_lo = dic->di_projid_lo; 79 buf->bs_projid_hi = dic->di_projid_hi; 80 buf->bs_ino = ino; 81 buf->bs_uid = dic->di_uid; 82 buf->bs_gid = dic->di_gid; 83 buf->bs_size = dic->di_size; 84 85 buf->bs_nlink = inode->i_nlink; 86 buf->bs_atime.tv_sec = inode->i_atime.tv_sec; 87 buf->bs_atime.tv_nsec = inode->i_atime.tv_nsec; 88 buf->bs_mtime.tv_sec = inode->i_mtime.tv_sec; 89 buf->bs_mtime.tv_nsec = inode->i_mtime.tv_nsec; 90 buf->bs_ctime.tv_sec = inode->i_ctime.tv_sec; 91 buf->bs_ctime.tv_nsec = inode->i_ctime.tv_nsec; 92 buf->bs_gen = inode->i_generation; 93 buf->bs_mode = inode->i_mode; 94 95 buf->bs_xflags = xfs_ip2xflags(ip); 96 buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog; 97 buf->bs_extents = dic->di_nextents; 98 memset(buf->bs_pad, 0, sizeof(buf->bs_pad)); 99 buf->bs_dmevmask = dic->di_dmevmask; 100 buf->bs_dmstate = dic->di_dmstate; 101 buf->bs_aextents = dic->di_anextents; 102 buf->bs_forkoff = XFS_IFORK_BOFF(ip); 103 104 if (dic->di_version == 3) { 105 if (dic->di_flags2 & XFS_DIFLAG2_COWEXTSIZE) 106 buf->bs_cowextsize = dic->di_cowextsize << 107 mp->m_sb.sb_blocklog; 108 } 109 110 switch (dic->di_format) { 111 case XFS_DINODE_FMT_DEV: 112 buf->bs_rdev = sysv_encode_dev(inode->i_rdev); 113 buf->bs_blksize = BLKDEV_IOSIZE; 114 buf->bs_blocks = 0; 115 break; 116 case XFS_DINODE_FMT_LOCAL: 117 buf->bs_rdev = 0; 118 buf->bs_blksize = mp->m_sb.sb_blocksize; 119 buf->bs_blocks = 0; 120 break; 121 case XFS_DINODE_FMT_EXTENTS: 122 case XFS_DINODE_FMT_BTREE: 123 buf->bs_rdev = 0; 124 buf->bs_blksize = mp->m_sb.sb_blocksize; 125 buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks; 126 break; 127 } 128 xfs_iunlock(ip, XFS_ILOCK_SHARED); 129 IRELE(ip); 130 131 error = formatter(buffer, ubsize, ubused, buf); 132 if (!error) 133 *stat = BULKSTAT_RV_DIDONE; 134 135 out_free: 136 kmem_free(buf); 137 return error; 138 } 139 140 /* Return 0 on success or positive error */ 141 STATIC int 142 xfs_bulkstat_one_fmt( 143 void __user *ubuffer, 144 int ubsize, 145 int *ubused, 146 const xfs_bstat_t *buffer) 147 { 148 if (ubsize < sizeof(*buffer)) 149 return -ENOMEM; 150 if (copy_to_user(ubuffer, buffer, sizeof(*buffer))) 151 return -EFAULT; 152 if (ubused) 153 *ubused = sizeof(*buffer); 154 return 0; 155 } 156 157 int 158 xfs_bulkstat_one( 159 xfs_mount_t *mp, /* mount point for filesystem */ 160 xfs_ino_t ino, /* inode number to get data for */ 161 void __user *buffer, /* buffer to place output in */ 162 int ubsize, /* size of buffer */ 163 int *ubused, /* bytes used by me */ 164 int *stat) /* BULKSTAT_RV_... */ 165 { 166 return xfs_bulkstat_one_int(mp, ino, buffer, ubsize, 167 xfs_bulkstat_one_fmt, ubused, stat); 168 } 169 170 /* 171 * Loop over all clusters in a chunk for a given incore inode allocation btree 172 * record. Do a readahead if there are any allocated inodes in that cluster. 173 */ 174 STATIC void 175 xfs_bulkstat_ichunk_ra( 176 struct xfs_mount *mp, 177 xfs_agnumber_t agno, 178 struct xfs_inobt_rec_incore *irec) 179 { 180 xfs_agblock_t agbno; 181 struct blk_plug plug; 182 int blks_per_cluster; 183 int inodes_per_cluster; 184 int i; /* inode chunk index */ 185 186 agbno = XFS_AGINO_TO_AGBNO(mp, irec->ir_startino); 187 blks_per_cluster = xfs_icluster_size_fsb(mp); 188 inodes_per_cluster = blks_per_cluster << mp->m_sb.sb_inopblog; 189 190 blk_start_plug(&plug); 191 for (i = 0; i < XFS_INODES_PER_CHUNK; 192 i += inodes_per_cluster, agbno += blks_per_cluster) { 193 if (xfs_inobt_maskn(i, inodes_per_cluster) & ~irec->ir_free) { 194 xfs_btree_reada_bufs(mp, agno, agbno, blks_per_cluster, 195 &xfs_inode_buf_ops); 196 } 197 } 198 blk_finish_plug(&plug); 199 } 200 201 /* 202 * Lookup the inode chunk that the given inode lives in and then get the record 203 * if we found the chunk. If the inode was not the last in the chunk and there 204 * are some left allocated, update the data for the pointed-to record as well as 205 * return the count of grabbed inodes. 206 */ 207 STATIC int 208 xfs_bulkstat_grab_ichunk( 209 struct xfs_btree_cur *cur, /* btree cursor */ 210 xfs_agino_t agino, /* starting inode of chunk */ 211 int *icount,/* return # of inodes grabbed */ 212 struct xfs_inobt_rec_incore *irec) /* btree record */ 213 { 214 int idx; /* index into inode chunk */ 215 int stat; 216 int error = 0; 217 218 /* Lookup the inode chunk that this inode lives in */ 219 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &stat); 220 if (error) 221 return error; 222 if (!stat) { 223 *icount = 0; 224 return error; 225 } 226 227 /* Get the record, should always work */ 228 error = xfs_inobt_get_rec(cur, irec, &stat); 229 if (error) 230 return error; 231 XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, stat == 1); 232 233 /* Check if the record contains the inode in request */ 234 if (irec->ir_startino + XFS_INODES_PER_CHUNK <= agino) { 235 *icount = 0; 236 return 0; 237 } 238 239 idx = agino - irec->ir_startino + 1; 240 if (idx < XFS_INODES_PER_CHUNK && 241 (xfs_inobt_maskn(idx, XFS_INODES_PER_CHUNK - idx) & ~irec->ir_free)) { 242 int i; 243 244 /* We got a right chunk with some left inodes allocated at it. 245 * Grab the chunk record. Mark all the uninteresting inodes 246 * free -- because they're before our start point. 247 */ 248 for (i = 0; i < idx; i++) { 249 if (XFS_INOBT_MASK(i) & ~irec->ir_free) 250 irec->ir_freecount++; 251 } 252 253 irec->ir_free |= xfs_inobt_maskn(0, idx); 254 *icount = irec->ir_count - irec->ir_freecount; 255 } 256 257 return 0; 258 } 259 260 #define XFS_BULKSTAT_UBLEFT(ubleft) ((ubleft) >= statstruct_size) 261 262 struct xfs_bulkstat_agichunk { 263 char __user **ac_ubuffer;/* pointer into user's buffer */ 264 int ac_ubleft; /* bytes left in user's buffer */ 265 int ac_ubelem; /* spaces used in user's buffer */ 266 }; 267 268 /* 269 * Process inodes in chunk with a pointer to a formatter function 270 * that will iget the inode and fill in the appropriate structure. 271 */ 272 static int 273 xfs_bulkstat_ag_ichunk( 274 struct xfs_mount *mp, 275 xfs_agnumber_t agno, 276 struct xfs_inobt_rec_incore *irbp, 277 bulkstat_one_pf formatter, 278 size_t statstruct_size, 279 struct xfs_bulkstat_agichunk *acp, 280 xfs_agino_t *last_agino) 281 { 282 char __user **ubufp = acp->ac_ubuffer; 283 int chunkidx; 284 int error = 0; 285 xfs_agino_t agino = irbp->ir_startino; 286 287 for (chunkidx = 0; chunkidx < XFS_INODES_PER_CHUNK; 288 chunkidx++, agino++) { 289 int fmterror; 290 int ubused; 291 292 /* inode won't fit in buffer, we are done */ 293 if (acp->ac_ubleft < statstruct_size) 294 break; 295 296 /* Skip if this inode is free */ 297 if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free) 298 continue; 299 300 /* Get the inode and fill in a single buffer */ 301 ubused = statstruct_size; 302 error = formatter(mp, XFS_AGINO_TO_INO(mp, agno, agino), 303 *ubufp, acp->ac_ubleft, &ubused, &fmterror); 304 305 if (fmterror == BULKSTAT_RV_GIVEUP || 306 (error && error != -ENOENT && error != -EINVAL)) { 307 acp->ac_ubleft = 0; 308 ASSERT(error); 309 break; 310 } 311 312 /* be careful not to leak error if at end of chunk */ 313 if (fmterror == BULKSTAT_RV_NOTHING || error) { 314 error = 0; 315 continue; 316 } 317 318 *ubufp += ubused; 319 acp->ac_ubleft -= ubused; 320 acp->ac_ubelem++; 321 } 322 323 /* 324 * Post-update *last_agino. At this point, agino will always point one 325 * inode past the last inode we processed successfully. Hence we 326 * substract that inode when setting the *last_agino cursor so that we 327 * return the correct cookie to userspace. On the next bulkstat call, 328 * the inode under the lastino cookie will be skipped as we have already 329 * processed it here. 330 */ 331 *last_agino = agino - 1; 332 333 return error; 334 } 335 336 /* 337 * Return stat information in bulk (by-inode) for the filesystem. 338 */ 339 int /* error status */ 340 xfs_bulkstat( 341 xfs_mount_t *mp, /* mount point for filesystem */ 342 xfs_ino_t *lastinop, /* last inode returned */ 343 int *ubcountp, /* size of buffer/count returned */ 344 bulkstat_one_pf formatter, /* func that'd fill a single buf */ 345 size_t statstruct_size, /* sizeof struct filling */ 346 char __user *ubuffer, /* buffer with inode stats */ 347 int *done) /* 1 if there are more stats to get */ 348 { 349 xfs_buf_t *agbp; /* agi header buffer */ 350 xfs_agino_t agino; /* inode # in allocation group */ 351 xfs_agnumber_t agno; /* allocation group number */ 352 xfs_btree_cur_t *cur; /* btree cursor for ialloc btree */ 353 xfs_inobt_rec_incore_t *irbuf; /* start of irec buffer */ 354 int nirbuf; /* size of irbuf */ 355 int ubcount; /* size of user's buffer */ 356 struct xfs_bulkstat_agichunk ac; 357 int error = 0; 358 359 /* 360 * Get the last inode value, see if there's nothing to do. 361 */ 362 agno = XFS_INO_TO_AGNO(mp, *lastinop); 363 agino = XFS_INO_TO_AGINO(mp, *lastinop); 364 if (agno >= mp->m_sb.sb_agcount || 365 *lastinop != XFS_AGINO_TO_INO(mp, agno, agino)) { 366 *done = 1; 367 *ubcountp = 0; 368 return 0; 369 } 370 371 ubcount = *ubcountp; /* statstruct's */ 372 ac.ac_ubuffer = &ubuffer; 373 ac.ac_ubleft = ubcount * statstruct_size; /* bytes */; 374 ac.ac_ubelem = 0; 375 376 *ubcountp = 0; 377 *done = 0; 378 379 irbuf = kmem_zalloc_large(PAGE_SIZE * 4, KM_SLEEP); 380 if (!irbuf) 381 return -ENOMEM; 382 nirbuf = (PAGE_SIZE * 4) / sizeof(*irbuf); 383 384 /* 385 * Loop over the allocation groups, starting from the last 386 * inode returned; 0 means start of the allocation group. 387 */ 388 while (agno < mp->m_sb.sb_agcount) { 389 struct xfs_inobt_rec_incore *irbp = irbuf; 390 struct xfs_inobt_rec_incore *irbufend = irbuf + nirbuf; 391 bool end_of_ag = false; 392 int icount = 0; 393 int stat; 394 395 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp); 396 if (error) 397 break; 398 /* 399 * Allocate and initialize a btree cursor for ialloc btree. 400 */ 401 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno, 402 XFS_BTNUM_INO); 403 if (agino > 0) { 404 /* 405 * In the middle of an allocation group, we need to get 406 * the remainder of the chunk we're in. 407 */ 408 struct xfs_inobt_rec_incore r; 409 410 error = xfs_bulkstat_grab_ichunk(cur, agino, &icount, &r); 411 if (error) 412 goto del_cursor; 413 if (icount) { 414 irbp->ir_startino = r.ir_startino; 415 irbp->ir_holemask = r.ir_holemask; 416 irbp->ir_count = r.ir_count; 417 irbp->ir_freecount = r.ir_freecount; 418 irbp->ir_free = r.ir_free; 419 irbp++; 420 } 421 /* Increment to the next record */ 422 error = xfs_btree_increment(cur, 0, &stat); 423 } else { 424 /* Start of ag. Lookup the first inode chunk */ 425 error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &stat); 426 } 427 if (error || stat == 0) { 428 end_of_ag = true; 429 goto del_cursor; 430 } 431 432 /* 433 * Loop through inode btree records in this ag, 434 * until we run out of inodes or space in the buffer. 435 */ 436 while (irbp < irbufend && icount < ubcount) { 437 struct xfs_inobt_rec_incore r; 438 439 error = xfs_inobt_get_rec(cur, &r, &stat); 440 if (error || stat == 0) { 441 end_of_ag = true; 442 goto del_cursor; 443 } 444 445 /* 446 * If this chunk has any allocated inodes, save it. 447 * Also start read-ahead now for this chunk. 448 */ 449 if (r.ir_freecount < r.ir_count) { 450 xfs_bulkstat_ichunk_ra(mp, agno, &r); 451 irbp->ir_startino = r.ir_startino; 452 irbp->ir_holemask = r.ir_holemask; 453 irbp->ir_count = r.ir_count; 454 irbp->ir_freecount = r.ir_freecount; 455 irbp->ir_free = r.ir_free; 456 irbp++; 457 icount += r.ir_count - r.ir_freecount; 458 } 459 error = xfs_btree_increment(cur, 0, &stat); 460 if (error || stat == 0) { 461 end_of_ag = true; 462 goto del_cursor; 463 } 464 cond_resched(); 465 } 466 467 /* 468 * Drop the btree buffers and the agi buffer as we can't hold any 469 * of the locks these represent when calling iget. If there is a 470 * pending error, then we are done. 471 */ 472 del_cursor: 473 xfs_btree_del_cursor(cur, error ? 474 XFS_BTREE_ERROR : XFS_BTREE_NOERROR); 475 xfs_buf_relse(agbp); 476 if (error) 477 break; 478 /* 479 * Now format all the good inodes into the user's buffer. The 480 * call to xfs_bulkstat_ag_ichunk() sets up the agino pointer 481 * for the next loop iteration. 482 */ 483 irbufend = irbp; 484 for (irbp = irbuf; 485 irbp < irbufend && ac.ac_ubleft >= statstruct_size; 486 irbp++) { 487 error = xfs_bulkstat_ag_ichunk(mp, agno, irbp, 488 formatter, statstruct_size, &ac, 489 &agino); 490 if (error) 491 break; 492 493 cond_resched(); 494 } 495 496 /* 497 * If we've run out of space or had a formatting error, we 498 * are now done 499 */ 500 if (ac.ac_ubleft < statstruct_size || error) 501 break; 502 503 if (end_of_ag) { 504 agno++; 505 agino = 0; 506 } 507 } 508 /* 509 * Done, we're either out of filesystem or space to put the data. 510 */ 511 kmem_free(irbuf); 512 *ubcountp = ac.ac_ubelem; 513 514 /* 515 * We found some inodes, so clear the error status and return them. 516 * The lastino pointer will point directly at the inode that triggered 517 * any error that occurred, so on the next call the error will be 518 * triggered again and propagated to userspace as there will be no 519 * formatted inodes in the buffer. 520 */ 521 if (ac.ac_ubelem) 522 error = 0; 523 524 /* 525 * If we ran out of filesystem, lastino will point off the end of 526 * the filesystem so the next call will return immediately. 527 */ 528 *lastinop = XFS_AGINO_TO_INO(mp, agno, agino); 529 if (agno >= mp->m_sb.sb_agcount) 530 *done = 1; 531 532 return error; 533 } 534 535 int 536 xfs_inumbers_fmt( 537 void __user *ubuffer, /* buffer to write to */ 538 const struct xfs_inogrp *buffer, /* buffer to read from */ 539 long count, /* # of elements to read */ 540 long *written) /* # of bytes written */ 541 { 542 if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer))) 543 return -EFAULT; 544 *written = count * sizeof(*buffer); 545 return 0; 546 } 547 548 /* 549 * Return inode number table for the filesystem. 550 */ 551 int /* error status */ 552 xfs_inumbers( 553 struct xfs_mount *mp,/* mount point for filesystem */ 554 xfs_ino_t *lastino,/* last inode returned */ 555 int *count,/* size of buffer/count returned */ 556 void __user *ubuffer,/* buffer with inode descriptions */ 557 inumbers_fmt_pf formatter) 558 { 559 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, *lastino); 560 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, *lastino); 561 struct xfs_btree_cur *cur = NULL; 562 struct xfs_buf *agbp = NULL; 563 struct xfs_inogrp *buffer; 564 int bcount; 565 int left = *count; 566 int bufidx = 0; 567 int error = 0; 568 569 *count = 0; 570 if (agno >= mp->m_sb.sb_agcount || 571 *lastino != XFS_AGINO_TO_INO(mp, agno, agino)) 572 return error; 573 574 bcount = MIN(left, (int)(PAGE_SIZE / sizeof(*buffer))); 575 buffer = kmem_zalloc(bcount * sizeof(*buffer), KM_SLEEP); 576 do { 577 struct xfs_inobt_rec_incore r; 578 int stat; 579 580 if (!agbp) { 581 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp); 582 if (error) 583 break; 584 585 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno, 586 XFS_BTNUM_INO); 587 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_GE, 588 &stat); 589 if (error) 590 break; 591 if (!stat) 592 goto next_ag; 593 } 594 595 error = xfs_inobt_get_rec(cur, &r, &stat); 596 if (error) 597 break; 598 if (!stat) 599 goto next_ag; 600 601 agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1; 602 buffer[bufidx].xi_startino = 603 XFS_AGINO_TO_INO(mp, agno, r.ir_startino); 604 buffer[bufidx].xi_alloccount = r.ir_count - r.ir_freecount; 605 buffer[bufidx].xi_allocmask = ~r.ir_free; 606 if (++bufidx == bcount) { 607 long written; 608 609 error = formatter(ubuffer, buffer, bufidx, &written); 610 if (error) 611 break; 612 ubuffer += written; 613 *count += bufidx; 614 bufidx = 0; 615 } 616 if (!--left) 617 break; 618 619 error = xfs_btree_increment(cur, 0, &stat); 620 if (error) 621 break; 622 if (stat) 623 continue; 624 625 next_ag: 626 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR); 627 cur = NULL; 628 xfs_buf_relse(agbp); 629 agbp = NULL; 630 agino = 0; 631 agno++; 632 } while (agno < mp->m_sb.sb_agcount); 633 634 if (!error) { 635 if (bufidx) { 636 long written; 637 638 error = formatter(ubuffer, buffer, bufidx, &written); 639 if (!error) 640 *count += bufidx; 641 } 642 *lastino = XFS_AGINO_TO_INO(mp, agno, agino); 643 } 644 645 kmem_free(buffer); 646 if (cur) 647 xfs_btree_del_cursor(cur, (error ? XFS_BTREE_ERROR : 648 XFS_BTREE_NOERROR)); 649 if (agbp) 650 xfs_buf_relse(agbp); 651 652 return error; 653 } 654