lib/libstand/hammer1.c

   1 /*
   2  * Copyright (c) 2008 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Simon Schubert <corecode@fs.ei.tum.de>
   6  * and Matthew Dillon <dillon@backplane.com>
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  *
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in
  16  *    the documentation and/or other materials provided with the
  17  *    distribution.
  18  * 3. Neither the name of The DragonFly Project nor the names of its
  19  *    contributors may be used to endorse or promote products derived
  20  *    from this software without specific, prior written permission.
  21  *
  22  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  23  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  24  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  25  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  26  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  27  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  28  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  29  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  30  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  31  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  32  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  33  * SUCH DAMAGE.
  34  */
  35
  36 /*
  37  * This file is being used by boot2 and libstand (loader).
  38  * Compile with -DTESTING to obtain a binary.
  39  */
  40
  41
  42 #if !defined(BOOT2) && !defined(TESTING)
  43 #define LIBSTAND        1
  44 #endif
  45
  46 #ifndef DEBUG
  47 #define DEBUG           0
  48 #endif
  49
  50 #ifdef BOOT2
  51 #include "boot2.h"
  52 #else
  53 #include <sys/param.h>
  54 #include <stddef.h>
  55 #include <stdint.h>
  56 #endif
  57
  58 #ifdef TESTING
  59 #include <sys/fcntl.h>
  60 #include <sys/stat.h>
  61 #include <unistd.h>
  62 #include <err.h>
  63 #include <stdio.h>
  64 #include <stdlib.h>
  65 #include <string.h>
  66 #include <errno.h>
  67 #include <dirent.h>
  68 #endif
  69
  70 #ifdef LIBSTAND
  71 #include "stand.h"
  72 #endif
  73
  74 #include <vfs/hammer/hammer_disk.h>
  75
  76 #ifndef BOOT2
  77 struct blockentry {
  78         hammer_off_t    off;
  79         int             use;
  80         char            *data;
  81 };
  82
  83 #ifdef TESTING
  84 #define NUMCACHE        16
  85 #else
  86 #define NUMCACHE        6
  87 #endif
  88
  89 struct hfs {
  90 #ifdef TESTING
  91         int             fd;
  92 #else   // libstand
  93         struct open_file *f;
  94 #endif
  95         hammer_off_t    root;
  96         int64_t         buf_beg;
  97         int64_t         last_dir_ino;
  98         u_int8_t        last_dir_cap_flags;
  99         int             lru;
 100         struct blockentry cache[NUMCACHE];
 101 };
 102
 103 static void *
 104 hread(struct hfs *hfs, hammer_off_t off)
 105 {
 106         hammer_off_t boff = off & ~HAMMER_BUFMASK64;
 107
 108         boff &= HAMMER_OFF_LONG_MASK;
 109
 110         if (HAMMER_ZONE_DECODE(off) != HAMMER_ZONE_RAW_VOLUME_INDEX)
 111                 boff += hfs->buf_beg;
 112
 113         struct blockentry *be = NULL;
 114         for (int i = 0; i < NUMCACHE; i++) {
 115                 if (be == NULL || be->use > hfs->cache[i].use)
 116                         be = &hfs->cache[i];
 117                 if (hfs->cache[i].off == boff) {
 118                         be = &hfs->cache[i];
 119                         break;
 120                 }
 121         }
 122         if (be->off != boff) {
 123                 // Didn't find any match
 124                 be->off = boff;
 125 #ifdef TESTING
 126                 ssize_t res = pread(hfs->fd, be->data, HAMMER_BUFSIZE,
 127                                     boff & HAMMER_OFF_SHORT_MASK);
 128                 if (res != HAMMER_BUFSIZE)
 129                         err(1, "short read on off %llx", boff);
 130 #else   // libstand
 131                 size_t rlen;
 132                 int rv = hfs->f->f_dev->dv_strategy(hfs->f->f_devdata, F_READ,
 133                         boff >> DEV_BSHIFT, HAMMER_BUFSIZE,
 134                         be->data, &rlen);
 135                 if (rv || rlen != HAMMER_BUFSIZE)
 136                         return (NULL);
 137 #endif
 138         }
 139
 140         be->use = ++hfs->lru;
 141         return &be->data[off & HAMMER_BUFMASK];
 142 }
 143
 144 #else   /* BOOT2 */
 145
 146 struct hammer_dmadat {
 147         struct boot2_dmadat boot2;
 148         char            buf[HAMMER_BUFSIZE];
 149 };
 150
 151 #define fsdmadat        ((struct hammer_dmadat *)boot2_dmadat)
 152
 153 struct hfs {
 154         hammer_off_t    root;
 155         int64_t         last_dir_ino;
 156         u_int8_t        last_dir_cap_flags;
 157         int64_t         buf_beg;
 158 };
 159
 160 static void *
 161 hread(struct hfs *hfs, hammer_off_t off)
 162 {
 163         char *buf = fsdmadat->buf;
 164
 165         hammer_off_t boff = off & ~HAMMER_BUFMASK64;
 166         boff &= HAMMER_OFF_LONG_MASK;
 167         if (HAMMER_ZONE_DECODE(off) != HAMMER_ZONE_RAW_VOLUME_INDEX)
 168                 boff += hfs->buf_beg;
 169         boff &= HAMMER_OFF_SHORT_MASK;
 170         boff >>= DEV_BSHIFT;
 171         if (dskread(buf, boff, HAMMER_BUFSIZE >> DEV_BSHIFT))
 172                 return (NULL);
 173         return (&buf[off & HAMMER_BUFMASK]);
 174 }
 175
 176 static void
 177 bzero(void *buf, size_t size)
 178 {
 179         for (size_t i = 0; i < size; i++)
 180                 ((char *)buf)[i] = 0;
 181 }
 182
 183 static void
 184 bcopy(void *src, void *dst, size_t size)
 185 {
 186         memcpy(dst, src, size);
 187 }
 188
 189 static size_t
 190 strlen(const char *s)
 191 {
 192         size_t l = 0;
 193         for (; *s != 0; s++)
 194                 l++;
 195         return (l);
 196 }
 197
 198 static int
 199 memcmp(const void *a, const void *b, size_t len)
 200 {
 201         for (size_t p = 0; p < len; p++) {
 202                 int r = ((const char *)a)[p] - ((const char *)b)[p];
 203                 if (r != 0)
 204                         return (r);
 205         }
 206
 207         return (0);
 208 }
 209
 210 #endif
 211
 212 /*
 213  * (from hammer_btree.c)
 214  *
 215  * Compare two B-Tree elements, return -N, 0, or +N (e.g. similar to strcmp).
 216  *
 217  * Note that for this particular function a return value of -1, 0, or +1
 218  * can denote a match if create_tid is otherwise discounted.  A create_tid
 219  * of zero is considered to be 'infinity' in comparisons.
 220  *
 221  * See also hammer_rec_rb_compare() and hammer_rec_cmp() in hammer_object.c.
 222  */
 223 static int
 224 hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2)
 225 {
 226         if (key1->localization < key2->localization)
 227                 return(-5);
 228         if (key1->localization > key2->localization)
 229                 return(5);
 230
 231         if (key1->obj_id < key2->obj_id)
 232                 return(-4);
 233         if (key1->obj_id > key2->obj_id)
 234                 return(4);
 235
 236         if (key1->rec_type < key2->rec_type)
 237                 return(-3);
 238         if (key1->rec_type > key2->rec_type)
 239                 return(3);
 240
 241         if (key1->key < key2->key)
 242                 return(-2);
 243         if (key1->key > key2->key)
 244                 return(2);
 245
 246         /*
 247          * A create_tid of zero indicates a record which is undeletable
 248          * and must be considered to have a value of positive infinity.
 249          */
 250         if (key1->create_tid == 0) {
 251                 if (key2->create_tid == 0)
 252                         return(0);
 253                 return(1);
 254         }
 255         if (key2->create_tid == 0)
 256                 return(-1);
 257         if (key1->create_tid < key2->create_tid)
 258                 return(-1);
 259         if (key1->create_tid > key2->create_tid)
 260                 return(1);
 261         return(0);
 262 }
 263
 264 /*
 265  * Heuristical search for the first element whos comparison is <= 1.  May
 266  * return an index whos compare result is > 1 but may only return an index
 267  * whos compare result is <= 1 if it is the first element with that result.
 268  */
 269 static int
 270 hammer_btree_search_node(hammer_base_elm_t elm, hammer_node_ondisk_t node)
 271 {
 272         int b;
 273         int s;
 274         int i;
 275         int r;
 276
 277         /*
 278          * Don't bother if the node does not have very many elements
 279          */
 280         b = 0;
 281         s = node->count;
 282         while (s - b > 4) {
 283                 i = b + (s - b) / 2;
 284                 r = hammer_btree_cmp(elm, &node->elms[i].leaf.base);
 285                 if (r <= 1) {
 286                         s = i;
 287                 } else {
 288                         b = i;
 289                 }
 290         }
 291         return(b);
 292 }
 293
 294 #if 0
 295 /*
 296  * (from hammer_subs.c)
 297  *
 298  * Return a namekey hash.   The 64 bit namekey hash consists of a 32 bit
 299  * crc in the MSB and 0 in the LSB.  The caller will use the low bits to
 300  * generate a unique key and will scan all entries with the same upper
 301  * 32 bits when issuing a lookup.
 302  *
 303  * We strip bit 63 in order to provide a positive key, this way a seek
 304  * offset of 0 will represent the base of the directory.
 305  *
 306  * This function can never return 0.  We use the MSB-0 space to synthesize
 307  * artificial directory entries such as "." and "..".
 308  */
 309 static int64_t
 310 hammer_directory_namekey(const void *name, int len)
 311 {
 312         int64_t key;
 313
 314         key = (int64_t)(crc32(name, len) & 0x7FFFFFFF) << 32;
 315         if (key == 0)
 316                 key |= 0x100000000LL;
 317         return(key);
 318 }
 319 #else
 320 static int64_t
 321 hammer_directory_namekey(const void *name __unused, int len __unused)
 322 {
 323         return (0);
 324 }
 325 #endif
 326
 327
 328 #ifndef BOOT2
 329 /*
 330  * Misc
 331  */
 332 static u_int32_t
 333 hammer_to_unix_xid(hammer_uuid_t *uuid)
 334 {
 335         return(*(u_int32_t *)&uuid->node[2]);
 336 }
 337
 338 static int
 339 hammer_get_dtype(u_int8_t obj_type)
 340 {
 341         switch(obj_type) {
 342         case HAMMER_OBJTYPE_DIRECTORY:
 343                 return(DT_DIR);
 344         case HAMMER_OBJTYPE_REGFILE:
 345                 return(DT_REG);
 346         case HAMMER_OBJTYPE_DBFILE:
 347                 return(DT_DBF);
 348         case HAMMER_OBJTYPE_FIFO:
 349                 return(DT_FIFO);
 350         case HAMMER_OBJTYPE_SOCKET:
 351                 return(DT_SOCK);
 352         case HAMMER_OBJTYPE_CDEV:
 353                 return(DT_CHR);
 354         case HAMMER_OBJTYPE_BDEV:
 355                 return(DT_BLK);
 356         case HAMMER_OBJTYPE_SOFTLINK:
 357                 return(DT_LNK);
 358         default:
 359                 return(DT_UNKNOWN);
 360         }
 361         /* not reached */
 362 }
 363
 364 static int
 365 hammer_get_mode(u_int8_t obj_type)
 366 {
 367         switch(obj_type) {
 368         case HAMMER_OBJTYPE_DIRECTORY:
 369                 return(S_IFDIR);
 370         case HAMMER_OBJTYPE_REGFILE:
 371                 return(S_IFREG);
 372         case HAMMER_OBJTYPE_DBFILE:
 373                 return(S_IFDB);
 374         case HAMMER_OBJTYPE_FIFO:
 375                 return(S_IFIFO);
 376         case HAMMER_OBJTYPE_SOCKET:
 377                 return(S_IFSOCK);
 378         case HAMMER_OBJTYPE_CDEV:
 379                 return(S_IFCHR);
 380         case HAMMER_OBJTYPE_BDEV:
 381                 return(S_IFBLK);
 382         case HAMMER_OBJTYPE_SOFTLINK:
 383                 return(S_IFLNK);
 384         default:
 385                 return(0);
 386         }
 387         /* not reached */
 388 }
 389
 390 #if DEBUG > 1
 391 static void
 392 hprintb(hammer_base_elm_t e)
 393 {
 394         printf("%d/", e->localization);
 395         if (e->obj_id >> 32 != 0)
 396                 printf("%lx%08lx",
 397                        (long)(e->obj_id >> 32),
 398                        (long)(e->obj_id & 0xffffffff));
 399         else
 400                 printf("%lx", (long)e->obj_id);
 401         printf("/%d/", e->rec_type);
 402         if (e->key >> 32 != 0)
 403                 printf("%lx%08lx",
 404                        (long)(e->key >> 32),
 405                        (long)(e->key & 0xffffffff));
 406         else
 407                 printf("%lx", (long)e->key);
 408 #ifdef TESTING
 409         printf("/%llx/%llx", e->create_tid, e->delete_tid);
 410 #endif
 411 }
 412 #endif /* DEBUG > 1 */
 413 #endif /* !BOOT2 */
 414
 415 static hammer_btree_leaf_elm_t
 416 hfind(struct hfs *hfs, hammer_base_elm_t key, hammer_base_elm_t end)
 417 {
 418 #if DEBUG > 1
 419         printf("searching for ");
 420         hprintb(key);
 421         printf(" end ");
 422         hprintb(end);
 423         printf("\n");
 424 #endif
 425
 426         int n;
 427         int r;
 428         struct hammer_base_elm search = *key;
 429         struct hammer_base_elm backtrack;
 430         hammer_off_t nodeoff = hfs->root;
 431         hammer_node_ondisk_t node;
 432         hammer_btree_elm_t e = NULL;
 433         int internal;
 434
 435 loop:
 436         node = hread(hfs, nodeoff);
 437         if (node == NULL)
 438                 return (NULL);
 439         internal = node->type == HAMMER_BTREE_TYPE_INTERNAL;
 440
 441 #if DEBUG > 3
 442         for (int i = 0; i < node->count; i++) {
 443                 printf("E: ");
 444                 hprintb(&node->elms[i].base);
 445                 printf("\n");
 446         }
 447         if (internal) {
 448                 printf("B: ");
 449                 hprintb(&node->elms[node->count].base);
 450                 printf("\n");
 451         }
 452 #endif
 453
 454         n = hammer_btree_search_node(&search, node);
 455
 456         // In internal nodes, we cover the right boundary as well.
 457         // If we hit it, we'll backtrack.
 458         for (; n < node->count + internal; n++) {
 459                 e = &node->elms[n];
 460                 r = hammer_btree_cmp(&search, &e->base);
 461
 462                 if (r < 0)
 463                         break;
 464         }
 465
 466         // unless we stopped right on the left side, we need to back off a bit
 467         if (n > 0)
 468                 e = &node->elms[--n];
 469
 470 #if DEBUG > 2
 471         printf("  found: ");
 472         hprintb(&e->base);
 473         printf("\n");
 474 #endif
 475
 476         if (internal) {
 477                 // If we hit the right boundary, backtrack to
 478                 // the next higher level.
 479                 if (n == node->count)
 480                         goto backtrack;
 481                 nodeoff = e->internal.subtree_offset;
 482                 backtrack = (e+1)->base;
 483                 goto loop;
 484         }
 485
 486         r = hammer_btree_cmp(key, &e->base);
 487         // If we're more off than the createtid, take the next elem
 488         if (r > 1) {
 489                 e++;
 490                 n++;
 491         }
 492
 493         // Skip deleted elements
 494         while (n < node->count && e->base.delete_tid != 0) {
 495                 e++;
 496                 n++;
 497         }
 498
 499         // In the unfortunate event when there is no next
 500         // element in this node, we repeat the search with
 501         // a key beyond the right boundary
 502         if (n == node->count) {
 503 backtrack:
 504                 search = backtrack;
 505                 nodeoff = hfs->root;
 506
 507 #if DEBUG > 2
 508                 printf("hit right boundary (%d), resetting search to ",
 509                        node->count);
 510                 hprintb(&search);
 511                 printf("\n");
 512 #endif
 513                 goto loop;
 514         }
 515
 516 #if DEBUG > 1
 517         printf("  result: ");
 518         hprintb(&e->base);
 519         printf("\n");
 520 #endif
 521
 522         if (end != NULL)
 523                 if (hammer_btree_cmp(end, &e->base) < -1)
 524                         goto fail;
 525
 526         return (&e->leaf);
 527
 528 fail:
 529 #if DEBUG > 1
 530         printf("  fail.\n");
 531 #endif
 532         return (NULL);
 533 }
 534
 535 /*
 536  * Returns the directory entry localization field based on the directory
 537  * inode's capabilities.
 538  */
 539 static u_int32_t
 540 hdirlocalization(struct hfs *hfs, ino_t ino)
 541 {
 542         struct hammer_base_elm key;
 543
 544         if (ino != hfs->last_dir_ino) {
 545                 bzero(&key, sizeof(key));
 546                 key.obj_id = ino;
 547                 key.localization = HAMMER_LOCALIZE_INODE;
 548                 key.rec_type = HAMMER_RECTYPE_INODE;
 549                 hammer_btree_leaf_elm_t e;
 550                 hammer_data_ondisk_t ed;
 551
 552                 e = hfind(hfs, &key, &key);
 553                 if (e) {
 554                         ed = hread(hfs, e->data_offset);
 555                         if (ed) {
 556                                 hfs->last_dir_ino = ino;
 557                                 hfs->last_dir_cap_flags = ed->inode.cap_flags;
 558                         } else {
 559                                 printf("hdirlocal: no inode data for %llx\n",
 560                                         (long long)ino);
 561                         }
 562                 } else {
 563                         printf("hdirlocal: no inode entry for %llx\n",
 564                                 (long long)ino);
 565                 }
 566         }
 567         if (hfs->last_dir_cap_flags & HAMMER_INODE_CAP_DIR_LOCAL_INO)
 568                 return(HAMMER_LOCALIZE_INODE);
 569         else
 570                 return(HAMMER_LOCALIZE_MISC);
 571 }
 572
 573 #ifndef BOOT2
 574 static int
 575 hreaddir(struct hfs *hfs, ino_t ino, int64_t *off, struct dirent *de)
 576 {
 577         struct hammer_base_elm key, end;
 578
 579 #if DEBUG > 2
 580         printf("%s(%llx, %lld)\n", __func__, (long long)ino, *off);
 581 #endif
 582
 583         bzero(&key, sizeof(key));
 584         key.obj_id = ino;
 585         key.localization = hdirlocalization(hfs, ino);
 586         key.rec_type = HAMMER_RECTYPE_DIRENTRY;
 587         key.key = *off;
 588
 589         end = key;
 590         end.key = HAMMER_MAX_KEY;
 591
 592         hammer_btree_leaf_elm_t e;
 593
 594         e = hfind(hfs, &key, &end);
 595         if (e == NULL) {
 596                 errno = ENOENT;
 597                 return (-1);
 598         }
 599
 600         *off = e->base.key + 1;         // remember next pos
 601
 602         de->d_namlen = e->data_len - HAMMER_ENTRY_NAME_OFF;
 603         de->d_type = hammer_get_dtype(e->base.obj_type);
 604         hammer_data_ondisk_t ed = hread(hfs, e->data_offset);
 605         if (ed == NULL)
 606                 return (-1);
 607         de->d_ino = ed->entry.obj_id;
 608         bcopy(ed->entry.name, de->d_name, de->d_namlen);
 609         de->d_name[de->d_namlen] = 0;
 610
 611         return (0);
 612 }
 613 #endif
 614
 615 static ino_t
 616 hresolve(struct hfs *hfs, ino_t dirino, const char *name)
 617 {
 618         struct hammer_base_elm key, end;
 619         size_t namel = strlen(name);
 620
 621 #if DEBUG > 2
 622         printf("%s(%llx, %s)\n", __func__, (long long)dirino, name);
 623 #endif
 624
 625         bzero(&key, sizeof(key));
 626         key.obj_id = dirino;
 627         key.localization = hdirlocalization(hfs, dirino);
 628         key.key = hammer_directory_namekey(name, namel);
 629         key.rec_type = HAMMER_RECTYPE_DIRENTRY;
 630         end = key;
 631         end.key = HAMMER_MAX_KEY;
 632
 633         hammer_btree_leaf_elm_t e;
 634         while ((e = hfind(hfs, &key, &end)) != NULL) {
 635                 key.key = e->base.key + 1;
 636
 637                 size_t elen = e->data_len - HAMMER_ENTRY_NAME_OFF;
 638                 hammer_data_ondisk_t ed = hread(hfs, e->data_offset);
 639                 if (ed == NULL)
 640                         return (-1);
 641 #ifdef BOOT2
 642                 if (ls) {
 643                         for (int i = 0; i < elen; i++)
 644                                 putchar(ed->entry.name[i]);
 645                         putchar(' ');
 646                         ls = 2;
 647                         continue;
 648                 }
 649 #endif
 650                 if (elen == namel && memcmp(ed->entry.name, name, MIN(elen, namel)) == 0)
 651                         return (ed->entry.obj_id);
 652         }
 653
 654 #ifdef BOOT2
 655         if (ls == 2)
 656                 printf("\n");
 657 #endif
 658
 659         return -1;
 660 }
 661
 662 static ino_t
 663 hlookup(struct hfs *hfs, const char *path)
 664 {
 665 #if DEBUG > 2
 666         printf("%s(%s)\n", __func__, path);
 667 #endif
 668
 669 #ifdef BOOT2
 670         ls = 0;
 671 #endif
 672         ino_t ino = 1;
 673         do {
 674                 char name[MAXPATHLEN + 1];
 675                 while (*path == '/')
 676                         path++;
 677                 if (*path == 0)
 678                         break;
 679                 for (char *n = name; *path != 0 && *path != '/'; path++, n++) {
 680                         n[0] = *path;
 681                         n[1] = 0;
 682                 }
 683
 684 #ifdef BOOT2
 685                 // A single ? means "list"
 686                 if (name[0] == '?' && name[1] == 0)
 687                         ls = 1;
 688 #endif
 689
 690                 ino = hresolve(hfs, ino, name);
 691         } while (ino != (ino_t)-1 && *path != 0);
 692
 693         return (ino);
 694 }
 695
 696
 697 #ifndef BOOT2
 698 static int
 699 hstat(struct hfs *hfs, ino_t ino, struct stat* st)
 700 {
 701         struct hammer_base_elm key;
 702
 703 #if DEBUG > 2
 704         printf("%s(%llx)\n", __func__, (long long)ino);
 705 #endif
 706
 707         bzero(&key, sizeof(key));
 708         key.obj_id = ino;
 709         key.localization = HAMMER_LOCALIZE_INODE;
 710         key.rec_type = HAMMER_RECTYPE_INODE;
 711
 712         hammer_btree_leaf_elm_t e = hfind(hfs, &key, &key);
 713         if (e == NULL) {
 714 #ifndef BOOT2
 715                 errno = ENOENT;
 716 #endif
 717                 return -1;
 718         }
 719
 720         hammer_data_ondisk_t ed = hread(hfs, e->data_offset);
 721         if (ed == NULL)
 722                 return (-1);
 723
 724         st->st_mode = ed->inode.mode | hammer_get_mode(ed->inode.obj_type);
 725         st->st_uid = hammer_to_unix_xid(&ed->inode.uid);
 726         st->st_gid = hammer_to_unix_xid(&ed->inode.gid);
 727         st->st_size = ed->inode.size;
 728
 729         return (0);
 730 }
 731 #endif
 732
 733 static ssize_t
 734 hreadf(struct hfs *hfs, ino_t ino, int64_t off, int64_t len, char *buf)
 735 {
 736         int64_t startoff = off;
 737         struct hammer_base_elm key, end;
 738
 739         bzero(&key, sizeof(key));
 740         key.obj_id = ino;
 741         key.localization = HAMMER_LOCALIZE_MISC;
 742         key.rec_type = HAMMER_RECTYPE_DATA;
 743         end = key;
 744         end.key = HAMMER_MAX_KEY;
 745
 746         while (len > 0) {
 747                 key.key = off + 1;
 748                 hammer_btree_leaf_elm_t e = hfind(hfs, &key, &end);
 749                 int64_t dlen;
 750
 751                 if (e == NULL || off > e->base.key) {
 752                         bzero(buf, len);
 753                         off += len;
 754                         len = 0;
 755                         break;
 756                 }
 757
 758                 int64_t doff = e->base.key - e->data_len;
 759                 if (off < doff) {
 760                         // sparse file, beginning
 761                         dlen = doff - off;
 762                         dlen = MIN(dlen, len);
 763                         bzero(buf, dlen);
 764                 } else {
 765                         int64_t boff = off - doff;
 766                         hammer_off_t roff = e->data_offset;
 767
 768                         dlen = e->data_len;
 769                         dlen -= boff;
 770                         dlen = MIN(dlen, len);
 771
 772                         while (boff >= HAMMER_BUFSIZE) {
 773                                 boff -= HAMMER_BUFSIZE;
 774                                 roff += HAMMER_BUFSIZE;
 775                         }
 776
 777                         /*
 778                          * boff - relative offset in disk buffer (not aligned)
 779                          * roff - base offset of disk buffer     (not aligned)
 780                          * dlen - amount of data we think we can copy
 781                          *
 782                          * hread only reads 16K aligned buffers, check for
 783                          * a length overflow and truncate dlen appropriately.
 784                          */
 785                         if ((roff & ~HAMMER_BUFMASK64) != ((roff + boff + dlen - 1) & ~HAMMER_BUFMASK64))
 786                                 dlen = HAMMER_BUFSIZE - ((boff + roff) & HAMMER_BUFMASK);
 787                         char *data = hread(hfs, roff);
 788                         if (data == NULL)
 789                                 return (-1);
 790                         bcopy(data + boff, buf, dlen);
 791                 }
 792
 793                 buf += dlen;
 794                 off += dlen;
 795                 len -= dlen;
 796         }
 797
 798         return (off - startoff);
 799 }
 800
 801 #ifdef BOOT2
 802 struct hfs hfs;
 803
 804 static int
 805 boot2_hammer_init(void)
 806 {
 807         hammer_volume_ondisk_t volhead;
 808
 809         volhead = hread(&hfs, HAMMER_ZONE_ENCODE(1, 0));
 810         if (volhead == NULL)
 811                 return (-1);
 812         if (volhead->vol_signature != HAMMER_FSBUF_VOLUME)
 813                 return (-1);
 814         hfs.root = volhead->vol0_btree_root;
 815         hfs.buf_beg = volhead->vol_buf_beg;
 816         return (0);
 817 }
 818
 819 static boot2_ino_t
 820 boot2_hammer_lookup(const char *path)
 821 {
 822         ino_t ino = hlookup(&hfs, path);
 823
 824         if (ino == -1)
 825                 ino = 0;
 826
 827         fs_off = 0;
 828
 829         return (ino);
 830 }
 831
 832 static ssize_t
 833 boot2_hammer_read(boot2_ino_t ino, void *buf, size_t len)
 834 {
 835         ssize_t rlen = hreadf(&hfs, ino, fs_off, len, buf);
 836         if (rlen != -1)
 837                 fs_off += rlen;
 838         return (rlen);
 839 }
 840
 841 const struct boot2_fsapi boot2_hammer_api = {
 842         .fsinit = boot2_hammer_init,
 843         .fslookup = boot2_hammer_lookup,
 844         .fsread = boot2_hammer_read
 845 };
 846
 847 #endif
 848
 849 #ifndef BOOT2
 850 static int
 851 hinit(struct hfs *hfs)
 852 {
 853 #if DEBUG
 854         printf("hinit\n");
 855 #endif
 856         for (int i = 0; i < NUMCACHE; i++) {
 857                 hfs->cache[i].data = malloc(HAMMER_BUFSIZE);
 858                 hfs->cache[i].off = -1; // invalid
 859                 hfs->cache[i].use = 0;
 860
 861 #if DEBUG
 862                 if (hfs->cache[i].data == NULL)
 863                         printf("malloc failed\n");
 864 #endif
 865         }
 866         hfs->lru = 0;
 867         hfs->last_dir_ino = -1;
 868
 869         hammer_volume_ondisk_t volhead = hread(hfs, HAMMER_ZONE_ENCODE(1, 0));
 870
 871 #ifdef TESTING
 872         if (volhead) {
 873                 printf("signature: %svalid\n",
 874                        volhead->vol_signature != HAMMER_FSBUF_VOLUME ?
 875                                 "in" :
 876                                 "");
 877                 printf("name: %s\n", volhead->vol_label);
 878         }
 879 #endif
 880
 881         if (volhead == NULL || volhead->vol_signature != HAMMER_FSBUF_VOLUME) {
 882                 for (int i = 0; i < NUMCACHE; i++) {
 883                         free(hfs->cache[i].data);
 884                         hfs->cache[i].data = NULL;
 885                 }
 886                 errno = ENODEV;
 887                 return (-1);
 888         }
 889
 890         hfs->root = volhead->vol0_btree_root;
 891         hfs->buf_beg = volhead->vol_buf_beg;
 892
 893         return (0);
 894 }
 895
 896 static void
 897 hclose(struct hfs *hfs)
 898 {
 899 #if DEBUG
 900         printf("hclose\n");
 901 #endif
 902         for (int i = 0; i < NUMCACHE; i++) {
 903                 if (hfs->cache[i].data) {
 904                         free(hfs->cache[i].data);
 905                         hfs->cache[i].data = NULL;
 906                 }
 907         }
 908 }
 909 #endif
 910
 911 #ifdef LIBSTAND
 912 struct hfile {
 913         struct hfs      hfs;
 914         ino_t           ino;
 915         int64_t         fsize;
 916 };
 917
 918 static int
 919 hammer_open(const char *path, struct open_file *f)
 920 {
 921         struct hfile *hf = malloc(sizeof(*hf));
 922
 923         bzero(hf, sizeof(*hf));
 924         f->f_fsdata = hf;
 925         hf->hfs.f = f;
 926         f->f_offset = 0;
 927
 928         int rv = hinit(&hf->hfs);
 929         if (rv) {
 930                 f->f_fsdata = NULL;
 931                 free(hf);
 932                 return (rv);
 933         }
 934
 935 #if DEBUG
 936         printf("hammer_open %s %p\n", path, f);
 937 #endif
 938
 939         hf->ino = hlookup(&hf->hfs, path);
 940         if (hf->ino == -1)
 941                 goto fail;
 942
 943         struct stat st;
 944         if (hstat(&hf->hfs, hf->ino, &st) == -1)
 945                 goto fail;
 946         hf->fsize = st.st_size;
 947
 948 #if DEBUG
 949         printf("        %ld\n", (long)hf->fsize);
 950 #endif
 951
 952         return (0);
 953
 954 fail:
 955 #if DEBUG
 956         printf("hammer_open fail\n");
 957 #endif
 958         f->f_fsdata = NULL;
 959         hclose(&hf->hfs);
 960         free(hf);
 961         return (ENOENT);
 962 }
 963
 964 static int
 965 hammer_close(struct open_file *f)
 966 {
 967         struct hfile *hf = f->f_fsdata;
 968
 969         f->f_fsdata = NULL;
 970         if (hf) {
 971             hclose(&hf->hfs);
 972             free(hf);
 973         }
 974         return (0);
 975 }
 976
 977 static int
 978 hammer_read(struct open_file *f, void *buf, size_t len, size_t *resid)
 979 {
 980         struct hfile *hf = f->f_fsdata;
 981
 982 #if DEBUG
 983         printf("hammer_read %p %ld %ld\n", f, f->f_offset, len);
 984 #endif
 985
 986         if (f->f_offset >= hf->fsize)
 987                 return (EINVAL);
 988
 989         size_t maxlen = len;
 990         if (f->f_offset + len > hf->fsize)
 991                 maxlen = hf->fsize - f->f_offset;
 992
 993         ssize_t rlen = hreadf(&hf->hfs, hf->ino, f->f_offset, maxlen, buf);
 994         if (rlen == -1)
 995                 return (EINVAL);
 996
 997         f->f_offset += rlen;
 998
 999         *resid = len - rlen;
1000         return (0);
1001 }
1002
1003 static off_t
1004 hammer_seek(struct open_file *f, off_t offset, int whence)
1005 {
1006         struct hfile *hf = f->f_fsdata;
1007
1008         switch (whence) {
1009         case SEEK_SET:
1010                 f->f_offset = offset;
1011                 break;
1012         case SEEK_CUR:
1013                 f->f_offset += offset;
1014                 break;
1015         case SEEK_END:
1016                 f->f_offset = hf->fsize - offset;
1017                 break;
1018         default:
1019                 return (-1);
1020         }
1021         return (f->f_offset);
1022 }
1023
1024 static int
1025 hammer_stat(struct open_file *f, struct stat *st)
1026 {
1027         struct hfile *hf = f->f_fsdata;
1028
1029         return (hstat(&hf->hfs, hf->ino, st));
1030 }
1031
1032 static int
1033 hammer_readdir(struct open_file *f, struct dirent *d)
1034 {
1035         struct hfile *hf = f->f_fsdata;
1036
1037         int64_t off = f->f_offset;
1038         int rv = hreaddir(&hf->hfs, hf->ino, &off, d);
1039         f->f_offset = off;
1040         return (rv);
1041 }
1042
1043 // libstand
1044 struct fs_ops hammer_fsops = {
1045         "hammer",
1046         hammer_open,
1047         hammer_close,
1048         hammer_read,
1049         null_write,
1050         hammer_seek,
1051         hammer_stat,
1052         hammer_readdir
1053 };
1054 #endif  // LIBSTAND
1055
1056 #ifdef TESTING
1057 int
1058 main(int argc, char **argv)
1059 {
1060         if (argc < 2) {
1061                 fprintf(stderr, "usage: hammerread <dev>\n");
1062                 return (1);
1063         }
1064
1065         struct hfs hfs;
1066         hfs.fd = open(argv[1], O_RDONLY);
1067         if (hfs.fd == -1)
1068                 err(1, "unable to open %s", argv[1]);
1069
1070         if (hinit(&hfs) == -1)
1071                 err(1, "invalid hammerfs");
1072
1073         for (int i = 2; i < argc; i++) {
1074                 ino_t ino = hlookup(&hfs, argv[i]);
1075                 if (ino == (ino_t)-1) {
1076                         warn("hlookup %s", argv[i]);
1077                         continue;
1078                 }
1079
1080                 struct stat st;
1081                 if (hstat(&hfs, ino, &st)) {
1082                         warn("hstat %s", argv[i]);
1083                         continue;
1084                 }
1085
1086                 printf("%s %d/%d %o %lld\n",
1087                        argv[i],
1088                        st.st_uid, st.st_gid,
1089                        st.st_mode, st.st_size);
1090
1091                 if (S_ISDIR(st.st_mode)) {
1092                         int64_t off = 0;
1093                         struct dirent de;
1094                         while (hreaddir(&hfs, ino, &off, &de) == 0) {
1095                                 printf("%s %d %llx\n",
1096                                        de.d_name, de.d_type, de.d_ino);
1097                         }
1098                 } else if (S_ISREG(st.st_mode)) {
1099                         char *buf = malloc(100000);
1100                         int64_t off = 0;
1101                         while (off < st.st_size) {
1102                                 int64_t len = MIN(100000, st.st_size - off);
1103                                 int64_t rl = hreadf(&hfs, ino, off, len, buf);
1104                                 fwrite(buf, rl, 1, stdout);
1105                                 off += rl;
1106                         }
1107                         free(buf);
1108                 }
1109         }
1110
1111         return 0;
1112 }
1113 #endif