static void
inhm_dimmlist(uint32_t node, nvlist_t *nvl)
{
nvlist_t **dimmlist;
nvlist_t **newchannel;
int nchannels = CHANNELS_PER_MEMORY_CONTROLLER;
int nd;
uint8_t i, j;
nhm_dimm_t **dimmpp;
nhm_dimm_t *dimmp;
dimmlist = kmem_zalloc(sizeof (nvlist_t *) * MAX_DIMMS_PER_CHANNEL,
KM_SLEEP);
newchannel = kmem_zalloc(sizeof (nvlist_t *) * nchannels, KM_SLEEP);
dimmpp = &nhm_dimms[node * CHANNELS_PER_MEMORY_CONTROLLER *
MAX_DIMMS_PER_CHANNEL];
(void) nvlist_add_string(nvl, "memory-policy",
closed_page ? "closed-page" : "open-page");
(void) nvlist_add_string(nvl, "memory-ecc",
ecc_enabled ? lockstep[node] ? "x8" : "x4" : "no");
for (i = 0; i < nchannels; i++) {
(void) nvlist_alloc(&newchannel[i], NV_UNIQUE_NAME, KM_SLEEP);
(void) nvlist_add_string(newchannel[i], "channel-mode",
CHANNEL_DISABLED(MC_STATUS_RD(node), i) ? "disabled" :
i != 2 && lockstep[node] ? "lockstep" :
i != 2 && mirror_mode[node] ?
REDUNDANCY_LOSS(MC_RAS_STATUS_RD(node)) ?
"redundancy-loss" : "mirror" :
i == 2 && spare_channel[node] &&
!REDUNDANCY_LOSS(MC_RAS_STATUS_RD(node)) ? "spare" :
"independent");
nd = 0;
for (j = 0; j < MAX_DIMMS_PER_CHANNEL; j++) {
dimmp = *dimmpp;
if (dimmp != NULL) {
dimmlist[nd] = inhm_dimm(dimmp, node, i,
(uint32_t)j);
nd++;
}
dimmpp++;
}
if (nd) {
(void) nvlist_add_nvlist_array(newchannel[i],
"memory-dimms", dimmlist, nd);
for (j = 0; j < nd; j++)
nvlist_free(dimmlist[j]);
}
}
(void) nvlist_add_nvlist_array(nvl, MCINTEL_NVLIST_MC, newchannel,
nchannels);
for (i = 0; i < nchannels; i++)
nvlist_free(newchannel[i]);
kmem_free(dimmlist, sizeof (nvlist_t *) * MAX_DIMMS_PER_CHANNEL);
kmem_free(newchannel, sizeof (nvlist_t *) * nchannels);
}
static int
ch_end(list_wrap_t **lw, boolean_t array, int argc, char **argv)
{
nvlist_t *parent;
char *name;
if (list_wrap_depth(*lw) < 2) {
(void) fprintf(stderr, "ERROR: not nested, cannot end.\n");
return (-1);
}
parent = (*lw)->lw_next->lw_nvl[(*lw)->lw_next->lw_pos];
name = (*lw)->lw_name;
if ((*lw)->lw_array) {
/*
* This was an array of objects.
*/
nvlist_t **children = (*lw)->lw_nvl;
int nelems = (*lw)->lw_pos + 1;
if (nvlist_add_nvlist_array(parent, name, children,
nelems) != 0) {
(void) fprintf(stderr, "fail at "
"nvlist_add_nvlist_array\n");
return (-1);
}
} else {
/*
* This was a single object.
*/
nvlist_t *child = (*lw)->lw_nvl[0];
if ((*lw)->lw_pos != 0)
abort();
if (nvlist_add_nvlist(parent, name, child) != 0) {
(void) fprintf(stderr, "fail at nvlist_add_nvlist\n");
return (-1);
}
}
*lw = list_wrap_pop_and_free(*lw);
return (0);
}
static void
inhm_rank(nvlist_t *newdimm, nhm_dimm_t *nhm_dimm, uint32_t node,
uint8_t channel, uint32_t dimm, uint64_t rank_size)
{
nvlist_t **newrank;
int num;
int i;
uint64_t dimm_base;
uint64_t vrank_sz;
uint64_t rank_addr;
uint64_t pa;
uint32_t sinterleave, cinterleave, rinterleave;
uint32_t sway, cway, rway;
newrank = kmem_zalloc(sizeof (nvlist_t *) * nhm_dimm->nranks, KM_SLEEP);
for (i = 0; i < nhm_dimm->nranks; i++) {
(void) nvlist_alloc(&newrank[i], NV_UNIQUE_NAME, KM_SLEEP);
rank_addr = 0;
num = 0;
while (rank_addr < rank_size) {
pa = dimm_to_addr(node, channel, dimm * 4 + i,
rank_addr, &dimm_base, &vrank_sz, &sinterleave,
&cinterleave, &rinterleave, &sway, &cway, &rway);
if (pa == -1)
break;
inhm_vrank(newrank[i], num, dimm_base,
vrank_sz * sinterleave * cinterleave * rinterleave,
sinterleave, cinterleave, rinterleave, sway, cway,
rway);
rank_addr += vrank_sz;
num++;
}
}
(void) nvlist_add_nvlist_array(newdimm, MCINTEL_NVLIST_RANKS, newrank,
nhm_dimm->nranks);
for (i = 0; i < nhm_dimm->nranks; i++)
nvlist_free(newrank[i]);
kmem_free(newrank, sizeof (nvlist_t *) * nhm_dimm->nranks);
}
/*
* Retrieve the command history of a pool.
*/
int
zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
{
char buf[HIS_BUF_LEN];
uint64_t off = 0;
nvlist_t **records = NULL;
uint_t numrecords = 0;
int err, i;
do {
uint64_t bytes_read = sizeof (buf);
uint64_t leftover;
if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
break;
/* if nothing else was read in, we're at EOF, just return */
if (!bytes_read)
break;
if ((err = zpool_history_unpack(buf, bytes_read,
&leftover, &records, &numrecords)) != 0)
break;
off -= leftover;
/* CONSTCOND */
} while (1);
if (!err) {
verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
records, numrecords) == 0);
}
for (i = 0; i < numrecords; i++)
nvlist_free(records[i]);
free(records);
return (err);
}
static
#endif
int
zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain, char **retdomain,
dmu_tx_t *tx)
{
fuid_domain_t searchnode, *findnode;
avl_index_t loc;
/*
* If the dummy "nobody" domain then return an index of 0
* to cause the created FUID to be a standard POSIX id
* for the user nobody.
*/
if (domain[0] == '\0') {
*retdomain = "";
return (0);
}
searchnode.f_ksid = ksid_lookupdomain(domain);
if (retdomain) {
*retdomain = searchnode.f_ksid->kd_name;
}
if (!zfsvfs->z_fuid_loaded)
zfs_fuid_init(zfsvfs, tx);
rw_enter(&zfsvfs->z_fuid_lock, RW_READER);
findnode = avl_find(&zfsvfs->z_fuid_domain, &searchnode, &loc);
rw_exit(&zfsvfs->z_fuid_lock);
if (findnode) {
ksiddomain_rele(searchnode.f_ksid);
return (findnode->f_idx);
} else {
fuid_domain_t *domnode;
nvlist_t *nvp;
nvlist_t **fuids;
uint64_t retidx;
size_t nvsize = 0;
char *packed;
dmu_buf_t *db;
int i = 0;
domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
domnode->f_ksid = searchnode.f_ksid;
rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
retidx = domnode->f_idx = avl_numnodes(&zfsvfs->z_fuid_idx) + 1;
avl_add(&zfsvfs->z_fuid_domain, domnode);
avl_add(&zfsvfs->z_fuid_idx, domnode);
/*
* Now resync the on-disk nvlist.
*/
VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
domnode = avl_first(&zfsvfs->z_fuid_domain);
fuids = kmem_alloc(retidx * sizeof (void *), KM_SLEEP);
while (domnode) {
VERIFY(nvlist_alloc(&fuids[i],
NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
domnode->f_idx) == 0);
VERIFY(nvlist_add_uint64(fuids[i],
FUID_OFFSET, 0) == 0);
VERIFY(nvlist_add_string(fuids[i++], FUID_DOMAIN,
domnode->f_ksid->kd_name) == 0);
domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode);
}
VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
fuids, retidx) == 0);
for (i = 0; i != retidx; i++)
nvlist_free(fuids[i]);
kmem_free(fuids, retidx * sizeof (void *));
VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
packed = kmem_alloc(nvsize, KM_SLEEP);
VERIFY(nvlist_pack(nvp, &packed, &nvsize,
NV_ENCODE_XDR, KM_SLEEP) == 0);
nvlist_free(nvp);
zfsvfs->z_fuid_size = nvsize;
dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
zfsvfs->z_fuid_size, packed, tx);
kmem_free(packed, zfsvfs->z_fuid_size);
VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
FTAG, &db));
dmu_buf_will_dirty(db, tx);
*(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
dmu_buf_rele(db, FTAG);
rw_exit(&zfsvfs->z_fuid_lock);
return (retidx);
}
}
/*
* 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);
}
/*
* 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);
}
void
fnvlist_add_nvlist_array(nvlist_t *nvl, const char *name,
nvlist_t **val, uint_t n)
{
VERIFY0(nvlist_add_nvlist_array(nvl, name, val, n));
}
/*
* sync out AVL trees to persistent storage.
*/
void
zfs_fuid_sync(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
{
#ifdef HAVE_ZPL
nvlist_t *nvp;
nvlist_t **fuids;
size_t nvsize = 0;
char *packed;
dmu_buf_t *db;
fuid_domain_t *domnode;
int numnodes;
int i;
if (!zfsvfs->z_fuid_dirty) {
return;
}
rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
/*
* First see if table needs to be created?
*/
if (zfsvfs->z_fuid_obj == 0) {
zfsvfs->z_fuid_obj = dmu_object_alloc(zfsvfs->z_os,
DMU_OT_FUID, 1 << 14, DMU_OT_FUID_SIZE,
sizeof (uint64_t), tx);
VERIFY(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
ZFS_FUID_TABLES, sizeof (uint64_t), 1,
&zfsvfs->z_fuid_obj, tx) == 0);
}
VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
numnodes = avl_numnodes(&zfsvfs->z_fuid_idx);
fuids = kmem_alloc(numnodes * sizeof (void *), KM_SLEEP);
for (i = 0, domnode = avl_first(&zfsvfs->z_fuid_domain); domnode; i++,
domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode)) {
VERIFY(nvlist_alloc(&fuids[i], NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
domnode->f_idx) == 0);
VERIFY(nvlist_add_uint64(fuids[i], FUID_OFFSET, 0) == 0);
VERIFY(nvlist_add_string(fuids[i], FUID_DOMAIN,
domnode->f_ksid->kd_name) == 0);
}
VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
fuids, numnodes) == 0);
for (i = 0; i != numnodes; i++)
nvlist_free(fuids[i]);
kmem_free(fuids, numnodes * sizeof (void *));
VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
packed = kmem_alloc(nvsize, KM_SLEEP);
VERIFY(nvlist_pack(nvp, &packed, &nvsize,
NV_ENCODE_XDR, KM_SLEEP) == 0);
nvlist_free(nvp);
zfsvfs->z_fuid_size = nvsize;
dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
zfsvfs->z_fuid_size, packed, tx);
kmem_free(packed, zfsvfs->z_fuid_size);
VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
FTAG, &db));
dmu_buf_will_dirty(db, tx);
*(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
dmu_buf_rele(db, FTAG);
zfsvfs->z_fuid_dirty = B_FALSE;
rw_exit(&zfsvfs->z_fuid_lock);
#endif /* HAVE_ZPL */
}
int
smb_get_bb_fmri(smbios_hdl_t *shp, nvlist_t *fmri, uint_t parent,
smbs_cnt_t *bbstypes)
{
int rc = 0;
int i, j, n, cnt;
int id, index;
nvlist_t *pairs[MAX_PAIRS];
smbios_bboard_t bb;
uint16_t chassis_inst, mch_inst;
char name[40];
char idstr[11];
bbindex_t bb_idx;
uint16_t bbid;
int chcnt = 0;
for (n = 0; n < MAX_PAIRS; n++) {
bb_idx.index[n] = 0;
pairs[n] = NULL;
}
bb_idx.count = 0;
get_bboard_index(bbstypes, parent, &bb_idx);
index = bb_idx.index[0];
bbid = bbstypes->ids[index]->id;
rc = get_chassis_inst(shp, &chassis_inst, bbid, &chcnt);
if (rc != 0) {
return (rc);
}
if ((bb_idx.count + chcnt) > MAX_PAIRS) {
return (-1);
}
i = 0;
if (chcnt > 1) {
/*
* create main chassis pair
*/
pairs[i] = fm_nvlist_create(NULL);
if (pairs[i] == NULL) {
return (-1);
}
mch_inst = 0;
(void) snprintf(idstr, sizeof (idstr), "%u", mch_inst);
if ((nvlist_add_string(pairs[i], FM_FMRI_HC_NAME,
"chassis") != 0) ||
(nvlist_add_string(pairs[i], FM_FMRI_HC_ID, idstr)) != 0) {
fm_nvlist_destroy(pairs[i], FM_NVA_FREE);
return (-1);
}
i++;
}
/*
* create chassis pair
*/
pairs[i] = fm_nvlist_create(NULL);
if (pairs[i] == NULL) {
for (n = 0; n < MAX_PAIRS; n++) {
if (pairs[n] != NULL)
fm_nvlist_destroy(pairs[n], FM_NVA_FREE);
}
return (-1);
}
(void) snprintf(idstr, sizeof (idstr), "%u", chassis_inst);
if ((nvlist_add_string(pairs[i], FM_FMRI_HC_NAME, "chassis") != 0) ||
(nvlist_add_string(pairs[i], FM_FMRI_HC_ID, idstr) != 0)) {
for (n = 0; n < MAX_PAIRS; n++) {
if (pairs[n] != NULL)
fm_nvlist_destroy(pairs[n], FM_NVA_FREE);
}
return (-1);
}
for (j = 0, i = chcnt, cnt = chcnt; j < bb_idx.count; j++) {
index = bb_idx.index[j];
bbid = bbstypes->ids[index]->id;
rc = smbios_info_bboard(shp, bbid, &bb);
if (rc != 0) {
rc = -1;
break;
}
pairs[i] = fm_nvlist_create(NULL);
if (pairs[i] == NULL) {
rc = -1;
break;
}
id = bbstypes->ids[index]->inst;
(void) snprintf(idstr, sizeof (idstr), "%u", id);
(void) strncpy(name, bbd_type[bb.smbb_type].name,
sizeof (name));
cnt++;
if (nvlist_add_string(pairs[i], FM_FMRI_HC_NAME, name) != 0 ||
nvlist_add_string(pairs[i], FM_FMRI_HC_ID, idstr)
!= 0) {
rc = -1;
break;
}
i++;
}
if (rc != -1) {
if (nvlist_add_nvlist_array(fmri, FM_FMRI_HC_LIST,
pairs, cnt) != 0) {
rc = -1;
}
}
for (n = 0; n < cnt; n++) {
if (pairs[n] != NULL)
fm_nvlist_destroy(pairs[n], FM_NVA_FREE);
}
return (rc);
}
/*
* 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 not set, 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.
*/
static void
zfs_process_add(zpool_handle_t *zhp, nvlist_t *vdev, boolean_t isdisk)
{
char *path;
vdev_state_t newstate;
nvlist_t *nvroot, *newvd;
uint64_t wholedisk = 0ULL;
uint64_t offline = 0ULL;
char *physpath = NULL;
char rawpath[PATH_MAX], fullpath[PATH_MAX];
size_t len;
if (nvlist_lookup_string(vdev, ZPOOL_CONFIG_PATH, &path) != 0)
return;
(void) nvlist_lookup_string(vdev, ZPOOL_CONFIG_PHYS_PATH, &physpath);
(void) nvlist_lookup_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK, &wholedisk);
(void) nvlist_lookup_uint64(vdev, ZPOOL_CONFIG_OFFLINE, &offline);
/*
* We should have a way to online a device by guid. With the current
* interface, we are forced to chop off the 's0' for whole disks.
*/
(void) strlcpy(fullpath, path, sizeof (fullpath));
if (wholedisk)
fullpath[strlen(fullpath) - 2] = '\0';
/*
* Attempt to online the device. It would be nice to online this by
* GUID, but the current interface only supports lookup by path.
*/
if (offline ||
(zpool_vdev_online(zhp, fullpath,
ZFS_ONLINE_CHECKREMOVE | ZFS_ONLINE_UNSPARE, &newstate) == 0 &&
(newstate == VDEV_STATE_HEALTHY ||
newstate == VDEV_STATE_DEGRADED)))
return;
/*
* If the pool doesn't have the autoreplace property set, then attempt a
* true online (without the unspare flag), which will trigger a FMA
* fault.
*/
if (!zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOREPLACE, NULL) ||
(isdisk && !wholedisk)) {
(void) zpool_vdev_online(zhp, fullpath, ZFS_ONLINE_FORCEFAULT,
&newstate);
return;
}
if (isdisk) {
/*
* 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
* access to a raw node. Ideally, we'd like to walk the devinfo
* tree and find a raw node from the corresponding parent node.
* This is overly complicated, and since we know how we labeled
* this device in the first place, we know it's save to switch
* from /dev/dsk to /dev/rdsk and append the backup slice.
*
* 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).
*/
if (strncmp(path, "/dev/dsk/", 9) != 0) {
(void) zpool_vdev_online(zhp, fullpath,
ZFS_ONLINE_FORCEFAULT, &newstate);
return;
}
(void) strlcpy(rawpath, path + 9, sizeof (rawpath));
len = strlen(rawpath);
rawpath[len - 2] = '\0';
if (zpool_label_disk(g_zfshdl, zhp, rawpath) != 0) {
(void) zpool_vdev_online(zhp, fullpath,
ZFS_ONLINE_FORCEFAULT, &newstate);
return;
}
}
/*
* Cosntruct 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)
return;
if (nvlist_alloc(&newvd, NV_UNIQUE_NAME, 0) != 0) {
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 ||
(physpath != NULL && nvlist_add_string(newvd,
ZPOOL_CONFIG_PHYS_PATH, physpath) != 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) {
nvlist_free(newvd);
nvlist_free(nvroot);
return;
}
nvlist_free(newvd);
(void) zpool_vdev_attach(zhp, fullpath, path, nvroot, B_TRUE);
nvlist_free(nvroot);
}
/*
* 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 not set, 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;
char rawpath[PATH_MAX], fullpath[PATH_MAX];
char devpath[PATH_MAX];
int ret;
if (nvlist_lookup_string(vdev, ZPOOL_CONFIG_PATH, &path) != 0)
return;
(void) nvlist_lookup_string(vdev, ZPOOL_CONFIG_PHYS_PATH, &physpath);
(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 */
zed_log_msg(LOG_INFO, "zfs_process_add: pool '%s' vdev '%s' (%llu)",
zpool_get_name(zhp), path, (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(g_zfshdl, fullpath);
(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;
}
/*
* If the pool doesn't have the autoreplace property set, then attempt
* a true online (without the unspare flag), which will trigger a FMA
* fault.
*/
if (!zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOREPLACE, NULL) ||
!wholedisk || physpath == NULL) {
(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;
}
/*
* convert physical path into its current device node
*/
(void) snprintf(rawpath, sizeof (rawpath), "%s%s", DEV_BYPATH_PATH,
physpath);
if (realpath(rawpath, devpath) == NULL) {
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;
}
/*
* we're auto-replacing a raw disk, so label it first
*/
if (!labeled) {
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, (long long unsigned int)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;
}
}
if (!found) {
/* unexpected partition slice encountered */
(void) zpool_vdev_online(zhp, fullpath,
ZFS_ONLINE_FORCEFAULT, &newstate);
return;
}
zed_log_msg(LOG_INFO, " zpool_label_disk: resume '%s' (%llu)",
physpath, (long long unsigned int)guid);
if (nvlist_lookup_string(vdev, "new_devid", &new_devid) != 0) {
zed_log_msg(LOG_INFO, " auto replace: missing devid!");
return;
}
(void) snprintf(devpath, sizeof (devpath), "%s%s",
DEV_BYID_PATH, new_devid);
path = devpath;
}
/*
* 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) ||
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);
/*
* auto replace a leaf disk at same physical location
*/
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 int
topo_prop_set(tnode_t *node, const char *pgname, const char *pname,
topo_type_t type, int flag, void *val, int nelems, int *err)
{
int ret;
topo_hdl_t *thp = node->tn_hdl;
nvlist_t *nvl;
if (topo_hdl_nvalloc(thp, &nvl, NV_UNIQUE_NAME) < 0) {
*err = ETOPO_PROP_NVL;
return (-1);
}
ret = nvlist_add_string(nvl, TOPO_PROP_VAL_NAME, pname);
ret |= nvlist_add_uint32(nvl, TOPO_PROP_VAL_TYPE, type);
switch (type) {
case TOPO_TYPE_INT32:
ret |= nvlist_add_int32(nvl, TOPO_PROP_VAL_VAL,
*(int32_t *)val);
break;
case TOPO_TYPE_UINT32:
ret |= nvlist_add_uint32(nvl, TOPO_PROP_VAL_VAL,
*(uint32_t *)val);
break;
case TOPO_TYPE_INT64:
ret |= nvlist_add_int64(nvl, TOPO_PROP_VAL_VAL,
*(int64_t *)val);
break;
case TOPO_TYPE_UINT64:
ret |= nvlist_add_uint64(nvl, TOPO_PROP_VAL_VAL,
*(uint64_t *)val);
break;
case TOPO_TYPE_DOUBLE:
ret |= nvlist_add_double(nvl, TOPO_PROP_VAL_VAL,
*(double *)val);
break;
case TOPO_TYPE_STRING:
ret |= nvlist_add_string(nvl, TOPO_PROP_VAL_VAL,
(char *)val);
break;
case TOPO_TYPE_FMRI:
ret |= nvlist_add_nvlist(nvl, TOPO_PROP_VAL_VAL,
(nvlist_t *)val);
break;
case TOPO_TYPE_INT32_ARRAY:
ret |= nvlist_add_int32_array(nvl,
TOPO_PROP_VAL_VAL, (int32_t *)val, nelems);
break;
case TOPO_TYPE_UINT32_ARRAY:
ret |= nvlist_add_uint32_array(nvl,
TOPO_PROP_VAL_VAL, (uint32_t *)val, nelems);
break;
case TOPO_TYPE_INT64_ARRAY:
ret |= nvlist_add_int64_array(nvl,
TOPO_PROP_VAL_VAL, (int64_t *)val, nelems);
break;
case TOPO_TYPE_UINT64_ARRAY:
ret |= nvlist_add_uint64_array(nvl,
TOPO_PROP_VAL_VAL, (uint64_t *)val, nelems);
break;
case TOPO_TYPE_STRING_ARRAY:
ret |= nvlist_add_string_array(nvl,
TOPO_PROP_VAL_VAL, (char **)val, nelems);
break;
case TOPO_TYPE_FMRI_ARRAY:
ret |= nvlist_add_nvlist_array(nvl,
TOPO_PROP_VAL_VAL, (nvlist_t **)val, nelems);
break;
default:
*err = ETOPO_PROP_TYPE;
return (-1);
}
if (ret != 0) {
nvlist_free(nvl);
if (ret == ENOMEM) {
*err = ETOPO_PROP_NOMEM;
return (-1);
} else {
*err = ETOPO_PROP_NVL;
return (-1);
}
}
if (topo_prop_setprop(node, pgname, nvl, flag, nvl, err) != 0) {
nvlist_free(nvl);
return (-1); /* err set */
}
nvlist_free(nvl);
return (ret);
}
nvlist_t *
cmd_mkboard_fru(fmd_hdl_t *hdl, char *frustr, char *serialstr, char *partstr) {
char *nac, *nac_name;
int n, i, len;
nvlist_t *fru, **hc_list;
if (frustr == NULL)
return (NULL);
if ((nac_name = strstr(frustr, "MB")) == NULL)
return (NULL);
len = strlen(nac_name) + 1;
nac = fmd_hdl_zalloc(hdl, len, FMD_SLEEP);
(void) strcpy(nac, nac_name);
n = cmd_count_components(nac, '/');
fmd_hdl_debug(hdl, "cmd_mkboard_fru: nac=%s components=%d\n", nac, n);
hc_list = fmd_hdl_zalloc(hdl, sizeof (nvlist_t *)*n, FMD_SLEEP);
for (i = 0; i < n; i++) {
(void) nvlist_alloc(&hc_list[i],
NV_UNIQUE_NAME|NV_UNIQUE_NAME_TYPE, 0);
}
if (cmd_breakup_components(nac, "/", hc_list) < 0) {
for (i = 0; i < n; i++) {
if (hc_list[i] != NULL)
nvlist_free(hc_list[i]);
}
fmd_hdl_free(hdl, hc_list, sizeof (nvlist_t *)*n);
fmd_hdl_free(hdl, nac, len);
return (NULL);
}
if (nvlist_alloc(&fru, NV_UNIQUE_NAME, 0) != 0) {
for (i = 0; i < n; i++) {
if (hc_list[i] != NULL)
nvlist_free(hc_list[i]);
}
fmd_hdl_free(hdl, hc_list, sizeof (nvlist_t *)*n);
fmd_hdl_free(hdl, nac, len);
return (NULL);
}
if (nvlist_add_uint8(fru, FM_VERSION, FM_HC_SCHEME_VERSION) != 0 ||
nvlist_add_string(fru, FM_FMRI_SCHEME, FM_FMRI_SCHEME_HC) != 0 ||
nvlist_add_string(fru, FM_FMRI_HC_ROOT, "") != 0 ||
nvlist_add_uint32(fru, FM_FMRI_HC_LIST_SZ, n) != 0 ||
nvlist_add_nvlist_array(fru, FM_FMRI_HC_LIST, hc_list, n) != 0) {
for (i = 0; i < n; i++) {
if (hc_list[i] != NULL)
nvlist_free(hc_list[i]);
}
fmd_hdl_free(hdl, hc_list, sizeof (nvlist_t *)*n);
fmd_hdl_free(hdl, nac, len);
nvlist_free(fru);
return (NULL);
}
for (i = 0; i < n; i++) {
if (hc_list[i] != NULL)
nvlist_free(hc_list[i]);
}
fmd_hdl_free(hdl, hc_list, sizeof (nvlist_t *)*n);
fmd_hdl_free(hdl, nac, len);
if ((serialstr != NULL &&
nvlist_add_string(fru, FM_FMRI_HC_SERIAL_ID, serialstr) != 0) ||
(partstr != NULL &&
nvlist_add_string(fru, FM_FMRI_HC_PART, partstr) != 0)) {
nvlist_free(fru);
return (NULL);
}
return (fru);
}
int
main(int argc, char *argv[])
{
fmd_msg_hdl_t *h;
pid_t pid;
int i, err = 0;
char *s;
nvlist_t *auth, *fmri, *list, *test_arr[TEST_ARR_SZ];
const char *code = "TEST-8000-08";
int64_t tod[] = { 0x9400000, 0 };
if (argc > 1) {
(void) fprintf(stderr, "Usage: %s\n", argv[0]);
return (2);
}
/*
* Build up a valid list.suspect event for a fictional diagnosis
* using a diagnosis code from our test dictionary so we can format
* messages.
*/
if (nvlist_alloc(&auth, NV_UNIQUE_NAME, 0) != 0 ||
nvlist_alloc(&fmri, NV_UNIQUE_NAME, 0) != 0 ||
nvlist_alloc(&list, NV_UNIQUE_NAME, 0) != 0) {
(void) fprintf(stderr, "%s: nvlist_alloc failed\n", argv[0]);
return (1);
}
err |= nvlist_add_uint8(auth, FM_VERSION, FM_FMRI_AUTH_VERSION);
err |= nvlist_add_string(auth, FM_FMRI_AUTH_PRODUCT, "product");
err |= nvlist_add_string(auth, FM_FMRI_AUTH_PRODUCT_SN, "product_sn");
err |= nvlist_add_string(auth, FM_FMRI_AUTH_CHASSIS, "chassis");
err |= nvlist_add_string(auth, FM_FMRI_AUTH_DOMAIN, "domain");
err |= nvlist_add_string(auth, FM_FMRI_AUTH_SERVER, "server");
if (err != 0) {
(void) fprintf(stderr, "%s: failed to build auth nvlist: %s\n",
argv[0], strerror(err));
return (1);
}
err |= nvlist_add_uint8(fmri, FM_VERSION, FM_FMD_SCHEME_VERSION);
err |= nvlist_add_string(fmri, FM_FMRI_SCHEME, FM_FMRI_SCHEME_FMD);
err |= nvlist_add_nvlist(fmri, FM_FMRI_AUTHORITY, auth);
err |= nvlist_add_string(fmri, FM_FMRI_FMD_NAME, "fmd_msg_test");
err |= nvlist_add_string(fmri, FM_FMRI_FMD_VERSION, "1.0");
if (err != 0) {
(void) fprintf(stderr, "%s: failed to build fmri nvlist: %s\n",
argv[0], strerror(err));
return (1);
}
err |= nvlist_add_uint8(list, FM_VERSION, FM_SUSPECT_VERSION);
err |= nvlist_add_string(list, FM_CLASS, FM_LIST_SUSPECT_CLASS);
err |= nvlist_add_string(list, FM_SUSPECT_UUID, "12345678");
err |= nvlist_add_string(list, FM_SUSPECT_DIAG_CODE, code);
err |= nvlist_add_int64_array(list, FM_SUSPECT_DIAG_TIME, tod, 2);
err |= nvlist_add_nvlist(list, FM_SUSPECT_DE, fmri);
err |= nvlist_add_uint32(list, FM_SUSPECT_FAULT_SZ, 0);
/*
* Add a contrived nvlist array to our list.suspect so that we can
* exercise the expansion syntax for dereferencing nvlist array members
*/
for (i = 0; i < TEST_ARR_SZ; i++) {
if (nvlist_alloc(&test_arr[i], NV_UNIQUE_NAME, 0) != 0) {
(void) fprintf(stderr, "%s: failed to alloc nvlist "
"array: %s\n", argv[0], strerror(err));
return (1);
}
err |= nvlist_add_uint8(test_arr[i], "index", i);
}
err |= nvlist_add_nvlist_array(list, "test_arr", test_arr, TEST_ARR_SZ);
if (err != 0) {
(void) fprintf(stderr, "%s: failed to build list nvlist: %s\n",
argv[0], strerror(err));
return (1);
}
/*
* Now initialize the libfmd_msg library for testing, using the message
* catalogs found in the proto area of the current workspace.
*/
if ((h = fmd_msg_init(getenv("ROOT"), FMD_MSG_VERSION)) == NULL) {
(void) fprintf(stderr, "%s: fmd_msg_init failed: %s\n",
argv[0], strerror(errno));
return (1);
}
/*
* Test 0: Verify that both fmd_msg_getitem_id and fmd_msg_gettext_id
* return NULL and EINVAL for an illegal message code, and NULL
* and ENOENT for a valid but not defined message code.
*/
s = fmd_msg_getitem_id(h, NULL, "I_AM_NOT_VALID", 0);
if (s != NULL || errno != EINVAL) {
(void) fprintf(stderr, "%s: test0 FAIL: illegal code returned "
"s = %p, errno = %d\n", argv[0], (void *)s, errno);
return (1);
}
s = fmd_msg_gettext_id(h, NULL, "I_AM_NOT_VALID");
if (s != NULL || errno != EINVAL) {
(void) fprintf(stderr, "%s: test0 FAIL: illegal code returned "
"s = %p, errno = %d\n", argv[0], (void *)s, errno);
return (1);
}
s = fmd_msg_getitem_id(h, NULL, "I_AM_NOT_HERE-0000-0000", 0);
if (s != NULL || errno != ENOENT) {
(void) fprintf(stderr, "%s: test0 FAIL: missing code returned "
"s = %p, errno = %d\n", argv[0], (void *)s, errno);
return (1);
}
s = fmd_msg_gettext_id(h, NULL, "I_AM_NOT_HERE-0000-0000");
if (s != NULL || errno != ENOENT) {
(void) fprintf(stderr, "%s: test0 FAIL: missing code returned "
"s = %p, errno = %d\n", argv[0], (void *)s, errno);
return (1);
}
/*
* Test 1: Use fmd_msg_getitem_id to retrieve the item strings for
* a known message code without having any actual event handle.
*/
for (i = 0; i < FMD_MSG_ITEM_MAX; i++) {
if ((s = fmd_msg_getitem_id(h, NULL, code, i)) == NULL) {
(void) fprintf(stderr, "%s: fmd_msg_getitem_id failed "
"for %s, item %d: %s\n",
argv[0], code, i, strerror(errno));
}
(void) printf("code %s item %d = <<%s>>\n", code, i, s);
free(s);
}
/*
* Test 2: Use fmd_msg_gettext_id to retrieve the complete message for
* a known message code without having any actual event handle.
*/
if ((s = fmd_msg_gettext_id(h, NULL, code)) == NULL) {
(void) fprintf(stderr, "%s: fmd_msg_gettext_id failed for %s: "
"%s\n", argv[0], code, strerror(errno));
return (1);
}
(void) printf("%s\n", s);
free(s);
/*
* Test 3: Use fmd_msg_getitem_nv to retrieve the item strings for
* our list.suspect event handle.
*/
for (i = 0; i < FMD_MSG_ITEM_MAX; i++) {
if ((s = fmd_msg_getitem_nv(h, NULL, list, i)) == NULL) {
(void) fprintf(stderr, "%s: fmd_msg_getitem_nv failed "
"for %s, item %d: %s\n",
argv[0], code, i, strerror(errno));
}
(void) printf("code %s item %d = <<%s>>\n", code, i, s);
free(s);
}
/*
* Test 4: Use fmd_msg_getitem_nv to retrieve the complete message for
* a known message code using our list.suspect event handle.
*/
if ((s = fmd_msg_gettext_nv(h, NULL, list)) == NULL) {
(void) fprintf(stderr, "%s: fmd_msg_gettext_nv failed for %s: "
"%s\n", argv[0], code, strerror(errno));
return (1);
}
(void) printf("%s\n", s);
free(s);
/*
* Test 5: Use fmd_msg_getitem_nv to retrieve the complete message for
* a known message code using our list.suspect event handle, but this
* time set the URL to our own customized URL. Our contrived message
* has been designed to exercise the key aspects of the variable
* expansion syntax.
*/
if (fmd_msg_url_set(h, "http://foo.bar.com/") != 0) {
(void) fprintf(stderr, "%s: fmd_msg_url_set failed: %s\n",
argv[0], strerror(errno));
}
if ((s = fmd_msg_gettext_nv(h, NULL, list)) == NULL) {
(void) fprintf(stderr, "%s: fmd_msg_gettext_nv failed for %s: "
"%s\n", argv[0], code, strerror(errno));
return (1);
}
(void) printf("%s\n", s);
free(s);
for (i = 0; i < TEST_ARR_SZ; i++)
nvlist_free(test_arr[i]);
nvlist_free(fmri);
nvlist_free(auth);
nvlist_free(list);
fmd_msg_fini(h); /* free library state before dumping core */
pid = fork(); /* fork into background to not bother make(1) */
switch (pid) {
case -1:
(void) fprintf(stderr, "FAIL (failed to fork)\n");
return (1);
case 0:
abort();
return (1);
}
if (waitpid(pid, &err, 0) == -1) {
(void) fprintf(stderr, "FAIL (failed to wait for %d: %s)\n",
(int)pid, strerror(errno));
return (1);
}
if (WIFSIGNALED(err) == 0 || WTERMSIG(err) != SIGABRT) {
(void) fprintf(stderr, "FAIL (child did not SIGABRT)\n");
return (1);
}
if (!WCOREDUMP(err)) {
(void) fprintf(stderr, "FAIL (no core generated)\n");
return (1);
}
(void) fprintf(stderr, "done\n");
return (0);
}
/**
* Create the root of the vdev tree according to the parameters (type and vdev)
* @param psz_type: type of zpool (""=raid0, "mirror or "raidz")
* @param ppsz_dev: the list of devices
* @param i_dev: the number of devices
* @param ppsz_error: return the error message if any
* @return the root vded or NULL in case of error
*/
nvlist_t *lzwu_make_root_vdev(const char *psz_type, const char **ppsz_dev, size_t i_dev, const char **ppsz_error)
{
nvlist_t *pnv_root, **ppnv_top;
int i_mindev, i_maxdev, i_top = 0;
size_t i;
/* Check the type and the required number of devices */
if(!strcmp(psz_type, "raidz", 5))
{
int i_parity;
const char *psz_parity = psz_type + 5;
if(*psz_parity == '\0')
i_parity = 1;
else if(*psz_parity == '0')
{
*ppsz_error = "raidz0 does not exist";
return NULL;
}
else
{
const char *psz_end;
i_parity = strtol(psz_parity, &psz_end, 10);
if(i_parity < 1 || i_parity > 255 || *psz_end != '\0')
{
*ppsz_error = "raidz only accept values in [1, 255]";
return NULL;
}
}
i_mindev = i_parity + 1;
i_maxdev = 255;
psz_type = "raidz";
}
else if(!strcmp(psz_type, "mirror"))
{
i_mindev = 2;
i_maxdev = INT_MAX;
}
else if(psz_type[0] == '\0')
{
i_mindev = 1;
i_maxdev = INT_MAX;
}
else
{
*ppsz_error = "unknown zpool type: only '', 'mirror' and 'raidz' are handled";
return NULL;
}
/* Check the number of devices */
if(i_dev < i_mindev || i_dev > i_maxdev)
{
*ppsz_error = i_dev < i_mindev ? "too few devices" :
"too much devices";
return NULL;
}
if(psz_type[0] == '\0')
{
ppnv_top = malloc(i_dev * sizeof(nvlist_t*));
i_top = i_dev;
for(i = 0; i < i_dev; i++)
{
ppnv_top[i] = lzwu_make_leaf_vdev(ppsz_dev[i]);
if(!ppnv_top[i])
{
size_t j;
for(j = 0; j < i; j++)
nvlist_free(ppnv_top[j]);
free(ppnv_top);
*ppsz_error = "unable to create the vdev array";
return NULL;
}
}
}
else
{
/* List all the devices */
nvlist_t **pp_children = malloc(i_dev * sizeof(nvlist_t *));
for(i = 0; i < i_dev; i++)
{
pp_children[i] = lzwu_make_leaf_vdev(ppsz_dev[i]);
if(!pp_children[i])
{
size_t j;
for(j = 0; j < i; j++)
nvlist_free(pp_children[j]);
*ppsz_error = "unable to create the vdev array";
return NULL;
}
}
// Build the list of devices
nvlist_t *pnv_fs;
assert(nvlist_alloc(&pnv_fs, NV_UNIQUE_NAME, 0) == 0);
assert(nvlist_add_string(pnv_fs, ZPOOL_CONFIG_TYPE, psz_type) == 0);
assert(nvlist_add_nvlist_array(pnv_fs, ZPOOL_CONFIG_CHILDREN, pp_children, i_dev) == 0);
if(!strncmp(psz_type, "raidz", 5))
nvlist_add_uint64(pnv_fs, ZPOOL_CONFIG_NPARITY, i_mindev - 1);
i_top = 1;
ppnv_top = malloc(sizeof(nvlist_t*));
ppnv_top[0] = pnv_fs;
for(i = 0; i < i_dev; i++)
nvlist_free(pp_children[i]);
free(pp_children);
}
/* Create the root tree */
assert(nvlist_alloc(&pnv_root, NV_UNIQUE_NAME, 0) == 0);
assert(nvlist_add_string(pnv_root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0);
assert(nvlist_add_nvlist_array(pnv_root, ZPOOL_CONFIG_CHILDREN, ppnv_top, i_top) == 0);
for(i = 0; i < i_top; i++)
nvlist_free(ppnv_top[i]);
free(ppnv_top);
return pnv_root;
}
