2 * Copyright (c) 2008 The DragonFly Project. All rights reserved.
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>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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
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.
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
35 * $DragonFly: src/lib/libstand/hammerread.c,v 1.2 2008/10/29 22:14:25 swildner Exp $
39 * This file is being used by boot2 and libstand (loader).
40 * Compile with -DTESTING to obtain a binary.
44 #if !defined(BOOT2) && !defined(TESTING)
51 #include <sys/param.h>
57 #include <sys/fcntl.h>
72 #include <vfs/hammer/hammer_disk.h>
96 u_int8_t last_dir_cap_flags;
98 struct blockentry cache[NUMCACHE];
102 hread(struct hfs *hfs, hammer_off_t off)
104 hammer_off_t boff = off & ~HAMMER_BUFMASK64;
106 boff &= HAMMER_OFF_LONG_MASK;
108 if (HAMMER_ZONE_DECODE(off) != HAMMER_ZONE_RAW_VOLUME_INDEX)
109 boff += hfs->buf_beg;
111 struct blockentry *be = NULL;
112 for (int i = 0; i < NUMCACHE; i++) {
113 if (be == NULL || be->use > hfs->cache[i].use)
115 if (hfs->cache[i].off == boff) {
120 if (be->off != boff) {
121 // Didn't find any match
124 ssize_t res = pread(hfs->fd, be->data, HAMMER_BUFSIZE,
125 boff & HAMMER_OFF_SHORT_MASK);
126 if (res != HAMMER_BUFSIZE)
127 err(1, "short read on off %llx", boff);
130 int rv = hfs->f->f_dev->dv_strategy(hfs->f->f_devdata, F_READ,
131 boff >> DEV_BSHIFT, HAMMER_BUFSIZE,
133 if (rv || rlen != HAMMER_BUFSIZE)
138 be->use = ++hfs->lru;
139 return &be->data[off & HAMMER_BUFMASK];
144 struct hammer_dmadat {
145 struct boot2_dmadat boot2;
146 char buf[HAMMER_BUFSIZE];
149 #define fsdmadat ((struct hammer_dmadat *)boot2_dmadat)
153 int64_t last_dir_ino;
154 u_int8_t last_dir_cap_flags;
159 hread(struct hfs *hfs, hammer_off_t off)
161 char *buf = fsdmadat->buf;
163 hammer_off_t boff = off & ~HAMMER_BUFMASK64;
164 boff &= HAMMER_OFF_LONG_MASK;
165 if (HAMMER_ZONE_DECODE(off) != HAMMER_ZONE_RAW_VOLUME_INDEX)
166 boff += hfs->buf_beg;
167 boff &= HAMMER_OFF_SHORT_MASK;
169 if (dskread(buf, boff, HAMMER_BUFSIZE >> DEV_BSHIFT))
171 return (&buf[off & HAMMER_BUFMASK]);
175 bzero(void *buf, size_t size)
177 for (size_t i = 0; i < size; i++)
178 ((char *)buf)[i] = 0;
182 bcopy(void *src, void *dst, size_t size)
184 memcpy(dst, src, size);
188 strlen(const char *s)
197 memcmp(const void *a, const void *b, size_t len)
199 for (size_t p = 0; p < len; p++) {
200 int r = ((const char *)a)[p] - ((const char *)b)[p];
211 * (from hammer_btree.c)
213 * Compare two B-Tree elements, return -N, 0, or +N (e.g. similar to strcmp).
215 * Note that for this particular function a return value of -1, 0, or +1
216 * can denote a match if create_tid is otherwise discounted. A create_tid
217 * of zero is considered to be 'infinity' in comparisons.
219 * See also hammer_rec_rb_compare() and hammer_rec_cmp() in hammer_object.c.
222 hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2)
224 if (key1->localization < key2->localization)
226 if (key1->localization > key2->localization)
229 if (key1->obj_id < key2->obj_id)
231 if (key1->obj_id > key2->obj_id)
234 if (key1->rec_type < key2->rec_type)
236 if (key1->rec_type > key2->rec_type)
239 if (key1->key < key2->key)
241 if (key1->key > key2->key)
245 * A create_tid of zero indicates a record which is undeletable
246 * and must be considered to have a value of positive infinity.
248 if (key1->create_tid == 0) {
249 if (key2->create_tid == 0)
253 if (key2->create_tid == 0)
255 if (key1->create_tid < key2->create_tid)
257 if (key1->create_tid > key2->create_tid)
263 * Heuristical search for the first element whos comparison is <= 1. May
264 * return an index whos compare result is > 1 but may only return an index
265 * whos compare result is <= 1 if it is the first element with that result.
268 hammer_btree_search_node(hammer_base_elm_t elm, hammer_node_ondisk_t node)
276 * Don't bother if the node does not have very many elements
282 r = hammer_btree_cmp(elm, &node->elms[i].leaf.base);
294 * (from hammer_subs.c)
296 * Return a namekey hash. The 64 bit namekey hash consists of a 32 bit
297 * crc in the MSB and 0 in the LSB. The caller will use the low bits to
298 * generate a unique key and will scan all entries with the same upper
299 * 32 bits when issuing a lookup.
301 * We strip bit 63 in order to provide a positive key, this way a seek
302 * offset of 0 will represent the base of the directory.
304 * This function can never return 0. We use the MSB-0 space to synthesize
305 * artificial directory entries such as "." and "..".
308 hammer_directory_namekey(const void *name, int len)
312 key = (int64_t)(crc32(name, len) & 0x7FFFFFFF) << 32;
314 key |= 0x100000000LL;
319 hammer_directory_namekey(const void *name __unused, int len __unused)
331 hammer_to_unix_xid(uuid_t *uuid)
333 return(*(u_int32_t *)&uuid->node[2]);
337 hammer_get_dtype(u_int8_t obj_type)
340 case HAMMER_OBJTYPE_DIRECTORY:
342 case HAMMER_OBJTYPE_REGFILE:
344 case HAMMER_OBJTYPE_DBFILE:
346 case HAMMER_OBJTYPE_FIFO:
348 case HAMMER_OBJTYPE_SOCKET:
350 case HAMMER_OBJTYPE_CDEV:
352 case HAMMER_OBJTYPE_BDEV:
354 case HAMMER_OBJTYPE_SOFTLINK:
363 hammer_get_mode(u_int8_t obj_type)
366 case HAMMER_OBJTYPE_DIRECTORY:
368 case HAMMER_OBJTYPE_REGFILE:
370 case HAMMER_OBJTYPE_DBFILE:
372 case HAMMER_OBJTYPE_FIFO:
374 case HAMMER_OBJTYPE_SOCKET:
376 case HAMMER_OBJTYPE_CDEV:
378 case HAMMER_OBJTYPE_BDEV:
380 case HAMMER_OBJTYPE_SOFTLINK:
390 hprintb(hammer_base_elm_t e)
392 printf("%d/", e->localization);
393 if (e->obj_id >> 32 != 0)
395 (long)(e->obj_id >> 32),
396 (long)(e->obj_id & 0xffffffff));
398 printf("%lx", (long)e->obj_id);
399 printf("/%d/", e->rec_type);
400 if (e->key >> 32 != 0)
402 (long)(e->key >> 32),
403 (long)(e->key & 0xffffffff));
405 printf("%lx", (long)e->key);
407 printf("/%llx/%llx", e->create_tid, e->delete_tid);
410 #endif /* DEBUG > 1 */
413 static hammer_btree_leaf_elm_t
414 hfind(struct hfs *hfs, hammer_base_elm_t key, hammer_base_elm_t end)
417 printf("searching for ");
426 struct hammer_base_elm search = *key;
427 struct hammer_base_elm backtrack;
428 hammer_off_t nodeoff = hfs->root;
429 hammer_node_ondisk_t node;
430 hammer_btree_elm_t e = NULL;
434 node = hread(hfs, nodeoff);
437 internal = node->type == HAMMER_BTREE_TYPE_INTERNAL;
440 for (int i = 0; i < node->count; i++) {
442 hprintb(&node->elms[i].base);
447 hprintb(&node->elms[node->count].base);
452 n = hammer_btree_search_node(&search, node);
454 // In internal nodes, we cover the right boundary as well.
455 // If we hit it, we'll backtrack.
456 for (; n < node->count + internal; n++) {
458 r = hammer_btree_cmp(&search, &e->base);
464 // unless we stopped right on the left side, we need to back off a bit
466 e = &node->elms[--n];
475 // If we hit the right boundary, backtrack to
476 // the next higher level.
477 if (n == node->count)
479 nodeoff = e->internal.subtree_offset;
480 backtrack = (e+1)->base;
484 r = hammer_btree_cmp(key, &e->base);
485 // If we're more off than the createtid, take the next elem
491 // Skip deleted elements
492 while (n < node->count && e->base.delete_tid != 0) {
497 // In the unfortunate event when there is no next
498 // element in this node, we repeat the search with
499 // a key beyond the right boundary
500 if (n == node->count) {
506 printf("hit right boundary (%d), resetting search to ",
521 if (hammer_btree_cmp(end, &e->base) < -1)
534 * Returns the directory entry localization field based on the directory
535 * inode's capabilities.
538 hdirlocalization(struct hfs *hfs, ino_t ino)
540 struct hammer_base_elm key;
542 if (ino != hfs->last_dir_ino) {
543 bzero(&key, sizeof(key));
545 key.localization = HAMMER_LOCALIZE_INODE;
546 key.rec_type = HAMMER_RECTYPE_INODE;
547 hammer_btree_leaf_elm_t e;
548 hammer_data_ondisk_t ed;
550 e = hfind(hfs, &key, &key);
552 ed = hread(hfs, e->data_offset);
554 hfs->last_dir_ino = ino;
555 hfs->last_dir_cap_flags = ed->inode.cap_flags;
557 printf("hdirlocal: no inode data for %llx\n",
561 printf("hdirlocal: no inode entry for %llx\n",
565 if (hfs->last_dir_cap_flags & HAMMER_INODE_CAP_DIR_LOCAL_INO)
566 return(HAMMER_LOCALIZE_INODE);
568 return(HAMMER_LOCALIZE_MISC);
573 hreaddir(struct hfs *hfs, ino_t ino, int64_t *off, struct dirent *de)
575 struct hammer_base_elm key, end;
578 printf("%s(%llx, %lld)\n", __FUNCTION__, (long long)ino, *off);
581 bzero(&key, sizeof(key));
583 key.localization = hdirlocalization(hfs, ino);
584 key.rec_type = HAMMER_RECTYPE_DIRENTRY;
588 end.key = HAMMER_MAX_KEY;
590 hammer_btree_leaf_elm_t e;
592 e = hfind(hfs, &key, &end);
598 *off = e->base.key + 1; // remember next pos
600 de->d_namlen = e->data_len - HAMMER_ENTRY_NAME_OFF;
601 de->d_type = hammer_get_dtype(e->base.obj_type);
602 hammer_data_ondisk_t ed = hread(hfs, e->data_offset);
605 de->d_ino = ed->entry.obj_id;
606 bcopy(ed->entry.name, de->d_name, de->d_namlen);
607 de->d_name[de->d_namlen] = 0;
614 hresolve(struct hfs *hfs, ino_t dirino, const char *name)
616 struct hammer_base_elm key, end;
617 size_t namel = strlen(name);
620 printf("%s(%llx, %s)\n", __FUNCTION__, (long long)dirino, name);
623 bzero(&key, sizeof(key));
625 key.localization = hdirlocalization(hfs, dirino);
626 key.key = hammer_directory_namekey(name, namel);
627 key.rec_type = HAMMER_RECTYPE_DIRENTRY;
629 end.key = HAMMER_MAX_KEY;
631 hammer_btree_leaf_elm_t e;
632 while ((e = hfind(hfs, &key, &end)) != NULL) {
633 key.key = e->base.key + 1;
635 size_t elen = e->data_len - HAMMER_ENTRY_NAME_OFF;
636 hammer_data_ondisk_t ed = hread(hfs, e->data_offset);
641 for (int i = 0; i < elen; i++)
642 putchar(ed->entry.name[i]);
648 if (elen == namel && memcmp(ed->entry.name, name, MIN(elen, namel)) == 0)
649 return (ed->entry.obj_id);
661 hlookup(struct hfs *hfs, const char *path)
664 printf("%s(%s)\n", __FUNCTION__, path);
672 char name[MAXPATHLEN + 1];
677 for (char *n = name; *path != 0 && *path != '/'; path++, n++) {
683 // A single ? means "list"
684 if (name[0] == '?' && name[1] == 0)
688 ino = hresolve(hfs, ino, name);
689 } while (ino != (ino_t)-1 && *path != 0);
697 hstat(struct hfs *hfs, ino_t ino, struct stat* st)
699 struct hammer_base_elm key;
702 printf("%s(%llx)\n", __FUNCTION__, (long long)ino);
705 bzero(&key, sizeof(key));
707 key.localization = HAMMER_LOCALIZE_INODE;
708 key.rec_type = HAMMER_RECTYPE_INODE;
710 hammer_btree_leaf_elm_t e = hfind(hfs, &key, &key);
718 hammer_data_ondisk_t ed = hread(hfs, e->data_offset);
722 st->st_mode = ed->inode.mode | hammer_get_mode(ed->inode.obj_type);
723 st->st_uid = hammer_to_unix_xid(&ed->inode.uid);
724 st->st_gid = hammer_to_unix_xid(&ed->inode.gid);
725 st->st_size = ed->inode.size;
732 hreadf(struct hfs *hfs, ino_t ino, int64_t off, int64_t len, char *buf)
734 int64_t startoff = off;
735 struct hammer_base_elm key, end;
737 bzero(&key, sizeof(key));
739 key.localization = HAMMER_LOCALIZE_MISC;
740 key.rec_type = HAMMER_RECTYPE_DATA;
742 end.key = HAMMER_MAX_KEY;
746 hammer_btree_leaf_elm_t e = hfind(hfs, &key, &end);
749 if (e == NULL || off > e->base.key) {
756 int64_t doff = e->base.key - e->data_len;
758 // sparse file, beginning
760 dlen = MIN(dlen, len);
763 int64_t boff = off - doff;
764 hammer_off_t roff = e->data_offset;
768 dlen = MIN(dlen, len);
770 while (boff >= HAMMER_BUFSIZE) {
771 boff -= HAMMER_BUFSIZE;
772 roff += HAMMER_BUFSIZE;
776 * boff - relative offset in disk buffer (not aligned)
777 * roff - base offset of disk buffer (not aligned)
778 * dlen - amount of data we think we can copy
780 * hread only reads 16K aligned buffers, check for
781 * a length overflow and truncate dlen appropriately.
783 if ((roff & ~HAMMER_BUFMASK64) != ((roff + boff + dlen - 1) & ~HAMMER_BUFMASK64))
784 dlen = HAMMER_BUFSIZE - ((boff + roff) & HAMMER_BUFMASK);
785 char *data = hread(hfs, roff);
788 bcopy(data + boff, buf, dlen);
796 return (off - startoff);
803 boot2_hammer_init(void)
805 hammer_volume_ondisk_t volhead;
807 volhead = hread(&hfs, HAMMER_ZONE_ENCODE(1, 0));
810 if (volhead->vol_signature != HAMMER_FSBUF_VOLUME)
812 hfs.root = volhead->vol0_btree_root;
813 hfs.buf_beg = volhead->vol_buf_beg;
818 boot2_hammer_lookup(const char *path)
820 ino_t ino = hlookup(&hfs, path);
831 boot2_hammer_read(boot2_ino_t ino, void *buf, size_t len)
833 ssize_t rlen = hreadf(&hfs, ino, fs_off, len, buf);
839 const struct boot2_fsapi boot2_hammer_api = {
840 .fsinit = boot2_hammer_init,
841 .fslookup = boot2_hammer_lookup,
842 .fsread = boot2_hammer_read
849 hinit(struct hfs *hfs)
854 for (int i = 0; i < NUMCACHE; i++) {
855 hfs->cache[i].data = malloc(HAMMER_BUFSIZE);
856 hfs->cache[i].off = -1; // invalid
857 hfs->cache[i].use = 0;
860 if (hfs->cache[i].data == NULL)
861 printf("malloc failed\n");
865 hfs->last_dir_ino = -1;
867 hammer_volume_ondisk_t volhead = hread(hfs, HAMMER_ZONE_ENCODE(1, 0));
871 printf("signature: %svalid\n",
872 volhead->vol_signature != HAMMER_FSBUF_VOLUME ?
875 printf("name: %s\n", volhead->vol_name);
879 if (volhead == NULL || volhead->vol_signature != HAMMER_FSBUF_VOLUME) {
880 for (int i = 0; i < NUMCACHE; i++) {
881 free(hfs->cache[i].data);
882 hfs->cache[i].data = NULL;
888 hfs->root = volhead->vol0_btree_root;
889 hfs->buf_beg = volhead->vol_buf_beg;
895 hclose(struct hfs *hfs)
900 for (int i = 0; i < NUMCACHE; i++) {
901 if (hfs->cache[i].data) {
902 free(hfs->cache[i].data);
903 hfs->cache[i].data = NULL;
917 hammer_open(const char *path, struct open_file *f)
919 struct hfile *hf = malloc(sizeof(*hf));
921 bzero(hf, sizeof(*hf));
926 int rv = hinit(&hf->hfs);
934 printf("hammer_open %s %p %ld\n", path, f);
937 hf->ino = hlookup(&hf->hfs, path);
942 if (hstat(&hf->hfs, hf->ino, &st) == -1)
944 hf->fsize = st.st_size;
947 printf(" %ld\n", (long)hf->fsize);
954 printf("hammer_open fail\n");
963 hammer_close(struct open_file *f)
965 struct hfile *hf = f->f_fsdata;
976 hammer_read(struct open_file *f, void *buf, size_t len, size_t *resid)
978 struct hfile *hf = f->f_fsdata;
981 printf("hammer_read %p %ld %ld\n", f, f->f_offset, len);
984 if (f->f_offset >= hf->fsize)
988 if (f->f_offset + len > hf->fsize)
989 maxlen = hf->fsize - f->f_offset;
991 ssize_t rlen = hreadf(&hf->hfs, hf->ino, f->f_offset, maxlen, buf);
1002 hammer_seek(struct open_file *f, off_t offset, int whence)
1004 struct hfile *hf = f->f_fsdata;
1008 f->f_offset = offset;
1011 f->f_offset += offset;
1014 f->f_offset = hf->fsize - offset;
1019 return (f->f_offset);
1023 hammer_stat(struct open_file *f, struct stat *st)
1025 struct hfile *hf = f->f_fsdata;
1027 return (hstat(&hf->hfs, hf->ino, st));
1031 hammer_readdir(struct open_file *f, struct dirent *d)
1033 struct hfile *hf = f->f_fsdata;
1035 int64_t off = f->f_offset;
1036 int rv = hreaddir(&hf->hfs, hf->ino, &off, d);
1042 struct fs_ops hammer_fsops = {
1056 main(int argc, char **argv)
1059 fprintf(stderr, "usage: hammerread <dev>\n");
1064 hfs.fd = open(argv[1], O_RDONLY);
1066 err(1, "unable to open %s", argv[1]);
1068 if (hinit(&hfs) == -1)
1069 err(1, "invalid hammerfs");
1071 for (int i = 2; i < argc; i++) {
1072 ino_t ino = hlookup(&hfs, argv[i]);
1073 if (ino == (ino_t)-1) {
1074 warn("hlookup %s", argv[i]);
1079 if (hstat(&hfs, ino, &st)) {
1080 warn("hstat %s", argv[i]);
1084 printf("%s %d/%d %o %lld\n",
1086 st.st_uid, st.st_gid,
1087 st.st_mode, st.st_size);
1089 if (S_ISDIR(st.st_mode)) {
1092 while (hreaddir(&hfs, ino, &off, &de) == 0) {
1093 printf("%s %d %llx\n",
1094 de.d_name, de.d_type, de.d_ino);
1096 } else if (S_ISREG(st.st_mode)) {
1097 char *buf = malloc(100000);
1099 while (off < st.st_size) {
1100 int64_t len = MIN(100000, st.st_size - off);
1101 int64_t rl = hreadf(&hfs, ino, off, len, buf);
1102 fwrite(buf, rl, 1, stdout);