/*
* "from" can be NULL, a snapshot, or a bookmark.
*
* If from is NULL, a full (non-incremental) stream will be estimated. This
* is calculated very efficiently.
*
* If from is a snapshot, lzc_send_space uses the deadlists attached to
* each snapshot to efficiently estimate the stream size.
*
* If from is a bookmark, the indirect blocks in the destination snapshot
* are traversed, looking for blocks with a birth time since the creation TXG of
* the snapshot this bookmark was created from. This will result in
* significantly more I/O and be less efficient than a send space estimation on
* an equivalent snapshot.
*/
int
lzc_send_space(const char *snapname, const char *from,
enum lzc_send_flags flags, uint64_t *spacep)
{
nvlist_t *args;
nvlist_t *result;
int err;
args = fnvlist_alloc();
if (from != NULL)
fnvlist_add_string(args, "from", from);
if (flags & LZC_SEND_FLAG_LARGE_BLOCK)
fnvlist_add_boolean(args, "largeblockok");
if (flags & LZC_SEND_FLAG_EMBED_DATA)
fnvlist_add_boolean(args, "embedok");
if (flags & LZC_SEND_FLAG_COMPRESS)
fnvlist_add_boolean(args, "compressok");
if (flags & LZC_SEND_FLAG_RAW)
fnvlist_add_boolean(args, "rawok");
err = lzc_ioctl(ZFS_IOC_SEND_SPACE, snapname, args, &result);
nvlist_free(args);
if (err == 0)
*spacep = fnvlist_lookup_uint64(result, "space");
nvlist_free(result);
return (err);
}
int
lzc_snaprange_space(const char *firstsnap, const char *lastsnap,
uint64_t *usedp)
{
nvlist_t *args;
nvlist_t *result;
int err;
char fs[MAXNAMELEN];
char *atp;
/* determine the fs name */
(void) strlcpy(fs, firstsnap, sizeof (fs));
atp = strchr(fs, '@');
if (atp == NULL)
return (EINVAL);
*atp = '\0';
args = fnvlist_alloc();
fnvlist_add_string(args, "firstsnap", firstsnap);
err = lzc_ioctl(ZFS_IOC_SPACE_SNAPS, lastsnap, args, &result);
nvlist_free(args);
if (err == 0)
*usedp = fnvlist_lookup_uint64(result, "used");
fnvlist_free(result);
return (err);
}
/*
* "from" can be NULL, a snapshot, or a bookmark.
*
* If from is NULL, a full (non-incremental) stream will be estimated. This
* is calculated very efficiently.
*
* If from is a snapshot, lzc_send_space uses the deadlists attached to
* each snapshot to efficiently estimate the stream size.
*
* If from is a bookmark, the indirect blocks in the destination snapshot
* are traversed, looking for blocks with a birth time since the creation TXG of
* the snapshot this bookmark was created from. This will result in
* significantly more I/O and be less efficient than a send space estimation on
* an equivalent snapshot. This process is also used if redact_snaps is
* non-null.
*/
int
lzc_send_space_resume_redacted(const char *snapname, const char *from,
enum lzc_send_flags flags, uint64_t resumeobj, uint64_t resumeoff,
uint64_t resume_bytes, const char *redactbook, int fd, uint64_t *spacep)
{
nvlist_t *args;
nvlist_t *result;
int err;
args = fnvlist_alloc();
if (from != NULL)
fnvlist_add_string(args, "from", from);
if (flags & LZC_SEND_FLAG_LARGE_BLOCK)
fnvlist_add_boolean(args, "largeblockok");
if (flags & LZC_SEND_FLAG_EMBED_DATA)
fnvlist_add_boolean(args, "embedok");
if (flags & LZC_SEND_FLAG_COMPRESS)
fnvlist_add_boolean(args, "compressok");
if (resumeobj != 0 || resumeoff != 0) {
fnvlist_add_uint64(args, "resume_object", resumeobj);
fnvlist_add_uint64(args, "resume_offset", resumeoff);
fnvlist_add_uint64(args, "bytes", resume_bytes);
}
if (redactbook != NULL)
fnvlist_add_string(args, "redactbook", redactbook);
if (fd != -1)
fnvlist_add_int32(args, "fd", fd);
err = lzc_ioctl(ZFS_IOC_SEND_SPACE, snapname, args, &result);
nvlist_free(args);
if (err == 0)
*spacep = fnvlist_lookup_uint64(result, "space");
nvlist_free(result);
return (err);
}
/*
* "from" can be NULL, a snapshot, or a bookmark.
*
* If from is NULL, a full (non-incremental) stream will be estimated. This
* is calculated very efficiently.
*
* If from is a snapshot, lzc_send_space uses the deadlists attached to
* each snapshot to efficiently estimate the stream size.
*
* If from is a bookmark, the indirect blocks in the destination snapshot
* are traversed, looking for blocks with a birth time since the creation TXG of
* the snapshot this bookmark was created from. This will result in
* significantly more I/O and be less efficient than a send space estimation on
* an equivalent snapshot.
*/
int
lzc_send_space(const char *snapname, const char *from, uint64_t *spacep)
{
nvlist_t *args;
nvlist_t *result;
int err;
args = fnvlist_alloc();
if (from != NULL)
fnvlist_add_string(args, "from", from);
err = lzc_ioctl(ZFS_IOC_SEND_SPACE, snapname, args, &result);
nvlist_free(args);
if (err == 0)
*spacep = fnvlist_lookup_uint64(result, "space");
nvlist_free(result);
return (err);
}
/*
* Convert our list of pools into the definitive set of configurations. We
* start by picking the best config for each toplevel vdev. Once that's done,
* we assemble the toplevel vdevs into a full config for the pool. We make a
* pass to fix up any incorrect paths, and then add it to the main list to
* return to the user.
*/
static nvlist_t *
get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok,
nvlist_t *policy)
{
pool_entry_t *pe;
vdev_entry_t *ve;
config_entry_t *ce;
nvlist_t *ret = NULL, *config = NULL, *tmp = NULL, *nvtop, *nvroot;
nvlist_t **spares, **l2cache;
uint_t i, nspares, nl2cache;
boolean_t config_seen;
uint64_t best_txg;
char *name, *hostname = NULL;
uint64_t guid;
uint_t children = 0;
nvlist_t **child = NULL;
uint_t holes;
uint64_t *hole_array, max_id;
uint_t c;
boolean_t isactive;
uint64_t hostid;
nvlist_t *nvl;
boolean_t valid_top_config = B_FALSE;
if (nvlist_alloc(&ret, 0, 0) != 0)
goto nomem;
for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
uint64_t id, max_txg = 0;
if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
goto nomem;
config_seen = B_FALSE;
/*
* Iterate over all toplevel vdevs. Grab the pool configuration
* from the first one we find, and then go through the rest and
* add them as necessary to the 'vdevs' member of the config.
*/
for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
/*
* Determine the best configuration for this vdev by
* selecting the config with the latest transaction
* group.
*/
best_txg = 0;
for (ce = ve->ve_configs; ce != NULL;
ce = ce->ce_next) {
if (ce->ce_txg > best_txg) {
tmp = ce->ce_config;
best_txg = ce->ce_txg;
}
}
/*
* We rely on the fact that the max txg for the
* pool will contain the most up-to-date information
* about the valid top-levels in the vdev namespace.
*/
if (best_txg > max_txg) {
(void) nvlist_remove(config,
ZPOOL_CONFIG_VDEV_CHILDREN,
DATA_TYPE_UINT64);
(void) nvlist_remove(config,
ZPOOL_CONFIG_HOLE_ARRAY,
DATA_TYPE_UINT64_ARRAY);
max_txg = best_txg;
hole_array = NULL;
holes = 0;
max_id = 0;
valid_top_config = B_FALSE;
if (nvlist_lookup_uint64(tmp,
ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
verify(nvlist_add_uint64(config,
ZPOOL_CONFIG_VDEV_CHILDREN,
max_id) == 0);
valid_top_config = B_TRUE;
}
if (nvlist_lookup_uint64_array(tmp,
ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
&holes) == 0) {
verify(nvlist_add_uint64_array(config,
ZPOOL_CONFIG_HOLE_ARRAY,
hole_array, holes) == 0);
}
}
if (!config_seen) {
/*
* Copy the relevant pieces of data to the pool
* configuration:
*
* version
* pool guid
* name
* pool txg (if available)
* comment (if available)
* pool state
* hostid (if available)
* hostname (if available)
*/
uint64_t state, version, pool_txg;
char *comment = NULL;
version = fnvlist_lookup_uint64(tmp,
ZPOOL_CONFIG_VERSION);
fnvlist_add_uint64(config,
ZPOOL_CONFIG_VERSION, version);
guid = fnvlist_lookup_uint64(tmp,
ZPOOL_CONFIG_POOL_GUID);
fnvlist_add_uint64(config,
ZPOOL_CONFIG_POOL_GUID, guid);
name = fnvlist_lookup_string(tmp,
ZPOOL_CONFIG_POOL_NAME);
fnvlist_add_string(config,
ZPOOL_CONFIG_POOL_NAME, name);
if (nvlist_lookup_uint64(tmp,
ZPOOL_CONFIG_POOL_TXG, &pool_txg) == 0)
fnvlist_add_uint64(config,
ZPOOL_CONFIG_POOL_TXG, pool_txg);
if (nvlist_lookup_string(tmp,
ZPOOL_CONFIG_COMMENT, &comment) == 0)
fnvlist_add_string(config,
ZPOOL_CONFIG_COMMENT, comment);
state = fnvlist_lookup_uint64(tmp,
ZPOOL_CONFIG_POOL_STATE);
fnvlist_add_uint64(config,
ZPOOL_CONFIG_POOL_STATE, state);
hostid = 0;
if (nvlist_lookup_uint64(tmp,
ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
fnvlist_add_uint64(config,
ZPOOL_CONFIG_HOSTID, hostid);
hostname = fnvlist_lookup_string(tmp,
ZPOOL_CONFIG_HOSTNAME);
fnvlist_add_string(config,
ZPOOL_CONFIG_HOSTNAME, hostname);
}
config_seen = B_TRUE;
}
/*
* Add this top-level vdev to the child array.
*/
verify(nvlist_lookup_nvlist(tmp,
ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
&id) == 0);
if (id >= children) {
nvlist_t **newchild;
newchild = zfs_alloc(hdl, (id + 1) *
sizeof (nvlist_t *));
if (newchild == NULL)
goto nomem;
for (c = 0; c < children; c++)
newchild[c] = child[c];
free(child);
child = newchild;
children = id + 1;
}
if (nvlist_dup(nvtop, &child[id], 0) != 0)
goto nomem;
}
/*
* If we have information about all the top-levels then
* clean up the nvlist which we've constructed. This
* means removing any extraneous devices that are
* beyond the valid range or adding devices to the end
* of our array which appear to be missing.
*/
if (valid_top_config) {
if (max_id < children) {
for (c = max_id; c < children; c++)
nvlist_free(child[c]);
children = max_id;
} else if (max_id > children) {
nvlist_t **newchild;
newchild = zfs_alloc(hdl, (max_id) *
sizeof (nvlist_t *));
if (newchild == NULL)
goto nomem;
for (c = 0; c < children; c++)
newchild[c] = child[c];
free(child);
child = newchild;
children = max_id;
}
}
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
&guid) == 0);
/*
* The vdev namespace may contain holes as a result of
* device removal. We must add them back into the vdev
* tree before we process any missing devices.
*/
if (holes > 0) {
ASSERT(valid_top_config);
for (c = 0; c < children; c++) {
nvlist_t *holey;
if (child[c] != NULL ||
!vdev_is_hole(hole_array, holes, c))
continue;
if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
0) != 0)
goto nomem;
/*
* Holes in the namespace are treated as
* "hole" top-level vdevs and have a
* special flag set on them.
*/
if (nvlist_add_string(holey,
ZPOOL_CONFIG_TYPE,
VDEV_TYPE_HOLE) != 0 ||
nvlist_add_uint64(holey,
ZPOOL_CONFIG_ID, c) != 0 ||
nvlist_add_uint64(holey,
ZPOOL_CONFIG_GUID, 0ULL) != 0) {
nvlist_free(holey);
goto nomem;
}
child[c] = holey;
}
}
/*
* Look for any missing top-level vdevs. If this is the case,
* create a faked up 'missing' vdev as a placeholder. We cannot
* simply compress the child array, because the kernel performs
* certain checks to make sure the vdev IDs match their location
* in the configuration.
*/
for (c = 0; c < children; c++) {
if (child[c] == NULL) {
nvlist_t *missing;
if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
0) != 0)
goto nomem;
if (nvlist_add_string(missing,
ZPOOL_CONFIG_TYPE,
VDEV_TYPE_MISSING) != 0 ||
nvlist_add_uint64(missing,
ZPOOL_CONFIG_ID, c) != 0 ||
nvlist_add_uint64(missing,
ZPOOL_CONFIG_GUID, 0ULL) != 0) {
nvlist_free(missing);
goto nomem;
}
child[c] = missing;
}
}
/*
* Put all of this pool's top-level vdevs into a root vdev.
*/
if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
goto nomem;
if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
VDEV_TYPE_ROOT) != 0 ||
nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
child, children) != 0) {
nvlist_free(nvroot);
goto nomem;
}
for (c = 0; c < children; c++)
nvlist_free(child[c]);
free(child);
children = 0;
child = NULL;
/*
* Go through and fix up any paths and/or devids based on our
* known list of vdev GUID -> path mappings.
*/
if (fix_paths(nvroot, pl->names) != 0) {
nvlist_free(nvroot);
goto nomem;
}
/*
* Add the root vdev to this pool's configuration.
*/
if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
nvroot) != 0) {
nvlist_free(nvroot);
goto nomem;
}
nvlist_free(nvroot);
/*
* zdb uses this path to report on active pools that were
* imported or created using -R.
*/
if (active_ok)
goto add_pool;
/*
* Determine if this pool is currently active, in which case we
* can't actually import it.
*/
verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
&name) == 0);
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
&guid) == 0);
if (pool_active(hdl, name, guid, &isactive) != 0)
goto error;
if (isactive) {
nvlist_free(config);
config = NULL;
continue;
}
if (policy != NULL) {
if (nvlist_add_nvlist(config, ZPOOL_LOAD_POLICY,
policy) != 0)
goto nomem;
}
if ((nvl = refresh_config(hdl, config)) == NULL) {
nvlist_free(config);
config = NULL;
continue;
}
nvlist_free(config);
config = nvl;
/*
* Go through and update the paths for spares, now that we have
* them.
*/
verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
&nvroot) == 0);
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
&spares, &nspares) == 0) {
for (i = 0; i < nspares; i++) {
if (fix_paths(spares[i], pl->names) != 0)
goto nomem;
}
}
/*
* Update the paths for l2cache devices.
*/
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
&l2cache, &nl2cache) == 0) {
for (i = 0; i < nl2cache; i++) {
if (fix_paths(l2cache[i], pl->names) != 0)
goto nomem;
}
}
/*
* Restore the original information read from the actual label.
*/
(void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
DATA_TYPE_UINT64);
(void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
DATA_TYPE_STRING);
if (hostid != 0) {
verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
hostid) == 0);
verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
hostname) == 0);
}
add_pool:
/*
* Add this pool to the list of configs.
*/
verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
&name) == 0);
if (nvlist_add_nvlist(ret, name, config) != 0)
goto nomem;
nvlist_free(config);
config = NULL;
}
return (ret);
nomem:
(void) no_memory(hdl);
error:
nvlist_free(config);
nvlist_free(ret);
for (c = 0; c < children; c++)
nvlist_free(child[c]);
free(child);
return (NULL);
}
/*
* Given a cache file, return the contents as a list of importable pools.
* poolname or guid (but not both) are provided by the caller when trying
* to import a specific pool.
*/
nvlist_t *
zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile,
char *poolname, uint64_t guid)
{
char *buf;
int fd;
struct stat statbuf;
nvlist_t *raw, *src, *dst;
nvlist_t *pools;
nvpair_t *elem;
char *name;
uint64_t this_guid;
boolean_t active;
verify(poolname == NULL || guid == 0);
if ((fd = open(cachefile, O_RDONLY)) < 0) {
zfs_error_aux(hdl, "%s", strerror(errno));
(void) zfs_error(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN, "failed to open cache file"));
return (NULL);
}
if (fstat(fd, &statbuf) != 0) {
zfs_error_aux(hdl, "%s", strerror(errno));
(void) close(fd);
(void) zfs_error(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
return (NULL);
}
if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
(void) close(fd);
return (NULL);
}
if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
(void) close(fd);
free(buf);
(void) zfs_error(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN,
"failed to read cache file contents"));
return (NULL);
}
(void) close(fd);
if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
free(buf);
(void) zfs_error(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN,
"invalid or corrupt cache file contents"));
return (NULL);
}
free(buf);
/*
* Go through and get the current state of the pools and refresh their
* state.
*/
if (nvlist_alloc(&pools, 0, 0) != 0) {
(void) no_memory(hdl);
nvlist_free(raw);
return (NULL);
}
elem = NULL;
while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
src = fnvpair_value_nvlist(elem);
name = fnvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME);
if (poolname != NULL && strcmp(poolname, name) != 0)
continue;
this_guid = fnvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID);
if (guid != 0 && guid != this_guid)
continue;
if (pool_active(hdl, name, this_guid, &active) != 0) {
nvlist_free(raw);
nvlist_free(pools);
return (NULL);
}
if (active)
continue;
if (nvlist_add_string(src, ZPOOL_CONFIG_CACHEFILE,
cachefile) != 0) {
(void) no_memory(hdl);
nvlist_free(raw);
nvlist_free(pools);
return (NULL);
}
if ((dst = refresh_config(hdl, src)) == NULL) {
nvlist_free(raw);
nvlist_free(pools);
return (NULL);
}
if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
(void) no_memory(hdl);
nvlist_free(dst);
nvlist_free(raw);
nvlist_free(pools);
return (NULL);
}
nvlist_free(dst);
}
nvlist_free(raw);
return (pools);
}
/*ARGSUSED*/
static void
spa_history_log_sync(void *arg, dmu_tx_t *tx)
{
nvlist_t *nvl = arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
objset_t *mos = spa->spa_meta_objset;
dmu_buf_t *dbp;
spa_history_phys_t *shpp;
size_t reclen;
uint64_t le_len;
char *record_packed = NULL;
int ret;
/*
* If we have an older pool that doesn't have a command
* history object, create it now.
*/
mutex_enter(&spa->spa_history_lock);
if (!spa->spa_history)
spa_history_create_obj(spa, tx);
mutex_exit(&spa->spa_history_lock);
/*
* Get the offset of where we need to write via the bonus buffer.
* Update the offset when the write completes.
*/
VERIFY0(dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp));
shpp = dbp->db_data;
dmu_buf_will_dirty(dbp, tx);
#ifdef ZFS_DEBUG
{
dmu_object_info_t doi;
dmu_object_info_from_db(dbp, &doi);
ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
}
#endif
fnvlist_add_uint64(nvl, ZPOOL_HIST_TIME, gethrestime_sec());
#ifdef _KERNEL
fnvlist_add_string(nvl, ZPOOL_HIST_HOST, utsname.nodename);
#endif
if (nvlist_exists(nvl, ZPOOL_HIST_CMD)) {
zfs_dbgmsg("command: %s",
fnvlist_lookup_string(nvl, ZPOOL_HIST_CMD));
} else if (nvlist_exists(nvl, ZPOOL_HIST_INT_NAME)) {
if (nvlist_exists(nvl, ZPOOL_HIST_DSNAME)) {
zfs_dbgmsg("txg %lld %s %s (id %llu) %s",
fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG),
fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME),
fnvlist_lookup_string(nvl, ZPOOL_HIST_DSNAME),
fnvlist_lookup_uint64(nvl, ZPOOL_HIST_DSID),
fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR));
} else {
zfs_dbgmsg("txg %lld %s %s",
fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG),
fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME),
fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR));
}
} else if (nvlist_exists(nvl, ZPOOL_HIST_IOCTL)) {
zfs_dbgmsg("ioctl %s",
fnvlist_lookup_string(nvl, ZPOOL_HIST_IOCTL));
}
record_packed = fnvlist_pack(nvl, &reclen);
mutex_enter(&spa->spa_history_lock);
/* write out the packed length as little endian */
le_len = LE_64((uint64_t)reclen);
ret = spa_history_write(spa, &le_len, sizeof (le_len), shpp, tx);
if (!ret)
ret = spa_history_write(spa, record_packed, reclen, shpp, tx);
/* The first command is the create, which we keep forever */
if (ret == 0 && shpp->sh_pool_create_len == 0 &&
nvlist_exists(nvl, ZPOOL_HIST_CMD)) {
shpp->sh_pool_create_len = shpp->sh_bof = shpp->sh_eof;
}
mutex_exit(&spa->spa_history_lock);
fnvlist_pack_free(record_packed, reclen);
dmu_buf_rele(dbp, FTAG);
fnvlist_free(nvl);
}
/*ARGSUSED*/
static void
spa_history_log_sync(void *arg, dmu_tx_t *tx)
{
nvlist_t *nvl = arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
objset_t *mos = spa->spa_meta_objset;
dmu_buf_t *dbp;
spa_history_phys_t *shpp;
size_t reclen;
uint64_t le_len;
char *record_packed = NULL;
int ret;
/*
* If we have an older pool that doesn't have a command
* history object, create it now.
*/
mutex_enter(&spa->spa_history_lock);
if (!spa->spa_history)
spa_history_create_obj(spa, tx);
mutex_exit(&spa->spa_history_lock);
/*
* Get the offset of where we need to write via the bonus buffer.
* Update the offset when the write completes.
*/
VERIFY0(dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp));
shpp = dbp->db_data;
dmu_buf_will_dirty(dbp, tx);
#ifdef ZFS_DEBUG
{
dmu_object_info_t doi;
dmu_object_info_from_db(dbp, &doi);
ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
}
#endif
fnvlist_add_uint64(nvl, ZPOOL_HIST_TIME, gethrestime_sec());
fnvlist_add_string(nvl, ZPOOL_HIST_HOST, utsname()->nodename);
if (nvlist_exists(nvl, ZPOOL_HIST_CMD)) {
zfs_dbgmsg("command: %s",
fnvlist_lookup_string(nvl, ZPOOL_HIST_CMD));
} else if (nvlist_exists(nvl, ZPOOL_HIST_INT_NAME)) {
if (nvlist_exists(nvl, ZPOOL_HIST_DSNAME)) {
zfs_dbgmsg("txg %lld %s %s (id %llu) %s",
fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG),
fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME),
fnvlist_lookup_string(nvl, ZPOOL_HIST_DSNAME),
fnvlist_lookup_uint64(nvl, ZPOOL_HIST_DSID),
fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR));
} else {
zfs_dbgmsg("txg %lld %s %s",
fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG),
fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME),
fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR));
}
/*
* The history sysevent is posted only for internal history
* messages to show what has happened, not how it happened. For
* example, the following command:
*
* # zfs destroy -r tank/foo
*
* will result in one sysevent posted per dataset that is
* destroyed as a result of the command - which could be more
* than one event in total. By contrast, if the sysevent was
* posted as a result of the ZPOOL_HIST_CMD key being present
* it would result in only one sysevent being posted with the
* full command line arguments, requiring the consumer to know
* how to parse and understand zfs(1M) command invocations.
*/
spa_history_log_notify(spa, nvl);
} else if (nvlist_exists(nvl, ZPOOL_HIST_IOCTL)) {
zfs_dbgmsg("ioctl %s",
fnvlist_lookup_string(nvl, ZPOOL_HIST_IOCTL));
}
VERIFY3U(nvlist_pack(nvl, &record_packed, &reclen, NV_ENCODE_NATIVE,
KM_SLEEP), ==, 0);
mutex_enter(&spa->spa_history_lock);
/* write out the packed length as little endian */
le_len = LE_64((uint64_t)reclen);
ret = spa_history_write(spa, &le_len, sizeof (le_len), shpp, tx);
if (!ret)
ret = spa_history_write(spa, record_packed, reclen, shpp, tx);
/* The first command is the create, which we keep forever */
if (ret == 0 && shpp->sh_pool_create_len == 0 &&
nvlist_exists(nvl, ZPOOL_HIST_CMD)) {
shpp->sh_pool_create_len = shpp->sh_bof = shpp->sh_eof;
}
mutex_exit(&spa->spa_history_lock);
fnvlist_pack_free(record_packed, reclen);
dmu_buf_rele(dbp, FTAG);
fnvlist_free(nvl);
}
static int
lzc_ioctl(zfs_ioc_t ioc, const char *name,
nvlist_t *source, nvlist_t **resultp)
{
zfs_cmd_t zc = {"\0"};
int error = 0;
char *packed = NULL;
size_t size = 0;
ASSERT3S(g_refcount, >, 0);
VERIFY3S(g_fd, !=, -1);
if (name != NULL)
(void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name));
if (source != NULL) {
packed = fnvlist_pack(source, &size);
zc.zc_nvlist_src = (uint64_t)(uintptr_t)packed;
zc.zc_nvlist_src_size = size;
}
if (resultp != NULL) {
*resultp = NULL;
if (ioc == ZFS_IOC_CHANNEL_PROGRAM) {
zc.zc_nvlist_dst_size = fnvlist_lookup_uint64(source,
ZCP_ARG_MEMLIMIT);
} else {
zc.zc_nvlist_dst_size = MAX(size * 2, 128 * 1024);
}
zc.zc_nvlist_dst = (uint64_t)(uintptr_t)
malloc(zc.zc_nvlist_dst_size);
if (zc.zc_nvlist_dst == (uint64_t)0) {
error = ENOMEM;
goto out;
}
}
while (ioctl(g_fd, ioc, &zc) != 0) {
/*
* If ioctl exited with ENOMEM, we retry the ioctl after
* increasing the size of the destination nvlist.
*
* Channel programs that exit with ENOMEM ran over the
* lua memory sandbox; they should not be retried.
*/
if (errno == ENOMEM && resultp != NULL &&
ioc != ZFS_IOC_CHANNEL_PROGRAM) {
free((void *)(uintptr_t)zc.zc_nvlist_dst);
zc.zc_nvlist_dst_size *= 2;
zc.zc_nvlist_dst = (uint64_t)(uintptr_t)
malloc(zc.zc_nvlist_dst_size);
if (zc.zc_nvlist_dst == (uint64_t)0) {
error = ENOMEM;
goto out;
}
} else {
error = errno;
break;
}
}
if (zc.zc_nvlist_dst_filled) {
*resultp = fnvlist_unpack((void *)(uintptr_t)zc.zc_nvlist_dst,
zc.zc_nvlist_dst_size);
}
out:
if (packed != NULL)
fnvlist_pack_free(packed, size);
free((void *)(uintptr_t)zc.zc_nvlist_dst);
return (error);
}
/*
* Active pool health status.
*
* To determine the status for a pool, we make several passes over the config,
* picking the most egregious error we find. In order of importance, we do the
* following:
*
* - Check for a complete and valid configuration
* - Look for any faulted or missing devices in a non-replicated config
* - Check for any data errors
* - Check for any faulted or missing devices in a replicated config
* - Look for any devices showing errors
* - Check for any resilvering devices
*
* There can obviously be multiple errors within a single pool, so this routine
* only picks the most damaging of all the current errors to report.
*/
static zpool_status_t
check_status(nvlist_t *config, boolean_t isimport, zpool_errata_t *erratap)
{
nvlist_t *nvroot;
vdev_stat_t *vs;
pool_scan_stat_t *ps = NULL;
uint_t vsc, psc;
uint64_t nerr;
uint64_t version;
uint64_t stateval;
uint64_t suspended;
uint64_t hostid = 0;
uint64_t errata = 0;
unsigned long system_hostid = get_system_hostid();
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
&version) == 0);
verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
&nvroot) == 0);
verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
(uint64_t **)&vs, &vsc) == 0);
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
&stateval) == 0);
/*
* Currently resilvering a vdev
*/
(void) nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_SCAN_STATS,
(uint64_t **)&ps, &psc);
if (ps != NULL && ps->pss_func == POOL_SCAN_RESILVER &&
ps->pss_state == DSS_SCANNING)
return (ZPOOL_STATUS_RESILVERING);
/*
* The multihost property is set and the pool may be active.
*/
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
vs->vs_aux == VDEV_AUX_ACTIVE) {
mmp_state_t mmp_state;
nvlist_t *nvinfo;
nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
mmp_state = fnvlist_lookup_uint64(nvinfo,
ZPOOL_CONFIG_MMP_STATE);
if (mmp_state == MMP_STATE_ACTIVE)
return (ZPOOL_STATUS_HOSTID_ACTIVE);
else if (mmp_state == MMP_STATE_NO_HOSTID)
return (ZPOOL_STATUS_HOSTID_REQUIRED);
else
return (ZPOOL_STATUS_HOSTID_MISMATCH);
}
/*
* Pool last accessed by another system.
*/
(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID, &hostid);
if (hostid != 0 && (unsigned long)hostid != system_hostid &&
stateval == POOL_STATE_ACTIVE)
return (ZPOOL_STATUS_HOSTID_MISMATCH);
/*
* Newer on-disk version.
*/
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
vs->vs_aux == VDEV_AUX_VERSION_NEWER)
return (ZPOOL_STATUS_VERSION_NEWER);
/*
* Unsupported feature(s).
*/
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
vs->vs_aux == VDEV_AUX_UNSUP_FEAT) {
nvlist_t *nvinfo;
verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO,
&nvinfo) == 0);
if (nvlist_exists(nvinfo, ZPOOL_CONFIG_CAN_RDONLY))
return (ZPOOL_STATUS_UNSUP_FEAT_WRITE);
return (ZPOOL_STATUS_UNSUP_FEAT_READ);
}
/*
* Check that the config is complete.
*/
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
vs->vs_aux == VDEV_AUX_BAD_GUID_SUM)
return (ZPOOL_STATUS_BAD_GUID_SUM);
/*
* Check whether the pool has suspended.
*/
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_SUSPENDED,
&suspended) == 0) {
uint64_t reason;
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_SUSPENDED_REASON,
&reason) == 0 && reason == ZIO_SUSPEND_MMP)
return (ZPOOL_STATUS_IO_FAILURE_MMP);
if (suspended == ZIO_FAILURE_MODE_CONTINUE)
return (ZPOOL_STATUS_IO_FAILURE_CONTINUE);
return (ZPOOL_STATUS_IO_FAILURE_WAIT);
}
/*
* Could not read a log.
*/
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
vs->vs_aux == VDEV_AUX_BAD_LOG) {
return (ZPOOL_STATUS_BAD_LOG);
}
/*
* Bad devices in non-replicated config.
*/
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
find_vdev_problem(nvroot, vdev_faulted))
return (ZPOOL_STATUS_FAULTED_DEV_NR);
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
find_vdev_problem(nvroot, vdev_missing))
return (ZPOOL_STATUS_MISSING_DEV_NR);
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
find_vdev_problem(nvroot, vdev_broken))
return (ZPOOL_STATUS_CORRUPT_LABEL_NR);
/*
* Corrupted pool metadata
*/
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
vs->vs_aux == VDEV_AUX_CORRUPT_DATA)
return (ZPOOL_STATUS_CORRUPT_POOL);
/*
* Persistent data errors.
*/
if (!isimport) {
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRCOUNT,
&nerr) == 0 && nerr != 0)
return (ZPOOL_STATUS_CORRUPT_DATA);
}
/*
* Missing devices in a replicated config.
*/
if (find_vdev_problem(nvroot, vdev_faulted))
return (ZPOOL_STATUS_FAULTED_DEV_R);
if (find_vdev_problem(nvroot, vdev_missing))
return (ZPOOL_STATUS_MISSING_DEV_R);
if (find_vdev_problem(nvroot, vdev_broken))
return (ZPOOL_STATUS_CORRUPT_LABEL_R);
/*
* Devices with errors
*/
if (!isimport && find_vdev_problem(nvroot, vdev_errors))
return (ZPOOL_STATUS_FAILING_DEV);
/*
* Offlined devices
*/
if (find_vdev_problem(nvroot, vdev_offlined))
return (ZPOOL_STATUS_OFFLINE_DEV);
/*
* Removed device
*/
if (find_vdev_problem(nvroot, vdev_removed))
return (ZPOOL_STATUS_REMOVED_DEV);
/*
* Informational errata available.
*/
(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRATA, &errata);
if (errata) {
*erratap = errata;
return (ZPOOL_STATUS_ERRATA);
}
/*
* Outdated, but usable, version
*/
if (SPA_VERSION_IS_SUPPORTED(version) && version != SPA_VERSION)
return (ZPOOL_STATUS_VERSION_OLDER);
/*
* Usable pool with disabled features
*/
if (version >= SPA_VERSION_FEATURES) {
int i;
nvlist_t *feat;
if (isimport) {
feat = fnvlist_lookup_nvlist(config,
ZPOOL_CONFIG_LOAD_INFO);
if (nvlist_exists(feat, ZPOOL_CONFIG_ENABLED_FEAT))
feat = fnvlist_lookup_nvlist(feat,
ZPOOL_CONFIG_ENABLED_FEAT);
} else {
feat = fnvlist_lookup_nvlist(config,
ZPOOL_CONFIG_FEATURE_STATS);
}
for (i = 0; i < SPA_FEATURES; i++) {
zfeature_info_t *fi = &spa_feature_table[i];
if (!nvlist_exists(feat, fi->fi_guid))
return (ZPOOL_STATUS_FEAT_DISABLED);
}
}
return (ZPOOL_STATUS_OK);
}
