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_format.h" 22 #include "xfs_log_format.h" 23 #include "xfs_trans_resv.h" 24 #include "xfs_bit.h" 25 #include "xfs_sb.h" 26 #include "xfs_ag.h" 27 #include "xfs_mount.h" 28 #include "xfs_da_format.h" 29 #include "xfs_da_btree.h" 30 #include "xfs_inode.h" 31 #include "xfs_trans.h" 32 #include "xfs_inode_item.h" 33 #include "xfs_bmap.h" 34 #include "xfs_attr.h" 35 #include "xfs_attr_sf.h" 36 #include "xfs_attr_remote.h" 37 #include "xfs_attr_leaf.h" 38 #include "xfs_error.h" 39 #include "xfs_trace.h" 40 #include "xfs_buf_item.h" 41 #include "xfs_cksum.h" 42 #include "xfs_dinode.h" 43 #include "xfs_dir2.h" 44 45 STATIC int 46 xfs_attr_shortform_compare(const void *a, const void *b) 47 { 48 xfs_attr_sf_sort_t *sa, *sb; 49 50 sa = (xfs_attr_sf_sort_t *)a; 51 sb = (xfs_attr_sf_sort_t *)b; 52 if (sa->hash < sb->hash) { 53 return(-1); 54 } else if (sa->hash > sb->hash) { 55 return(1); 56 } else { 57 return(sa->entno - sb->entno); 58 } 59 } 60 61 #define XFS_ISRESET_CURSOR(cursor) \ 62 (!((cursor)->initted) && !((cursor)->hashval) && \ 63 !((cursor)->blkno) && !((cursor)->offset)) 64 /* 65 * Copy out entries of shortform attribute lists for attr_list(). 66 * Shortform attribute lists are not stored in hashval sorted order. 67 * If the output buffer is not large enough to hold them all, then we 68 * we have to calculate each entries' hashvalue and sort them before 69 * we can begin returning them to the user. 70 */ 71 int 72 xfs_attr_shortform_list(xfs_attr_list_context_t *context) 73 { 74 attrlist_cursor_kern_t *cursor; 75 xfs_attr_sf_sort_t *sbuf, *sbp; 76 xfs_attr_shortform_t *sf; 77 xfs_attr_sf_entry_t *sfe; 78 xfs_inode_t *dp; 79 int sbsize, nsbuf, count, i; 80 int error; 81 82 ASSERT(context != NULL); 83 dp = context->dp; 84 ASSERT(dp != NULL); 85 ASSERT(dp->i_afp != NULL); 86 sf = (xfs_attr_shortform_t *)dp->i_afp->if_u1.if_data; 87 ASSERT(sf != NULL); 88 if (!sf->hdr.count) 89 return(0); 90 cursor = context->cursor; 91 ASSERT(cursor != NULL); 92 93 trace_xfs_attr_list_sf(context); 94 95 /* 96 * If the buffer is large enough and the cursor is at the start, 97 * do not bother with sorting since we will return everything in 98 * one buffer and another call using the cursor won't need to be 99 * made. 100 * Note the generous fudge factor of 16 overhead bytes per entry. 101 * If bufsize is zero then put_listent must be a search function 102 * and can just scan through what we have. 103 */ 104 if (context->bufsize == 0 || 105 (XFS_ISRESET_CURSOR(cursor) && 106 (dp->i_afp->if_bytes + sf->hdr.count * 16) < context->bufsize)) { 107 for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) { 108 error = context->put_listent(context, 109 sfe->flags, 110 sfe->nameval, 111 (int)sfe->namelen, 112 (int)sfe->valuelen, 113 &sfe->nameval[sfe->namelen]); 114 115 /* 116 * Either search callback finished early or 117 * didn't fit it all in the buffer after all. 118 */ 119 if (context->seen_enough) 120 break; 121 122 if (error) 123 return error; 124 sfe = XFS_ATTR_SF_NEXTENTRY(sfe); 125 } 126 trace_xfs_attr_list_sf_all(context); 127 return(0); 128 } 129 130 /* do no more for a search callback */ 131 if (context->bufsize == 0) 132 return 0; 133 134 /* 135 * It didn't all fit, so we have to sort everything on hashval. 136 */ 137 sbsize = sf->hdr.count * sizeof(*sbuf); 138 sbp = sbuf = kmem_alloc(sbsize, KM_SLEEP | KM_NOFS); 139 140 /* 141 * Scan the attribute list for the rest of the entries, storing 142 * the relevant info from only those that match into a buffer. 143 */ 144 nsbuf = 0; 145 for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) { 146 if (unlikely( 147 ((char *)sfe < (char *)sf) || 148 ((char *)sfe >= ((char *)sf + dp->i_afp->if_bytes)))) { 149 XFS_CORRUPTION_ERROR("xfs_attr_shortform_list", 150 XFS_ERRLEVEL_LOW, 151 context->dp->i_mount, sfe); 152 kmem_free(sbuf); 153 return XFS_ERROR(EFSCORRUPTED); 154 } 155 156 sbp->entno = i; 157 sbp->hash = xfs_da_hashname(sfe->nameval, sfe->namelen); 158 sbp->name = sfe->nameval; 159 sbp->namelen = sfe->namelen; 160 /* These are bytes, and both on-disk, don't endian-flip */ 161 sbp->valuelen = sfe->valuelen; 162 sbp->flags = sfe->flags; 163 sfe = XFS_ATTR_SF_NEXTENTRY(sfe); 164 sbp++; 165 nsbuf++; 166 } 167 168 /* 169 * Sort the entries on hash then entno. 170 */ 171 xfs_sort(sbuf, nsbuf, sizeof(*sbuf), xfs_attr_shortform_compare); 172 173 /* 174 * Re-find our place IN THE SORTED LIST. 175 */ 176 count = 0; 177 cursor->initted = 1; 178 cursor->blkno = 0; 179 for (sbp = sbuf, i = 0; i < nsbuf; i++, sbp++) { 180 if (sbp->hash == cursor->hashval) { 181 if (cursor->offset == count) { 182 break; 183 } 184 count++; 185 } else if (sbp->hash > cursor->hashval) { 186 break; 187 } 188 } 189 if (i == nsbuf) { 190 kmem_free(sbuf); 191 return(0); 192 } 193 194 /* 195 * Loop putting entries into the user buffer. 196 */ 197 for ( ; i < nsbuf; i++, sbp++) { 198 if (cursor->hashval != sbp->hash) { 199 cursor->hashval = sbp->hash; 200 cursor->offset = 0; 201 } 202 error = context->put_listent(context, 203 sbp->flags, 204 sbp->name, 205 sbp->namelen, 206 sbp->valuelen, 207 &sbp->name[sbp->namelen]); 208 if (error) 209 return error; 210 if (context->seen_enough) 211 break; 212 cursor->offset++; 213 } 214 215 kmem_free(sbuf); 216 return(0); 217 } 218 219 STATIC int 220 xfs_attr_node_list(xfs_attr_list_context_t *context) 221 { 222 attrlist_cursor_kern_t *cursor; 223 xfs_attr_leafblock_t *leaf; 224 xfs_da_intnode_t *node; 225 struct xfs_attr3_icleaf_hdr leafhdr; 226 struct xfs_da3_icnode_hdr nodehdr; 227 struct xfs_da_node_entry *btree; 228 int error, i; 229 struct xfs_buf *bp; 230 struct xfs_inode *dp = context->dp; 231 232 trace_xfs_attr_node_list(context); 233 234 cursor = context->cursor; 235 cursor->initted = 1; 236 237 /* 238 * Do all sorts of validation on the passed-in cursor structure. 239 * If anything is amiss, ignore the cursor and look up the hashval 240 * starting from the btree root. 241 */ 242 bp = NULL; 243 if (cursor->blkno > 0) { 244 error = xfs_da3_node_read(NULL, dp, cursor->blkno, -1, 245 &bp, XFS_ATTR_FORK); 246 if ((error != 0) && (error != EFSCORRUPTED)) 247 return(error); 248 if (bp) { 249 struct xfs_attr_leaf_entry *entries; 250 251 node = bp->b_addr; 252 switch (be16_to_cpu(node->hdr.info.magic)) { 253 case XFS_DA_NODE_MAGIC: 254 case XFS_DA3_NODE_MAGIC: 255 trace_xfs_attr_list_wrong_blk(context); 256 xfs_trans_brelse(NULL, bp); 257 bp = NULL; 258 break; 259 case XFS_ATTR_LEAF_MAGIC: 260 case XFS_ATTR3_LEAF_MAGIC: 261 leaf = bp->b_addr; 262 xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf); 263 entries = xfs_attr3_leaf_entryp(leaf); 264 if (cursor->hashval > be32_to_cpu( 265 entries[leafhdr.count - 1].hashval)) { 266 trace_xfs_attr_list_wrong_blk(context); 267 xfs_trans_brelse(NULL, bp); 268 bp = NULL; 269 } else if (cursor->hashval <= be32_to_cpu( 270 entries[0].hashval)) { 271 trace_xfs_attr_list_wrong_blk(context); 272 xfs_trans_brelse(NULL, bp); 273 bp = NULL; 274 } 275 break; 276 default: 277 trace_xfs_attr_list_wrong_blk(context); 278 xfs_trans_brelse(NULL, bp); 279 bp = NULL; 280 } 281 } 282 } 283 284 /* 285 * We did not find what we expected given the cursor's contents, 286 * so we start from the top and work down based on the hash value. 287 * Note that start of node block is same as start of leaf block. 288 */ 289 if (bp == NULL) { 290 cursor->blkno = 0; 291 for (;;) { 292 __uint16_t magic; 293 294 error = xfs_da3_node_read(NULL, dp, 295 cursor->blkno, -1, &bp, 296 XFS_ATTR_FORK); 297 if (error) 298 return(error); 299 node = bp->b_addr; 300 magic = be16_to_cpu(node->hdr.info.magic); 301 if (magic == XFS_ATTR_LEAF_MAGIC || 302 magic == XFS_ATTR3_LEAF_MAGIC) 303 break; 304 if (magic != XFS_DA_NODE_MAGIC && 305 magic != XFS_DA3_NODE_MAGIC) { 306 XFS_CORRUPTION_ERROR("xfs_attr_node_list(3)", 307 XFS_ERRLEVEL_LOW, 308 context->dp->i_mount, 309 node); 310 xfs_trans_brelse(NULL, bp); 311 return XFS_ERROR(EFSCORRUPTED); 312 } 313 314 dp->d_ops->node_hdr_from_disk(&nodehdr, node); 315 btree = dp->d_ops->node_tree_p(node); 316 for (i = 0; i < nodehdr.count; btree++, i++) { 317 if (cursor->hashval 318 <= be32_to_cpu(btree->hashval)) { 319 cursor->blkno = be32_to_cpu(btree->before); 320 trace_xfs_attr_list_node_descend(context, 321 btree); 322 break; 323 } 324 } 325 if (i == nodehdr.count) { 326 xfs_trans_brelse(NULL, bp); 327 return 0; 328 } 329 xfs_trans_brelse(NULL, bp); 330 } 331 } 332 ASSERT(bp != NULL); 333 334 /* 335 * Roll upward through the blocks, processing each leaf block in 336 * order. As long as there is space in the result buffer, keep 337 * adding the information. 338 */ 339 for (;;) { 340 leaf = bp->b_addr; 341 error = xfs_attr3_leaf_list_int(bp, context); 342 if (error) { 343 xfs_trans_brelse(NULL, bp); 344 return error; 345 } 346 xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf); 347 if (context->seen_enough || leafhdr.forw == 0) 348 break; 349 cursor->blkno = leafhdr.forw; 350 xfs_trans_brelse(NULL, bp); 351 error = xfs_attr3_leaf_read(NULL, dp, cursor->blkno, -1, &bp); 352 if (error) 353 return error; 354 } 355 xfs_trans_brelse(NULL, bp); 356 return 0; 357 } 358 359 /* 360 * Copy out attribute list entries for attr_list(), for leaf attribute lists. 361 */ 362 int 363 xfs_attr3_leaf_list_int( 364 struct xfs_buf *bp, 365 struct xfs_attr_list_context *context) 366 { 367 struct attrlist_cursor_kern *cursor; 368 struct xfs_attr_leafblock *leaf; 369 struct xfs_attr3_icleaf_hdr ichdr; 370 struct xfs_attr_leaf_entry *entries; 371 struct xfs_attr_leaf_entry *entry; 372 int retval; 373 int i; 374 375 trace_xfs_attr_list_leaf(context); 376 377 leaf = bp->b_addr; 378 xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); 379 entries = xfs_attr3_leaf_entryp(leaf); 380 381 cursor = context->cursor; 382 cursor->initted = 1; 383 384 /* 385 * Re-find our place in the leaf block if this is a new syscall. 386 */ 387 if (context->resynch) { 388 entry = &entries[0]; 389 for (i = 0; i < ichdr.count; entry++, i++) { 390 if (be32_to_cpu(entry->hashval) == cursor->hashval) { 391 if (cursor->offset == context->dupcnt) { 392 context->dupcnt = 0; 393 break; 394 } 395 context->dupcnt++; 396 } else if (be32_to_cpu(entry->hashval) > 397 cursor->hashval) { 398 context->dupcnt = 0; 399 break; 400 } 401 } 402 if (i == ichdr.count) { 403 trace_xfs_attr_list_notfound(context); 404 return 0; 405 } 406 } else { 407 entry = &entries[0]; 408 i = 0; 409 } 410 context->resynch = 0; 411 412 /* 413 * We have found our place, start copying out the new attributes. 414 */ 415 retval = 0; 416 for (; i < ichdr.count; entry++, i++) { 417 if (be32_to_cpu(entry->hashval) != cursor->hashval) { 418 cursor->hashval = be32_to_cpu(entry->hashval); 419 cursor->offset = 0; 420 } 421 422 if (entry->flags & XFS_ATTR_INCOMPLETE) 423 continue; /* skip incomplete entries */ 424 425 if (entry->flags & XFS_ATTR_LOCAL) { 426 xfs_attr_leaf_name_local_t *name_loc = 427 xfs_attr3_leaf_name_local(leaf, i); 428 429 retval = context->put_listent(context, 430 entry->flags, 431 name_loc->nameval, 432 (int)name_loc->namelen, 433 be16_to_cpu(name_loc->valuelen), 434 &name_loc->nameval[name_loc->namelen]); 435 if (retval) 436 return retval; 437 } else { 438 xfs_attr_leaf_name_remote_t *name_rmt = 439 xfs_attr3_leaf_name_remote(leaf, i); 440 441 int valuelen = be32_to_cpu(name_rmt->valuelen); 442 443 if (context->put_value) { 444 xfs_da_args_t args; 445 446 memset((char *)&args, 0, sizeof(args)); 447 args.geo = context->dp->i_mount->m_attr_geo; 448 args.dp = context->dp; 449 args.whichfork = XFS_ATTR_FORK; 450 args.valuelen = valuelen; 451 args.rmtvaluelen = 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 uint lock_mode; 513 514 XFS_STATS_INC(xs_attr_list); 515 516 if (XFS_FORCED_SHUTDOWN(dp->i_mount)) 517 return EIO; 518 519 /* 520 * Decide on what work routines to call based on the inode size. 521 */ 522 lock_mode = xfs_ilock_attr_map_shared(dp); 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 xfs_iunlock(dp, lock_mode); 533 return error; 534 } 535 536 #define ATTR_ENTBASESIZE /* minimum bytes used by an attr */ \ 537 (((struct attrlist_ent *) 0)->a_name - (char *) 0) 538 #define ATTR_ENTSIZE(namelen) /* actual bytes used by an attr */ \ 539 ((ATTR_ENTBASESIZE + (namelen) + 1 + sizeof(u_int32_t)-1) \ 540 & ~(sizeof(u_int32_t)-1)) 541 542 /* 543 * Format an attribute and copy it out to the user's buffer. 544 * Take care to check values and protect against them changing later, 545 * we may be reading them directly out of a user buffer. 546 */ 547 STATIC int 548 xfs_attr_put_listent( 549 xfs_attr_list_context_t *context, 550 int flags, 551 unsigned char *name, 552 int namelen, 553 int valuelen, 554 unsigned char *value) 555 { 556 struct attrlist *alist = (struct attrlist *)context->alist; 557 attrlist_ent_t *aep; 558 int arraytop; 559 560 ASSERT(!(context->flags & ATTR_KERNOVAL)); 561 ASSERT(context->count >= 0); 562 ASSERT(context->count < (ATTR_MAX_VALUELEN/8)); 563 ASSERT(context->firstu >= sizeof(*alist)); 564 ASSERT(context->firstu <= context->bufsize); 565 566 /* 567 * Only list entries in the right namespace. 568 */ 569 if (((context->flags & ATTR_SECURE) == 0) != 570 ((flags & XFS_ATTR_SECURE) == 0)) 571 return 0; 572 if (((context->flags & ATTR_ROOT) == 0) != 573 ((flags & XFS_ATTR_ROOT) == 0)) 574 return 0; 575 576 arraytop = sizeof(*alist) + 577 context->count * sizeof(alist->al_offset[0]); 578 context->firstu -= ATTR_ENTSIZE(namelen); 579 if (context->firstu < arraytop) { 580 trace_xfs_attr_list_full(context); 581 alist->al_more = 1; 582 context->seen_enough = 1; 583 return 1; 584 } 585 586 aep = (attrlist_ent_t *)&context->alist[context->firstu]; 587 aep->a_valuelen = valuelen; 588 memcpy(aep->a_name, name, namelen); 589 aep->a_name[namelen] = 0; 590 alist->al_offset[context->count++] = context->firstu; 591 alist->al_count = context->count; 592 trace_xfs_attr_list_add(context); 593 return 0; 594 } 595 596 /* 597 * Generate a list of extended attribute names and optionally 598 * also value lengths. Positive return value follows the XFS 599 * convention of being an error, zero or negative return code 600 * is the length of the buffer returned (negated), indicating 601 * success. 602 */ 603 int 604 xfs_attr_list( 605 xfs_inode_t *dp, 606 char *buffer, 607 int bufsize, 608 int flags, 609 attrlist_cursor_kern_t *cursor) 610 { 611 xfs_attr_list_context_t context; 612 struct attrlist *alist; 613 int error; 614 615 /* 616 * Validate the cursor. 617 */ 618 if (cursor->pad1 || cursor->pad2) 619 return(XFS_ERROR(EINVAL)); 620 if ((cursor->initted == 0) && 621 (cursor->hashval || cursor->blkno || cursor->offset)) 622 return XFS_ERROR(EINVAL); 623 624 /* 625 * Check for a properly aligned buffer. 626 */ 627 if (((long)buffer) & (sizeof(int)-1)) 628 return XFS_ERROR(EFAULT); 629 if (flags & ATTR_KERNOVAL) 630 bufsize = 0; 631 632 /* 633 * Initialize the output buffer. 634 */ 635 memset(&context, 0, sizeof(context)); 636 context.dp = dp; 637 context.cursor = cursor; 638 context.resynch = 1; 639 context.flags = flags; 640 context.alist = buffer; 641 context.bufsize = (bufsize & ~(sizeof(int)-1)); /* align */ 642 context.firstu = context.bufsize; 643 context.put_listent = xfs_attr_put_listent; 644 645 alist = (struct attrlist *)context.alist; 646 alist->al_count = 0; 647 alist->al_more = 0; 648 alist->al_offset[0] = context.bufsize; 649 650 error = xfs_attr_list_int(&context); 651 ASSERT(error >= 0); 652 return error; 653 } 654