return rc;
}
+static int
+lustre_lnet_config_healthv(int value, bool all, lnet_nid_t nid,
+ enum lnet_health_type type, char *name,
+ int seq_no, struct cYAML **err_rc)
+{
+ struct lnet_ioctl_reset_health_cfg data;
+ int rc = LUSTRE_CFG_RC_NO_ERR;
+ char err_str[LNET_MAX_STR_LEN];
+
+ snprintf(err_str, sizeof(err_str), "\"success\"");
+
+ LIBCFS_IOC_INIT_V2(data, rh_hdr);
+ data.rh_type = type;
+ data.rh_all = all;
+ data.rh_value = value;
+ data.rh_nid = nid;
+
+ rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_SET_HEALHV, &data);
+ if (rc != 0) {
+ rc = -errno;
+ snprintf(err_str,
+ sizeof(err_str), "Can not configure health value: %s",
+ strerror(errno));
+ }
+
+ cYAML_build_error(rc, seq_no, ADD_CMD, name, err_str, err_rc);
+
+ return rc;
+}
+
+int lustre_lnet_config_ni_healthv(int value, bool all, char *ni_nid, int seq_no,
+ struct cYAML **err_rc)
+{
+ lnet_nid_t nid;
+ if (ni_nid)
+ nid = libcfs_str2nid(ni_nid);
+ else
+ nid = LNET_NID_ANY;
+ return lustre_lnet_config_healthv(value, all, nid,
+ LNET_HEALTH_TYPE_LOCAL_NI,
+ "ni healthv", seq_no, err_rc);
+}
+
+int lustre_lnet_config_peer_ni_healthv(int value, bool all, char *lpni_nid,
+ int seq_no, struct cYAML **err_rc)
+{
+ lnet_nid_t nid;
+ if (lpni_nid)
+ nid = libcfs_str2nid(lpni_nid);
+ else
+ nid = LNET_NID_ANY;
+ return lustre_lnet_config_healthv(value, all, nid,
+ LNET_HEALTH_TYPE_PEER_NI,
+ "peer_ni healthv", seq_no, err_rc);
+}
+
static bool
add_msg_stats_to_yaml_blk(struct cYAML *yaml,
struct lnet_ioctl_comm_count *counts)
return rc;
}
+int lustre_lnet_config_recov_intrv(int intrv, int seq_no, struct cYAML **err_rc)
+{
+ int rc = LUSTRE_CFG_RC_NO_ERR;
+ char err_str[LNET_MAX_STR_LEN];
+ char val[LNET_MAX_STR_LEN];
+
+ snprintf(err_str, sizeof(err_str), "\"success\"");
+
+ snprintf(val, sizeof(val), "%d", intrv);
+
+ rc = write_sysfs_file(modparam_path, "lnet_recovery_interval", val,
+ 1, strlen(val) + 1);
+ if (rc)
+ snprintf(err_str, sizeof(err_str),
+ "\"cannot configure recovery interval: %s\"",
+ strerror(errno));
+
+ cYAML_build_error(rc, seq_no, ADD_CMD, "recovery_interval", err_str, err_rc);
+
+ return rc;
+}
+
int lustre_lnet_config_hsensitivity(int sen, int seq_no, struct cYAML **err_rc)
{
int rc = LUSTRE_CFG_RC_NO_ERR;
struct lnet_peer_ni_credit_info *lpni_cri;
struct lnet_ioctl_element_stats *lpni_stats;
struct lnet_ioctl_element_msg_stats *msg_stats;
+ struct lnet_ioctl_peer_ni_hstats *hstats;
lnet_nid_t *nidp;
int rc = LUSTRE_CFG_RC_OUT_OF_MEM;
int i, j, k;
__u32 size;
struct cYAML *root = NULL, *peer = NULL, *peer_ni = NULL,
*first_seq = NULL, *peer_root = NULL, *tmp = NULL,
- *msg_statistics = NULL, *statistics = NULL;
+ *msg_statistics = NULL, *statistics = NULL,
+ *yhstats;
char err_str[LNET_MAX_STR_LEN];
struct lnet_process_id *list = NULL;
void *data = NULL;
void *lpni_data;
+ bool exist = false;
snprintf(err_str, sizeof(err_str),
"\"out of memory\"");
l_errno = ENOMEM;
goto out;
}
+
for (i = 0; i < count; i++) {
for (;;) {
memset(&peer_info, 0, sizeof(peer_info));
goto out;
}
}
+ exist = true;
peer = cYAML_create_seq_item(peer_root);
if (peer == NULL)
lpni_cri = (void*)nidp + sizeof(nidp);
lpni_stats = (void *)lpni_cri + sizeof(*lpni_cri);
msg_stats = (void *)lpni_stats + sizeof(*lpni_stats);
- lpni_data = (void *)msg_stats + sizeof(*msg_stats);
+ hstats = (void *)msg_stats + sizeof(*msg_stats);
+ lpni_data = (void *)hstats + sizeof(*hstats);
peer_ni = cYAML_create_seq_item(tmp);
if (peer_ni == NULL)
goto out;
}
+ yhstats = cYAML_create_object(peer_ni, "health stats");
+ if (!yhstats)
+ goto out;
+ if (cYAML_create_number(yhstats, "health value",
+ hstats->hlpni_health_value)
+ == NULL)
+ goto out;
+ if (cYAML_create_number(yhstats, "dropped",
+ hstats->hlpni_remote_dropped)
+ == NULL)
+ goto out;
+ if (cYAML_create_number(yhstats, "timeout",
+ hstats->hlpni_remote_timeout)
+ == NULL)
+ goto out;
+ if (cYAML_create_number(yhstats, "error",
+ hstats->hlpni_remote_error)
+ == NULL)
+ goto out;
+ if (cYAML_create_number(yhstats, "network timeout",
+ hstats->hlpni_network_timeout)
+ == NULL)
+ goto out;
}
}
out:
free(list);
free(data);
- if (show_rc == NULL || rc != LUSTRE_CFG_RC_NO_ERR) {
+ if (show_rc == NULL || rc != LUSTRE_CFG_RC_NO_ERR || !exist) {
cYAML_free_tree(root);
} else if (show_rc != NULL && *show_rc != NULL) {
struct cYAML *show_node;
data.sv_value, show_rc, err_rc, l_errno);
}
+int lustre_lnet_show_recov_intrv(int seq_no, struct cYAML **show_rc,
+ struct cYAML **err_rc)
+{
+ int rc = LUSTRE_CFG_RC_OUT_OF_MEM;
+ char val[LNET_MAX_STR_LEN];
+ int intrv = -1, l_errno = 0;
+ char err_str[LNET_MAX_STR_LEN];
+
+ snprintf(err_str, sizeof(err_str), "\"out of memory\"");
+
+ rc = read_sysfs_file(modparam_path, "lnet_recovery_interval", val,
+ 1, sizeof(val));
+ if (rc) {
+ l_errno = -errno;
+ snprintf(err_str, sizeof(err_str),
+ "\"cannot get recovery interval: %d\"", rc);
+ } else {
+ intrv = atoi(val);
+ }
+
+ return build_global_yaml_entry(err_str, sizeof(err_str), seq_no,
+ "recovery_interval", intrv, show_rc,
+ err_rc, l_errno);
+}
+
int lustre_lnet_show_hsensitivity(int seq_no, struct cYAML **show_rc,
struct cYAML **err_rc)
{
err_rc, l_errno);
}
+int show_recovery_queue(enum lnet_health_type type, char *name, int seq_no,
+ struct cYAML **show_rc, struct cYAML **err_rc)
+{
+ struct lnet_ioctl_recovery_list nid_list;
+ struct cYAML *root = NULL, *nids = NULL;
+ int rc, i;
+ char err_str[LNET_MAX_STR_LEN];
+
+ snprintf(err_str, sizeof(err_str), "failed to print recovery queue\n");
+
+ LIBCFS_IOC_INIT_V2(nid_list, rlst_hdr);
+ nid_list.rlst_type = type;
+
+ rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_GET_RECOVERY_QUEUE, &nid_list);
+ if (rc) {
+ rc = errno;
+ goto out;
+ }
+
+ if (nid_list.rlst_num_nids == 0)
+ goto out;
+
+ root = cYAML_create_object(NULL, NULL);
+ if (root == NULL)
+ goto out;
+
+ nids = cYAML_create_object(root, name);
+ if (nids == NULL)
+ goto out;
+
+ rc = -EINVAL;
+
+ for (i = 0; i < nid_list.rlst_num_nids; i++) {
+ char nidenum[LNET_MAX_STR_LEN];
+ snprintf(nidenum, sizeof(nidenum), "nid-%d", i);
+ if (!cYAML_create_string(nids, nidenum,
+ libcfs_nid2str(nid_list.rlst_nid_array[i])))
+ goto out;
+ }
+
+ snprintf(err_str, sizeof(err_str), "success\n");
+
+ rc = 0;
+
+out:
+ if (show_rc == NULL || rc != LUSTRE_CFG_RC_NO_ERR) {
+ cYAML_free_tree(root);
+ } else if (show_rc != NULL && *show_rc != NULL) {
+ struct cYAML *show_node;
+ /* find the net node, if one doesn't exist
+ * then insert one. Otherwise add to the one there
+ */
+ show_node = cYAML_get_object_item(*show_rc, name);
+ if (show_node != NULL && cYAML_is_sequence(show_node)) {
+ cYAML_insert_child(show_node, nids);
+ free(nids);
+ free(root);
+ } else if (show_node == NULL) {
+ cYAML_insert_sibling((*show_rc)->cy_child,
+ nids);
+ free(root);
+ } else {
+ cYAML_free_tree(root);
+ }
+ } else {
+ *show_rc = root;
+ }
+
+ cYAML_build_error(rc, seq_no, SHOW_CMD, name, err_str, err_rc);
+
+ return rc;
+}
+
+int lustre_lnet_show_local_ni_recovq(int seq_no, struct cYAML **show_rc,
+ struct cYAML **err_rc)
+{
+ return show_recovery_queue(LNET_HEALTH_TYPE_LOCAL_NI, "local NI recovery",
+ seq_no, show_rc, err_rc);
+}
+
+int lustre_lnet_show_peer_ni_recovq(int seq_no, struct cYAML **show_rc,
+ struct cYAML **err_rc)
+{
+ return show_recovery_queue(LNET_HEALTH_TYPE_PEER_NI, "peer NI recovery",
+ seq_no, show_rc, err_rc);
+}
+
int lustre_lnet_show_max_intf(int seq_no, struct cYAML **show_rc,
struct cYAML **err_rc)
{
struct cYAML **err_rc)
{
struct lnet_ioctl_lnet_stats data;
+ struct lnet_counters *cntrs;
int rc;
int l_errno;
char err_str[LNET_MAX_STR_LEN];
LIBCFS_IOC_INIT_V2(data, st_hdr);
rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_GET_LNET_STATS, &data);
- if (rc != 0) {
+ if (rc) {
l_errno = errno;
snprintf(err_str,
sizeof(err_str),
rc = LUSTRE_CFG_RC_OUT_OF_MEM;
+ cntrs = &data.st_cntrs;
+
root = cYAML_create_object(NULL, NULL);
- if (root == NULL)
+ if (!root)
goto out;
stats = cYAML_create_object(root, "statistics");
- if (stats == NULL)
+ if (!stats)
goto out;
- if (cYAML_create_number(stats, "msgs_alloc",
- data.st_cntrs.msgs_alloc) == NULL)
+ if (!cYAML_create_number(stats, "msgs_alloc",
+ cntrs->lct_common.lcc_msgs_alloc))
goto out;
- if (cYAML_create_number(stats, "msgs_max",
- data.st_cntrs.msgs_max) == NULL)
+ if (!cYAML_create_number(stats, "msgs_max",
+ cntrs->lct_common.lcc_msgs_max))
goto out;
- if (cYAML_create_number(stats, "errors",
- data.st_cntrs.errors) == NULL)
+ if (!cYAML_create_number(stats, "rst_alloc",
+ cntrs->lct_health.lch_rst_alloc))
goto out;
- if (cYAML_create_number(stats, "send_count",
- data.st_cntrs.send_count) == NULL)
+ if (!cYAML_create_number(stats, "errors",
+ cntrs->lct_common.lcc_errors))
goto out;
- if (cYAML_create_number(stats, "recv_count",
- data.st_cntrs.recv_count) == NULL)
+ if (!cYAML_create_number(stats, "send_count",
+ cntrs->lct_common.lcc_send_count))
goto out;
- if (cYAML_create_number(stats, "route_count",
- data.st_cntrs.route_count) == NULL)
+ if (!cYAML_create_number(stats, "resend_count",
+ cntrs->lct_health.lch_resend_count))
goto out;
- if (cYAML_create_number(stats, "drop_count",
- data.st_cntrs.drop_count) == NULL)
+ if (!cYAML_create_number(stats, "response_timeout_count",
+ cntrs->lct_health.lch_response_timeout_count))
goto out;
- if (cYAML_create_number(stats, "send_length",
- data.st_cntrs.send_length) == NULL)
+ if (!cYAML_create_number(stats, "local_interrupt_count",
+ cntrs->lct_health.lch_local_interrupt_count))
goto out;
- if (cYAML_create_number(stats, "recv_length",
- data.st_cntrs.recv_length) == NULL)
+ if (!cYAML_create_number(stats, "local_dropped_count",
+ cntrs->lct_health.lch_local_dropped_count))
goto out;
- if (cYAML_create_number(stats, "route_length",
- data.st_cntrs.route_length) == NULL)
+ if (!cYAML_create_number(stats, "local_aborted_count",
+ cntrs->lct_health.lch_local_aborted_count))
goto out;
- if (cYAML_create_number(stats, "drop_length",
- data.st_cntrs.drop_length) == NULL)
+ if (!cYAML_create_number(stats, "local_no_route_count",
+ cntrs->lct_health.lch_local_no_route_count))
goto out;
- if (show_rc == NULL)
+ if (!cYAML_create_number(stats, "local_timeout_count",
+ cntrs->lct_health.lch_local_timeout_count))
+ goto out;
+
+ if (!cYAML_create_number(stats, "local_error_count",
+ cntrs->lct_health.lch_local_error_count))
+ goto out;
+
+ if (!cYAML_create_number(stats, "remote_dropped_count",
+ cntrs->lct_health.lch_remote_dropped_count))
+ goto out;
+
+ if (!cYAML_create_number(stats, "remote_error_count",
+ cntrs->lct_health.lch_remote_error_count))
+ goto out;
+
+ if (!cYAML_create_number(stats, "remote_timeout_count",
+ cntrs->lct_health.lch_remote_timeout_count))
+ goto out;
+
+ if (!cYAML_create_number(stats, "network_timeout_count",
+ cntrs->lct_health.lch_network_timeout_count))
+ goto out;
+
+ if (!cYAML_create_number(stats, "recv_count",
+ cntrs->lct_common.lcc_recv_count))
+ goto out;
+
+ if (!cYAML_create_number(stats, "route_count",
+ cntrs->lct_common.lcc_route_count))
+ goto out;
+
+ if (!cYAML_create_number(stats, "drop_count",
+ cntrs->lct_common.lcc_drop_count))
+ goto out;
+
+ if (!cYAML_create_number(stats, "send_length",
+ cntrs->lct_common.lcc_send_length))
+ goto out;
+
+ if (!cYAML_create_number(stats, "recv_length",
+ cntrs->lct_common.lcc_recv_length))
+ goto out;
+
+ if (!cYAML_create_number(stats, "route_length",
+ cntrs->lct_common.lcc_route_length))
+ goto out;
+
+ if (!cYAML_create_number(stats, "drop_length",
+ cntrs->lct_common.lcc_drop_length))
+ goto out;
+
+ if (!show_rc)
cYAML_print_tree(root);
snprintf(err_str, sizeof(err_str), "\"success\"");
{
char **nids = NULL;
int num, rc;
- struct cYAML *seq_no, *prim_nid, *non_mr, *ip2nets, *peer_nis;
+ struct cYAML *seq_no, *prim_nid, *mr, *ip2nets, *peer_nis;
char err_str[LNET_MAX_STR_LEN];
+ bool mr_value;
seq_no = cYAML_get_object_item(tree, "seq_no");
prim_nid = cYAML_get_object_item(tree, "primary nid");
- non_mr = cYAML_get_object_item(tree, "non_mr");
+ mr = cYAML_get_object_item(tree, "Multi-Rail");
ip2nets = cYAML_get_object_item(tree, "ip2nets");
peer_nis = cYAML_get_object_item(tree, "peer ni");
return rc;
}
+ if (!mr)
+ mr_value = true;
+ else {
+ if (!mr->cy_valuestring || !strcmp(mr->cy_valuestring, "True"))
+ mr_value = true;
+ else if (!strcmp(mr->cy_valuestring, "False"))
+ mr_value = false;
+ else {
+ rc = LUSTRE_CFG_RC_BAD_PARAM;
+ snprintf(err_str, sizeof(err_str), "Bad MR value");
+ cYAML_build_error(rc, (seq_no) ? seq_no->cy_valueint : -1,
+ ADD_CMD, "peer", err_str, err_rc);
+ return rc;
+ }
+ }
+
num = yaml_copy_peer_nids((ip2nets) ? ip2nets : peer_nis, &nids,
(prim_nid) ? prim_nid->cy_valuestring : NULL,
false);
}
rc = lustre_lnet_config_peer_nid((prim_nid) ? prim_nid->cy_valuestring : NULL,
- nids, num,
- (non_mr) ? false : true,
+ nids, num, mr_value,
(ip2nets) ? true : false,
(seq_no) ? seq_no->cy_valueint : -1,
err_rc);
struct cYAML **err_rc)
{
struct cYAML *max_intf, *numa, *discovery, *retry, *tto, *seq_no,
- *sen;
+ *sen, *recov;
int rc = 0;
seq_no = cYAML_get_object_item(tree, "seq_no");
: -1,
err_rc);
+ recov = cYAML_get_object_item(tree, "recovery_interval");
+ if (recov)
+ rc = lustre_lnet_config_recov_intrv(recov->cy_valueint,
+ seq_no ? seq_no->cy_valueint
+ : -1,
+ err_rc);
+
return rc;
}
struct cYAML **err_rc)
{
struct cYAML *max_intf, *numa, *discovery, *retry, *tto, *seq_no,
- *sen;
+ *sen, *recov;
int rc = 0;
seq_no = cYAML_get_object_item(tree, "seq_no");
: -1,
show_rc, err_rc);
+ recov = cYAML_get_object_item(tree, "recovery_interval");
+ if (recov)
+ rc = lustre_lnet_show_recov_intrv(seq_no ? seq_no->cy_valueint
+ : -1,
+ show_rc, err_rc);
+
return rc;
}