void lzwu_print_dedup_stats(nvlist_t *config)
{
ddt_histogram_t *ddh;
ddt_stat_t *dds;
ddt_object_t *ddo;
uint_t c;
/*
* If the pool was faulted then we may not have been able to
* obtain the config. Otherwise, if have anything in the dedup
* table continue processing the stats.
*/
if (nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_OBJ_STATS,
(uint64_t **)&ddo, &c) != 0 || ddo->ddo_count == 0)
return;
printf("\n");
printf("DDT entries %llu, size %llu on disk, %llu in core\n",
(u_longlong_t)ddo->ddo_count,
(u_longlong_t)ddo->ddo_dspace,
(u_longlong_t)ddo->ddo_mspace);
verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_STATS,
(uint64_t **)&dds, &c) == 0);
verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_HISTOGRAM,
(uint64_t **)&ddh, &c) == 0);
zpool_dump_ddt(dds, ddh);
}
static int
zpool_find_load_time(zpool_handle_t *zhp, void *arg)
{
struct load_time_arg *lta = arg;
uint64_t pool_guid;
uint64_t *tod;
nvlist_t *config;
uint_t nelem;
if (lta->lt_found)
return (0);
pool_guid = zpool_get_prop_int(zhp, ZPOOL_PROP_GUID, NULL);
if (pool_guid != lta->lt_guid)
return (0);
if ((config = zpool_get_config(zhp, NULL)) == NULL) {
zpool_close(zhp);
return (-1);
}
if (nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_LOADED_TIME,
&tod, &nelem) == 0 && nelem == 2) {
lta->lt_found = B_TRUE;
lta->lt_time->ertv_sec = tod[0];
lta->lt_time->ertv_nsec = tod[1];
}
return (0);
}
/*
* Set the pool-wide health based on the vdev state of the root vdev.
*/
int
set_pool_health(nvlist_t *config)
{
nvlist_t *nvroot;
vdev_stat_t *vs;
uint_t vsc;
char *health;
verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
&nvroot) == 0);
verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_STATS,
(uint64_t **)&vs, &vsc) == 0);
switch (vs->vs_state) {
case VDEV_STATE_CLOSED:
case VDEV_STATE_CANT_OPEN:
case VDEV_STATE_OFFLINE:
health = dgettext(TEXT_DOMAIN, "FAULTED");
break;
case VDEV_STATE_DEGRADED:
health = dgettext(TEXT_DOMAIN, "DEGRADED");
break;
case VDEV_STATE_HEALTHY:
health = dgettext(TEXT_DOMAIN, "ONLINE");
break;
default:
abort();
}
return (nvlist_add_string(config, ZPOOL_CONFIG_POOL_HEALTH, health));
}
static int
add_pool_to_xml(nvlist_t *config, void *data)
{
char *c;
char *name;
uint64_t guid;
uint64_t version;
uint64_t state;
nvlist_t *devices;
uint_t n;
vdev_stat_t *vs;
xmlNodePtr pool;
xmlNodePtr importable = *((xmlNodePtr *)data);
if (nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, &name) ||
nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) ||
nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &version) ||
nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, &state) ||
nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &devices) ||
nvlist_lookup_uint64_array(
devices, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &n)) {
return (-1);
}
pool = xmlNewChild(importable, NULL, (xmlChar *)ELEMENT_POOL, NULL);
(void) xmlSetProp(pool, (xmlChar *)ATTR_POOL_NAME, (xmlChar *)name);
set_uint64_prop(pool, ATTR_POOL_ID, guid);
set_uint64_prop(pool, ATTR_POOL_VERSION, version);
set_uint64_prop(pool, ATTR_POOL_USED, vs->vs_alloc);
set_uint64_prop(pool, ATTR_POOL_SIZE, vs->vs_space);
set_uint64_prop(pool, ATTR_POOL_REPLACEMENT_SIZE, vs->vs_rsize);
set_uint64_prop(pool, ATTR_POOL_READ_BYTES,
vs->vs_bytes[ZIO_TYPE_READ]);
set_uint64_prop(pool, ATTR_POOL_WRITE_BYTES,
vs->vs_bytes[ZIO_TYPE_WRITE]);
set_uint64_prop(pool, ATTR_POOL_READ_OPERATIONS,
vs->vs_ops[ZIO_TYPE_READ]);
set_uint64_prop(pool, ATTR_POOL_WRITE_OPERATIONS,
vs->vs_ops[ZIO_TYPE_WRITE]);
set_uint64_prop(pool, ATTR_POOL_READ_ERRORS, vs->vs_read_errors);
set_uint64_prop(pool, ATTR_POOL_WRITE_ERRORS, vs->vs_write_errors);
set_uint64_prop(pool, ATTR_POOL_CHECKSUM_ERRORS,
vs->vs_checksum_errors);
(void) xmlSetProp(pool, (xmlChar *)ATTR_DEVICE_STATE,
(xmlChar *)zjni_vdev_state_to_str(vs->vs_state));
(void) xmlSetProp(pool, (xmlChar *)ATTR_DEVICE_STATUS,
(xmlChar *)zjni_vdev_aux_to_str(vs->vs_aux));
(void) xmlSetProp(pool, (xmlChar *)ATTR_POOL_STATE,
(xmlChar *)zjni_pool_state_to_str(state));
(void) xmlSetProp(pool, (xmlChar *)ATTR_POOL_STATUS, (xmlChar *)
zjni_pool_status_to_str(zpool_import_status(config, &c)));
return (0);
}
/*
* Print out detailed scrub status.
* @param pnv_root: the root tree
*/
void lzwu_zpool_print_scrub_status(nvlist_t *pnv_root)
{
vdev_stat_t *p_vs;
unsigned vs_count;
time_t start, end, now;
double fraction_done;
uint64_t examined, total, minutes_left, minutes_taken;
const char *psz_scrub_type;
assert(nvlist_lookup_uint64_array(pnv_root, ZPOOL_CONFIG_STATS,
(uint64_t **)&p_vs, &vs_count) == 0);
/*
* If there's never been a scrub, there's not much to say.
*/
if(p_vs->vs_scrub_end == 0 && p_vs->vs_scrub_type == POOL_SCRUB_NONE)
{
printf("none requested\n");
return;
}
psz_scrub_type = (p_vs->vs_scrub_type == POOL_SCRUB_RESILVER) ? "resilver" : "scrub";
start = p_vs->vs_scrub_start;
end = p_vs->vs_scrub_end;
now = time(NULL);
examined = p_vs->vs_scrub_examined;
total = p_vs->vs_alloc;
if(end != 0)
{
minutes_taken = (uint64_t)((end - start) / 60);
printf("%s %s after %lluh%um with %llu errors on %s",
psz_scrub_type, p_vs->vs_scrub_complete ? "completed" : "stopped",
(u_longlong_t)(minutes_taken / 60),
(uint_t)(minutes_taken % 60),
(u_longlong_t)p_vs->vs_scrub_errors, ctime(&end));
return;
}
if(examined == 0)
examined = 1;
if(examined > total)
total = examined;
fraction_done = (double)examined / total;
minutes_left = (uint64_t)((now - start) *
(1 - fraction_done) / fraction_done / 60);
minutes_taken = (uint64_t)((now - start) / 60);
printf("%s in progress for %lluh%um, %.2f%% done, %lluh%um to go\n",
psz_scrub_type, (u_longlong_t)(minutes_taken / 60),
(uint_t)(minutes_taken % 60), 100 * fraction_done,
(u_longlong_t)(minutes_left / 60), (uint_t)(minutes_left % 60));
}
/*ARGSUSED*/
static int
zfs_mark_pool(zpool_handle_t *zhp, void *unused)
{
zfs_case_t *zcp;
uint64_t pool_guid;
uint64_t *tod;
er_timeval_t loaded = { 0 };
nvlist_t *config, *vd;
uint_t nelem = 0;
int ret;
pool_guid = zpool_get_prop_int(zhp, ZPOOL_PROP_GUID, NULL);
/*
* Mark any cases associated with just this pool.
*/
for (zcp = uu_list_first(zfs_cases); zcp != NULL;
zcp = uu_list_next(zfs_cases, zcp)) {
if (zcp->zc_data.zc_pool_guid == pool_guid &&
zcp->zc_data.zc_vdev_guid == 0)
zcp->zc_present = B_TRUE;
}
if ((config = zpool_get_config(zhp, NULL)) == NULL) {
zpool_close(zhp);
return (-1);
}
(void) nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_LOADED_TIME,
&tod, &nelem);
if (nelem == 2) {
loaded.ertv_sec = tod[0];
loaded.ertv_nsec = tod[1];
for (zcp = uu_list_first(zfs_cases); zcp != NULL;
zcp = uu_list_next(zfs_cases, zcp)) {
if (zcp->zc_data.zc_pool_guid == pool_guid &&
zcp->zc_data.zc_vdev_guid == 0) {
zcp->zc_when = loaded;
}
}
}
ret = nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &vd);
if (ret) {
zpool_close(zhp);
return (-1);
}
zfs_mark_vdev(pool_guid, vd, &loaded);
zpool_close(zhp);
return (0);
}
static int
zfs_toplevel_state(zpool_handle_t *zhp)
{
nvlist_t *nvroot;
vdev_stat_t *vs;
unsigned int c;
verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
(uint64_t **)&vs, &c) == 0);
return (vs->vs_state);
}
/*ARGSUSED*/
static void
zfs_ereport_when(fmd_hdl_t *hdl, nvlist_t *nvl, er_timeval_t *when)
{
uint64_t *tod;
uint_t nelem;
if (nvlist_lookup_uint64_array(nvl, FMD_EVN_TOD, &tod, &nelem) == 0 &&
nelem == 2) {
when->ertv_sec = tod[0];
when->ertv_nsec = tod[1];
} else {
when->ertv_sec = when->ertv_nsec = UINT64_MAX;
}
}
/*
* Return the total space in the pool.
*/
uint64_t
zpool_get_space_total(zpool_handle_t *zhp)
{
nvlist_t *nvroot;
vdev_stat_t *vs;
uint_t vsc;
verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
&nvroot) == 0);
verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_STATS,
(uint64_t **)&vs, &vsc) == 0);
return (vs->vs_space);
}
/*
* Detect if any leaf devices that have seen errors or could not be opened.
*/
static boolean_t
find_vdev_problem(nvlist_t *vdev, int (*func)(vdev_stat_t *, uint_t),
boolean_t ignore_replacing)
{
nvlist_t **child;
vdev_stat_t *vs;
uint_t c, vsc, children;
/*
* Ignore problems within a 'replacing' vdev, since we're presumably in
* the process of repairing any such errors, and don't want to call them
* out again. We'll pick up the fact that a resilver is happening
* later.
*/
if (ignore_replacing == B_TRUE) {
char *type;
verify(nvlist_lookup_string(vdev, ZPOOL_CONFIG_TYPE,
&type) == 0);
if (strcmp(type, VDEV_TYPE_REPLACING) == 0)
return (B_FALSE);
}
if (nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, &child,
&children) == 0) {
for (c = 0; c < children; c++)
if (find_vdev_problem(child[c], func, ignore_replacing))
return (B_TRUE);
} else {
verify(nvlist_lookup_uint64_array(vdev, ZPOOL_CONFIG_VDEV_STATS,
(uint64_t **)&vs, &vsc) == 0);
if (func(vs, vsc) != 0)
return (B_TRUE);
}
/*
* Check any L2 cache devs
*/
if (nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_L2CACHE, &child,
&children) == 0) {
for (c = 0; c < children; c++)
if (find_vdev_problem(child[c], func, ignore_replacing))
return (B_TRUE);
}
return (B_FALSE);
}
/* ARGSUSED */
int
fab_prep_basic_erpt(fmd_hdl_t *hdl, nvlist_t *nvl, nvlist_t *erpt,
boolean_t isRC)
{
uint64_t *now;
uint64_t ena;
uint_t nelem;
nvlist_t *detector, *new_detector;
char rcpath[255];
int err = 0;
/* Grab the tod, ena and detector(FMRI) */
err |= nvlist_lookup_uint64_array(nvl, "__tod", &now, &nelem);
err |= nvlist_lookup_uint64(nvl, "ena", &ena);
err |= nvlist_lookup_nvlist(nvl, FM_EREPORT_DETECTOR, &detector);
if (err)
return (err);
/* Make a copy of the detector */
err = nvlist_dup(detector, &new_detector, NV_UNIQUE_NAME);
if (err)
return (err);
/* Copy the tod and ena to erpt */
(void) nvlist_add_uint64(erpt, FM_EREPORT_ENA, ena);
(void) nvlist_add_uint64_array(erpt, "__tod", now, nelem);
/*
* Create the correct ROOT FMRI from PCIe leaf fabric ereports. Used
* only by fab_prep_fake_rc_erpt. See the fab_pciex_fake_rc_erpt_tbl
* comments for more information.
*/
if (isRC && fab_get_rcpath(hdl, nvl, rcpath)) {
/* Create the correct PCIe RC new_detector aka FMRI */
(void) nvlist_remove(new_detector, FM_FMRI_DEV_PATH,
DATA_TYPE_STRING);
(void) nvlist_add_string(new_detector, FM_FMRI_DEV_PATH,
rcpath);
}
/* Copy the FMRI to erpt */
(void) nvlist_add_nvlist(erpt, FM_EREPORT_DETECTOR, new_detector);
nvlist_free(new_detector);
return (err);
}
/*
* Collects zpool stats of the pool configured in the configData var.
* It opens a handler to the pool, load a tree of vdevs and then the
* stats of the root vdev. These stats are temporarily saved and then
* inserted in the SQL database.
*/
int collectZpoolStats() {
// Open a handle to the zfs filesystems
libzfs_handle_t *g_zfs = libzfs_init();
// Open a handle to the defined zpool
zpool_handle_t *zhp;
zhp = zpool_open_canfail(g_zfs, configData.poolname->value);
// Declarations and such...
nvlist_t *configuration, *vdevroot;
vdev_stat_t *vdevstats;
iostatcollection statisticscollection;
int nelem;
configuration = zpool_get_config(zhp, NULL);
// Now we have the config, we can release the handle to the zpool
libzfs_fini(g_zfs);
free(zhp);
// Put the vdev tree belonging to the pool in newconfig in newnvroot.
verify(nvlist_lookup_nvlist(configuration, ZPOOL_CONFIG_VDEV_TREE, &vdevroot) == 0);
// Load the new vdev stats in newvdevstat
verify(nvlist_lookup_uint64_array(vdevroot, ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vdevstats, &nelem) == 0);
// Place the collected statistics in the collection
statisticscollection.readops = vdevstats->vs_ops[ZIO_TYPE_READ];
statisticscollection.writeops = vdevstats->vs_ops[ZIO_TYPE_WRITE];
statisticscollection.readbytes = vdevstats->vs_bytes[ZIO_TYPE_READ];
statisticscollection.writebytes = vdevstats->vs_bytes[ZIO_TYPE_WRITE];
statisticscollection.checksum_errors = vdevstats->vs_checksum_errors;
statisticscollection.state = vdevstats->vs_state;
statisticscollection.space_alloc = vdevstats->vs_alloc;
statisticscollection.space = vdevstats->vs_space;
// Create the query and post it to MySQL
char queryString[1024];
snprintf(queryString, 1024, "INSERT INTO io_stats (date, %s, %s, %s, %s, %s, %s, %s, %s) VALUES (NOW(), '%llu', '%llu', '%llu', '%llu', '%llu', '%llu', '%llu', '%llu')", "iop_read", "iop_write", "bandwidth_read", "bandwidth_write", "space", "space_alloc", "checksum_errors", "state", statisticscollection.readops, statisticscollection.writeops, statisticscollection.readbytes, statisticscollection.writebytes, statisticscollection.space, statisticscollection.space_alloc, statisticscollection.checksum_errors, statisticscollection.state);
if (executeQuery(queryString)) return 1;
return 0;
}
int
zpool_get_stats(zpool_handle_t * zhp, void * data) {
config_t * cnf = (config_t *)data;
uint_t c;
boolean_t missing;
nvlist_t * nv, * config;
vdev_stat_t * vs;
if (zpool_refresh_stats(zhp, &missing) != 0)
return 1;
config = zpool_get_config(zhp, NULL);
if (nvlist_lookup_nvlist(config,
ZPOOL_CONFIG_VDEV_TREE, &nv) != 0) {
return 1;
}
if (nvlist_lookup_uint64_array(nv,
ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &c) != 0) {
return 1;
}
if (!strcmp(zpool_get_name(zhp), cnf->zname)) {
cnf->zpool.read_ops = vs->vs_ops[ZIO_TYPE_READ];
cnf->zpool.write_ops = vs->vs_ops[ZIO_TYPE_WRITE];
cnf->zpool.read_bts = vs->vs_bytes[ZIO_TYPE_READ];
cnf->zpool.write_bts = vs->vs_bytes[ZIO_TYPE_WRITE];
cnf->zpool.alloc = vs->vs_alloc;
cnf->zpool.free = vs->vs_space - vs->vs_alloc;
cnf->zpool.health = zpool_get_health(zhp);
cnf->zpool.dedupratio = zpool_get_dedupratio(zhp);
cnf->zpool.name = zpool_get_poolname(zhp);
cnf->zpool.ddt_memory = get_dedup_stats(config);
}
zpool_close(zhp);
return 0;
}
static void
zfs_ioctl_compat_fix_stats_nvlist(nvlist_t *nvl)
{
nvlist_t **child;
nvlist_t *nvroot = NULL;
vdev_stat_t *vs;
uint_t c, children, nelem;
if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN,
&child, &children) == 0) {
for (c = 0; c < children; c++) {
zfs_ioctl_compat_fix_stats_nvlist(child[c]);
}
}
if (nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_VDEV_TREE,
&nvroot) == 0)
zfs_ioctl_compat_fix_stats_nvlist(nvroot);
#ifdef _KERNEL
if ((nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_VDEV_STATS,
#else
if ((nvlist_lookup_uint64_array(nvl, "stats",
#endif
(uint64_t **)&vs, &nelem) == 0)) {
nvlist_add_uint64_array(nvl,
#ifdef _KERNEL
"stats",
#else
ZPOOL_CONFIG_VDEV_STATS,
#endif
(uint64_t *)vs, nelem);
#ifdef _KERNEL
nvlist_remove(nvl, ZPOOL_CONFIG_VDEV_STATS,
#else
nvlist_remove(nvl, "stats",
#endif
DATA_TYPE_UINT64_ARRAY);
}
int
fmd_fmri_unusable(nvlist_t *nvl)
{
uint64_t pool_guid, vdev_guid;
cbdata_t cb;
nvlist_t *vd;
int ret;
(void) nvlist_lookup_uint64(nvl, FM_FMRI_ZFS_POOL, &pool_guid);
cb.cb_guid = pool_guid;
cb.cb_pool = NULL;
if (zpool_iter(g_zfs, find_pool, &cb) != 1)
return (1);
if (nvlist_lookup_uint64(nvl, FM_FMRI_ZFS_VDEV, &vdev_guid) != 0) {
ret = (zpool_get_state(cb.cb_pool) == POOL_STATE_UNAVAIL);
zpool_close(cb.cb_pool);
return (ret);
}
vd = find_vdev(cb.cb_pool, vdev_guid);
if (vd == NULL) {
ret = 1;
} else {
vdev_stat_t *vs;
uint_t c;
(void) nvlist_lookup_uint64_array(vd, ZPOOL_CONFIG_STATS,
(uint64_t **)&vs, &c);
ret = (vs->vs_state < VDEV_STATE_DEGRADED);
}
zpool_close(cb.cb_pool);
return (ret);
}
/*
* Detect if any leaf devices that have seen errors or could not be opened.
*/
static boolean_t
find_vdev_problem(nvlist_t *vdev, int (*func)(uint64_t, uint64_t, uint64_t))
{
nvlist_t **child;
vdev_stat_t *vs;
uint_t c, children;
char *type;
/*
* Ignore problems within a 'replacing' vdev, since we're presumably in
* the process of repairing any such errors, and don't want to call them
* out again. We'll pick up the fact that a resilver is happening
* later.
*/
verify(nvlist_lookup_string(vdev, ZPOOL_CONFIG_TYPE, &type) == 0);
if (strcmp(type, VDEV_TYPE_REPLACING) == 0)
return (B_FALSE);
if (nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, &child,
&children) == 0) {
for (c = 0; c < children; c++)
if (find_vdev_problem(child[c], func))
return (B_TRUE);
} else {
verify(nvlist_lookup_uint64_array(vdev, ZPOOL_CONFIG_STATS,
(uint64_t **)&vs, &c) == 0);
if (func(vs->vs_state, vs->vs_aux,
vs->vs_read_errors +
vs->vs_write_errors +
vs->vs_checksum_errors))
return (B_TRUE);
}
return (B_FALSE);
}
/*
* 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);
}
/*
* Print out configuration state as requested by status_callback.
*/
void lzwu_zpool_print_status_config(libzfs_handle_t *p_zhd, zpool_handle_t *zhp,
const char *name, nvlist_t *nv, int namewidth,
int depth, boolean_t isspare)
{
nvlist_t **child;
uint_t children;
unsigned c;
vdev_stat_t *vs;
char rbuf[6], wbuf[6], cbuf[6], repaired[7];
char *vname;
uint64_t notpresent;
spare_cbdata_t cb;
char *state;
verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_STATS,
(uint64_t **)&vs, &c) == 0);
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0)
children = 0;
state = zpool_state_to_name(vs->vs_state, vs->vs_aux);
if(isspare)
{
/*
* For hot spares, we use the terms 'INUSE' and 'AVAILABLE' for
* online drives.
*/
if(vs->vs_aux == VDEV_AUX_SPARED)
state = "INUSE";
else if(vs->vs_state == VDEV_STATE_HEALTHY)
state = "AVAIL";
}
printf("\t%*s%-*s %-8s", depth, "", namewidth - depth,
name, state);
if(!isspare)
{
zfs_nicenum(vs->vs_read_errors, rbuf, sizeof (rbuf));
zfs_nicenum(vs->vs_write_errors, wbuf, sizeof (wbuf));
zfs_nicenum(vs->vs_checksum_errors, cbuf, sizeof (cbuf));
printf(" %5s %5s %5s", rbuf, wbuf, cbuf);
}
if(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, ¬present) == 0)
{
char *path;
verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
printf(" was %s", path);
}
else if(vs->vs_aux != 0)
{
printf(" ");
switch (vs->vs_aux)
{
case VDEV_AUX_OPEN_FAILED:
printf("cannot open");
break;
case VDEV_AUX_BAD_GUID_SUM:
printf("missing device");
break;
case VDEV_AUX_NO_REPLICAS:
printf("insufficient replicas");
break;
case VDEV_AUX_VERSION_NEWER:
printf("newer version");
break;
case VDEV_AUX_SPARED:
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
&cb.cb_guid) == 0);
if(zpool_iter(p_zhd, lzwu_find_spare, &cb) == 1)
{
if(strcmp(zpool_get_name(cb.cb_zhp),
zpool_get_name(zhp)) == 0)
printf("currently in use");
else
printf("in use by pool '%s'", zpool_get_name(cb.cb_zhp));
zpool_close(cb.cb_zhp);
}
else
printf("currently in use");
break;
case VDEV_AUX_ERR_EXCEEDED:
printf("too many errors");
break;
case VDEV_AUX_IO_FAILURE:
printf("experienced I/O failures");
break;
case VDEV_AUX_BAD_LOG:
printf("bad intent log");
break;
case VDEV_AUX_EXTERNAL:
printf("external device fault");
break;
case VDEV_AUX_SPLIT_POOL:
printf("split into new pool");
break;
default:
printf("corrupted data");
break;
}
}
else if(vs->vs_scrub_repaired != 0 && children == 0)
{
/*
* Report bytes resilvered/repaired on leaf devices.
*/
zfs_nicenum(vs->vs_scrub_repaired, repaired, sizeof (repaired));
printf(" %s %s", repaired,
(vs->vs_scrub_type == POOL_SCRUB_RESILVER) ?
"resilvered" : "repaired");
}
printf("\n");
for(unsigned c = 0; c < children; c++)
{
uint64_t islog = B_FALSE, ishole = B_FALSE;
/* Don't print logs or holes here */
nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, &islog);
nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, &ishole);
if(islog || ishole)
continue;
vname = zpool_vdev_name(p_zhd, zhp, child[c], B_TRUE);
lzwu_zpool_print_status_config(p_zhd, zhp, vname, child[c],
namewidth, depth + 2, isspare);
free(vname);
}
}
/*
* Print the configuration of an exported pool. Iterate over all vdevs in the
* pool, printing out the name and status for each one.
*/
static void lzwu_print_import_config(libzfs_handle_t *p_zhd, const char *name, nvlist_t *nv, int namewidth, int depth)
{
nvlist_t **child;
uint_t c, children;
vdev_stat_t *vs;
char *type, *vname;
verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
if (strcmp(type, VDEV_TYPE_MISSING) == 0 ||
strcmp(type, VDEV_TYPE_HOLE) == 0)
return;
verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_STATS,
(uint64_t **)&vs, &c) == 0);
(void) printf("\t%*s%-*s", depth, "", namewidth - depth, name);
(void) printf(" %s", zpool_state_to_name(vs->vs_state, vs->vs_aux));
if (vs->vs_aux != 0) {
(void) printf(" ");
switch (vs->vs_aux) {
case VDEV_AUX_OPEN_FAILED:
printf("cannot open");
break;
case VDEV_AUX_BAD_GUID_SUM:
printf("missing device");
break;
case VDEV_AUX_NO_REPLICAS:
printf("insufficient replicas");
break;
case VDEV_AUX_VERSION_NEWER:
printf("newer version");
break;
case VDEV_AUX_ERR_EXCEEDED:
printf("too many errors");
break;
default:
printf("corrupted data");
break;
}
}
(void) printf("\n");
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0)
return;
for (c = 0; c < children; c++) {
uint64_t is_log = B_FALSE;
(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
&is_log);
if (is_log)
continue;
vname = zpool_vdev_name(p_zhd, NULL, child[c], B_TRUE);
lzwu_print_import_config(p_zhd, vname, child[c], namewidth, depth + 2);
free(vname);
}
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
&child, &children) == 0) {
printf("\tcache\n");
for (c = 0; c < children; c++) {
vname = zpool_vdev_name(p_zhd, NULL, child[c], B_FALSE);
(void) printf("\t %s\n", vname);
free(vname);
}
}
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
&child, &children) == 0) {
printf("\tspares\n");
for (c = 0; c < children; c++) {
vname = zpool_vdev_name(p_zhd, NULL, child[c], B_FALSE);
(void) printf("\t %s\n", vname);
free(vname);
}
}
}
/*
* 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)
{
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;
unsigned long system_hostid = gethostid() & 0xffffffff;
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 && ps->pss_func == POOL_SCAN_RESILVER &&
ps->pss_state == DSS_SCANNING)
return (ZPOOL_STATUS_RESILVERING);
/*
* 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);
/*
* 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 due to failed I/O.
*/
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_SUSPENDED,
&suspended) == 0) {
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);
/*
* Outdated, but usable, version
*/
if (version < SPA_VERSION)
return (ZPOOL_STATUS_VERSION_OLDER);
return (ZPOOL_STATUS_OK);
}
/*
* 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 missing devices in a non-replicated config
* - Check for any data errors
* - Check for any 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)
{
nvlist_t *nvroot;
vdev_stat_t *vs;
uint_t vsc;
uint64_t nerr;
uint64_t version;
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_STATS,
(uint64_t **)&vs, &vsc) == 0);
/*
* Newer on-disk version.
*/
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
vs->vs_aux == VDEV_AUX_VERSION_NEWER)
return (ZPOOL_STATUS_VERSION_NEWER);
/*
* 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);
/*
* Missing devices in non-replicated config.
*/
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_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);
/*
* Currently resilvering
*/
if (!vs->vs_scrub_complete && vs->vs_scrub_type == POOL_SCRUB_RESILVER)
return (ZPOOL_STATUS_RESILVERING);
/*
* Outdated, but usable, version
*/
if (version < ZFS_VERSION)
return (ZPOOL_STATUS_VERSION_OLDER);
return (ZPOOL_STATUS_OK);
}
/*
* 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(zpool_handle_t *zhp, nvlist_t *config, boolean_t isimport)
{
nvlist_t *nvroot;
vdev_stat_t *vs;
uint_t vsc;
uint64_t nerr;
uint64_t version;
uint64_t stateval;
uint64_t hostid = 0;
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_STATS,
(uint64_t **)&vs, &vsc) == 0);
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
&stateval) == 0);
(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID, &hostid);
/*
* Pool last accessed by another system.
*/
if (hostid != 0 && (unsigned long)hostid != gethostid() &&
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);
/*
* 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);
/*
* Pool has experienced failed I/O.
*/
if (stateval == POOL_STATE_IO_FAILURE) {
zpool_handle_t *tmp_zhp = NULL;
libzfs_handle_t *hdl = NULL;
char property[ZPOOL_MAXPROPLEN];
char *failmode = NULL;
if (zhp == NULL) {
char *poolname;
verify(nvlist_lookup_string(config,
ZPOOL_CONFIG_POOL_NAME, &poolname) == 0);
if ((hdl = libzfs_init()) == NULL)
return (ZPOOL_STATUS_IO_FAILURE_WAIT);
tmp_zhp = zpool_open_canfail(hdl, poolname);
if (tmp_zhp == NULL) {
libzfs_fini(hdl);
return (ZPOOL_STATUS_IO_FAILURE_WAIT);
}
}
if (zpool_get_prop(zhp ? zhp : tmp_zhp, ZPOOL_PROP_FAILUREMODE,
property, sizeof (property), NULL) == 0)
failmode = property;
if (tmp_zhp != NULL)
zpool_close(tmp_zhp);
if (hdl != NULL)
libzfs_fini(hdl);
if (failmode == NULL)
return (ZPOOL_STATUS_IO_FAILURE_WAIT);
if (strncmp(failmode, "continue", strlen("continue")) == 0)
return (ZPOOL_STATUS_IO_FAILURE_CONTINUE);
else
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);
/*
* Currently resilvering
*/
if (!vs->vs_scrub_complete && vs->vs_scrub_type == POOL_SCRUB_RESILVER)
return (ZPOOL_STATUS_RESILVERING);
/*
* Outdated, but usable, version
*/
if (version < SPA_VERSION)
return (ZPOOL_STATUS_VERSION_OLDER);
return (ZPOOL_STATUS_OK);
}
/*
* 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)
{
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;
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 && ps->pss_func == POOL_SCAN_RESILVER &&
ps->pss_state == DSS_SCANNING)
return (ZPOOL_STATUS_RESILVERING);
/*
* Pool last accessed by another system.
*/
(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID, &hostid);
if (hostid != 0 && (unsigned long)hostid != gethostid() &&
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 due to failed I/O.
*/
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_SUSPENDED,
&suspended) == 0) {
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, B_TRUE))
return (ZPOOL_STATUS_FAULTED_DEV_NR);
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
find_vdev_problem(nvroot, vdev_missing, B_TRUE))
return (ZPOOL_STATUS_MISSING_DEV_NR);
if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
find_vdev_problem(nvroot, vdev_broken, B_TRUE))
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, B_TRUE))
return (ZPOOL_STATUS_FAULTED_DEV_R);
if (find_vdev_problem(nvroot, vdev_missing, B_TRUE))
return (ZPOOL_STATUS_MISSING_DEV_R);
if (find_vdev_problem(nvroot, vdev_broken, B_TRUE))
return (ZPOOL_STATUS_CORRUPT_LABEL_R);
/*
* Devices with errors
*/
if (!isimport && find_vdev_problem(nvroot, vdev_errors, B_TRUE))
return (ZPOOL_STATUS_FAILING_DEV);
/*
* Offlined devices
*/
if (find_vdev_problem(nvroot, vdev_offlined, B_TRUE))
return (ZPOOL_STATUS_OFFLINE_DEV);
/*
* Removed device
*/
if (find_vdev_problem(nvroot, vdev_removed, B_TRUE))
return (ZPOOL_STATUS_REMOVED_DEV);
/*
* Suboptimal, but usable, ashift configuration.
*/
if (find_vdev_problem(nvroot, vdev_non_native_ashift, B_FALSE))
return (ZPOOL_STATUS_NON_NATIVE_ASHIFT);
/*
* 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);
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);
}
static int
prop_getval(tnode_t *node, const char *pgname, const char *pname, void *val,
topo_type_t type, uint_t *nelems, int *err)
{
int i, j, ret = 0;
topo_hdl_t *thp = node->tn_hdl;
topo_propval_t *pv;
topo_node_lock(node);
if ((pv = prop_get(node, pgname, pname, NULL, err))
== NULL)
return (get_properror(node, err, *err));
if (pv->tp_type != type)
return (get_properror(node, err, ETOPO_PROP_TYPE));
switch (type) {
case TOPO_TYPE_INT32:
ret = nvlist_lookup_int32(pv->tp_val, TOPO_PROP_VAL_VAL,
(int32_t *)val);
break;
case TOPO_TYPE_UINT32:
ret = nvlist_lookup_uint32(pv->tp_val,
TOPO_PROP_VAL_VAL, (uint32_t *)val);
break;
case TOPO_TYPE_INT64:
ret = nvlist_lookup_int64(pv->tp_val, TOPO_PROP_VAL_VAL,
(int64_t *)val);
break;
case TOPO_TYPE_UINT64:
ret = nvlist_lookup_uint64(pv->tp_val,
TOPO_PROP_VAL_VAL, (uint64_t *)val);
break;
case TOPO_TYPE_DOUBLE:
ret = nvlist_lookup_double(pv->tp_val,
TOPO_PROP_VAL_VAL, (double *)val);
break;
case TOPO_TYPE_STRING: {
char *str;
ret = nvlist_lookup_string(pv->tp_val,
TOPO_PROP_VAL_VAL, &str);
if (ret == 0) {
char *s2;
if ((s2 = topo_hdl_strdup(thp, str)) == NULL)
ret = -1;
else
*(char **)val = s2;
}
break;
}
case TOPO_TYPE_FMRI: {
nvlist_t *nvl;
ret = nvlist_lookup_nvlist(pv->tp_val,
TOPO_PROP_VAL_VAL, &nvl);
if (ret == 0)
ret = topo_hdl_nvdup(thp, nvl,
(nvlist_t **)val);
break;
}
case TOPO_TYPE_INT32_ARRAY: {
int32_t *a1, *a2;
if ((ret = nvlist_lookup_int32_array(pv->tp_val,
TOPO_PROP_VAL_VAL, &a2, nelems)) != 0)
break;
if ((a1 = topo_hdl_alloc(thp, sizeof (int32_t) *
*nelems)) == NULL) {
ret = ETOPO_NOMEM;
break;
}
for (i = 0; i < *nelems; ++i)
a1[i] = a2[i];
*(int32_t **)val = a1;
break;
}
case TOPO_TYPE_UINT32_ARRAY: {
uint32_t *a1, *a2;
if ((ret = nvlist_lookup_uint32_array(pv->tp_val,
TOPO_PROP_VAL_VAL, &a2, nelems)) != 0)
break;
if ((a1 = topo_hdl_alloc(thp, sizeof (uint32_t) *
*nelems)) == NULL) {
ret = ETOPO_NOMEM;
break;
}
for (i = 0; i < *nelems; ++i)
a1[i] = a2[i];
*(uint32_t **)val = a1;
break;
}
case TOPO_TYPE_INT64_ARRAY: {
int64_t *a1, *a2;
if ((ret = nvlist_lookup_int64_array(pv->tp_val,
TOPO_PROP_VAL_VAL, &a2, nelems)) != 0)
break;
if ((a1 = topo_hdl_alloc(thp, sizeof (int64_t) *
*nelems)) == NULL) {
ret = ETOPO_NOMEM;
break;
}
for (i = 0; i < *nelems; ++i)
a1[i] = a2[i];
*(int64_t **)val = a1;
break;
}
case TOPO_TYPE_UINT64_ARRAY: {
uint64_t *a1, *a2;
if ((ret = nvlist_lookup_uint64_array(pv->tp_val,
TOPO_PROP_VAL_VAL, &a2, nelems)) != 0)
break;
if ((a1 = topo_hdl_alloc(thp, sizeof (uint64_t) *
*nelems)) == NULL) {
ret = ETOPO_NOMEM;
break;
}
for (i = 0; i < *nelems; ++i)
a1[i] = a2[i];
*(uint64_t **)val = a1;
break;
}
case TOPO_TYPE_STRING_ARRAY: {
char **a1, **a2;
if ((ret = nvlist_lookup_string_array(pv->tp_val,
TOPO_PROP_VAL_VAL, &a2, nelems)) != 0)
break;
if ((a1 = topo_hdl_alloc(thp, sizeof (char *) *
*nelems)) == NULL) {
ret = ETOPO_NOMEM;
break;
}
for (i = 0; i < *nelems; ++i) {
if ((a1[i] = topo_hdl_strdup(thp, a2[i]))
== NULL) {
for (j = 0; j < i; ++j)
topo_hdl_free(thp, a1[j],
sizeof (char *));
topo_hdl_free(thp, a1,
sizeof (char *) * *nelems);
break;
}
}
*(char ***)val = a1;
break;
}
case TOPO_TYPE_FMRI_ARRAY: {
nvlist_t **a1, **a2;
if ((ret = nvlist_lookup_nvlist_array(pv->tp_val,
TOPO_PROP_VAL_VAL, &a2, nelems)) != 0)
break;
if ((a1 = topo_hdl_alloc(thp, sizeof (nvlist_t *) *
*nelems)) == NULL) {
ret = ETOPO_NOMEM;
break;
}
for (i = 0; i < *nelems; ++i) {
if (topo_hdl_nvdup(thp, a2[i], &a1[i]) < 0) {
for (j = 0; j < i; ++j)
nvlist_free(a1[j]);
topo_hdl_free(thp, a1,
sizeof (nvlist_t *) * *nelems);
break;
}
}
*(nvlist_t ***)val = a1;
break;
}
default:
ret = ETOPO_PROP_NOENT;
}
if (ret != 0)
if (ret == ENOENT)
return (get_properror(node, err, ETOPO_PROP_NOENT));
else if (ret < ETOPO_UNKNOWN)
return (get_properror(node, err, ETOPO_PROP_NVL));
else
return (get_properror(node, err, ret));
topo_node_unlock(node);
return (0);
}
int
amd_rank_create(topo_mod_t *mod, tnode_t *pnode, nvlist_t *dimmnvl,
nvlist_t *auth)
{
uint64_t *csnumarr;
char **csnamearr;
uint_t ncs, ncsname;
tnode_t *ranknode;
nvlist_t *fmri, *pfmri = NULL;
uint64_t dsz, rsz;
int nerr = 0;
int err;
int i;
if (nvlist_lookup_uint64_array(dimmnvl, "csnums", &csnumarr,
&ncs) != 0 || nvlist_lookup_string_array(dimmnvl, "csnames",
&csnamearr, &ncsname) != 0 || ncs != ncsname) {
whinge(mod, &nerr, "amd_rank_create: "
"csnums/csnames extraction failed\n");
return (nerr);
}
if (topo_node_resource(pnode, &pfmri, &err) < 0) {
whinge(mod, &nerr, "amd_rank_create: parent fmri lookup "
"failed\n");
return (nerr);
}
if (topo_node_range_create(mod, pnode, RANK_NODE_NAME, 0, ncs) < 0) {
whinge(mod, &nerr, "amd_rank_create: range create failed\n");
nvlist_free(pfmri);
return (nerr);
}
if (topo_prop_get_uint64(pnode, PGNAME(DIMM), "size", &dsz,
&err) == 0) {
rsz = dsz / ncs;
} else {
whinge(mod, &nerr, "amd_rank_create: parent dimm has no "
"size\n");
return (nerr);
}
for (i = 0; i < ncs; i++) {
if (mkrsrc(mod, pnode, RANK_NODE_NAME, i, auth, &fmri) < 0) {
whinge(mod, &nerr, "amd_rank_create: mkrsrc failed\n");
continue;
}
if ((ranknode = topo_node_bind(mod, pnode, RANK_NODE_NAME, i,
fmri)) == NULL) {
nvlist_free(fmri);
whinge(mod, &nerr, "amd_rank_create: node bind "
"failed\n");
continue;
}
nvlist_free(fmri);
if (FM_AWARE_SMBIOS(mod))
(void) topo_node_fru_set(ranknode, NULL, 0, &err);
else
(void) topo_node_fru_set(ranknode, pfmri, 0, &err);
/*
* If a rank is faulted the asru is the associated
* chip-select, but if a page within a rank is faulted
* the asru is just that page. Hence the dual preconstructed
* and computed ASRU.
*/
if (topo_method_register(mod, ranknode, rank_methods) < 0)
whinge(mod, &nerr, "amd_rank_create: "
"topo_method_register failed");
if (! is_xpv() && topo_method_register(mod, ranknode,
ntv_page_retire_methods) < 0)
whinge(mod, &nerr, "amd_rank_create: "
"topo_method_register failed");
(void) topo_node_asru_set(ranknode, cs_fmri[csnumarr[i]],
TOPO_ASRU_COMPUTE, &err);
(void) topo_pgroup_create(ranknode, &rank_pgroup, &err);
(void) topo_prop_set_uint64(ranknode, PGNAME(RANK), "size",
TOPO_PROP_IMMUTABLE, rsz, &err);
(void) topo_prop_set_string(ranknode, PGNAME(RANK), "csname",
TOPO_PROP_IMMUTABLE, csnamearr[i], &err);
(void) topo_prop_set_uint64(ranknode, PGNAME(RANK), "csnum",
TOPO_PROP_IMMUTABLE, csnumarr[i], &err);
}
nvlist_free(pfmri);
return (nerr);
}
/* Calls the client callback function, if one is registered */
static void
notifyClient(sysevent_t *ev)
{
nvlist_t *attr_list = NULL;
uint64_t *val = NULL;
int32_t *instance = NULL;
int32_t *major = NULL;
int valAllocated = 0;
uint_t nelem = 0;
int i = 0;
int eventType = 0;
int index = -1;
void *pCallerData = NULL;
char subClassName[256];
MP_BOOL becomingVisible = MP_FALSE;
MP_OID_LIST *oidList = NULL;
log(LOG_INFO, "notifyClient()", "- enter");
(void) strncpy(subClassName, sysevent_get_subclass_name(ev), 256);
if (strstr(subClassName, "change")) {
eventType = PROP_CHANGE;
log(LOG_INFO, "notifyClient()", "- got a change event");
log(LOG_INFO, "notifyClient()", ": [%s]",
subClassName);
if (strncmp(subClassName, ESC_SUN_MP_PLUGIN_CHANGE, 255)
== 0) {
index = MP_OBJECT_TYPE_PLUGIN;
} else if (strncmp(subClassName, ESC_SUN_MP_LU_CHANGE, 255)
== 0) {
index = MP_OBJECT_TYPE_MULTIPATH_LU;
} else if (strncmp(subClassName, ESC_SUN_MP_PATH_CHANGE, 255)
== 0) {
index = MP_OBJECT_TYPE_PATH_LU;
} else if (strncmp(subClassName, ESC_SUN_MP_INIT_PORT_CHANGE,
255) == 0) {
index = MP_OBJECT_TYPE_INITIATOR_PORT;
} else if (strncmp(subClassName, ESC_SUN_MP_TPG_CHANGE, 255)
== 0) {
index = MP_OBJECT_TYPE_TARGET_PORT_GROUP;
} else if (strncmp(subClassName, ESC_SUN_MP_TARGET_PORT_CHANGE,
255) == 0) {
index = MP_OBJECT_TYPE_TARGET_PORT;
} else if (strncmp(subClassName, ESC_SUN_MP_DEV_PROD_CHANGE,
255) == 0) {
index = MP_OBJECT_TYPE_DEVICE_PRODUCT;
}
} else if ((strstr(subClassName, "add")) ||
(strstr(subClassName, "initiator_register"))) {
eventType = VISA_CHANGE;
becomingVisible = MP_TRUE;
log(LOG_INFO, "notifyClient()", "- got a visibility"
" add event");
log(LOG_INFO, "notifyClient()", ": [%s]",
subClassName);
if (strncmp(subClassName, ESC_SUN_MP_LU_ADD, 255) == 0) {
index = MP_OBJECT_TYPE_MULTIPATH_LU;
} else if (strncmp(subClassName, ESC_SUN_MP_PATH_ADD, 255)
== 0) {
index = MP_OBJECT_TYPE_PATH_LU;
} else if (strncmp(subClassName, ESC_DDI_INITIATOR_REGISTER,
244) == 0) {
index = MP_OBJECT_TYPE_INITIATOR_PORT;
} else if (strncmp(subClassName, ESC_SUN_MP_TPG_ADD,
255) == 0) {
index = MP_OBJECT_TYPE_TARGET_PORT_GROUP;
} else if (strncmp(subClassName, ESC_SUN_MP_TARGET_PORT_ADD,
255) == 0) {
index = MP_OBJECT_TYPE_TARGET_PORT;
} else if (strncmp(subClassName, ESC_SUN_MP_DEV_PROD_ADD, 255)
== 0) {
index = MP_OBJECT_TYPE_DEVICE_PRODUCT;
}
} else if ((strstr(subClassName, "remove")) ||
(strstr(subClassName, "initiator_unregister"))) {
eventType = VISA_CHANGE;
becomingVisible = MP_FALSE;
log(LOG_INFO, "notifyClient()", "- got a visibility"
" remove event");
log(LOG_INFO, "notifyClient()", ": [%s]",
subClassName);
if (strncmp(subClassName, ESC_SUN_MP_LU_REMOVE, 255) == 0) {
index = MP_OBJECT_TYPE_MULTIPATH_LU;
} else if (strncmp(subClassName, ESC_SUN_MP_PATH_REMOVE, 255)
== 0) {
index = MP_OBJECT_TYPE_PATH_LU;
} else if (strncmp(subClassName, ESC_DDI_INITIATOR_UNREGISTER,
255) == 0) {
index = MP_OBJECT_TYPE_INITIATOR_PORT;
} else if (strncmp(subClassName, ESC_SUN_MP_TPG_REMOVE, 255)
== 0) {
index = MP_OBJECT_TYPE_TARGET_PORT_GROUP;
} else if (strncmp(subClassName, ESC_SUN_MP_TARGET_PORT_REMOVE,
255) == 0) {
index = MP_OBJECT_TYPE_TARGET_PORT;
} else if (strncmp(subClassName, ESC_SUN_MP_DEV_PROD_REMOVE,
255) == 0) {
index = MP_OBJECT_TYPE_DEVICE_PRODUCT;
}
} else {
log(LOG_INFO, "notifyClient()", "- got an unsupported event");
return;
}
if (index < 0) {
log(LOG_INFO, "notifyClient()", "- index is less than zero");
return;
}
if (eventType == VISA_CHANGE) {
(void) pthread_mutex_lock(&g_visa_mutex);
if (NULL == g_Visibility_Callback_List[index].pClientFn) {
log(LOG_INFO, "notifyClient()",
"- no visibility change callback to notify");
(void) pthread_mutex_unlock(&g_visa_mutex);
return;
}
(void) pthread_mutex_unlock(&g_visa_mutex);
}
if (eventType == PROP_CHANGE) {
(void) pthread_mutex_lock(&g_prop_mutex);
if (NULL == g_Property_Callback_List[index].pClientFn) {
log(LOG_INFO, "notifyClient()",
"- no property change callback to notify");
(void) pthread_mutex_unlock(&g_prop_mutex);
return;
}
(void) pthread_mutex_unlock(&g_prop_mutex);
}
(void) sysevent_get_attr_list(ev, &attr_list);
if (NULL != attr_list) {
if ((VISA_CHANGE == eventType) &&
(MP_OBJECT_TYPE_PLUGIN == index)) {
val = (uint64_t *)malloc(sizeof (uint64_t));
valAllocated = 1;
/*
* We have no well-defined way to determine our OSN.
* Currently the common library uses 0 as OSN for every
* plugin, so just use 0. If the OSN assigned by the
* common library changed, this code would have to be
* updated.
*/
*val = 0;
nelem = 1;
} else if ((VISA_CHANGE == eventType) &&
(MP_OBJECT_TYPE_INITIATOR_PORT == index)) {
(void) nvlist_lookup_int32_array(attr_list,
DDI_INSTANCE, &instance, &nelem);
log(LOG_INFO, "notifyClient()",
"- event (PHCI_INSTANCE) has [%d] elements",
nelem);
(void) nvlist_lookup_int32_array(attr_list,
DDI_DRIVER_MAJOR, &major, &nelem);
log(LOG_INFO, "notifyClient()",
"- event (PHCI_DRIVER_MAJOR) has [%d] elements",
nelem);
if ((NULL != instance) & (NULL != major)) {
val = (uint64_t *)malloc(sizeof (uint64_t));
valAllocated = 1;
*val = 0;
*val = MP_STORE_INST_TO_ID(*instance, *val);
*val = MP_STORE_MAJOR_TO_ID(*major, *val);
nelem = 1;
} else {
nelem = 0;
}
} else {
(void) nvlist_lookup_uint64_array(attr_list, OIDLIST,
&val, &nelem);
log(LOG_INFO, "notifyClient()",
"- event has [%d] elements",
nelem);
}
if (nelem > 0) {
for (i = 0; i < nelem; i++) {
log(LOG_INFO, "notifyClient()",
"- event [%d] = %llx",
i, val[i]);
}
oidList = createOidList(nelem);
if (NULL == oidList) {
log(LOG_INFO, "notifyClient()",
"- unable to create MP_OID_LIST");
log(LOG_INFO, "notifyClient()",
"- error exit");
nvlist_free(attr_list);
return;
}
oidList->oidCount = nelem;
for (i = 0; i < nelem; i++) {
oidList->oids[i].objectType = index;
oidList->oids[i].ownerId = g_pluginOwnerID;
oidList->oids[i].objectSequenceNumber = val[i];
}
if (valAllocated) {
free(val);
}
for (i = 0; i < oidList->oidCount; i++) {
log(LOG_INFO, "notifyClient()",
"oidList->oids[%d].objectType"
" = %d",
i, oidList->oids[i].objectType);
log(LOG_INFO, "notifyClient()",
"oidList->oids[%d].ownerId"
" = %d",
i, oidList->oids[i].ownerId);
log(LOG_INFO, "notifyClient()",
"oidList->oids[%d].objectSequenceNumber"
" = %llx",
i, oidList->oids[i].objectSequenceNumber);
}
if (eventType == PROP_CHANGE) {
(void) pthread_mutex_lock(&g_prop_mutex);
pCallerData = g_Property_Callback_List[index].
pCallerData;
(g_Property_Callback_List[index].pClientFn)
(oidList, pCallerData);
(void) pthread_mutex_unlock(&g_prop_mutex);
} else if (eventType == VISA_CHANGE) {
(void) pthread_mutex_lock(&g_visa_mutex);
pCallerData = g_Visibility_Callback_List[index].
pCallerData;
(g_Visibility_Callback_List[index].pClientFn)
(becomingVisible, oidList, pCallerData);
(void) pthread_mutex_unlock(&g_visa_mutex);
}
}
nvlist_free(attr_list);
}
log(LOG_INFO, "notifyClient()", "- exit");
}
/*
* The device associated with the given vdev (either by devid or physical path)
* has been added to the system. If 'isdisk' is set, then we only attempt a
* replacement if it's a whole disk. This also implies that we should label the
* disk first.
*
* First, we attempt to online the device (making sure to undo any spare
* operation when finished). If this succeeds, then we're done. If it fails,
* and the new state is VDEV_CANT_OPEN, it indicates that the device was opened,
* but that the label was not what we expected. If the 'autoreplace' property
* is enabled, then we relabel the disk (if specified), and attempt a 'zpool
* replace'. If the online is successful, but the new state is something else
* (REMOVED or FAULTED), it indicates that we're out of sync or in some sort of
* race, and we should avoid attempting to relabel the disk.
*
* Also can arrive here from a ESC_ZFS_VDEV_CHECK event
*/
static void
zfs_process_add(zpool_handle_t *zhp, nvlist_t *vdev, boolean_t labeled)
{
char *path;
vdev_state_t newstate;
nvlist_t *nvroot, *newvd;
pendingdev_t *device;
uint64_t wholedisk = 0ULL;
uint64_t offline = 0ULL;
uint64_t guid = 0ULL;
char *physpath = NULL, *new_devid = NULL, *enc_sysfs_path = NULL;
char rawpath[PATH_MAX], fullpath[PATH_MAX];
char devpath[PATH_MAX];
int ret;
int is_dm = 0;
int is_sd = 0;
uint_t c;
vdev_stat_t *vs;
if (nvlist_lookup_string(vdev, ZPOOL_CONFIG_PATH, &path) != 0)
return;
/* Skip healthy disks */
verify(nvlist_lookup_uint64_array(vdev, ZPOOL_CONFIG_VDEV_STATS,
(uint64_t **)&vs, &c) == 0);
if (vs->vs_state == VDEV_STATE_HEALTHY) {
zed_log_msg(LOG_INFO, "%s: %s is already healthy, skip it.",
__func__, path);
return;
}
(void) nvlist_lookup_string(vdev, ZPOOL_CONFIG_PHYS_PATH, &physpath);
(void) nvlist_lookup_string(vdev, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH,
&enc_sysfs_path);
(void) nvlist_lookup_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK, &wholedisk);
(void) nvlist_lookup_uint64(vdev, ZPOOL_CONFIG_OFFLINE, &offline);
(void) nvlist_lookup_uint64(vdev, ZPOOL_CONFIG_GUID, &guid);
if (offline)
return; /* don't intervene if it was taken offline */
is_dm = zfs_dev_is_dm(path);
zed_log_msg(LOG_INFO, "zfs_process_add: pool '%s' vdev '%s', phys '%s'"
" wholedisk %d, dm %d (%llu)", zpool_get_name(zhp), path,
physpath ? physpath : "NULL", wholedisk, is_dm,
(long long unsigned int)guid);
/*
* The VDEV guid is preferred for identification (gets passed in path)
*/
if (guid != 0) {
(void) snprintf(fullpath, sizeof (fullpath), "%llu",
(long long unsigned int)guid);
} else {
/*
* otherwise use path sans partition suffix for whole disks
*/
(void) strlcpy(fullpath, path, sizeof (fullpath));
if (wholedisk) {
char *spath = zfs_strip_partition(fullpath);
if (!spath) {
zed_log_msg(LOG_INFO, "%s: Can't alloc",
__func__);
return;
}
(void) strlcpy(fullpath, spath, sizeof (fullpath));
free(spath);
}
}
/*
* Attempt to online the device.
*/
if (zpool_vdev_online(zhp, fullpath,
ZFS_ONLINE_CHECKREMOVE | ZFS_ONLINE_UNSPARE, &newstate) == 0 &&
(newstate == VDEV_STATE_HEALTHY ||
newstate == VDEV_STATE_DEGRADED)) {
zed_log_msg(LOG_INFO, " zpool_vdev_online: vdev %s is %s",
fullpath, (newstate == VDEV_STATE_HEALTHY) ?
"HEALTHY" : "DEGRADED");
return;
}
/*
* vdev_id alias rule for using scsi_debug devices (FMA automated
* testing)
*/
if (physpath != NULL && strcmp("scsidebug", physpath) == 0)
is_sd = 1;
/*
* If the pool doesn't have the autoreplace property set, then use
* vdev online to trigger a FMA fault by posting an ereport.
*/
if (!zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOREPLACE, NULL) ||
!(wholedisk || is_dm) || (physpath == NULL)) {
(void) zpool_vdev_online(zhp, fullpath, ZFS_ONLINE_FORCEFAULT,
&newstate);
zed_log_msg(LOG_INFO, "Pool's autoreplace is not enabled or "
"not a whole disk for '%s'", fullpath);
return;
}
/*
* Convert physical path into its current device node. Rawpath
* needs to be /dev/disk/by-vdev for a scsi_debug device since
* /dev/disk/by-path will not be present.
*/
(void) snprintf(rawpath, sizeof (rawpath), "%s%s",
is_sd ? DEV_BYVDEV_PATH : DEV_BYPATH_PATH, physpath);
if (realpath(rawpath, devpath) == NULL && !is_dm) {
zed_log_msg(LOG_INFO, " realpath: %s failed (%s)",
rawpath, strerror(errno));
(void) zpool_vdev_online(zhp, fullpath, ZFS_ONLINE_FORCEFAULT,
&newstate);
zed_log_msg(LOG_INFO, " zpool_vdev_online: %s FORCEFAULT (%s)",
fullpath, libzfs_error_description(g_zfshdl));
return;
}
/* Only autoreplace bad disks */
if ((vs->vs_state != VDEV_STATE_DEGRADED) &&
(vs->vs_state != VDEV_STATE_FAULTED) &&
(vs->vs_state != VDEV_STATE_CANT_OPEN)) {
return;
}
nvlist_lookup_string(vdev, "new_devid", &new_devid);
if (is_dm) {
/* Don't label device mapper or multipath disks. */
} else if (!labeled) {
/*
* we're auto-replacing a raw disk, so label it first
*/
char *leafname;
/*
* If this is a request to label a whole disk, then attempt to
* write out the label. Before we can label the disk, we need
* to map the physical string that was matched on to the under
* lying device node.
*
* If any part of this process fails, then do a force online
* to trigger a ZFS fault for the device (and any hot spare
* replacement).
*/
leafname = strrchr(devpath, '/') + 1;
/*
* If this is a request to label a whole disk, then attempt to
* write out the label.
*/
if (zpool_label_disk(g_zfshdl, zhp, leafname) != 0) {
zed_log_msg(LOG_INFO, " zpool_label_disk: could not "
"label '%s' (%s)", leafname,
libzfs_error_description(g_zfshdl));
(void) zpool_vdev_online(zhp, fullpath,
ZFS_ONLINE_FORCEFAULT, &newstate);
return;
}
/*
* The disk labeling is asynchronous on Linux. Just record
* this label request and return as there will be another
* disk add event for the partition after the labeling is
* completed.
*/
device = malloc(sizeof (pendingdev_t));
(void) strlcpy(device->pd_physpath, physpath,
sizeof (device->pd_physpath));
list_insert_tail(&g_device_list, device);
zed_log_msg(LOG_INFO, " zpool_label_disk: async '%s' (%llu)",
leafname, (u_longlong_t)guid);
return; /* resumes at EC_DEV_ADD.ESC_DISK for partition */
} else /* labeled */ {
boolean_t found = B_FALSE;
/*
* match up with request above to label the disk
*/
for (device = list_head(&g_device_list); device != NULL;
device = list_next(&g_device_list, device)) {
if (strcmp(physpath, device->pd_physpath) == 0) {
list_remove(&g_device_list, device);
free(device);
found = B_TRUE;
break;
}
zed_log_msg(LOG_INFO, "zpool_label_disk: %s != %s",
physpath, device->pd_physpath);
}
if (!found) {
/* unexpected partition slice encountered */
zed_log_msg(LOG_INFO, "labeled disk %s unexpected here",
fullpath);
(void) zpool_vdev_online(zhp, fullpath,
ZFS_ONLINE_FORCEFAULT, &newstate);
return;
}
zed_log_msg(LOG_INFO, " zpool_label_disk: resume '%s' (%llu)",
physpath, (u_longlong_t)guid);
(void) snprintf(devpath, sizeof (devpath), "%s%s",
DEV_BYID_PATH, new_devid);
}
/*
* Construct the root vdev to pass to zpool_vdev_attach(). While adding
* the entire vdev structure is harmless, we construct a reduced set of
* path/physpath/wholedisk to keep it simple.
*/
if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0) {
zed_log_msg(LOG_WARNING, "zfs_mod: nvlist_alloc out of memory");
return;
}
if (nvlist_alloc(&newvd, NV_UNIQUE_NAME, 0) != 0) {
zed_log_msg(LOG_WARNING, "zfs_mod: nvlist_alloc out of memory");
nvlist_free(nvroot);
return;
}
if (nvlist_add_string(newvd, ZPOOL_CONFIG_TYPE, VDEV_TYPE_DISK) != 0 ||
nvlist_add_string(newvd, ZPOOL_CONFIG_PATH, path) != 0 ||
nvlist_add_string(newvd, ZPOOL_CONFIG_DEVID, new_devid) != 0 ||
(physpath != NULL && nvlist_add_string(newvd,
ZPOOL_CONFIG_PHYS_PATH, physpath) != 0) ||
(enc_sysfs_path != NULL && nvlist_add_string(newvd,
ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH, enc_sysfs_path) != 0) ||
nvlist_add_uint64(newvd, ZPOOL_CONFIG_WHOLE_DISK, wholedisk) != 0 ||
nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) != 0 ||
nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &newvd,
1) != 0) {
zed_log_msg(LOG_WARNING, "zfs_mod: unable to add nvlist pairs");
nvlist_free(newvd);
nvlist_free(nvroot);
return;
}
nvlist_free(newvd);
/*
* Wait for udev to verify the links exist, then auto-replace
* the leaf disk at same physical location.
*/
if (zpool_label_disk_wait(path, 3000) != 0) {
zed_log_msg(LOG_WARNING, "zfs_mod: expected replacement "
"disk %s is missing", path);
nvlist_free(nvroot);
return;
}
ret = zpool_vdev_attach(zhp, fullpath, path, nvroot, B_TRUE);
zed_log_msg(LOG_INFO, " zpool_vdev_replace: %s with %s (%s)",
fullpath, path, (ret == 0) ? "no errors" :
libzfs_error_description(g_zfshdl));
nvlist_free(nvroot);
}
static void
show_vdev_stats(const char *desc, const char *ctype, nvlist_t *nv, int indent)
{
vdev_stat_t *vs;
vdev_stat_t *v0 = { 0 };
uint64_t sec;
uint64_t is_log = 0;
nvlist_t **child;
uint_t c, children;
char used[6], avail[6];
char rops[6], wops[6], rbytes[6], wbytes[6], rerr[6], werr[6], cerr[6];
v0 = umem_zalloc(sizeof (*v0), UMEM_NOFAIL);
if (indent == 0 && desc != NULL) {
(void) printf(" "
" capacity operations bandwidth ---- errors ----\n");
(void) printf("description "
"used avail read write read write read write cksum\n");
}
if (desc != NULL) {
char *suffix = "", *bias = NULL;
char bias_suffix[32];
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &is_log);
(void) nvlist_lookup_string(nv, ZPOOL_CONFIG_ALLOCATION_BIAS,
&bias);
if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
(uint64_t **)&vs, &c) != 0)
vs = v0;
if (bias != NULL) {
(void) snprintf(bias_suffix, sizeof (bias_suffix),
" (%s)", bias);
suffix = bias_suffix;
} else if (is_log) {
suffix = " (log)";
}
sec = MAX(1, vs->vs_timestamp / NANOSEC);
nicenum(vs->vs_alloc, used, sizeof (used));
nicenum(vs->vs_space - vs->vs_alloc, avail, sizeof (avail));
nicenum(vs->vs_ops[ZIO_TYPE_READ] / sec, rops, sizeof (rops));
nicenum(vs->vs_ops[ZIO_TYPE_WRITE] / sec, wops, sizeof (wops));
nicenum(vs->vs_bytes[ZIO_TYPE_READ] / sec, rbytes,
sizeof (rbytes));
nicenum(vs->vs_bytes[ZIO_TYPE_WRITE] / sec, wbytes,
sizeof (wbytes));
nicenum(vs->vs_read_errors, rerr, sizeof (rerr));
nicenum(vs->vs_write_errors, werr, sizeof (werr));
nicenum(vs->vs_checksum_errors, cerr, sizeof (cerr));
(void) printf("%*s%s%*s%*s%*s %5s %5s %5s %5s %5s %5s %5s\n",
indent, "",
desc,
(int)(indent+strlen(desc)-25-(vs->vs_space ? 0 : 12)),
suffix,
vs->vs_space ? 6 : 0, vs->vs_space ? used : "",
vs->vs_space ? 6 : 0, vs->vs_space ? avail : "",
rops, wops, rbytes, wbytes, rerr, werr, cerr);
}
free(v0);
if (nvlist_lookup_nvlist_array(nv, ctype, &child, &children) != 0)
return;
for (c = 0; c < children; c++) {
nvlist_t *cnv = child[c];
char *cname = NULL, *tname;
uint64_t np;
int len;
if (nvlist_lookup_string(cnv, ZPOOL_CONFIG_PATH, &cname) &&
nvlist_lookup_string(cnv, ZPOOL_CONFIG_TYPE, &cname))
cname = "<unknown>";
len = strlen(cname) + 2;
tname = umem_zalloc(len, UMEM_NOFAIL);
(void) strlcpy(tname, cname, len);
if (nvlist_lookup_uint64(cnv, ZPOOL_CONFIG_NPARITY, &np) == 0)
tname[strlen(tname)] = '0' + np;
show_vdev_stats(tname, ctype, cnv, indent + 2);
free(tname);
}
}
