/*
* Return the number of logs in supplied nvlist
*/
uint_t lzwu_num_logs(nvlist_t *nv)
{
uint_t nlogs = 0;
uint_t c, children;
nvlist_t **child;
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0)
return (0);
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)
nlogs++;
}
return (nlogs);
}
boolean_t
find_guid(nvlist_t *nv, uint64_t guid)
{
uint64_t tmp;
nvlist_t **child;
uint_t c, children;
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
if (tmp == guid)
return (B_TRUE);
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) == 0) {
for (c = 0; c < children; c++)
if (find_guid(child[c], guid))
return (B_TRUE);
}
return (B_FALSE);
}
/* ARGSUSED */
static kv_status_t
iser_declare_key_values(idm_conn_t *ic, nvlist_t *config_nvl,
nvlist_t *outgoing_nvl)
{
kv_status_t kvrc;
int nvrc = 0;
int rc;
uint64_t uint64_val;
if ((rc = nvlist_lookup_uint64(config_nvl,
ISER_KV_KEY_NAME_MAX_OUTSTANDING_PDU, &uint64_val)) != ENOENT) {
ASSERT(rc == 0);
if (outgoing_nvl) {
nvrc = nvlist_add_uint64(outgoing_nvl,
ISER_KV_KEY_NAME_MAX_OUTSTANDING_PDU, uint64_val);
}
}
kvrc = idm_nvstat_to_kvstat(nvrc);
return (kvrc);
}
static diskmon_t *
dm_fmristring_to_diskmon(char *str)
{
diskmon_t *p = NULL;
uint64_t u64val;
char ch;
char *lastsl = strrchr(str, '/');
ch = *lastsl;
*lastsl = 0;
if (nvlist_lookup_uint64(g_topo2diskmon, str, &u64val) == 0) {
p = (diskmon_t *)(uintptr_t)u64val;
}
*lastsl = ch;
return (p);
}
boolean_t lzwu_zpool_find_vdev(nvlist_t *pnv_root, uint64_t search)
{
uint64_t guid;
nvlist_t **ppnv_child;
uint_t i_children;
if(nvlist_lookup_uint64(pnv_root, ZPOOL_CONFIG_GUID, &guid) == 0 &&
search == guid)
return B_TRUE;
if(nvlist_lookup_nvlist_array(pnv_root, ZPOOL_CONFIG_CHILDREN,
&ppnv_child, &i_children) == 0)
{
for(unsigned c = 0; c < i_children; c++)
if(lzwu_zpool_find_vdev(ppnv_child[c], search))
return B_TRUE;
}
return B_FALSE;
}
static boolean_t
encryption_feature_is_enabled(zpool_handle_t *zph)
{
nvlist_t *features;
uint64_t feat_refcount;
/* check that features can be enabled */
if (zpool_get_prop_int(zph, ZPOOL_PROP_VERSION, NULL)
< SPA_VERSION_FEATURES)
return (B_FALSE);
/* check for crypto feature */
features = zpool_get_features(zph);
if (!features || nvlist_lookup_uint64(features,
spa_feature_table[SPA_FEATURE_ENCRYPTION].fi_guid,
&feat_refcount) != 0)
return (B_FALSE);
return (B_TRUE);
}
int
ses2_element_setdef(ses_node_t *np, ses2_diag_page_t page, void *data)
{
const ses2_ctl_desc_t *dp;
nvlist_t *props = ses_node_props(np);
uint64_t type;
VERIFY(nvlist_lookup_uint64(props, SES_PROP_ELEMENT_TYPE, &type) == 0);
for (dp = &ctl_descs[0]; dp->scd_et != -1; dp++)
if (dp->scd_et == type)
break;
if (dp->scd_et == -1)
return (0);
if (dp->scd_setdef(np, page, data) != 0)
return (-1);
return (0);
}
/*
* Are there allocatable vdevs?
*/
boolean_t
zfs_allocatable_devs(nvlist_t *nv)
{
uint64_t is_log;
uint_t c;
nvlist_t **child;
uint_t children;
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0) {
return (B_FALSE);
}
for (c = 0; c < children; c++) {
is_log = 0;
(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
&is_log);
if (!is_log)
return (B_TRUE);
}
return (B_FALSE);
}
/*ARGSUSED2*/
static void
frudr_completion_handler(char *ename, void *earg, size_t size)
{
picl_nodehdl_t fruh;
if (strcmp(ename, PICL_FRU_REMOVED) == 0) {
/*
* now frudata has been notified that the node is to be
* removed, we can actually remove it
*/
fruh = NULL;
(void) nvlist_lookup_uint64(earg,
PICLEVENTARG_FRUHANDLE, &fruh);
if (fruh != NULL) {
(void) remove_picl_node(fruh);
}
}
nvlist_free(earg);
free(earg);
free(ename);
}
static nvlist_t *
find_vdev_iter(nvlist_t *nv, uint64_t search)
{
uint_t c, children;
nvlist_t **child;
uint64_t guid;
nvlist_t *ret;
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid);
if (search == guid)
return (nv);
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) == 0) {
for (c = 0; c < children; c++)
if ((ret = find_vdev_iter(child[c], search)) != 0)
return (ret);
}
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
&child, &children) == 0) {
for (c = 0; c < children; c++)
if ((ret = find_vdev_iter(child[c], search)) != 0)
return (ret);
}
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
&child, &children) == 0) {
for (c = 0; c < children; c++)
if ((ret = find_vdev_iter(child[c], search)) != 0)
return (ret);
}
return (NULL);
}
/**
* Remove the given vdev from the pool
* @param p_zpool: the zpool handler
* @param psz_name: the name of the device to remove
* @param ppsz_error: the error message if any
* @return 0 in case of success, the error code overwise
*/
int libzfs_zpool_vdev_remove(zpool_handle_t *p_zpool, const char *psz_name, const char **ppsz_error)
{
nvlist_t *pnv_tgt;
spa_t *p_spa;
boolean_t avail_spare, l2cache, islog;
uint64_t guid;
int i_error;
if((pnv_tgt = zpool_find_vdev(p_zpool, psz_name,
&avail_spare, &l2cache, &islog)) == 0)
{
*ppsz_error = "no vdev corresponding to the one given";
return ENOENT;
}
assert(nvlist_lookup_uint64(pnv_tgt, ZPOOL_CONFIG_GUID, &guid) == 0);
if((i_error = spa_open(p_zpool->zpool_name, &p_spa, FTAG)))
{
*ppsz_error = "unable to open the spa";
return i_error;
}
i_error = spa_vdev_remove(p_spa, guid, B_FALSE);
spa_close(p_spa, FTAG);
switch(i_error)
{
case 0:
return 0;
case ENOTSUP:
*ppsz_error = "only spares, slogs, and level 2 ARC devices can be removed";
break;
case ENOENT:
*ppsz_error = "no vdev corresponding to the one given";
break;
}
return i_error;
}
/*
* Take the specified vdev offline
*/
int
zpool_vdev_offline(zpool_handle_t *zhp, const char *path, int istmp)
{
zfs_cmd_t zc = { 0 };
char msg[1024];
nvlist_t *tgt;
boolean_t avail_spare;
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if ((tgt = zpool_find_vdev(zhp, path, &avail_spare)) == NULL)
return (zfs_error(hdl, EZFS_NODEVICE, msg));
verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
if (avail_spare || is_spare(zhp, zc.zc_guid) == B_TRUE)
return (zfs_error(hdl, EZFS_ISSPARE, msg));
zc.zc_cookie = istmp;
if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_OFFLINE, &zc) == 0)
return (0);
switch (errno) {
case EBUSY:
/*
* There are no other replicas of this device.
*/
return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
default:
return (zpool_standard_error(hdl, errno, msg));
}
}
int
ses2_element_ctl(ses_plugin_t *sp, ses_node_t *np, const char *op,
nvlist_t *nvl)
{
const ses2_ctl_desc_t *dp;
nvlist_t *props = ses_node_props(np);
uint64_t type;
if (strcmp(op, SES_CTL_OP_SETPROP) != 0)
return (0);
VERIFY(nvlist_lookup_uint64(props, SES_PROP_ELEMENT_TYPE,
&type) == 0);
for (dp = &ctl_descs[0]; dp->scd_et != -1; dp++)
if (dp->scd_et == type)
break;
if (dp->scd_et == -1)
return (0);
return (ses2_setprop(sp, np, dp->scd_props, nvl));
}
/*
* Find a vdev within a tree with a matching GUID.
*/
static nvlist_t *
find_vdev(nvlist_t *nv, uint64_t search)
{
uint64_t guid;
nvlist_t **child;
uint_t c, children;
nvlist_t *ret;
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0 &&
guid == search)
return (nv);
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0)
return (NULL);
for (c = 0; c < children; c++) {
if ((ret = find_vdev(child[c], search)) != NULL)
return (ret);
}
return (NULL);
}
nvlist_t *
get_cpu_fault_resource(fmd_hdl_t *hdl, nvlist_t *asru)
{
uint32_t cpu;
uint64_t serint;
char serial[64];
nvlist_t *rsc = NULL;
cpuid_t cpuid;
char strid[10];
if (nvlist_lookup_uint64(asru, FM_FMRI_CPU_SERIAL_ID, &serint) != 0 ||
nvlist_lookup_uint32(asru, FM_FMRI_CPU_ID, &cpu) != 0)
return (rsc);
(void) snprintf(serial, sizeof (serial), "%llx", serint);
(void) snprintf(strid, sizeof (strid), "%d", cpu);
(void) strcpy(cpuid.serial, serial);
(void) strcpy(cpuid.id, strid);
rsc = cmd_find_cpu_rsc_by_sn(hdl, &cpuid);
return (rsc);
}
/*
* Returns TRUE if the given guid corresponds to a spare (INUSE or not).
*/
static boolean_t
is_spare(zpool_handle_t *zhp, uint64_t guid)
{
uint64_t spare_guid;
nvlist_t *nvroot;
nvlist_t **spares;
uint_t nspares;
int i;
verify(nvlist_lookup_nvlist(zhp->zpool_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++) {
verify(nvlist_lookup_uint64(spares[i],
ZPOOL_CONFIG_GUID, &spare_guid) == 0);
if (guid == spare_guid)
return (B_TRUE);
}
}
return (B_FALSE);
}
/*
* This function handles the ESC_DEV_DLE device change event. Use the
* provided vdev guid when looking up a disk or partition, when the guid
* is not present assume the entire disk is owned by ZFS and append the
* expected -part1 partition information then lookup by physical path.
*/
static int
zfs_deliver_dle(nvlist_t *nvl)
{
char *devname, name[MAXPATHLEN];
uint64_t guid;
if (nvlist_lookup_uint64(nvl, ZFS_EV_VDEV_GUID, &guid) == 0) {
sprintf(name, "%llu", (u_longlong_t)guid);
} else if (nvlist_lookup_string(nvl, DEV_PHYS_PATH, &devname) == 0) {
strlcpy(name, devname, MAXPATHLEN);
zfs_append_partition(name, MAXPATHLEN);
} else {
zed_log_msg(LOG_INFO, "zfs_deliver_dle: no guid or physpath");
}
if (zpool_iter(g_zfshdl, zfsdle_vdev_online, name) != 1) {
zed_log_msg(LOG_INFO, "zfs_deliver_dle: device '%s' not "
"found", name);
return (1);
}
return (0);
}
/*
* Returns true if the named pool matches the given GUID.
*/
static int
pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
boolean_t *isactive)
{
zpool_handle_t *zhp;
uint64_t theguid;
if (zpool_open_silent(hdl, name, &zhp) != 0)
return (-1);
if (zhp == NULL) {
*isactive = B_FALSE;
return (0);
}
verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
&theguid) == 0);
zpool_close(zhp);
*isactive = (theguid == guid);
return (0);
}
void
gmem_page_serd_create(fmd_hdl_t *hdl, gmem_page_t *page, nvlist_t *nvl)
{
uint32_t erpt_serdn, serd_n;
uint64_t erpt_serdt, serd_t;
serd_n = gmem.gm_ce_n;
serd_t = gmem.gm_ce_t;
if (serd_n == DEFAULT_SERDN && serd_t == DEFAULT_SERDT) {
if (nvlist_lookup_uint32(nvl, GMEM_ERPT_PAYLOAD_SERDN,
&erpt_serdn) == 0)
serd_n = erpt_serdn;
if (nvlist_lookup_uint64(nvl, GMEM_ERPT_PAYLOAD_SERDT,
&erpt_serdt) == 0)
serd_t = erpt_serdt;
}
page->page_case.cc_serdnm = gmem_page_serdnm_create(hdl, "page",
page->page_physbase);
fmd_serd_create(hdl, page->page_case.cc_serdnm, serd_n, serd_t);
}
static int
zfs_iter_pool(zpool_handle_t *zhp, void *data)
{
nvlist_t *config, *nvl;
dev_data_t *dp = data;
uint64_t pool_guid;
unavailpool_t *pool;
if ((config = zpool_get_config(zhp, NULL)) != NULL) {
if (dp->dd_pool_guid == 0 ||
(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
&pool_guid) == 0 && pool_guid == dp->dd_pool_guid)) {
(void) nvlist_lookup_nvlist(config,
ZPOOL_CONFIG_VDEV_TREE, &nvl);
zfs_iter_vdev(zhp, nvl, data);
}
}
if (g_enumeration_done) {
for (pool = list_head(&g_pool_list); pool != NULL;
pool = list_next(&g_pool_list, pool)) {
if (strcmp(zpool_get_name(zhp),
zpool_get_name(pool->uap_zhp)))
continue;
if (zfs_toplevel_state(zhp) >= VDEV_STATE_DEGRADED) {
list_remove(&g_pool_list, pool);
(void) tpool_dispatch(g_tpool, zfs_enable_ds,
pool);
break;
}
}
}
zpool_close(zhp);
return (0);
}
static int
zfsdle_vdev_online(zpool_handle_t *zhp, void *data)
{
char *devname = data;
boolean_t avail_spare, l2cache;
nvlist_t *tgt;
int error;
zed_log_msg(LOG_INFO, "zfsdle_vdev_online: searching for '%s' in '%s'",
devname, zpool_get_name(zhp));
if ((tgt = zpool_find_vdev_by_physpath(zhp, devname,
&avail_spare, &l2cache, NULL)) != NULL) {
char *path, fullpath[MAXPATHLEN];
uint64_t wholedisk;
error = nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &path);
if (error) {
zpool_close(zhp);
return (0);
}
error = nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
&wholedisk);
if (error)
wholedisk = 0;
if (wholedisk) {
path = strrchr(path, '/');
if (path != NULL) {
path = zfs_strip_partition(path + 1);
if (path == NULL) {
zpool_close(zhp);
return (0);
}
} else {
zpool_close(zhp);
return (0);
}
(void) strlcpy(fullpath, path, sizeof (fullpath));
free(path);
/*
* We need to reopen the pool associated with this
* device so that the kernel can update the size of
* the expanded device. When expanding there is no
* need to restart the scrub from the beginning.
*/
boolean_t scrub_restart = B_FALSE;
(void) zpool_reopen_one(zhp, &scrub_restart);
} else {
(void) strlcpy(fullpath, path, sizeof (fullpath));
}
if (zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
vdev_state_t newstate;
if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL) {
error = zpool_vdev_online(zhp, fullpath, 0,
&newstate);
zed_log_msg(LOG_INFO, "zfsdle_vdev_online: "
"setting device '%s' to ONLINE state "
"in pool '%s': %d", fullpath,
zpool_get_name(zhp), error);
}
}
zpool_close(zhp);
return (1);
}
zpool_close(zhp);
return (0);
}
static void
zfs_iter_vdev(zpool_handle_t *zhp, nvlist_t *nvl, void *data)
{
dev_data_t *dp = data;
char *path = NULL;
uint_t c, children;
nvlist_t **child;
/*
* First iterate over any children.
*/
if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN,
&child, &children) == 0) {
for (c = 0; c < children; c++)
zfs_iter_vdev(zhp, child[c], data);
}
/*
* Iterate over any spares and cache devices
*/
if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_SPARES,
&child, &children) == 0) {
for (c = 0; c < children; c++)
zfs_iter_vdev(zhp, child[c], data);
}
if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_L2CACHE,
&child, &children) == 0) {
for (c = 0; c < children; c++)
zfs_iter_vdev(zhp, child[c], data);
}
/* once a vdev was matched and processed there is nothing left to do */
if (dp->dd_found)
return;
/*
* Match by GUID if available otherwise fallback to devid or physical
*/
if (dp->dd_vdev_guid != 0) {
uint64_t guid;
if (nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_GUID,
&guid) != 0 || guid != dp->dd_vdev_guid) {
return;
}
zed_log_msg(LOG_INFO, " zfs_iter_vdev: matched on %llu", guid);
dp->dd_found = B_TRUE;
} else if (dp->dd_compare != NULL) {
/*
* NOTE: On Linux there is an event for partition, so unlike
* illumos, substring matching is not required to accommodate
* the partition suffix. An exact match will be present in
* the dp->dd_compare value.
*/
if (nvlist_lookup_string(nvl, dp->dd_prop, &path) != 0 ||
strcmp(dp->dd_compare, path) != 0)
return;
zed_log_msg(LOG_INFO, " zfs_iter_vdev: matched %s on %s",
dp->dd_prop, path);
dp->dd_found = B_TRUE;
/* pass the new devid for use by replacing code */
if (dp->dd_new_devid != NULL) {
(void) nvlist_add_string(nvl, "new_devid",
dp->dd_new_devid);
}
}
(dp->dd_func)(zhp, nvl, dp->dd_islabeled);
}
/*
* 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);
}
/*
* Given a file descriptor, read the label information and return an nvlist
* describing the configuration, if there is one.
* Return 0 on success, or -1 on failure
*/
int
zpool_read_label(int fd, nvlist_t **config, int *num_labels)
{
struct stat statbuf;
int l, count = 0;
vdev_label_t *label;
nvlist_t *expected_config = NULL;
uint64_t expected_guid = 0, size;
*config = NULL;
if (fstat(fd, &statbuf) == -1)
return (-1);
size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
if ((label = malloc(sizeof (vdev_label_t))) == NULL)
return (-1);
for (l = 0; l < VDEV_LABELS; l++) {
uint64_t state, guid, txg;
if (pread(fd, label, sizeof (vdev_label_t),
label_offset(size, l)) != sizeof (vdev_label_t))
continue;
if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
continue;
if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_GUID,
&guid) != 0 || guid == 0) {
nvlist_free(*config);
continue;
}
if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
&state) != 0 || state > POOL_STATE_L2CACHE) {
nvlist_free(*config);
continue;
}
if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
(nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
&txg) != 0 || txg == 0)) {
nvlist_free(*config);
continue;
}
if (expected_guid) {
if (expected_guid == guid)
count++;
nvlist_free(*config);
} else {
expected_config = *config;
expected_guid = guid;
count++;
}
}
if (num_labels != NULL)
*num_labels = count;
free(label);
*config = expected_config;
return (0);
}
/*
* 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);
}
/*
* Add the given configuration to the list of known devices.
*/
static int
add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
int order, int num_labels, nvlist_t *config)
{
uint64_t pool_guid, vdev_guid, top_guid, txg, state;
pool_entry_t *pe;
vdev_entry_t *ve;
config_entry_t *ce;
name_entry_t *ne;
/*
* If this is a hot spare not currently in use or level 2 cache
* device, add it to the list of names to translate, but don't do
* anything else.
*/
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
&state) == 0 &&
(state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
return (-1);
if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
free(ne);
return (-1);
}
ne->ne_guid = vdev_guid;
ne->ne_order = order;
ne->ne_num_labels = num_labels;
ne->ne_next = pl->names;
pl->names = ne;
return (0);
}
/*
* If we have a valid config but cannot read any of these fields, then
* it means we have a half-initialized label. In vdev_label_init()
* we write a label with txg == 0 so that we can identify the device
* in case the user refers to the same disk later on. If we fail to
* create the pool, we'll be left with a label in this state
* which should not be considered part of a valid pool.
*/
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
&pool_guid) != 0 ||
nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
&vdev_guid) != 0 ||
nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
&top_guid) != 0 ||
nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
&txg) != 0 || txg == 0) {
nvlist_free(config);
return (0);
}
/*
* First, see if we know about this pool. If not, then add it to the
* list of known pools.
*/
for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
if (pe->pe_guid == pool_guid)
break;
}
if (pe == NULL) {
if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
nvlist_free(config);
return (-1);
}
pe->pe_guid = pool_guid;
pe->pe_next = pl->pools;
pl->pools = pe;
}
/*
* Second, see if we know about this toplevel vdev. Add it if its
* missing.
*/
for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
if (ve->ve_guid == top_guid)
break;
}
if (ve == NULL) {
if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
nvlist_free(config);
return (-1);
}
ve->ve_guid = top_guid;
ve->ve_next = pe->pe_vdevs;
pe->pe_vdevs = ve;
}
/*
* Third, see if we have a config with a matching transaction group. If
* so, then we do nothing. Otherwise, add it to the list of known
* configs.
*/
for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
if (ce->ce_txg == txg)
break;
}
if (ce == NULL) {
if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
nvlist_free(config);
return (-1);
}
ce->ce_txg = txg;
ce->ce_config = config;
ce->ce_next = ve->ve_configs;
ve->ve_configs = ce;
} else {
nvlist_free(config);
}
/*
* At this point we've successfully added our config to the list of
* known configs. The last thing to do is add the vdev guid -> path
* mappings so that we can fix up the configuration as necessary before
* doing the import.
*/
if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
return (-1);
if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
free(ne);
return (-1);
}
ne->ne_guid = vdev_guid;
ne->ne_order = order;
ne->ne_num_labels = num_labels;
ne->ne_next = pl->names;
pl->names = ne;
return (0);
}
/*
* Determines if the pool is in use. If so, it returns true and the state of
* the pool as well as the name of the pool. Both strings are allocated and
* must be freed by the caller.
*/
int
zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
boolean_t *inuse)
{
nvlist_t *config;
char *name;
boolean_t ret;
uint64_t guid, vdev_guid;
zpool_handle_t *zhp;
nvlist_t *pool_config;
uint64_t stateval, isspare;
aux_cbdata_t cb = { 0 };
boolean_t isactive;
*inuse = B_FALSE;
if (zpool_read_label(fd, &config, NULL) != 0 && errno == ENOMEM) {
(void) no_memory(hdl);
return (-1);
}
if (config == NULL)
return (0);
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
&stateval) == 0);
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
&vdev_guid) == 0);
if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
&name) == 0);
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
&guid) == 0);
}
switch (stateval) {
case POOL_STATE_EXPORTED:
/*
* A pool with an exported state may in fact be imported
* read-only, so check the in-core state to see if it's
* active and imported read-only. If it is, set
* its state to active.
*/
if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
(zhp = zpool_open_canfail(hdl, name)) != NULL) {
if (zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
stateval = POOL_STATE_ACTIVE;
/*
* All we needed the zpool handle for is the
* readonly prop check.
*/
zpool_close(zhp);
}
ret = B_TRUE;
break;
case POOL_STATE_ACTIVE:
/*
* For an active pool, we have to determine if it's really part
* of a currently active pool (in which case the pool will exist
* and the guid will be the same), or whether it's part of an
* active pool that was disconnected without being explicitly
* exported.
*/
if (pool_active(hdl, name, guid, &isactive) != 0) {
nvlist_free(config);
return (-1);
}
if (isactive) {
/*
* Because the device may have been removed while
* offlined, we only report it as active if the vdev is
* still present in the config. Otherwise, pretend like
* it's not in use.
*/
if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
(pool_config = zpool_get_config(zhp, NULL))
!= NULL) {
nvlist_t *nvroot;
verify(nvlist_lookup_nvlist(pool_config,
ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
ret = find_guid(nvroot, vdev_guid);
} else {
ret = B_FALSE;
}
/*
* If this is an active spare within another pool, we
* treat it like an unused hot spare. This allows the
* user to create a pool with a hot spare that currently
* in use within another pool. Since we return B_TRUE,
* libdiskmgt will continue to prevent generic consumers
* from using the device.
*/
if (ret && nvlist_lookup_uint64(config,
ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
stateval = POOL_STATE_SPARE;
if (zhp != NULL)
zpool_close(zhp);
} else {
stateval = POOL_STATE_POTENTIALLY_ACTIVE;
ret = B_TRUE;
}
break;
case POOL_STATE_SPARE:
/*
* For a hot spare, it can be either definitively in use, or
* potentially active. To determine if it's in use, we iterate
* over all pools in the system and search for one with a spare
* with a matching guid.
*
* Due to the shared nature of spares, we don't actually report
* the potentially active case as in use. This means the user
* can freely create pools on the hot spares of exported pools,
* but to do otherwise makes the resulting code complicated, and
* we end up having to deal with this case anyway.
*/
cb.cb_zhp = NULL;
cb.cb_guid = vdev_guid;
cb.cb_type = ZPOOL_CONFIG_SPARES;
if (zpool_iter(hdl, find_aux, &cb) == 1) {
name = (char *)zpool_get_name(cb.cb_zhp);
ret = B_TRUE;
} else {
ret = B_FALSE;
}
break;
case POOL_STATE_L2CACHE:
/*
* Check if any pool is currently using this l2cache device.
*/
cb.cb_zhp = NULL;
cb.cb_guid = vdev_guid;
cb.cb_type = ZPOOL_CONFIG_L2CACHE;
if (zpool_iter(hdl, find_aux, &cb) == 1) {
name = (char *)zpool_get_name(cb.cb_zhp);
ret = B_TRUE;
} else {
ret = B_FALSE;
}
break;
default:
ret = B_FALSE;
}
if (ret) {
if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
if (cb.cb_zhp)
zpool_close(cb.cb_zhp);
nvlist_free(config);
return (-1);
}
*state = (pool_state_t)stateval;
}
if (cb.cb_zhp)
zpool_close(cb.cb_zhp);
nvlist_free(config);
*inuse = ret;
return (0);
}
/*
* Given a list of directories to search, find all pools stored on disk. This
* includes partial pools which are not available to import. If no args are
* given (argc is 0), then the default directory (/dev/dsk) is searched.
* poolname or guid (but not both) are provided by the caller when trying
* to import a specific pool.
*/
static nvlist_t *
zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
{
int i, dirs = iarg->paths;
struct dirent *dp;
char path[MAXPATHLEN];
char *end, **dir = iarg->path;
size_t pathleft;
nvlist_t *ret = NULL;
pool_list_t pools = { 0 };
pool_entry_t *pe, *penext;
vdev_entry_t *ve, *venext;
config_entry_t *ce, *cenext;
name_entry_t *ne, *nenext;
avl_tree_t slice_cache;
rdsk_node_t *slice;
void *cookie;
verify(iarg->poolname == NULL || iarg->guid == 0);
if (dirs == 0) {
#ifdef HAVE_LIBBLKID
/* Use libblkid to scan all device for their type */
if (zpool_find_import_blkid(hdl, &pools) == 0)
goto skip_scanning;
(void) zfs_error_fmt(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN, "blkid failure falling back "
"to manual probing"));
#endif /* HAVE_LIBBLKID */
dir = zpool_default_import_path;
dirs = DEFAULT_IMPORT_PATH_SIZE;
}
/*
* Go through and read the label configuration information from every
* possible device, organizing the information according to pool GUID
* and toplevel GUID.
*/
for (i = 0; i < dirs; i++) {
taskq_t *t;
char rdsk[MAXPATHLEN];
int dfd;
boolean_t config_failed = B_FALSE;
DIR *dirp;
/* use realpath to normalize the path */
if (realpath(dir[i], path) == 0) {
/* it is safe to skip missing search paths */
if (errno == ENOENT)
continue;
zfs_error_aux(hdl, strerror(errno));
(void) zfs_error_fmt(hdl, EZFS_BADPATH,
dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
goto error;
}
end = &path[strlen(path)];
*end++ = '/';
*end = 0;
pathleft = &path[sizeof (path)] - end;
/*
* Using raw devices instead of block devices when we're
* reading the labels skips a bunch of slow operations during
* close(2) processing, so we replace /dev/dsk with /dev/rdsk.
*/
if (strcmp(path, ZFS_DISK_ROOTD) == 0)
(void) strlcpy(rdsk, ZFS_RDISK_ROOTD, sizeof (rdsk));
else
(void) strlcpy(rdsk, path, sizeof (rdsk));
if ((dfd = open(rdsk, O_RDONLY)) < 0 ||
(dirp = fdopendir(dfd)) == NULL) {
if (dfd >= 0)
(void) close(dfd);
zfs_error_aux(hdl, strerror(errno));
(void) zfs_error_fmt(hdl, EZFS_BADPATH,
dgettext(TEXT_DOMAIN, "cannot open '%s'"),
rdsk);
goto error;
}
avl_create(&slice_cache, slice_cache_compare,
sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node));
/*
* This is not MT-safe, but we have no MT consumers of libzfs
*/
while ((dp = readdir(dirp)) != NULL) {
const char *name = dp->d_name;
if (name[0] == '.' &&
(name[1] == 0 || (name[1] == '.' && name[2] == 0)))
continue;
slice = zfs_alloc(hdl, sizeof (rdsk_node_t));
slice->rn_name = zfs_strdup(hdl, name);
slice->rn_avl = &slice_cache;
slice->rn_dfd = dfd;
slice->rn_hdl = hdl;
slice->rn_nozpool = B_FALSE;
avl_add(&slice_cache, slice);
}
/*
* create a thread pool to do all of this in parallel;
* rn_nozpool is not protected, so this is racy in that
* multiple tasks could decide that the same slice can
* not hold a zpool, which is benign. Also choose
* double the number of processors; we hold a lot of
* locks in the kernel, so going beyond this doesn't
* buy us much.
*/
t = taskq_create("z_import", 2 * max_ncpus, defclsyspri,
2 * max_ncpus, INT_MAX, TASKQ_PREPOPULATE);
for (slice = avl_first(&slice_cache); slice;
(slice = avl_walk(&slice_cache, slice,
AVL_AFTER)))
(void) taskq_dispatch(t, zpool_open_func, slice,
TQ_SLEEP);
taskq_wait(t);
taskq_destroy(t);
cookie = NULL;
while ((slice = avl_destroy_nodes(&slice_cache,
&cookie)) != NULL) {
if (slice->rn_config != NULL && !config_failed) {
nvlist_t *config = slice->rn_config;
boolean_t matched = B_TRUE;
if (iarg->poolname != NULL) {
char *pname;
matched = nvlist_lookup_string(config,
ZPOOL_CONFIG_POOL_NAME,
&pname) == 0 &&
strcmp(iarg->poolname, pname) == 0;
} else if (iarg->guid != 0) {
uint64_t this_guid;
matched = nvlist_lookup_uint64(config,
ZPOOL_CONFIG_POOL_GUID,
&this_guid) == 0 &&
iarg->guid == this_guid;
}
if (!matched) {
nvlist_free(config);
} else {
/*
* use the non-raw path for the config
*/
(void) strlcpy(end, slice->rn_name,
pathleft);
if (add_config(hdl, &pools, path, i+1,
slice->rn_num_labels, config) != 0)
config_failed = B_TRUE;
}
}
free(slice->rn_name);
free(slice);
}
avl_destroy(&slice_cache);
(void) closedir(dirp);
if (config_failed)
goto error;
}
#ifdef HAVE_LIBBLKID
skip_scanning:
#endif
ret = get_configs(hdl, &pools, iarg->can_be_active, iarg->policy);
error:
for (pe = pools.pools; pe != NULL; pe = penext) {
penext = pe->pe_next;
for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
venext = ve->ve_next;
for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
cenext = ce->ce_next;
if (ce->ce_config)
nvlist_free(ce->ce_config);
free(ce);
}
free(ve);
}
free(pe);
}
for (ne = pools.names; ne != NULL; ne = nenext) {
nenext = ne->ne_next;
free(ne->ne_name);
free(ne);
}
return (ret);
}
/*
* Go through and fix up any path and/or devid information for the given vdev
* configuration.
*/
static int
fix_paths(nvlist_t *nv, name_entry_t *names)
{
nvlist_t **child;
uint_t c, children;
uint64_t guid;
name_entry_t *ne, *best;
char *path, *devid;
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) == 0) {
for (c = 0; c < children; c++)
if (fix_paths(child[c], names) != 0)
return (-1);
return (0);
}
/*
* This is a leaf (file or disk) vdev. In either case, go through
* the name list and see if we find a matching guid. If so, replace
* the path and see if we can calculate a new devid.
*
* There may be multiple names associated with a particular guid, in
* which case we have overlapping partitions or multiple paths to the
* same disk. In this case we prefer to use the path name which
* matches the ZPOOL_CONFIG_PATH. If no matching entry is found we
* use the lowest order device which corresponds to the first match
* while traversing the ZPOOL_IMPORT_PATH search path.
*/
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
path = NULL;
best = NULL;
for (ne = names; ne != NULL; ne = ne->ne_next) {
if (ne->ne_guid == guid) {
if (path == NULL) {
best = ne;
break;
}
if ((strlen(path) == strlen(ne->ne_name)) &&
strncmp(path, ne->ne_name, strlen(path)) == 0) {
best = ne;
break;
}
if (best == NULL) {
best = ne;
continue;
}
/* Prefer paths with move vdev labels. */
if (ne->ne_num_labels > best->ne_num_labels) {
best = ne;
continue;
}
/* Prefer paths earlier in the search order. */
if (ne->ne_num_labels == best->ne_num_labels &&
ne->ne_order < best->ne_order) {
best = ne;
continue;
}
}
}
if (best == NULL)
return (0);
if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
return (-1);
if ((devid = get_devid(best->ne_name)) == NULL) {
(void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
} else {
if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) {
devid_str_free(devid);
return (-1);
}
devid_str_free(devid);
}
return (0);
}
/*
* Attach new_disk (fully described by nvroot) to old_disk.
* If 'replacing' is specified, tne new disk will replace the old one.
*/
int
zpool_vdev_attach(zpool_handle_t *zhp,
const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
{
zfs_cmd_t zc = { 0 };
char msg[1024];
char *packed;
int ret;
size_t len;
nvlist_t *tgt;
boolean_t avail_spare;
uint64_t val;
char *path;
nvlist_t **child;
uint_t children;
nvlist_t *config_root;
libzfs_handle_t *hdl = zhp->zpool_hdl;
if (replacing)
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot replace %s with %s"), old_disk, new_disk);
else
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot attach %s to %s"), new_disk, old_disk);
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare)) == 0)
return (zfs_error(hdl, EZFS_NODEVICE, msg));
if (avail_spare)
return (zfs_error(hdl, EZFS_ISSPARE, msg));
verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
zc.zc_cookie = replacing;
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0 || children != 1) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"new device must be a single disk"));
return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
}
verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
/*
* If the target is a hot spare that has been swapped in, we can only
* replace it with another hot spare.
*/
if (replacing &&
nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
nvlist_lookup_string(child[0], ZPOOL_CONFIG_PATH, &path) == 0 &&
(zpool_find_vdev(zhp, path, &avail_spare) == NULL ||
!avail_spare) && is_replacing_spare(config_root, tgt, 1)) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"can only be replaced by another hot spare"));
return (zfs_error(hdl, EZFS_BADTARGET, msg));
}
/*
* If we are attempting to replace a spare, it canot be applied to an
* already spared device.
*/
if (replacing &&
nvlist_lookup_string(child[0], ZPOOL_CONFIG_PATH, &path) == 0 &&
zpool_find_vdev(zhp, path, &avail_spare) != NULL && avail_spare &&
is_replacing_spare(config_root, tgt, 0)) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"device has already been replaced with a spare"));
return (zfs_error(hdl, EZFS_BADTARGET, msg));
}
verify(nvlist_size(nvroot, &len, NV_ENCODE_NATIVE) == 0);
if ((packed = zfs_alloc(zhp->zpool_hdl, len)) == NULL)
return (-1);
verify(nvlist_pack(nvroot, &packed, &len, NV_ENCODE_NATIVE, 0) == 0);
zc.zc_config_src = (uint64_t)(uintptr_t)packed;
zc.zc_config_src_size = len;
ret = ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_ATTACH, &zc);
free(packed);
if (ret == 0)
return (0);
switch (errno) {
case ENOTSUP:
/*
* Can't attach to or replace this type of vdev.
*/
if (replacing)
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"cannot replace a replacing device"));
else
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"can only attach to mirrors and top-level "
"disks"));
(void) zfs_error(hdl, EZFS_BADTARGET, msg);
break;
case EINVAL:
/*
* The new device must be a single disk.
*/
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"new device must be a single disk"));
(void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
break;
case EBUSY:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
new_disk);
(void) zfs_error(hdl, EZFS_BADDEV, msg);
break;
case EOVERFLOW:
/*
* The new device is too small.
*/
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"device is too small"));
(void) zfs_error(hdl, EZFS_BADDEV, msg);
break;
case EDOM:
/*
* The new device has a different alignment requirement.
*/
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"devices have different sector alignment"));
(void) zfs_error(hdl, EZFS_BADDEV, msg);
break;
case ENAMETOOLONG:
/*
* The resulting top-level vdev spec won't fit in the label.
*/
(void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
break;
default:
(void) zpool_standard_error(hdl, errno, msg);
}
return (-1);
}
