1 /* 2 * Copyright (c) 2000-2005 Silicon Graphics, Inc. 3 * Copyright (c) 2013 Red Hat, Inc. 4 * All Rights Reserved. 5 * 6 * This program is free software; you can redistribute it and/or 7 * modify it under the terms of the GNU General Public License as 8 * published by the Free Software Foundation. 9 * 10 * This program is distributed in the hope that it would be useful, 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13 * GNU General Public License for more details. 14 * 15 * You should have received a copy of the GNU General Public License 16 * along with this program; if not, write the Free Software Foundation, 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA 18 */ 19 #include "xfs.h" 20 #include "xfs_fs.h" 21 #include "xfs_types.h" 22 #include "xfs_bit.h" 23 #include "xfs_log.h" 24 #include "xfs_trans.h" 25 #include "xfs_sb.h" 26 #include "xfs_ag.h" 27 #include "xfs_mount.h" 28 #include "xfs_da_btree.h" 29 #include "xfs_bmap_btree.h" 30 #include "xfs_alloc_btree.h" 31 #include "xfs_ialloc_btree.h" 32 #include "xfs_alloc.h" 33 #include "xfs_btree.h" 34 #include "xfs_attr_sf.h" 35 #include "xfs_attr_remote.h" 36 #include "xfs_dinode.h" 37 #include "xfs_inode.h" 38 #include "xfs_inode_item.h" 39 #include "xfs_bmap.h" 40 #include "xfs_attr.h" 41 #include "xfs_attr_leaf.h" 42 #include "xfs_error.h" 43 #include "xfs_trace.h" 44 #include "xfs_buf_item.h" 45 #include "xfs_cksum.h" 46 47 STATIC int 48 xfs_attr_shortform_compare(const void *a, const void *b) 49 { 50 xfs_attr_sf_sort_t *sa, *sb; 51 52 sa = (xfs_attr_sf_sort_t *)a; 53 sb = (xfs_attr_sf_sort_t *)b; 54 if (sa->hash < sb->hash) { 55 return(-1); 56 } else if (sa->hash > sb->hash) { 57 return(1); 58 } else { 59 return(sa->entno - sb->entno); 60 } 61 } 62 63 #define XFS_ISRESET_CURSOR(cursor) \ 64 (!((cursor)->initted) && !((cursor)->hashval) && \ 65 !((cursor)->blkno) && !((cursor)->offset)) 66 /* 67 * Copy out entries of shortform attribute lists for attr_list(). 68 * Shortform attribute lists are not stored in hashval sorted order. 69 * If the output buffer is not large enough to hold them all, then we 70 * we have to calculate each entries' hashvalue and sort them before 71 * we can begin returning them to the user. 72 */ 73 int 74 xfs_attr_shortform_list(xfs_attr_list_context_t *context) 75 { 76 attrlist_cursor_kern_t *cursor; 77 xfs_attr_sf_sort_t *sbuf, *sbp; 78 xfs_attr_shortform_t *sf; 79 xfs_attr_sf_entry_t *sfe; 80 xfs_inode_t *dp; 81 int sbsize, nsbuf, count, i; 82 int error; 83 84 ASSERT(context != NULL); 85 dp = context->dp; 86 ASSERT(dp != NULL); 87 ASSERT(dp->i_afp != NULL); 88 sf = (xfs_attr_shortform_t *)dp->i_afp->if_u1.if_data; 89 ASSERT(sf != NULL); 90 if (!sf->hdr.count) 91 return(0); 92 cursor = context->cursor; 93 ASSERT(cursor != NULL); 94 95 trace_xfs_attr_list_sf(context); 96 97 /* 98 * If the buffer is large enough and the cursor is at the start, 99 * do not bother with sorting since we will return everything in 100 * one buffer and another call using the cursor won't need to be 101 * made. 102 * Note the generous fudge factor of 16 overhead bytes per entry. 103 * If bufsize is zero then put_listent must be a search function 104 * and can just scan through what we have. 105 */ 106 if (context->bufsize == 0 || 107 (XFS_ISRESET_CURSOR(cursor) && 108 (dp->i_afp->if_bytes + sf->hdr.count * 16) < context->bufsize)) { 109 for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) { 110 error = context->put_listent(context, 111 sfe->flags, 112 sfe->nameval, 113 (int)sfe->namelen, 114 (int)sfe->valuelen, 115 &sfe->nameval[sfe->namelen]); 116 117 /* 118 * Either search callback finished early or 119 * didn't fit it all in the buffer after all. 120 */ 121 if (context->seen_enough) 122 break; 123 124 if (error) 125 return error; 126 sfe = XFS_ATTR_SF_NEXTENTRY(sfe); 127 } 128 trace_xfs_attr_list_sf_all(context); 129 return(0); 130 } 131 132 /* do no more for a search callback */ 133 if (context->bufsize == 0) 134 return 0; 135 136 /* 137 * It didn't all fit, so we have to sort everything on hashval. 138 */ 139 sbsize = sf->hdr.count * sizeof(*sbuf); 140 sbp = sbuf = kmem_alloc(sbsize, KM_SLEEP | KM_NOFS); 141 142 /* 143 * Scan the attribute list for the rest of the entries, storing 144 * the relevant info from only those that match into a buffer. 145 */ 146 nsbuf = 0; 147 for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) { 148 if (unlikely( 149 ((char *)sfe < (char *)sf) || 150 ((char *)sfe >= ((char *)sf + dp->i_afp->if_bytes)))) { 151 XFS_CORRUPTION_ERROR("xfs_attr_shortform_list", 152 XFS_ERRLEVEL_LOW, 153 context->dp->i_mount, sfe); 154 kmem_free(sbuf); 155 return XFS_ERROR(EFSCORRUPTED); 156 } 157 158 sbp->entno = i; 159 sbp->hash = xfs_da_hashname(sfe->nameval, sfe->namelen); 160 sbp->name = sfe->nameval; 161 sbp->namelen = sfe->namelen; 162 /* These are bytes, and both on-disk, don't endian-flip */ 163 sbp->valuelen = sfe->valuelen; 164 sbp->flags = sfe->flags; 165 sfe = XFS_ATTR_SF_NEXTENTRY(sfe); 166 sbp++; 167 nsbuf++; 168 } 169 170 /* 171 * Sort the entries on hash then entno. 172 */ 173 xfs_sort(sbuf, nsbuf, sizeof(*sbuf), xfs_attr_shortform_compare); 174 175 /* 176 * Re-find our place IN THE SORTED LIST. 177 */ 178 count = 0; 179 cursor->initted = 1; 180 cursor->blkno = 0; 181 for (sbp = sbuf, i = 0; i < nsbuf; i++, sbp++) { 182 if (sbp->hash == cursor->hashval) { 183 if (cursor->offset == count) { 184 break; 185 } 186 count++; 187 } else if (sbp->hash > cursor->hashval) { 188 break; 189 } 190 } 191 if (i == nsbuf) { 192 kmem_free(sbuf); 193 return(0); 194 } 195 196 /* 197 * Loop putting entries into the user buffer. 198 */ 199 for ( ; i < nsbuf; i++, sbp++) { 200 if (cursor->hashval != sbp->hash) { 201 cursor->hashval = sbp->hash; 202 cursor->offset = 0; 203 } 204 error = context->put_listent(context, 205 sbp->flags, 206 sbp->name, 207 sbp->namelen, 208 sbp->valuelen, 209 &sbp->name[sbp->namelen]); 210 if (error) 211 return error; 212 if (context->seen_enough) 213 break; 214 cursor->offset++; 215 } 216 217 kmem_free(sbuf); 218 return(0); 219 } 220 221 STATIC int 222 xfs_attr_node_list(xfs_attr_list_context_t *context) 223 { 224 attrlist_cursor_kern_t *cursor; 225 xfs_attr_leafblock_t *leaf; 226 xfs_da_intnode_t *node; 227 struct xfs_attr3_icleaf_hdr leafhdr; 228 struct xfs_da3_icnode_hdr nodehdr; 229 struct xfs_da_node_entry *btree; 230 int error, i; 231 struct xfs_buf *bp; 232 233 trace_xfs_attr_node_list(context); 234 235 cursor = context->cursor; 236 cursor->initted = 1; 237 238 /* 239 * Do all sorts of validation on the passed-in cursor structure. 240 * If anything is amiss, ignore the cursor and look up the hashval 241 * starting from the btree root. 242 */ 243 bp = NULL; 244 if (cursor->blkno > 0) { 245 error = xfs_da3_node_read(NULL, context->dp, cursor->blkno, -1, 246 &bp, XFS_ATTR_FORK); 247 if ((error != 0) && (error != EFSCORRUPTED)) 248 return(error); 249 if (bp) { 250 struct xfs_attr_leaf_entry *entries; 251 252 node = bp->b_addr; 253 switch (be16_to_cpu(node->hdr.info.magic)) { 254 case XFS_DA_NODE_MAGIC: 255 case XFS_DA3_NODE_MAGIC: 256 trace_xfs_attr_list_wrong_blk(context); 257 xfs_trans_brelse(NULL, bp); 258 bp = NULL; 259 break; 260 case XFS_ATTR_LEAF_MAGIC: 261 case XFS_ATTR3_LEAF_MAGIC: 262 leaf = bp->b_addr; 263 xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf); 264 entries = xfs_attr3_leaf_entryp(leaf); 265 if (cursor->hashval > be32_to_cpu( 266 entries[leafhdr.count - 1].hashval)) { 267 trace_xfs_attr_list_wrong_blk(context); 268 xfs_trans_brelse(NULL, bp); 269 bp = NULL; 270 } else if (cursor->hashval <= be32_to_cpu( 271 entries[0].hashval)) { 272 trace_xfs_attr_list_wrong_blk(context); 273 xfs_trans_brelse(NULL, bp); 274 bp = NULL; 275 } 276 break; 277 default: 278 trace_xfs_attr_list_wrong_blk(context); 279 xfs_trans_brelse(NULL, bp); 280 bp = NULL; 281 } 282 } 283 } 284 285 /* 286 * We did not find what we expected given the cursor's contents, 287 * so we start from the top and work down based on the hash value. 288 * Note that start of node block is same as start of leaf block. 289 */ 290 if (bp == NULL) { 291 cursor->blkno = 0; 292 for (;;) { 293 __uint16_t magic; 294 295 error = xfs_da3_node_read(NULL, context->dp, 296 cursor->blkno, -1, &bp, 297 XFS_ATTR_FORK); 298 if (error) 299 return(error); 300 node = bp->b_addr; 301 magic = be16_to_cpu(node->hdr.info.magic); 302 if (magic == XFS_ATTR_LEAF_MAGIC || 303 magic == XFS_ATTR3_LEAF_MAGIC) 304 break; 305 if (magic != XFS_DA_NODE_MAGIC && 306 magic != XFS_DA3_NODE_MAGIC) { 307 XFS_CORRUPTION_ERROR("xfs_attr_node_list(3)", 308 XFS_ERRLEVEL_LOW, 309 context->dp->i_mount, 310 node); 311 xfs_trans_brelse(NULL, bp); 312 return XFS_ERROR(EFSCORRUPTED); 313 } 314 315 xfs_da3_node_hdr_from_disk(&nodehdr, node); 316 btree = xfs_da3_node_tree_p(node); 317 for (i = 0; i < nodehdr.count; btree++, i++) { 318 if (cursor->hashval 319 <= be32_to_cpu(btree->hashval)) { 320 cursor->blkno = be32_to_cpu(btree->before); 321 trace_xfs_attr_list_node_descend(context, 322 btree); 323 break; 324 } 325 } 326 if (i == nodehdr.count) { 327 xfs_trans_brelse(NULL, bp); 328 return 0; 329 } 330 xfs_trans_brelse(NULL, bp); 331 } 332 } 333 ASSERT(bp != NULL); 334 335 /* 336 * Roll upward through the blocks, processing each leaf block in 337 * order. As long as there is space in the result buffer, keep 338 * adding the information. 339 */ 340 for (;;) { 341 leaf = bp->b_addr; 342 error = xfs_attr3_leaf_list_int(bp, context); 343 if (error) { 344 xfs_trans_brelse(NULL, bp); 345 return error; 346 } 347 xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf); 348 if (context->seen_enough || leafhdr.forw == 0) 349 break; 350 cursor->blkno = leafhdr.forw; 351 xfs_trans_brelse(NULL, bp); 352 error = xfs_attr3_leaf_read(NULL, context->dp, cursor->blkno, -1, 353 &bp); 354 if (error) 355 return error; 356 } 357 xfs_trans_brelse(NULL, bp); 358 return 0; 359 } 360 361 /* 362 * Copy out attribute list entries for attr_list(), for leaf attribute lists. 363 */ 364 int 365 xfs_attr3_leaf_list_int( 366 struct xfs_buf *bp, 367 struct xfs_attr_list_context *context) 368 { 369 struct attrlist_cursor_kern *cursor; 370 struct xfs_attr_leafblock *leaf; 371 struct xfs_attr3_icleaf_hdr ichdr; 372 struct xfs_attr_leaf_entry *entries; 373 struct xfs_attr_leaf_entry *entry; 374 int retval; 375 int i; 376 377 trace_xfs_attr_list_leaf(context); 378 379 leaf = bp->b_addr; 380 xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); 381 entries = xfs_attr3_leaf_entryp(leaf); 382 383 cursor = context->cursor; 384 cursor->initted = 1; 385 386 /* 387 * Re-find our place in the leaf block if this is a new syscall. 388 */ 389 if (context->resynch) { 390 entry = &entries[0]; 391 for (i = 0; i < ichdr.count; entry++, i++) { 392 if (be32_to_cpu(entry->hashval) == cursor->hashval) { 393 if (cursor->offset == context->dupcnt) { 394 context->dupcnt = 0; 395 break; 396 } 397 context->dupcnt++; 398 } else if (be32_to_cpu(entry->hashval) > 399 cursor->hashval) { 400 context->dupcnt = 0; 401 break; 402 } 403 } 404 if (i == ichdr.count) { 405 trace_xfs_attr_list_notfound(context); 406 return 0; 407 } 408 } else { 409 entry = &entries[0]; 410 i = 0; 411 } 412 context->resynch = 0; 413 414 /* 415 * We have found our place, start copying out the new attributes. 416 */ 417 retval = 0; 418 for (; i < ichdr.count; entry++, i++) { 419 if (be32_to_cpu(entry->hashval) != cursor->hashval) { 420 cursor->hashval = be32_to_cpu(entry->hashval); 421 cursor->offset = 0; 422 } 423 424 if (entry->flags & XFS_ATTR_INCOMPLETE) 425 continue; /* skip incomplete entries */ 426 427 if (entry->flags & XFS_ATTR_LOCAL) { 428 xfs_attr_leaf_name_local_t *name_loc = 429 xfs_attr3_leaf_name_local(leaf, i); 430 431 retval = context->put_listent(context, 432 entry->flags, 433 name_loc->nameval, 434 (int)name_loc->namelen, 435 be16_to_cpu(name_loc->valuelen), 436 &name_loc->nameval[name_loc->namelen]); 437 if (retval) 438 return retval; 439 } else { 440 xfs_attr_leaf_name_remote_t *name_rmt = 441 xfs_attr3_leaf_name_remote(leaf, i); 442 443 int valuelen = be32_to_cpu(name_rmt->valuelen); 444 445 if (context->put_value) { 446 xfs_da_args_t args; 447 448 memset((char *)&args, 0, sizeof(args)); 449 args.dp = context->dp; 450 args.whichfork = XFS_ATTR_FORK; 451 args.valuelen = valuelen; 452 args.value = kmem_alloc(valuelen, KM_SLEEP | KM_NOFS); 453 args.rmtblkno = be32_to_cpu(name_rmt->valueblk); 454 args.rmtblkcnt = xfs_attr3_rmt_blocks( 455 args.dp->i_mount, valuelen); 456 retval = xfs_attr_rmtval_get(&args); 457 if (retval) 458 return retval; 459 retval = context->put_listent(context, 460 entry->flags, 461 name_rmt->name, 462 (int)name_rmt->namelen, 463 valuelen, 464 args.value); 465 kmem_free(args.value); 466 } else { 467 retval = context->put_listent(context, 468 entry->flags, 469 name_rmt->name, 470 (int)name_rmt->namelen, 471 valuelen, 472 NULL); 473 } 474 if (retval) 475 return retval; 476 } 477 if (context->seen_enough) 478 break; 479 cursor->offset++; 480 } 481 trace_xfs_attr_list_leaf_end(context); 482 return retval; 483 } 484 485 /* 486 * Copy out attribute entries for attr_list(), for leaf attribute lists. 487 */ 488 STATIC int 489 xfs_attr_leaf_list(xfs_attr_list_context_t *context) 490 { 491 int error; 492 struct xfs_buf *bp; 493 494 trace_xfs_attr_leaf_list(context); 495 496 context->cursor->blkno = 0; 497 error = xfs_attr3_leaf_read(NULL, context->dp, 0, -1, &bp); 498 if (error) 499 return XFS_ERROR(error); 500 501 error = xfs_attr3_leaf_list_int(bp, context); 502 xfs_trans_brelse(NULL, bp); 503 return XFS_ERROR(error); 504 } 505 506 int 507 xfs_attr_list_int( 508 xfs_attr_list_context_t *context) 509 { 510 int error; 511 xfs_inode_t *dp = context->dp; 512 513 XFS_STATS_INC(xs_attr_list); 514 515 if (XFS_FORCED_SHUTDOWN(dp->i_mount)) 516 return EIO; 517 518 xfs_ilock(dp, XFS_ILOCK_SHARED); 519 520 /* 521 * Decide on what work routines to call based on the inode size. 522 */ 523 if (!xfs_inode_hasattr(dp)) { 524 error = 0; 525 } else if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) { 526 error = xfs_attr_shortform_list(context); 527 } else if (xfs_bmap_one_block(dp, XFS_ATTR_FORK)) { 528 error = xfs_attr_leaf_list(context); 529 } else { 530 error = xfs_attr_node_list(context); 531 } 532 533 xfs_iunlock(dp, XFS_ILOCK_SHARED); 534 535 return error; 536 } 537 538 #define ATTR_ENTBASESIZE /* minimum bytes used by an attr */ \ 539 (((struct attrlist_ent *) 0)->a_name - (char *) 0) 540 #define ATTR_ENTSIZE(namelen) /* actual bytes used by an attr */ \ 541 ((ATTR_ENTBASESIZE + (namelen) + 1 + sizeof(u_int32_t)-1) \ 542 & ~(sizeof(u_int32_t)-1)) 543 544 /* 545 * Format an attribute and copy it out to the user's buffer. 546 * Take care to check values and protect against them changing later, 547 * we may be reading them directly out of a user buffer. 548 */ 549 STATIC int 550 xfs_attr_put_listent( 551 xfs_attr_list_context_t *context, 552 int flags, 553 unsigned char *name, 554 int namelen, 555 int valuelen, 556 unsigned char *value) 557 { 558 struct attrlist *alist = (struct attrlist *)context->alist; 559 attrlist_ent_t *aep; 560 int arraytop; 561 562 ASSERT(!(context->flags & ATTR_KERNOVAL)); 563 ASSERT(context->count >= 0); 564 ASSERT(context->count < (ATTR_MAX_VALUELEN/8)); 565 ASSERT(context->firstu >= sizeof(*alist)); 566 ASSERT(context->firstu <= context->bufsize); 567 568 /* 569 * Only list entries in the right namespace. 570 */ 571 if (((context->flags & ATTR_SECURE) == 0) != 572 ((flags & XFS_ATTR_SECURE) == 0)) 573 return 0; 574 if (((context->flags & ATTR_ROOT) == 0) != 575 ((flags & XFS_ATTR_ROOT) == 0)) 576 return 0; 577 578 arraytop = sizeof(*alist) + 579 context->count * sizeof(alist->al_offset[0]); 580 context->firstu -= ATTR_ENTSIZE(namelen); 581 if (context->firstu < arraytop) { 582 trace_xfs_attr_list_full(context); 583 alist->al_more = 1; 584 context->seen_enough = 1; 585 return 1; 586 } 587 588 aep = (attrlist_ent_t *)&context->alist[context->firstu]; 589 aep->a_valuelen = valuelen; 590 memcpy(aep->a_name, name, namelen); 591 aep->a_name[namelen] = 0; 592 alist->al_offset[context->count++] = context->firstu; 593 alist->al_count = context->count; 594 trace_xfs_attr_list_add(context); 595 return 0; 596 } 597 598 /* 599 * Generate a list of extended attribute names and optionally 600 * also value lengths. Positive return value follows the XFS 601 * convention of being an error, zero or negative return code 602 * is the length of the buffer returned (negated), indicating 603 * success. 604 */ 605 int 606 xfs_attr_list( 607 xfs_inode_t *dp, 608 char *buffer, 609 int bufsize, 610 int flags, 611 attrlist_cursor_kern_t *cursor) 612 { 613 xfs_attr_list_context_t context; 614 struct attrlist *alist; 615 int error; 616 617 /* 618 * Validate the cursor. 619 */ 620 if (cursor->pad1 || cursor->pad2) 621 return(XFS_ERROR(EINVAL)); 622 if ((cursor->initted == 0) && 623 (cursor->hashval || cursor->blkno || cursor->offset)) 624 return XFS_ERROR(EINVAL); 625 626 /* 627 * Check for a properly aligned buffer. 628 */ 629 if (((long)buffer) & (sizeof(int)-1)) 630 return XFS_ERROR(EFAULT); 631 if (flags & ATTR_KERNOVAL) 632 bufsize = 0; 633 634 /* 635 * Initialize the output buffer. 636 */ 637 memset(&context, 0, sizeof(context)); 638 context.dp = dp; 639 context.cursor = cursor; 640 context.resynch = 1; 641 context.flags = flags; 642 context.alist = buffer; 643 context.bufsize = (bufsize & ~(sizeof(int)-1)); /* align */ 644 context.firstu = context.bufsize; 645 context.put_listent = xfs_attr_put_listent; 646 647 alist = (struct attrlist *)context.alist; 648 alist->al_count = 0; 649 alist->al_more = 0; 650 alist->al_offset[0] = context.bufsize; 651 652 error = xfs_attr_list_int(&context); 653 ASSERT(error >= 0); 654 return error; 655 } 656