2 * Copyright (c) 2007 Doug Rabson
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
33 * Stand-alone file reading package.
38 #include <sys/param.h>
40 #include <sys/queue.h>
45 #include <bootstrap.h>
51 /* Define the range of indexes to be populated with ZFS Boot Environments */
52 #define ZFS_BE_FIRST 4
55 static int zfs_open(const char *path, struct open_file *f);
56 static int zfs_close(struct open_file *f);
57 static int zfs_read(struct open_file *f, void *buf, size_t size, size_t *resid);
58 static off_t zfs_seek(struct open_file *f, off_t offset, int where);
59 static int zfs_stat(struct open_file *f, struct stat *sb);
60 static int zfs_readdir(struct open_file *f, struct dirent *d);
61 static int zfs_mount(const char *dev, const char *path, void **data);
62 static int zfs_unmount(const char *dev, void *data);
64 static void zfs_bootenv_initial(const char *envname, spa_t *spa,
65 const char *name, const char *dsname, int checkpoint);
66 static void zfs_checkpoints_initial(spa_t *spa, const char *name,
69 static int zfs_parsedev(struct devdesc **idev, const char *devspec,
74 struct fs_ops zfs_fsops = {
77 .fo_close = zfs_close,
79 .fo_write = null_write,
82 .fo_readdir = zfs_readdir,
83 .fo_mount = zfs_mount,
84 .fo_unmount = zfs_unmount
91 off_t f_seekp; /* seek pointer */
93 uint64_t f_zap_type; /* zap type for readdir */
94 uint64_t f_num_leafs; /* number of fzap leaf blocks */
95 zap_leaf_phys_t *f_zap_leaf; /* zap leaf buffer */
98 static int zfs_env_index;
99 static int zfs_env_count;
101 SLIST_HEAD(zfs_be_list, zfs_be_entry) zfs_be_head = SLIST_HEAD_INITIALIZER(zfs_be_head);
102 struct zfs_be_list *zfs_be_headp;
103 struct zfs_be_entry {
105 SLIST_ENTRY(zfs_be_entry) entries;
106 } *zfs_be, *zfs_be_tmp;
112 zfs_open(const char *upath, struct open_file *f)
114 struct devdesc *dev = f->f_devdata;
115 struct zfsmount *mount = dev->d_opendata;
119 if (f->f_dev != &zfs_dev)
122 /* allocate file system specific data structure */
123 fp = calloc(1, sizeof(struct file));
128 rc = zfs_lookup(mount, upath, &fp->f_dnode);
138 zfs_close(struct open_file *f)
140 struct file *fp = (struct file *)f->f_fsdata;
142 dnode_cache_obj = NULL;
150 * Copy a portion of a file into kernel memory.
151 * Cross block boundaries when necessary.
154 zfs_read(struct open_file *f, void *start, size_t size, size_t *resid /* out */)
156 struct devdesc *dev = f->f_devdata;
157 const spa_t *spa = ((struct zfsmount *)dev->d_opendata)->spa;
158 struct file *fp = (struct file *)f->f_fsdata;
163 rc = zfs_stat(f, &sb);
167 if (fp->f_seekp + n > sb.st_size)
168 n = sb.st_size - fp->f_seekp;
170 rc = dnode_read(spa, &fp->f_dnode, fp->f_seekp, start, n);
176 for (i = 0; i < n; i++)
177 putchar(((char*) start)[i]);
187 zfs_seek(struct open_file *f, off_t offset, int where)
189 struct file *fp = (struct file *)f->f_fsdata;
193 fp->f_seekp = offset;
196 fp->f_seekp += offset;
203 error = zfs_stat(f, &sb);
208 fp->f_seekp = sb.st_size - offset;
215 return (fp->f_seekp);
219 zfs_stat(struct open_file *f, struct stat *sb)
221 struct devdesc *dev = f->f_devdata;
222 const spa_t *spa = ((struct zfsmount *)dev->d_opendata)->spa;
223 struct file *fp = (struct file *)f->f_fsdata;
225 return (zfs_dnode_stat(spa, &fp->f_dnode, sb));
229 zfs_readdir(struct open_file *f, struct dirent *d)
231 struct devdesc *dev = f->f_devdata;
232 const spa_t *spa = ((struct zfsmount *)dev->d_opendata)->spa;
233 struct file *fp = (struct file *)f->f_fsdata;
236 size_t bsize = fp->f_dnode.dn_datablkszsec << SPA_MINBLOCKSHIFT;
239 rc = zfs_stat(f, &sb);
242 if (!S_ISDIR(sb.st_mode))
246 * If this is the first read, get the zap type.
248 if (fp->f_seekp == 0) {
249 rc = dnode_read(spa, &fp->f_dnode,
250 0, &fp->f_zap_type, sizeof(fp->f_zap_type));
254 if (fp->f_zap_type == ZBT_MICRO) {
255 fp->f_seekp = offsetof(mzap_phys_t, mz_chunk);
257 rc = dnode_read(spa, &fp->f_dnode,
258 offsetof(zap_phys_t, zap_num_leafs),
260 sizeof(fp->f_num_leafs));
265 fp->f_zap_leaf = malloc(bsize);
266 if (fp->f_zap_leaf == NULL)
268 rc = dnode_read(spa, &fp->f_dnode,
277 if (fp->f_zap_type == ZBT_MICRO) {
279 if (fp->f_seekp >= bsize)
282 rc = dnode_read(spa, &fp->f_dnode,
283 fp->f_seekp, &mze, sizeof(mze));
286 fp->f_seekp += sizeof(mze);
288 if (!mze.mze_name[0])
291 d->d_fileno = ZFS_DIRENT_OBJ(mze.mze_value);
292 d->d_type = ZFS_DIRENT_TYPE(mze.mze_value);
293 strcpy(d->d_name, mze.mze_name);
294 d->d_namlen = strlen(d->d_name);
298 zap_leaf_chunk_t *zc, *nc;
305 * Initialise this so we can use the ZAP size
306 * calculating macros.
308 zl.l_bs = ilog2(bsize);
309 zl.l_phys = fp->f_zap_leaf;
312 * Figure out which chunk we are currently looking at
313 * and consider seeking to the next leaf. We use the
314 * low bits of f_seekp as a simple chunk index.
317 chunk = fp->f_seekp & (bsize - 1);
318 if (chunk == ZAP_LEAF_NUMCHUNKS(&zl)) {
319 fp->f_seekp = rounddown2(fp->f_seekp, bsize) + bsize;
323 * Check for EOF and read the new leaf.
325 if (fp->f_seekp >= bsize * fp->f_num_leafs)
328 rc = dnode_read(spa, &fp->f_dnode,
336 zc = &ZAP_LEAF_CHUNK(&zl, chunk);
338 if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY)
341 namelen = zc->l_entry.le_name_numints;
342 if (namelen > sizeof(d->d_name))
343 namelen = sizeof(d->d_name);
346 * Paste the name back together.
348 nc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_name_chunk);
350 while (namelen > 0) {
353 if (len > ZAP_LEAF_ARRAY_BYTES)
354 len = ZAP_LEAF_ARRAY_BYTES;
355 memcpy(p, nc->l_array.la_array, len);
358 nc = &ZAP_LEAF_CHUNK(&zl, nc->l_array.la_next);
360 d->d_name[sizeof(d->d_name) - 1] = 0;
363 * Assume the first eight bytes of the value are
366 value = fzap_leaf_value(&zl, zc);
368 d->d_fileno = ZFS_DIRENT_OBJ(value);
369 d->d_type = ZFS_DIRENT_TYPE(value);
370 d->d_namlen = strlen(d->d_name);
377 spa_find_by_dev(struct zfs_devdesc *dev)
380 if (dev->dd.d_dev->dv_type != DEVT_ZFS)
383 if (dev->pool_guid == 0)
384 return (STAILQ_FIRST(&zfs_pools));
386 return (spa_find_by_guid(dev->pool_guid));
390 * if path is NULL, create mount structure, but do not add it to list.
393 zfs_mount(const char *dev, const char *path, void **data)
395 struct zfs_devdesc *zfsdev = NULL;
397 struct zfsmount *mnt = NULL;
401 rv = zfs_parsedev((struct devdesc **)&zfsdev, dev, NULL);
406 spa = spa_find_by_dev(zfsdev);
412 mnt = calloc(1, sizeof(*mnt));
418 if (mnt->path != NULL) {
419 mnt->path = strdup(path);
420 if (mnt->path == NULL) {
426 rv = zfs_mount_impl(spa, zfsdev->root_guid, mnt);
428 if (rv == 0 && mnt->objset.os_type != DMU_OST_ZFS) {
429 printf("Unexpected object set type %ju\n",
430 (uintmax_t)mnt->objset.os_type);
444 STAILQ_INSERT_TAIL(&zfsmount, mnt, next);
452 zfs_unmount(const char *dev, void *data)
454 struct zfsmount *mnt = data;
456 STAILQ_REMOVE(&zfsmount, mnt, zfsmount, next);
463 vdev_read(vdev_t *vdev, void *priv, off_t offset, void *buf, size_t bytes)
466 size_t res, head, tail, total_size, full_sec_size;
467 unsigned secsz, do_tail_read;
469 char *outbuf, *bouncebuf;
471 fd = (uintptr_t) priv;
472 outbuf = (char *) buf;
475 ret = ioctl(fd, DIOCGSECTORSIZE, &secsz);
480 * Handling reads of arbitrary offset and size - multi-sector case
481 * and single-sector case.
484 * (do_tail_read = true if tail > 0)
486 * |<----------------------total_size--------------------->|
488 * |<--head-->|<--------------bytes------------>|<--tail-->|
490 * | | |<~full_sec_size~>| | |
491 * +------------------+ +------------------+
492 * | |0101010| . . . |0101011| |
493 * +------------------+ +------------------+
494 * start_sec start_sec + n
498 * (do_tail_read = false)
500 * |<------total_size = secsz----->|
502 * |<-head->|<---bytes--->|<-tail->|
503 * +-------------------------------+
504 * | |0101010101010| |
505 * +-------------------------------+
508 start_sec = offset / secsz;
509 head = offset % secsz;
510 total_size = roundup2(head + bytes, secsz);
511 tail = total_size - (head + bytes);
512 do_tail_read = ((tail > 0) && (head + bytes > secsz));
513 full_sec_size = total_size;
515 full_sec_size -= secsz;
517 full_sec_size -= secsz;
519 /* Return of partial sector data requires a bounce buffer. */
520 if ((head > 0) || do_tail_read || bytes < secsz) {
521 bouncebuf = malloc(secsz);
522 if (bouncebuf == NULL) {
523 printf("vdev_read: out of memory\n");
528 if (lseek(fd, start_sec * secsz, SEEK_SET) == -1) {
533 /* Partial data return from first sector */
535 res = read(fd, bouncebuf, secsz);
540 memcpy(outbuf, bouncebuf + head, min(secsz - head, bytes));
541 outbuf += min(secsz - head, bytes);
545 * Full data return from read sectors.
546 * Note, there is still corner case where we read
547 * from sector boundary, but less than sector size, e.g. reading 512B
550 if (full_sec_size > 0) {
551 if (bytes < full_sec_size) {
552 res = read(fd, bouncebuf, secsz);
557 memcpy(outbuf, bouncebuf, bytes);
559 res = read(fd, outbuf, full_sec_size);
560 if (res != full_sec_size) {
564 outbuf += full_sec_size;
568 /* Partial data return from last sector */
570 res = read(fd, bouncebuf, secsz);
575 memcpy(outbuf, bouncebuf, secsz - tail);
585 vdev_write(vdev_t *vdev, off_t offset, void *buf, size_t bytes)
588 size_t head, tail, total_size, full_sec_size;
589 unsigned secsz, do_tail_write;
592 char *outbuf, *bouncebuf;
594 fd = (uintptr_t)vdev->v_priv;
595 outbuf = (char *)buf;
598 ret = ioctl(fd, DIOCGSECTORSIZE, &secsz);
602 start_sec = offset / secsz;
603 head = offset % secsz;
604 total_size = roundup2(head + bytes, secsz);
605 tail = total_size - (head + bytes);
606 do_tail_write = ((tail > 0) && (head + bytes > secsz));
607 full_sec_size = total_size;
609 full_sec_size -= secsz;
611 full_sec_size -= secsz;
613 /* Partial sector write requires a bounce buffer. */
614 if ((head > 0) || do_tail_write || bytes < secsz) {
615 bouncebuf = malloc(secsz);
616 if (bouncebuf == NULL) {
617 printf("vdev_write: out of memory\n");
622 if (lseek(fd, start_sec * secsz, SEEK_SET) == -1) {
627 /* Partial data for first sector */
629 res = read(fd, bouncebuf, secsz);
630 if ((unsigned)res != secsz) {
634 memcpy(bouncebuf + head, outbuf, min(secsz - head, bytes));
635 (void) lseek(fd, -secsz, SEEK_CUR);
636 res = write(fd, bouncebuf, secsz);
637 if ((unsigned)res != secsz) {
641 outbuf += min(secsz - head, bytes);
645 * Full data write to sectors.
646 * Note, there is still corner case where we write
647 * to sector boundary, but less than sector size, e.g. write 512B
650 if (full_sec_size > 0) {
651 if (bytes < full_sec_size) {
652 res = read(fd, bouncebuf, secsz);
653 if ((unsigned)res != secsz) {
657 memcpy(bouncebuf, outbuf, bytes);
658 (void) lseek(fd, -secsz, SEEK_CUR);
659 res = write(fd, bouncebuf, secsz);
660 if ((unsigned)res != secsz) {
665 res = write(fd, outbuf, full_sec_size);
666 if ((unsigned)res != full_sec_size) {
670 outbuf += full_sec_size;
674 /* Partial data write to last sector */
676 res = read(fd, bouncebuf, secsz);
677 if ((unsigned)res != secsz) {
681 memcpy(bouncebuf, outbuf, secsz - tail);
682 (void) lseek(fd, -secsz, SEEK_CUR);
683 res = write(fd, bouncebuf, secsz);
684 if ((unsigned)res != secsz) {
704 if (archsw.arch_zfs_probe == NULL)
706 archsw.arch_zfs_probe();
709 spa = STAILQ_FIRST(&zfs_pools);
710 while (spa != NULL) {
711 next = STAILQ_NEXT(spa, spa_link);
712 if (zfs_spa_init(spa)) {
714 STAILQ_REMOVE_HEAD(&zfs_pools, spa_link);
716 STAILQ_REMOVE_AFTER(&zfs_pools, prev, spa_link);
724 struct zfs_probe_args {
732 zfs_diskread(void *arg, void *buf, size_t blocks, uint64_t offset)
734 struct zfs_probe_args *ppa;
736 ppa = (struct zfs_probe_args *)arg;
737 return (vdev_read(NULL, (void *)(uintptr_t)ppa->fd,
738 offset * ppa->secsz, buf, blocks * ppa->secsz));
742 zfs_probe(int fd, uint64_t *pool_guid)
748 ret = vdev_probe(vdev_read, vdev_write, (void *)(uintptr_t)fd, &spa);
749 if (ret == 0 && pool_guid != NULL)
751 *pool_guid = spa->spa_guid;
756 zfs_probe_partition(void *arg, const char *partname,
757 const struct ptable_entry *part)
759 struct zfs_probe_args *ppa, pa;
760 struct ptable *table;
764 /* Probe only freebsd-zfs and freebsd partitions */
765 if (part->type != PART_FREEBSD &&
766 part->type != PART_FREEBSD_ZFS)
769 ppa = (struct zfs_probe_args *)arg;
770 strncpy(devname, ppa->devname, strlen(ppa->devname) - 1);
771 devname[strlen(ppa->devname) - 1] = '\0';
772 snprintf(devname, sizeof(devname), "%s%s:", devname, partname);
773 pa.fd = open(devname, O_RDWR);
776 ret = zfs_probe(pa.fd, ppa->pool_guid);
779 /* Do we have BSD label here? */
780 if (part->type == PART_FREEBSD) {
781 pa.devname = devname;
782 pa.pool_guid = ppa->pool_guid;
783 pa.secsz = ppa->secsz;
784 table = ptable_open(&pa, part->end - part->start + 1,
785 ppa->secsz, zfs_diskread);
787 ptable_iterate(table, &pa, zfs_probe_partition);
796 * Return bootenv nvlist from pool label.
799 zfs_get_bootenv(void *vdev, nvlist_t **benvp)
803 if ((spa = spa_find_by_dev((struct zfs_devdesc *)vdev)) == NULL)
806 return (zfs_get_bootenv_spa(spa, benvp));
810 * Store nvlist to pool label bootenv area. Also updates cached pointer in spa.
813 zfs_set_bootenv(void *vdev, nvlist_t *benv)
817 if ((spa = spa_find_by_dev((struct zfs_devdesc *)vdev)) == NULL)
820 return (zfs_set_bootenv_spa(spa, benv));
824 * Get bootonce value by key. The bootonce <key, value> pair is removed
825 * from the bootenv nvlist and the remaining nvlist is committed back to disk.
828 zfs_get_bootonce(void *vdev, const char *key, char *buf, size_t size)
832 if ((spa = spa_find_by_dev((struct zfs_devdesc *)vdev)) == NULL)
835 return (zfs_get_bootonce_spa(spa, key, buf, size));
842 static int zfs_nvstore_setter(void *, int, const char *,
843 const void *, size_t);
844 static int zfs_nvstore_setter_str(void *, const char *, const char *,
846 static int zfs_nvstore_unset_impl(void *, const char *, bool);
847 static int zfs_nvstore_setenv(void *, void *);
850 * nvstore is only present for current rootfs pool.
853 zfs_nvstore_sethook(struct env_var *ev, int flags __unused, const void *value)
855 struct zfs_devdesc *dev;
858 archsw.arch_getdev((void **)&dev, NULL, NULL);
862 rv = zfs_nvstore_setter_str(dev, NULL, ev->ev_name, value);
869 * nvstore is only present for current rootfs pool.
872 zfs_nvstore_unsethook(struct env_var *ev)
874 struct zfs_devdesc *dev;
877 archsw.arch_getdev((void **)&dev, NULL, NULL);
881 rv = zfs_nvstore_unset_impl(dev, ev->ev_name, false);
888 zfs_nvstore_getter(void *vdev, const char *name, void **data)
890 struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev;
897 if (dev->dd.d_dev->dv_type != DEVT_ZFS)
900 if ((spa = spa_find_by_dev(dev)) == NULL)
903 if (spa->spa_bootenv == NULL)
906 if (nvlist_find(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST,
907 NULL, &nv, NULL) != 0)
910 rv = nvlist_find(nv, name, DATA_TYPE_STRING, NULL, &str, &size);
913 asprintf(ptr, "%.*s", size, str);
922 zfs_nvstore_setter(void *vdev, int type, const char *name,
923 const void *data, size_t size)
925 struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev;
931 if (dev->dd.d_dev->dv_type != DEVT_ZFS)
934 if ((spa = spa_find_by_dev(dev)) == NULL)
937 if (spa->spa_bootenv == NULL)
940 if (nvlist_find(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST,
941 NULL, &nv, NULL) != 0) {
942 nv = nvlist_create(NV_UNIQUE_NAME);
950 if (size != sizeof (int8_t)) {
954 rv = nvlist_add_int8(nv, name, *(int8_t *)data);
957 case DATA_TYPE_INT16:
958 if (size != sizeof (int16_t)) {
962 rv = nvlist_add_int16(nv, name, *(int16_t *)data);
965 case DATA_TYPE_INT32:
966 if (size != sizeof (int32_t)) {
970 rv = nvlist_add_int32(nv, name, *(int32_t *)data);
973 case DATA_TYPE_INT64:
974 if (size != sizeof (int64_t)) {
978 rv = nvlist_add_int64(nv, name, *(int64_t *)data);
982 if (size != sizeof (uint8_t)) {
986 rv = nvlist_add_byte(nv, name, *(int8_t *)data);
989 case DATA_TYPE_UINT8:
990 if (size != sizeof (uint8_t)) {
994 rv = nvlist_add_uint8(nv, name, *(int8_t *)data);
997 case DATA_TYPE_UINT16:
998 if (size != sizeof (uint16_t)) {
1002 rv = nvlist_add_uint16(nv, name, *(uint16_t *)data);
1005 case DATA_TYPE_UINT32:
1006 if (size != sizeof (uint32_t)) {
1010 rv = nvlist_add_uint32(nv, name, *(uint32_t *)data);
1013 case DATA_TYPE_UINT64:
1014 if (size != sizeof (uint64_t)) {
1018 rv = nvlist_add_uint64(nv, name, *(uint64_t *)data);
1021 case DATA_TYPE_STRING:
1022 rv = nvlist_add_string(nv, name, data);
1025 case DATA_TYPE_BOOLEAN_VALUE:
1026 if (size != sizeof (boolean_t)) {
1030 rv = nvlist_add_boolean_value(nv, name, *(boolean_t *)data);
1039 rv = nvlist_add_nvlist(spa->spa_bootenv, OS_NVSTORE, nv);
1041 rv = zfs_set_bootenv(vdev, spa->spa_bootenv);
1045 rv = zfs_nvstore_setenv(vdev,
1046 nvpair_find(nv, name));
1048 env_discard(env_getenv(name));
1059 get_int64(const char *data, int64_t *ip)
1065 val = strtoll(data, &end, 0);
1066 if (errno != 0 || *data == '\0' || *end != '\0')
1074 get_uint64(const char *data, uint64_t *ip)
1080 val = strtoull(data, &end, 0);
1081 if (errno != 0 || *data == '\0' || *end != '\0')
1089 * Translate textual data to data type. If type is not set, and we are
1090 * creating new pair, use DATA_TYPE_STRING.
1093 zfs_nvstore_setter_str(void *vdev, const char *type, const char *name,
1096 struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev;
1104 if (dev->dd.d_dev->dv_type != DEVT_ZFS)
1107 if ((spa = spa_find_by_dev(dev)) == NULL)
1110 if (spa->spa_bootenv == NULL)
1113 if (nvlist_find(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST,
1114 NULL, &nv, NULL) != 0) {
1122 * if there is no existing pair, default to string.
1123 * Otherwise, use type from existing pair.
1125 nvh = nvpair_find(nv, name);
1127 dt = DATA_TYPE_STRING;
1129 nv_string_t *nvp_name;
1130 nv_pair_data_t *nvp_data;
1132 nvp_name = (nv_string_t *)(nvh + 1);
1133 nvp_data = (nv_pair_data_t *)(&nvp_name->nv_data[0] +
1134 NV_ALIGN4(nvp_name->nv_size));
1135 dt = nvp_data->nv_type;
1138 dt = nvpair_type_from_name(type);
1144 case DATA_TYPE_INT8:
1145 rv = get_int64(data, &val);
1149 rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v));
1152 case DATA_TYPE_INT16:
1153 rv = get_int64(data, &val);
1157 rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v));
1160 case DATA_TYPE_INT32:
1161 rv = get_int64(data, &val);
1165 rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v));
1168 case DATA_TYPE_INT64:
1169 rv = get_int64(data, &val);
1171 rv = zfs_nvstore_setter(vdev, dt, name, &val,
1176 case DATA_TYPE_BYTE:
1177 rv = get_uint64(data, &uval);
1181 rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v));
1185 case DATA_TYPE_UINT8:
1186 rv = get_uint64(data, &uval);
1190 rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v));
1194 case DATA_TYPE_UINT16:
1195 rv = get_uint64(data, &uval);
1199 rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v));
1203 case DATA_TYPE_UINT32:
1204 rv = get_uint64(data, &uval);
1208 rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v));
1212 case DATA_TYPE_UINT64:
1213 rv = get_uint64(data, &uval);
1215 rv = zfs_nvstore_setter(vdev, dt, name, &uval,
1220 case DATA_TYPE_STRING:
1221 rv = zfs_nvstore_setter(vdev, dt, name, data, strlen(data) + 1);
1224 case DATA_TYPE_BOOLEAN_VALUE:
1225 rv = get_int64(data, &val);
1229 rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v));
1239 zfs_nvstore_unset_impl(void *vdev, const char *name, bool unset_env)
1241 struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev;
1246 if (dev->dd.d_dev->dv_type != DEVT_ZFS)
1249 if ((spa = spa_find_by_dev(dev)) == NULL)
1252 if (spa->spa_bootenv == NULL)
1255 if (nvlist_find(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST,
1256 NULL, &nv, NULL) != 0)
1259 rv = nvlist_remove(nv, name, DATA_TYPE_UNKNOWN);
1261 if (nvlist_next_nvpair(nv, NULL) == NULL) {
1262 rv = nvlist_remove(spa->spa_bootenv, OS_NVSTORE,
1265 rv = nvlist_add_nvlist(spa->spa_bootenv,
1269 rv = zfs_set_bootenv(vdev, spa->spa_bootenv);
1273 env_discard(env_getenv(name));
1278 zfs_nvstore_unset(void *vdev, const char *name)
1280 return (zfs_nvstore_unset_impl(vdev, name, true));
1284 zfs_nvstore_print(void *vdev __unused, void *ptr)
1287 nvpair_print(ptr, 0);
1292 * Create environment variable from nvpair.
1293 * set hook will update nvstore with new value, unset hook will remove
1294 * variable from nvstore.
1297 zfs_nvstore_setenv(void *vdev __unused, void *ptr)
1299 nvp_header_t *nvh = ptr;
1300 nv_string_t *nvp_name, *nvp_value;
1301 nv_pair_data_t *nvp_data;
1308 nvp_name = (nv_string_t *)(nvh + 1);
1309 nvp_data = (nv_pair_data_t *)(&nvp_name->nv_data[0] +
1310 NV_ALIGN4(nvp_name->nv_size));
1312 if ((name = nvstring_get(nvp_name)) == NULL)
1316 switch (nvp_data->nv_type) {
1317 case DATA_TYPE_BYTE:
1318 case DATA_TYPE_UINT8:
1319 (void) asprintf(&value, "%uc",
1320 *(unsigned *)&nvp_data->nv_data[0]);
1325 case DATA_TYPE_INT8:
1326 (void) asprintf(&value, "%c", *(int *)&nvp_data->nv_data[0]);
1331 case DATA_TYPE_INT16:
1332 (void) asprintf(&value, "%hd", *(short *)&nvp_data->nv_data[0]);
1337 case DATA_TYPE_UINT16:
1338 (void) asprintf(&value, "%hu",
1339 *(unsigned short *)&nvp_data->nv_data[0]);
1344 case DATA_TYPE_BOOLEAN_VALUE:
1345 case DATA_TYPE_INT32:
1346 (void) asprintf(&value, "%d", *(int *)&nvp_data->nv_data[0]);
1351 case DATA_TYPE_UINT32:
1352 (void) asprintf(&value, "%u",
1353 *(unsigned *)&nvp_data->nv_data[0]);
1358 case DATA_TYPE_INT64:
1359 (void) asprintf(&value, "%jd",
1360 (intmax_t)*(int64_t *)&nvp_data->nv_data[0]);
1365 case DATA_TYPE_UINT64:
1366 (void) asprintf(&value, "%ju",
1367 (uintmax_t)*(uint64_t *)&nvp_data->nv_data[0]);
1372 case DATA_TYPE_STRING:
1373 nvp_value = (nv_string_t *)&nvp_data->nv_data[0];
1374 if ((value = nvstring_get(nvp_value)) == NULL) {
1385 if (value != NULL) {
1386 rv = env_setenv(name, EV_VOLATILE | EV_NOHOOK, value,
1387 zfs_nvstore_sethook, zfs_nvstore_unsethook);
1395 zfs_nvstore_iterate(void *vdev, int (*cb)(void *, void *))
1397 struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev;
1403 if (dev->dd.d_dev->dv_type != DEVT_ZFS)
1406 if ((spa = spa_find_by_dev(dev)) == NULL)
1409 if (spa->spa_bootenv == NULL)
1412 if (nvlist_find(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST,
1413 NULL, &nv, NULL) != 0)
1418 while ((nvh = nvlist_next_nvpair(nv, nvh)) != NULL) {
1426 nvs_callbacks_t nvstore_zfs_cb = {
1427 .nvs_getter = zfs_nvstore_getter,
1428 .nvs_setter = zfs_nvstore_setter,
1429 .nvs_setter_str = zfs_nvstore_setter_str,
1430 .nvs_unset = zfs_nvstore_unset,
1431 .nvs_print = zfs_nvstore_print,
1432 .nvs_iterate = zfs_nvstore_iterate
1436 zfs_attach_nvstore(void *vdev)
1438 struct zfs_devdesc *dev = vdev;
1443 if (dev->dd.d_dev->dv_type != DEVT_ZFS)
1446 if ((spa = spa_find_by_dev(dev)) == NULL)
1449 rv = nvlist_find(spa->spa_bootenv, BOOTENV_VERSION, DATA_TYPE_UINT64,
1450 NULL, &version, NULL);
1452 if (rv != 0 || version != VB_NVLIST) {
1456 dev = malloc(sizeof (*dev));
1459 memcpy(dev, vdev, sizeof (*dev));
1461 rv = nvstore_init(spa->spa_name, &nvstore_zfs_cb, dev);
1465 rv = zfs_nvstore_iterate(dev, zfs_nvstore_setenv);
1470 zfs_probe_dev(const char *devname, uint64_t *pool_guid, bool parts_too)
1472 struct ptable *table;
1473 struct zfs_probe_args pa;
1479 pa.fd = open(devname, O_RDWR);
1482 /* Probe the whole disk */
1483 ret = zfs_probe(pa.fd, pool_guid);
1489 /* Probe each partition */
1490 ret = ioctl(pa.fd, DIOCGMEDIASIZE, &mediasz);
1492 ret = ioctl(pa.fd, DIOCGSECTORSIZE, &pa.secsz);
1494 pa.devname = devname;
1495 pa.pool_guid = pool_guid;
1496 table = ptable_open(&pa, mediasz / pa.secsz, pa.secsz,
1498 if (table != NULL) {
1499 ptable_iterate(table, &pa, zfs_probe_partition);
1500 ptable_close(table);
1504 if (pool_guid && *pool_guid == 0)
1510 * Print information about ZFS pools
1513 zfs_dev_print(int verbose)
1519 if (STAILQ_EMPTY(&zfs_pools))
1522 printf("%s devices:", zfs_dev.dv_name);
1523 if ((ret = pager_output("\n")) != 0)
1527 return (spa_all_status());
1529 STAILQ_FOREACH(spa, &zfs_pools, spa_link) {
1530 snprintf(line, sizeof(line), " zfs:%s\n", spa->spa_name);
1531 ret = pager_output(line);
1539 * Attempt to open the pool described by (dev) for use by (f).
1542 zfs_dev_open(struct open_file *f, ...)
1545 struct zfs_devdesc *dev;
1546 struct zfsmount *mount;
1551 dev = va_arg(args, struct zfs_devdesc *);
1554 if ((spa = spa_find_by_dev(dev)) == NULL)
1557 STAILQ_FOREACH(mount, &zfsmount, next) {
1558 if (spa->spa_guid == mount->spa->spa_guid)
1563 /* This device is not set as currdev, mount us private copy. */
1565 rv = zfs_mount(devformat(&dev->dd), NULL, (void **)&mount);
1568 dev->dd.d_opendata = mount;
1574 zfs_dev_close(struct open_file *f)
1576 struct devdesc *dev;
1577 struct zfsmount *mnt, *mount;
1580 mnt = dev->d_opendata;
1582 STAILQ_FOREACH(mount, &zfsmount, next) {
1583 if (mnt->spa->spa_guid == mount->spa->spa_guid)
1592 zfs_dev_strategy(void *devdata, int rw, daddr_t dblk, size_t size, char *buf, size_t *rsize)
1598 struct devsw zfs_dev = {
1600 .dv_type = DEVT_ZFS,
1601 .dv_init = zfs_dev_init,
1602 .dv_strategy = zfs_dev_strategy,
1603 .dv_open = zfs_dev_open,
1604 .dv_close = zfs_dev_close,
1605 .dv_ioctl = noioctl,
1606 .dv_print = zfs_dev_print,
1607 .dv_cleanup = nullsys,
1608 .dv_fmtdev = zfs_fmtdev,
1609 .dv_parsedev = zfs_parsedev,
1613 zfs_parsedev(struct devdesc **idev, const char *devspec, const char **path)
1615 static char rootname[ZFS_MAXNAMELEN];
1616 static char poolname[ZFS_MAXNAMELEN];
1622 struct zfs_devdesc *dev;
1624 np = devspec + 3; /* Skip the leading 'zfs' */
1628 end = strrchr(np, ':');
1631 sep = strchr(np, '/');
1632 if (sep == NULL || sep >= end)
1634 memcpy(poolname, np, sep - np);
1635 poolname[sep - np] = '\0';
1638 memcpy(rootname, sep, end - sep);
1639 rootname[end - sep] = '\0';
1644 spa = spa_find_by_name(poolname);
1647 dev = malloc(sizeof(*dev));
1650 dev->pool_guid = spa->spa_guid;
1651 rv = zfs_lookup_dataset(spa, rootname, &dev->root_guid);
1657 *path = (*end == '\0') ? end : end + 1;
1658 dev->dd.d_dev = &zfs_dev;
1664 zfs_fmtdev(struct devdesc *vdev)
1666 static char rootname[ZFS_MAXNAMELEN];
1667 static char buf[2 * ZFS_MAXNAMELEN + 8];
1668 struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev;
1672 if (vdev->d_dev->dv_type != DEVT_ZFS)
1675 /* Do we have any pools? */
1676 spa = STAILQ_FIRST(&zfs_pools);
1680 if (dev->pool_guid == 0)
1681 dev->pool_guid = spa->spa_guid;
1683 spa = spa_find_by_guid(dev->pool_guid);
1686 printf("ZFS: can't find pool by guid\n");
1689 if (dev->root_guid == 0 && zfs_get_root(spa, &dev->root_guid)) {
1690 printf("ZFS: can't find root filesystem\n");
1693 if (zfs_rlookup(spa, dev->root_guid, rootname)) {
1694 printf("ZFS: can't find filesystem by guid\n");
1698 if (rootname[0] == '\0')
1699 snprintf(buf, sizeof(buf), "%s:%s:", dev->dd.d_dev->dv_name,
1702 snprintf(buf, sizeof(buf), "%s:%s/%s:", dev->dd.d_dev->dv_name,
1703 spa->spa_name, rootname);
1708 split_devname(const char *name, char *poolname, size_t size,
1709 const char **dsnamep)
1714 ASSERT(name != NULL);
1715 ASSERT(poolname != NULL);
1718 dsname = strchr(name, '/');
1719 if (dsname != NULL) {
1720 len = dsname - name;
1728 strlcpy(poolname, name, len + 1);
1730 if (dsnamep != NULL)
1737 zfs_list(const char *name)
1739 static char poolname[ZFS_MAXNAMELEN];
1745 if (split_devname(name, poolname, sizeof(poolname), &dsname) != 0)
1748 spa = spa_find_by_name(poolname);
1751 rv = zfs_lookup_dataset(spa, dsname, &objid);
1755 return (zfs_list_dataset(spa, objid));
1759 init_zfs_boot_options(const char *currdev_in)
1761 char poolname[ZFS_MAXNAMELEN];
1762 char *beroot, *currdev;
1768 currdev_len = strlen(currdev_in);
1769 if (currdev_len == 0)
1771 if (strncmp(currdev_in, "zfs:", 4) != 0)
1773 currdev = strdup(currdev_in);
1774 if (currdev == NULL)
1776 /* Remove the trailing : */
1777 currdev[currdev_len - 1] = '\0';
1779 setenv("zfs_be_active", currdev, 1);
1780 setenv("zfs_be_currpage", "1", 1);
1781 /* Remove the last element (current bootenv) */
1782 beroot = strrchr(currdev, '/');
1785 beroot = strchr(currdev, ':') + 1;
1786 setenv("zfs_be_root", beroot, 1);
1788 if (split_devname(beroot, poolname, sizeof(poolname), &dsname) != 0)
1791 spa = spa_find_by_name(poolname);
1795 zfs_bootenv_initial("bootenvs", spa, beroot, dsname, 0);
1796 zfs_checkpoints_initial(spa, beroot, dsname);
1802 zfs_checkpoints_initial(spa_t *spa, const char *name, const char *dsname)
1806 if (spa->spa_uberblock_checkpoint.ub_checkpoint_txg != 0) {
1807 snprintf(envname, sizeof(envname), "zpool_checkpoint");
1808 setenv(envname, name, 1);
1810 spa->spa_uberblock = &spa->spa_uberblock_checkpoint;
1811 spa->spa_mos = &spa->spa_mos_checkpoint;
1813 zfs_bootenv_initial("bootenvs_check", spa, name, dsname, 1);
1815 spa->spa_uberblock = &spa->spa_uberblock_master;
1816 spa->spa_mos = &spa->spa_mos_master;
1821 zfs_bootenv_initial(const char *envprefix, spa_t *spa, const char *rootname,
1822 const char *dsname, int checkpoint)
1824 char envname[32], envval[256];
1826 int bootenvs_idx, rv;
1828 SLIST_INIT(&zfs_be_head);
1831 rv = zfs_lookup_dataset(spa, dsname, &objid);
1835 rv = zfs_callback_dataset(spa, objid, zfs_belist_add);
1837 /* Populate the initial environment variables */
1838 SLIST_FOREACH_SAFE(zfs_be, &zfs_be_head, entries, zfs_be_tmp) {
1839 /* Enumerate all bootenvs for general usage */
1840 snprintf(envname, sizeof(envname), "%s[%d]",
1841 envprefix, bootenvs_idx);
1842 snprintf(envval, sizeof(envval), "zfs:%s%s/%s",
1843 checkpoint ? "!" : "", rootname, zfs_be->name);
1844 rv = setenv(envname, envval, 1);
1849 snprintf(envname, sizeof(envname), "%s_count", envprefix);
1850 snprintf(envval, sizeof(envval), "%d", bootenvs_idx);
1851 setenv(envname, envval, 1);
1853 /* Clean up the SLIST of ZFS BEs */
1854 while (!SLIST_EMPTY(&zfs_be_head)) {
1855 zfs_be = SLIST_FIRST(&zfs_be_head);
1856 SLIST_REMOVE_HEAD(&zfs_be_head, entries);
1863 zfs_bootenv(const char *name)
1865 char poolname[ZFS_MAXNAMELEN], *root;
1870 int rv, pages, perpage, currpage;
1874 if ((root = getenv("zfs_be_root")) == NULL)
1877 if (strcmp(name, root) != 0) {
1878 if (setenv("zfs_be_root", name, 1) != 0)
1882 SLIST_INIT(&zfs_be_head);
1885 if (split_devname(name, poolname, sizeof(poolname), &dsname) != 0)
1888 spa = spa_find_by_name(poolname);
1891 rv = zfs_lookup_dataset(spa, dsname, &objid);
1894 rv = zfs_callback_dataset(spa, objid, zfs_belist_add);
1896 /* Calculate and store the number of pages of BEs */
1897 perpage = (ZFS_BE_LAST - ZFS_BE_FIRST + 1);
1898 pages = (zfs_env_count / perpage) + ((zfs_env_count % perpage) > 0 ? 1 : 0);
1899 snprintf(becount, 4, "%d", pages);
1900 if (setenv("zfs_be_pages", becount, 1) != 0)
1903 /* Roll over the page counter if it has exceeded the maximum */
1904 currpage = strtol(getenv("zfs_be_currpage"), NULL, 10);
1905 if (currpage > pages) {
1906 if (setenv("zfs_be_currpage", "1", 1) != 0)
1910 /* Populate the menu environment variables */
1913 /* Clean up the SLIST of ZFS BEs */
1914 while (!SLIST_EMPTY(&zfs_be_head)) {
1915 zfs_be = SLIST_FIRST(&zfs_be_head);
1916 SLIST_REMOVE_HEAD(&zfs_be_head, entries);
1925 zfs_belist_add(const char *name, uint64_t value __unused)
1928 /* Skip special datasets that start with a $ character */
1929 if (strncmp(name, "$", 1) == 0) {
1932 /* Add the boot environment to the head of the SLIST */
1933 zfs_be = malloc(sizeof(struct zfs_be_entry));
1934 if (zfs_be == NULL) {
1937 zfs_be->name = strdup(name);
1938 if (zfs_be->name == NULL) {
1942 SLIST_INSERT_HEAD(&zfs_be_head, zfs_be, entries);
1951 char envname[32], envval[256];
1952 char *beroot, *pagenum;
1955 beroot = getenv("zfs_be_root");
1956 if (beroot == NULL) {
1960 pagenum = getenv("zfs_be_currpage");
1961 if (pagenum != NULL) {
1962 page = strtol(pagenum, NULL, 10);
1969 zfs_env_index = ZFS_BE_FIRST;
1970 SLIST_FOREACH_SAFE(zfs_be, &zfs_be_head, entries, zfs_be_tmp) {
1971 /* Skip to the requested page number */
1972 if (ctr <= ((ZFS_BE_LAST - ZFS_BE_FIRST + 1) * (page - 1))) {
1977 snprintf(envname, sizeof(envname), "bootenvmenu_caption[%d]", zfs_env_index);
1978 snprintf(envval, sizeof(envval), "%s", zfs_be->name);
1979 rv = setenv(envname, envval, 1);
1984 snprintf(envname, sizeof(envname), "bootenvansi_caption[%d]", zfs_env_index);
1985 rv = setenv(envname, envval, 1);
1990 snprintf(envname, sizeof(envname), "bootenvmenu_command[%d]", zfs_env_index);
1991 rv = setenv(envname, "set_bootenv", 1);
1996 snprintf(envname, sizeof(envname), "bootenv_root[%d]", zfs_env_index);
1997 snprintf(envval, sizeof(envval), "zfs:%s/%s", beroot, zfs_be->name);
1998 rv = setenv(envname, envval, 1);
2004 if (zfs_env_index > ZFS_BE_LAST) {
2010 for (; zfs_env_index <= ZFS_BE_LAST; zfs_env_index++) {
2011 snprintf(envname, sizeof(envname), "bootenvmenu_caption[%d]", zfs_env_index);
2012 (void)unsetenv(envname);
2013 snprintf(envname, sizeof(envname), "bootenvansi_caption[%d]", zfs_env_index);
2014 (void)unsetenv(envname);
2015 snprintf(envname, sizeof(envname), "bootenvmenu_command[%d]", zfs_env_index);
2016 (void)unsetenv(envname);
2017 snprintf(envname, sizeof(envname), "bootenv_root[%d]", zfs_env_index);
2018 (void)unsetenv(envname);