X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lnet%2Futils%2Flnetconfig%2Fliblnetconfig.c;h=3f24797d37eb9f49c7a5733ca05e96c33f57ce94;hb=4b2a760fc4747c748ce460289854b76ed17b9ca2;hp=ce460cf20b4520cc3932e1a491cede50271e7aab;hpb=15020fd977af68620e862ad999eaab17688933e2;p=fs%2Flustre-release.git diff --git a/lnet/utils/lnetconfig/liblnetconfig.c b/lnet/utils/lnetconfig/liblnetconfig.c index ce460cf..3f24797 100644 --- a/lnet/utils/lnetconfig/liblnetconfig.c +++ b/lnet/utils/lnetconfig/liblnetconfig.c @@ -50,7 +50,10 @@ #include #include #include +#include #include "liblnetconfig.h" +#include +#include #define CONFIG_CMD "configure" #define UNCONFIG_CMD "unconfigure" @@ -63,8 +66,13 @@ #define MAX_NUM_IPS 128 #define modparam_path "/sys/module/lnet/parameters/" +#define o2ib_modparam_path "/sys/module/ko2iblnd/parameters/" #define gni_nid_path "/proc/cray_xt/" +#ifndef HAVE_USRSPC_RDMA_PS_TCP +#define RDMA_PS_TCP 0x0106 +#endif + const char *gmsg_stat_names[] = {"sent_stats", "received_stats", "dropped_stats"}; @@ -248,102 +256,60 @@ void lustre_lnet_init_nw_descr(struct lnet_dlc_network_descr *nw_descr) } } -static char *get_next_delimiter_in_nid(char *str, char sep) +int lustre_lnet_parse_nidstr(char *nidstr, lnet_nid_t *lnet_nidlist, + int max_nids, char *err_str) { - char *at, *comma; - - /* first find the '@' */ - at = strchr(str, '@'); - if (!at) - return str; - - /* now that you found the at find the sep after */ - comma = strchr(at, sep); - return comma; -} + int rc, num_nids = 0; + struct list_head nidlist; -int lustre_lnet_parse_nids(char *nids, char **array, int size, - char ***out_array) -{ - int num_nids = 0; - char *comma = nids, *cur, *entry; - char **new_array; - int i, len, start = 0, finish = 0; - - if (nids == NULL || strlen(nids) == 0) - return size; - - /* count the number or new nids, by counting the number of comma*/ - while (comma) { - comma = get_next_delimiter_in_nid(comma, ','); - if (comma) { - comma++; - num_nids++; - } else { - num_nids++; - } + if (!nidstr) { + snprintf(err_str, LNET_MAX_STR_LEN, "supplied nidstr is NULL"); + return LUSTRE_CFG_RC_BAD_PARAM; } - /* - * if the array is not NULL allocate a large enough array to house - * the old and new entries - */ - new_array = calloc(sizeof(char*), - (size > 0) ? size + num_nids : num_nids); - - if (!new_array) - goto failed; - - /* parse our the new nids and add them to the tail of the array */ - comma = nids; - cur = nids; - start = (size > 0) ? size: 0; - finish = (size > 0) ? size + num_nids : num_nids; - for (i = start; i < finish; i++) { - comma = get_next_delimiter_in_nid(comma, ','); - if (!comma) - /* - * the length of the string to be parsed out is - * from cur to end of string. So it's good enough - * to strlen(cur) - */ - len = strlen(cur) + 1; - else - /* length of the string is comma - cur */ - len = (comma - cur) + 1; + if (strchr(nidstr, '*')) { + snprintf(err_str, LNET_MAX_STR_LEN, + "asterisk not allowed in nidstring \"%s\"", nidstr); + return LUSTRE_CFG_RC_BAD_PARAM; + } - entry = calloc(1, len); - if (!entry) { - finish = i > 0 ? i - 1: 0; - goto failed; - } - strncpy(entry, cur, len - 1); - entry[len] = '\0'; - new_array[i] = entry; - if (comma) { - comma++; - cur = comma; - } + INIT_LIST_HEAD(&nidlist); + rc = cfs_parse_nidlist(nidstr, strlen(nidstr), &nidlist); + if (rc == 0) { + snprintf(err_str, LNET_MAX_STR_LEN, + "Unable to parse nidlist from: %s\n", nidstr); + return LUSTRE_CFG_RC_BAD_PARAM; } - /* add the old entries in the array and delete the old array*/ - for (i = 0; i < size; i++) - new_array[i] = array[i]; + if (list_empty(&nidlist)) { + snprintf(err_str, LNET_MAX_STR_LEN, + "\"%s\" does not specify any valid nid lists", nidstr); + return LUSTRE_CFG_RC_BAD_PARAM; + } - if (array) - free(array); + num_nids = cfs_expand_nidlist(&nidlist, lnet_nidlist, max_nids); + cfs_free_nidlist(&nidlist); - *out_array = new_array; + if (num_nids == -1) { + snprintf(err_str, LNET_MAX_STR_LEN, + "\"%s\" specifies more than the %d NIDs allowed by this operation.", + nidstr, max_nids); + return LUSTRE_CFG_RC_BAD_PARAM; + } - return finish; + if (num_nids < 0) { + snprintf(err_str, LNET_MAX_STR_LEN, + "Failed to expand nidstr: %s", strerror(num_nids)); + return LUSTRE_CFG_RC_OUT_OF_MEM; + } -failed: - for (i = start; i < finish; i++) - free(new_array[i]); - if (new_array) - free(new_array); + if (num_nids == 0) { + snprintf(err_str, LNET_MAX_STR_LEN, + "\"%s\" did not expand to any nids", nidstr); + return LUSTRE_CFG_RC_BAD_PARAM; + } - return size; + return num_nids; } /* @@ -427,9 +393,7 @@ int lustre_lnet_config_ni_system(bool up, bool load_ni_from_mod, struct libcfs_ioctl_data data; unsigned int opc; int rc; - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"Success\""); + char err_str[LNET_MAX_STR_LEN] = "\"Success\""; LIBCFS_IOC_INIT(data); @@ -454,23 +418,17 @@ int lustre_lnet_config_ni_system(bool up, bool load_ni_from_mod, return rc; } -static int dispatch_peer_ni_cmd(lnet_nid_t pnid, lnet_nid_t nid, __u32 cmd, - struct lnet_ioctl_peer_cfg *data, +static int dispatch_peer_ni_cmd(__u32 cmd, struct lnet_ioctl_peer_cfg *data, char *err_str, char *cmd_str) { int rc; - data->prcfg_prim_nid = pnid; - data->prcfg_cfg_nid = nid; - rc = l_ioctl(LNET_DEV_ID, cmd, data); - if (rc != 0) { + if (rc) { rc = -errno; - snprintf(err_str, - LNET_MAX_STR_LEN, - "\"cannot %s peer ni: %s\"", - (cmd_str) ? cmd_str : "add", strerror(errno)); - err_str[LNET_MAX_STR_LEN - 1] = '\0'; + snprintf(err_str, LNET_MAX_STR_LEN, + "\"%s peer ni operation failed: %s\"", + cmd_str, strerror(errno)); } return rc; @@ -602,8 +560,7 @@ static int infra_ping_nid(char *ping_nids, char *oper, int param, int ioc_call, goto out; for (i = 0; i < ping.ping_count; i++) { - if (!strcmp(libcfs_nid2str(ping.ping_buf[i].nid), - "0@lo")) + if (ping.ping_buf[i].nid == LNET_NID_LO_0) continue; peer_ni = cYAML_create_seq_item(tmp); if (peer_ni == NULL) @@ -668,341 +625,176 @@ int lustre_lnet_discover_nid(char *ping_nids, int force, int seq_no, return rc; } -static void lustre_lnet_clean_ip2nets(struct lustre_lnet_ip2nets *ip2nets) +static int lustre_lnet_handle_peer_nidlist(lnet_nid_t *nidlist, int num_nids, + bool is_mr, __u32 cmd, + char *cmd_type, char *err_str) { - struct lustre_lnet_ip_range_descr *ipr, *tmp; - struct cfs_expr_list *el, *el_tmp; + struct lnet_ioctl_peer_cfg data; + int rc, nid_idx; - list_for_each_entry_safe(ipr, tmp, - &ip2nets->ip2nets_ip_ranges, - ipr_entry) { - list_del(&ipr->ipr_entry); - list_for_each_entry_safe(el, el_tmp, &ipr->ipr_expr, - el_link) { - list_del(&el->el_link); - cfs_expr_list_free(el); - } - free(ipr); - } -} + if (cmd == IOC_LIBCFS_ADD_PEER_NI) { + /* When adding a peer we first need to create the peer using the + * specified (or implied) primary nid. Then we can add + * additional nids to this peer using the primary nid as a key + */ + LIBCFS_IOC_INIT_V2(data, prcfg_hdr); + data.prcfg_mr = is_mr; + data.prcfg_prim_nid = nidlist[0]; + data.prcfg_cfg_nid = LNET_NID_ANY; -/* - * returns an rc < 0 if there is an error - * otherwise it returns the number IPs generated - * it also has out params: net - network name - */ -static int lnet_expr2ips(char *nidstr, __u32 *ip_list, - struct lustre_lnet_ip2nets *ip2nets, - __u32 *net, char *err_str) -{ - struct lustre_lnet_ip_range_descr *ipr; - char *comp1, *comp2; - int ip_idx = MAX_NUM_IPS - 1; - int ip_range_len, rc = LUSTRE_CFG_RC_NO_ERR; - __u32 net_type; - char ip_range[LNET_MAX_STR_LEN]; - - /* separate the two components of the NID */ - comp1 = nidstr; - comp2 = strchr(nidstr, '@'); - if (comp2 == NULL) { - snprintf(err_str, - LNET_MAX_STR_LEN, - "\"cannot parse NID %s\"", nidstr); - err_str[LNET_MAX_STR_LEN - 1] = '\0'; - rc = LUSTRE_CFG_RC_BAD_PARAM; - goto out; - } + rc = dispatch_peer_ni_cmd(cmd, &data, err_str, cmd_type); - /* length of the expected ip-range */ - ip_range_len = comp2 - comp1; - if (ip_range_len >= LNET_MAX_STR_LEN) { - snprintf(err_str, - LNET_MAX_STR_LEN, - "\"cannot parse ip_range '%s'\"", ip_range); - err_str[LNET_MAX_STR_LEN - 1] = '\0'; - rc = LUSTRE_CFG_RC_BAD_PARAM; - goto out; + if (rc) + return rc; } - /* move beyond '@' */ - comp2++; - - /* - * if the net component is either o2ib or tcp then we expect - * an IP range which could only be a single IP address. - * Parse that. + /* Add or delete any specified NIs associated with the specified + * (or implied) primary nid */ - *net = libcfs_str2net(comp2); - net_type = LNET_NETTYP(*net); - /* expression support is for o2iblnd and socklnd only */ - if (net_type != O2IBLND && net_type != SOCKLND) - return LUSTRE_CFG_RC_SKIP; - - strncpy(ip_range, comp1, ip_range_len); - ip_range[ip_range_len] = '\0'; - ip2nets->ip2nets_net.nw_id = *net; + for (nid_idx = 1; nid_idx < num_nids; nid_idx++) { + LIBCFS_IOC_INIT_V2(data, prcfg_hdr); + data.prcfg_mr = is_mr; + data.prcfg_prim_nid = nidlist[0]; + data.prcfg_cfg_nid = nidlist[nid_idx]; - rc = lustre_lnet_add_ip_range(&ip2nets->ip2nets_ip_ranges, ip_range); - if (rc != LUSTRE_CFG_RC_NO_ERR) { - snprintf(err_str, - LNET_MAX_STR_LEN, - "\"cannot parse ip_range '%s'\"", ip_range); - err_str[LNET_MAX_STR_LEN - 1] = '\0'; - rc = LUSTRE_CFG_RC_BAD_PARAM; - goto out; - } + rc = dispatch_peer_ni_cmd(cmd, &data, err_str, cmd_type); - /* - * Generate all the IP Addresses from the parsed range. For sanity - * we allow only a max of MAX_NUM_IPS nids to be configured for - * a single peer. - */ - list_for_each_entry(ipr, &ip2nets->ip2nets_ip_ranges, ipr_entry) - ip_idx = cfs_ip_addr_range_gen(ip_list, MAX_NUM_IPS, - &ipr->ipr_expr); - - if (ip_idx == MAX_NUM_IPS - 1) { - snprintf(err_str, LNET_MAX_STR_LEN, - "no NIDs provided for configuration"); - err_str[LNET_MAX_STR_LEN - 1] = '\0'; - rc = LUSTRE_CFG_RC_NO_MATCH; - goto out; - } else if (ip_idx == -1) { - rc = LUSTRE_CFG_RC_LAST_ELEM; - } else { - rc = ip_idx; + if (rc) + return rc; } -out: - return rc; -} - -static int lustre_lnet_handle_peer_ip2nets(char **nid, int num_nids, bool mr, - bool range, __u32 cmd, - char *cmd_type, char *err_str) -{ - __u32 net = LNET_NIDNET(LNET_NID_ANY); - int ip_idx; - int i, j, rc = LUSTRE_CFG_RC_NO_ERR; - __u32 ip_list[MAX_NUM_IPS]; - struct lustre_lnet_ip2nets ip2nets; - struct lnet_ioctl_peer_cfg data; - lnet_nid_t peer_nid; - lnet_nid_t prim_nid = LNET_NID_ANY; - - /* initialize all lists */ - INIT_LIST_HEAD(&ip2nets.ip2nets_ip_ranges); - INIT_LIST_HEAD(&ip2nets.ip2nets_net.network_on_rule); - INIT_LIST_HEAD(&ip2nets.ip2nets_net.nw_intflist); - - /* each nid entry is an expression */ - for (i = 0; i < num_nids; i++) { - if (!range && i == 0) - prim_nid = libcfs_str2nid(nid[0]); - else if (range) - prim_nid = LNET_NID_ANY; - - rc = lnet_expr2ips(nid[i], ip_list, &ip2nets, &net, err_str); - if (rc == LUSTRE_CFG_RC_SKIP) - continue; - else if (rc == LUSTRE_CFG_RC_LAST_ELEM) - rc = -1; - else if (rc < LUSTRE_CFG_RC_NO_ERR) - goto out; - - ip_idx = rc; - - for (j = MAX_NUM_IPS - 1; j > ip_idx; j--) { - peer_nid = LNET_MKNID(net, ip_list[j]); - if (peer_nid == LNET_NID_ANY) { - snprintf(err_str, - LNET_MAX_STR_LEN, - "\"cannot parse NID\""); - err_str[LNET_MAX_STR_LEN - 1] = '\0'; - rc = LUSTRE_CFG_RC_BAD_PARAM; - goto out; - } - - LIBCFS_IOC_INIT_V2(data, prcfg_hdr); - data.prcfg_mr = mr; - - if (prim_nid == LNET_NID_ANY && j == MAX_NUM_IPS - 1) { - prim_nid = peer_nid; - peer_nid = LNET_NID_ANY; - } - - if (!range && num_nids > 1 && i == 0 && - cmd == IOC_LIBCFS_DEL_PEER_NI) - continue; - else if (!range && i == 0) - peer_nid = LNET_NID_ANY; - - /* - * If prim_nid is not provided then the first nid in the - * list becomes the prim_nid. First time round the loop - * use LNET_NID_ANY for the first parameter, then use - * nid[0] as the key nid after wards - */ - rc = dispatch_peer_ni_cmd(prim_nid, peer_nid, cmd, - &data, err_str, cmd_type); - if (rc != 0) - goto out; + if (cmd == IOC_LIBCFS_DEL_PEER_NI && num_nids == 1) { + /* In the delete case we may have been given just the + * primary nid of the peer. This tells us to delete the peer + * completely (rather than just delete some of its NIs) + */ + LIBCFS_IOC_INIT_V2(data, prcfg_hdr); + data.prcfg_prim_nid = nidlist[0]; + data.prcfg_cfg_nid = LNET_NID_ANY; - /* - * we just deleted the entire peer using the - * primary_nid. So don't bother iterating through - * the rest of the nids - */ - if (prim_nid != LNET_NID_ANY && - peer_nid == LNET_NID_ANY && - cmd == IOC_LIBCFS_DEL_PEER_NI) - goto next_nid; - } -next_nid: - lustre_lnet_clean_ip2nets(&ip2nets); + rc = dispatch_peer_ni_cmd(cmd, &data, err_str, cmd_type); } -out: - lustre_lnet_clean_ip2nets(&ip2nets); return rc; } -int lustre_lnet_config_peer_nid(char *pnid, char **nid, int num_nids, - bool mr, bool ip2nets, int seq_no, - struct cYAML **err_rc) +static int +lustre_lnet_mod_peer_nidlist(lnet_nid_t pnid, lnet_nid_t *lnet_nidlist, + int cmd, int num_nids, bool is_mr, int seq_no, + struct cYAML **err_rc) { int rc = LUSTRE_CFG_RC_NO_ERR; - char err_str[LNET_MAX_STR_LEN] = {0}; - char **nid_array = NULL; - - snprintf(err_str, sizeof(err_str), "\"Success\""); + char err_str[LNET_MAX_STR_LEN]; + lnet_nid_t *lnet_nidlist2 = NULL; + int ioc_cmd = (cmd == LNETCTL_ADD_CMD) ? IOC_LIBCFS_ADD_PEER_NI : + IOC_LIBCFS_DEL_PEER_NI; + char *cmd_str = (cmd == LNETCTL_ADD_CMD) ? ADD_CMD : DEL_CMD; - if (ip2nets) { - rc = lustre_lnet_handle_peer_ip2nets(nid, num_nids, mr, - ip2nets, IOC_LIBCFS_ADD_PEER_NI, - ADD_CMD, err_str); + num_nids++; + lnet_nidlist2 = calloc(sizeof(*lnet_nidlist2), num_nids); + if (!lnet_nidlist2) { + snprintf(err_str, LNET_MAX_STR_LEN, "out of memory"); + rc = LUSTRE_CFG_RC_OUT_OF_MEM; goto out; } + lnet_nidlist2[0] = pnid; + memcpy(&lnet_nidlist2[1], lnet_nidlist, sizeof(*lnet_nidlist) * + (num_nids - 1)); - if (pnid) { - if (libcfs_str2nid(pnid) == LNET_NID_ANY) { - snprintf(err_str, sizeof(err_str), - "bad primary NID: '%s'", - pnid); - rc = LUSTRE_CFG_RC_MISSING_PARAM; - goto out; - } - - num_nids++; - - nid_array = calloc(sizeof(*nid_array), num_nids); - if (!nid_array) { - snprintf(err_str, sizeof(err_str), - "out of memory"); - rc = LUSTRE_CFG_RC_OUT_OF_MEM; - goto out; - } - nid_array[0] = pnid; - memcpy(&nid_array[1], nid, sizeof(*nid) * (num_nids - 1)); - } - - rc = lustre_lnet_handle_peer_ip2nets((pnid) ? nid_array : nid, - num_nids, mr, ip2nets, - IOC_LIBCFS_ADD_PEER_NI, ADD_CMD, - err_str); - if (rc) - goto out; - + rc = lustre_lnet_handle_peer_nidlist(lnet_nidlist2, + num_nids, is_mr, ioc_cmd, + cmd_str, err_str); out: - if (nid_array) - free(nid_array); + if (lnet_nidlist2) + free(lnet_nidlist2); - cYAML_build_error(rc, seq_no, ADD_CMD, "peer_ni", err_str, err_rc); + cYAML_build_error(rc, seq_no, cmd_str, "peer_ni", err_str, err_rc); return rc; } -int lustre_lnet_del_peer_nid(char *pnid, char **nid, int num_nids, - bool ip2nets, int seq_no, struct cYAML **err_rc) +static void +replace_sep(char *str, char sep, char newsep) { - int rc = LUSTRE_CFG_RC_NO_ERR; - char err_str[LNET_MAX_STR_LEN] = {0}; - char **nid_array = NULL; + int bracket = 0; + int i; + if (!str) + return; + for (i = 0; i < strlen(str); i++) { + /* don't replace ',' within [] */ + if (str[i] == '[') + bracket++; + else if (str[i] == ']') + bracket--; + else if (str[i] == sep && bracket == 0) + str[i] = newsep; + } +} - snprintf(err_str, sizeof(err_str), "\"Success\""); +int lustre_lnet_modify_peer(char *prim_nid, char *nids, bool is_mr, + int cmd, int seq_no, struct cYAML **err_rc) +{ + int num_nids, rc; + char err_str[LNET_MAX_STR_LEN] = "Error"; + lnet_nid_t lnet_nidlist[LNET_MAX_NIDS_PER_PEER]; + lnet_nid_t pnid = LNET_NID_ANY; - if (ip2nets) { - rc = lustre_lnet_handle_peer_ip2nets(nid, num_nids, false, - ip2nets, IOC_LIBCFS_DEL_PEER_NI, - DEL_CMD, err_str); + if (!prim_nid) { + rc = LUSTRE_CFG_RC_BAD_PARAM; + snprintf(err_str, LNET_MAX_STR_LEN, + "--prim_nid must be specified"); goto out; } - if (pnid == NULL) { - snprintf(err_str, sizeof(err_str), - "\"Primary nid is not provided\""); - rc = LUSTRE_CFG_RC_MISSING_PARAM; + pnid = libcfs_str2nid(prim_nid); + if (pnid == LNET_NID_ANY) { + rc = LUSTRE_CFG_RC_BAD_PARAM; + snprintf(err_str, LNET_MAX_STR_LEN, + "badly formatted primary NID: %s", prim_nid); goto out; - } else if (!ip2nets) { - if (libcfs_str2nid(pnid) == LNET_NID_ANY) { - rc = LUSTRE_CFG_RC_BAD_PARAM; - snprintf(err_str, sizeof(err_str), - "bad key NID: '%s'", - pnid); - goto out; - } } - num_nids++; - nid_array = calloc(sizeof(*nid_array), num_nids); - if (!nid_array) { - snprintf(err_str, sizeof(err_str), - "out of memory"); - rc = LUSTRE_CFG_RC_OUT_OF_MEM; - goto out; + num_nids = 0; + if (nids) { + /* + * if there is no primary nid we need to make the first nid in the + * nids list the primary nid + */ + replace_sep(nids, ',', ' '); + rc = lustre_lnet_parse_nidstr(nids, lnet_nidlist, + LNET_MAX_NIDS_PER_PEER, err_str); + if (rc < 0) + goto out; + + num_nids = rc; } - nid_array[0] = pnid; - memcpy(&nid_array[1], nid, sizeof(*nid) * (num_nids - 1)); - rc = lustre_lnet_handle_peer_ip2nets(nid_array, num_nids, false, - ip2nets, IOC_LIBCFS_DEL_PEER_NI, - DEL_CMD, err_str); - if (rc) - goto out; + rc = lustre_lnet_mod_peer_nidlist(pnid, lnet_nidlist, + cmd, num_nids, is_mr, + -1, err_rc); out: - if (nid_array) - free(nid_array); + if (rc != LUSTRE_CFG_RC_NO_ERR) + cYAML_build_error(rc, -1, "peer", + cmd == LNETCTL_ADD_CMD ? "add" : "del", + err_str, err_rc); - cYAML_build_error(rc, seq_no, DEL_CMD, "peer_ni", err_str, err_rc); return rc; } -int lustre_lnet_config_route(char *nw, char *gw, int hops, int prio, - int seq_no, struct cYAML **err_rc) +int lustre_lnet_route_common(char *nw, char *nidstr, int hops, int prio, + int sen, int seq_no, struct cYAML **err_rc, + int cmd) { + int rc, num_nids, idx; + __u32 rnet; + char err_str[LNET_MAX_STR_LEN] = "\"generic error\""; struct lnet_ioctl_config_data data; - lnet_nid_t gateway_nid; - int rc = LUSTRE_CFG_RC_NO_ERR; - int ip_idx, i; - __u32 rnet = LNET_NIDNET(LNET_NID_ANY); - __u32 net = LNET_NIDNET(LNET_NID_ANY); - char err_str[LNET_MAX_STR_LEN]; - __u32 ip_list[MAX_NUM_IPS]; - struct lustre_lnet_ip2nets ip2nets; - - /* initialize all lists */ - INIT_LIST_HEAD(&ip2nets.ip2nets_ip_ranges); - INIT_LIST_HEAD(&ip2nets.ip2nets_net.network_on_rule); - INIT_LIST_HEAD(&ip2nets.ip2nets_net.nw_intflist); + lnet_nid_t lnet_nidlist[LNET_MAX_NIDS_PER_PEER]; - snprintf(err_str, sizeof(err_str), "\"Success\""); - - if (nw == NULL || gw == NULL) { - snprintf(err_str, - sizeof(err_str), - "\"missing mandatory parameter in route config:'%s'\"", - (nw == NULL && gw == NULL) ? "network, gateway" : + if (nw == NULL || nidstr == NULL) { + snprintf(err_str, LNET_MAX_STR_LEN, + "\"missing mandatory parameter:'%s'\"", + (nw == NULL && nidstr == NULL) ? "network, gateway" : (nw == NULL) ? "network" : "gateway"); rc = LUSTRE_CFG_RC_MISSING_PARAM; goto out; @@ -1010,157 +802,114 @@ int lustre_lnet_config_route(char *nw, char *gw, int hops, int prio, rnet = libcfs_str2net(nw); if (rnet == LNET_NIDNET(LNET_NID_ANY)) { - snprintf(err_str, - sizeof(err_str), + snprintf(err_str, LNET_MAX_STR_LEN, "\"cannot parse remote net %s\"", nw); rc = LUSTRE_CFG_RC_BAD_PARAM; goto out; } - if (hops == -1) { - /* hops is undefined */ - hops = LNET_UNDEFINED_HOPS; - } else if (hops < 1 || hops > 255) { - snprintf(err_str, - sizeof(err_str), - "\"invalid hop count %d, must be between 1 and 255\"", - hops); - rc = LUSTRE_CFG_RC_OUT_OF_RANGE_PARAM; + replace_sep(nidstr, ',', ' '); + rc = lustre_lnet_parse_nidstr(nidstr, lnet_nidlist, + LNET_MAX_NIDS_PER_PEER, err_str); + if (rc < 0) goto out; - } - if (prio == -1) { - prio = 0; - } else if (prio < 0) { - snprintf(err_str, - sizeof(err_str), - "\"invalid priority %d, must be greater than 0\"", - prio); - rc = LUSTRE_CFG_RC_OUT_OF_RANGE_PARAM; - goto out; - } - - rc = lnet_expr2ips(gw, ip_list, - &ip2nets, &net, err_str); - if (rc == LUSTRE_CFG_RC_LAST_ELEM) - rc = -1; - else if (rc < LUSTRE_CFG_RC_NO_ERR) - goto out; + num_nids = rc; - ip_idx = rc; + for (idx = 0; idx < num_nids; idx++) { + LIBCFS_IOC_INIT_V2(data, cfg_hdr); + data.cfg_net = rnet; + if (cmd == LNETCTL_ADD_CMD) { + data.cfg_config_u.cfg_route.rtr_hop = hops; + data.cfg_config_u.cfg_route.rtr_priority = prio; + data.cfg_config_u.cfg_route.rtr_sensitivity = sen; + } - LIBCFS_IOC_INIT_V2(data, cfg_hdr); - data.cfg_net = rnet; - data.cfg_config_u.cfg_route.rtr_hop = hops; - data.cfg_config_u.cfg_route.rtr_priority = prio; + data.cfg_nid = lnet_nidlist[idx]; - for (i = MAX_NUM_IPS - 1; i > ip_idx; i--) { - gateway_nid = LNET_MKNID(net, ip_list[i]); - if (gateway_nid == LNET_NID_ANY) { - snprintf(err_str, - LNET_MAX_STR_LEN, - "\"cannot form gateway NID: %u\"", - ip_list[i]); - err_str[LNET_MAX_STR_LEN - 1] = '\0'; - rc = LUSTRE_CFG_RC_BAD_PARAM; - goto out; - } - data.cfg_nid = gateway_nid; + if (cmd == LNETCTL_ADD_CMD) + rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_ADD_ROUTE, + &data); + else + rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_DEL_ROUTE, + &data); - rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_ADD_ROUTE, &data); - if (rc != 0) { + if (rc != 0 && errno != EEXIST && + errno != EHOSTUNREACH) { rc = -errno; - snprintf(err_str, - sizeof(err_str), - "\"cannot add route: %s\"", strerror(errno)); + snprintf(err_str, LNET_MAX_STR_LEN, + "route operation failed: %s", + strerror(errno)); goto out; + } else if (errno == EEXIST) { + /* + * continue chugging along if one of the + * routes already exists + */ + rc = 0; } } + out: - cYAML_build_error(rc, seq_no, ADD_CMD, "route", err_str, err_rc); + cYAML_build_error(rc, seq_no, + cmd == LNETCTL_ADD_CMD ? ADD_CMD : DEL_CMD, "route", + err_str, err_rc); return rc; } -int lustre_lnet_del_route(char *nw, char *gw, - int seq_no, struct cYAML **err_rc) +int lustre_lnet_config_route(char *nw, char *nidstr, int hops, int prio, + int sen, int seq_no, struct cYAML **err_rc) { - struct lnet_ioctl_config_data data; - lnet_nid_t gateway_nid; - int rc = LUSTRE_CFG_RC_NO_ERR; - __u32 rnet = LNET_NIDNET(LNET_NID_ANY); - __u32 net = LNET_NIDNET(LNET_NID_ANY); - char err_str[LNET_MAX_STR_LEN]; - int ip_idx, i; - __u32 ip_list[MAX_NUM_IPS]; - struct lustre_lnet_ip2nets ip2nets; - - /* initialize all lists */ - INIT_LIST_HEAD(&ip2nets.ip2nets_ip_ranges); - INIT_LIST_HEAD(&ip2nets.ip2nets_net.network_on_rule); - INIT_LIST_HEAD(&ip2nets.ip2nets_net.nw_intflist); - - snprintf(err_str, sizeof(err_str), "\"Success\""); + int rc; + char err_str[LNET_MAX_STR_LEN] = "\"generic error\""; - if (nw == NULL || gw == NULL) { - snprintf(err_str, - sizeof(err_str), - "\"missing mandatory parameter in route delete: '%s'\"", - (nw == NULL && gw == NULL) ? "network, gateway" : - (nw == NULL) ? "network" : "gateway"); - rc = LUSTRE_CFG_RC_MISSING_PARAM; + if (hops == -1) { + hops = LNET_UNDEFINED_HOPS; + } else if (hops < 1 || hops > 255) { + snprintf(err_str, LNET_MAX_STR_LEN, + "\"invalid hop count %d, must be between 1 and 255\"", + hops); + rc = LUSTRE_CFG_RC_OUT_OF_RANGE_PARAM; goto out; } - rnet = libcfs_str2net(nw); - if (rnet == LNET_NIDNET(LNET_NID_ANY)) { - snprintf(err_str, - sizeof(err_str), - "\"cannot parse remote net '%s'\"", nw); - rc = LUSTRE_CFG_RC_BAD_PARAM; + if (prio == -1) { + prio = 0; + } else if (prio < 0) { + snprintf(err_str, LNET_MAX_STR_LEN, + "\"invalid priority %d, must be greater than 0\"", + prio); + rc = LUSTRE_CFG_RC_OUT_OF_RANGE_PARAM; goto out; } - rc = lnet_expr2ips(gw, ip_list, - &ip2nets, &net, err_str); - if (rc == LUSTRE_CFG_RC_LAST_ELEM) - rc = -1; - else if (rc < LUSTRE_CFG_RC_NO_ERR) + if (sen == -1) { + sen = 1; + } else if (sen < 1) { + snprintf(err_str, LNET_MAX_STR_LEN, + "\"invalid health sensitivity %d, must be 1 or greater\"", + sen); + rc = LUSTRE_CFG_RC_OUT_OF_RANGE_PARAM; goto out; - - ip_idx = rc; - - LIBCFS_IOC_INIT_V2(data, cfg_hdr); - data.cfg_net = rnet; - - for (i = MAX_NUM_IPS - 1; i > ip_idx; i--) { - gateway_nid = LNET_MKNID(net, ip_list[i]); - if (gateway_nid == LNET_NID_ANY) { - snprintf(err_str, - LNET_MAX_STR_LEN, - "\"cannot form gateway NID: %u\"", - ip_list[i]); - err_str[LNET_MAX_STR_LEN - 1] = '\0'; - rc = LUSTRE_CFG_RC_BAD_PARAM; - goto out; - } - data.cfg_nid = gateway_nid; - - rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_DEL_ROUTE, &data); - if (rc != 0) { - rc = -errno; - snprintf(err_str, - sizeof(err_str), - "\"cannot delete route: %s\"", strerror(errno)); - goto out; - } } + + rc = lustre_lnet_route_common(nw, nidstr, hops, prio, sen, seq_no, + err_rc, LNETCTL_ADD_CMD); + return rc; out: - cYAML_build_error(rc, seq_no, DEL_CMD, "route", err_str, err_rc); + cYAML_build_error(rc, seq_no, ADD_CMD, "route", err_str, err_rc); return rc; } +int lustre_lnet_del_route(char *nw, char *nidstr, int seq_no, + struct cYAML **err_rc) +{ + return lustre_lnet_route_common(nw, nidstr, 0, 0, 0, seq_no, err_rc, + LNETCTL_DEL_CMD); +} + int lustre_lnet_show_route(char *nw, char *gw, int hops, int prio, int detail, int seq_no, struct cYAML **show_rc, struct cYAML **err_rc, bool backup) @@ -1173,12 +922,9 @@ int lustre_lnet_show_route(char *nw, char *gw, int hops, int prio, int detail, int i; struct cYAML *root = NULL, *route = NULL, *item = NULL; struct cYAML *first_seq = NULL; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; bool exist = false; - snprintf(err_str, sizeof(err_str), - "\"out of memory\""); - if (nw != NULL) { net = libcfs_str2net(nw); if (net == LNET_NIDNET(LNET_NID_ANY)) { @@ -1226,6 +972,9 @@ int lustre_lnet_show_route(char *nw, char *gw, int hops, int prio, int detail, goto out; for (i = 0;; i++) { + __u32 rt_alive; + __u32 rt_multi_hop; + LIBCFS_IOC_INIT_V2(data, cfg_hdr); data.cfg_count = i; @@ -1283,12 +1032,27 @@ int lustre_lnet_show_route(char *nw, char *gw, int hops, int prio, int detail, cfg_route.rtr_priority) == NULL) goto out; + if (cYAML_create_number(item, "health_sensitivity", + data.cfg_config_u. + cfg_route.rtr_sensitivity) == NULL) + goto out; + + rt_alive = data.cfg_config_u.cfg_route.rtr_flags & + LNET_RT_ALIVE; + rt_multi_hop = data.cfg_config_u.cfg_route.rtr_flags & + LNET_RT_MULTI_HOP; + if (!backup && cYAML_create_string(item, "state", - data.cfg_config_u.cfg_route. - rtr_flags ? + rt_alive ? "up" : "down") == NULL) goto out; + + if (!backup && + cYAML_create_string(item, "type", + rt_multi_hop? + "multi-hop" : "single-hop") == NULL) + goto out; } } @@ -1397,12 +1161,13 @@ static int lustre_lnet_intf2nids(struct lnet_dlc_network_descr *nw, char val[LNET_MAX_STR_LEN]; __u32 ip; int gni_num; + char *endp; + unsigned int num; if (nw == NULL || nids == NULL) { snprintf(err_str, str_len, "\"unexpected parameters to lustre_lnet_intf2nids()\""); - err_str[str_len - 1] = '\0'; return LUSTRE_CFG_RC_BAD_PARAM; } @@ -1417,7 +1182,6 @@ static int lustre_lnet_intf2nids(struct lnet_dlc_network_descr *nw, if (*nids == NULL) { snprintf(err_str, str_len, "\"out of memory\""); - err_str[str_len - 1] = '\0'; return LUSTRE_CFG_RC_OUT_OF_MEM; } /* @@ -1431,7 +1195,6 @@ static int lustre_lnet_intf2nids(struct lnet_dlc_network_descr *nw, if (rc) { snprintf(err_str, str_len, "\"cannot read gni nid\""); - err_str[str_len - 1] = '\0'; goto failed; } gni_num = atoi(val); @@ -1443,15 +1206,30 @@ static int lustre_lnet_intf2nids(struct lnet_dlc_network_descr *nw, /* look at the other interfaces */ list_for_each_entry(intf, &nw->nw_intflist, intf_on_network) { - rc = lustre_lnet_queryip(intf, &ip); - if (rc != LUSTRE_CFG_RC_NO_ERR) { - snprintf(err_str, str_len, - "\"couldn't query intf %s\"", intf->intf_name); - err_str[str_len - 1] = '\0'; - goto failed; + if (LNET_NETTYP(nw->nw_id) == PTL4LND) { + /* handle LNDs with numeric interface name */ + num = strtoul(intf->intf_name, &endp, 0); + if (endp == intf->intf_name || *endp != '\0') { + rc = LUSTRE_CFG_RC_BAD_PARAM; + snprintf(err_str, str_len, + "\"couldn't query intf %s\"", + intf->intf_name); + goto failed; + } + (*nids)[i] = LNET_MKNID(nw->nw_id, num); + i++; + } else { + /* handle LNDs with ip interface name */ + rc = lustre_lnet_queryip(intf, &ip); + if (rc != LUSTRE_CFG_RC_NO_ERR) { + snprintf(err_str, str_len, + "\"couldn't query intf %s\"", + intf->intf_name); + goto failed; + } + (*nids)[i] = LNET_MKNID(nw->nw_id, ip); + i++; } - (*nids)[i] = LNET_MKNID(nw->nw_id, ip); - i++; } out: @@ -1634,7 +1412,6 @@ static int lustre_lnet_resolve_ip2nets_rule(struct lustre_lnet_ip2nets *ip2nets, if (rc < 0) { snprintf(err_str, str_len, "\"failed to get interface addresses: %d\"", -errno); - err_str[str_len - 1] = '\0'; return -errno; } @@ -1644,13 +1421,12 @@ static int lustre_lnet_resolve_ip2nets_rule(struct lustre_lnet_ip2nets *ip2nets, if (rc != LUSTRE_CFG_RC_MATCH) { snprintf(err_str, str_len, "\"couldn't match ip to existing interfaces\""); - err_str[str_len - 1] = '\0'; freeifaddrs(ifa); return rc; } rc = lustre_lnet_intf2nids(&ip2nets->ip2nets_net, nids, nnids, - err_str, sizeof(err_str)); + err_str, str_len); if (rc != LUSTRE_CFG_RC_NO_ERR) { *nids = NULL; *nnids = 0; @@ -1751,9 +1527,7 @@ lustre_lnet_config_ip2nets(struct lustre_lnet_ip2nets *ip2nets, lnet_nid_t *nids = NULL; __u32 nnids = 0; int rc; - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"success\""); + char err_str[LNET_MAX_STR_LEN] = "\"success\""; if (!ip2nets) { snprintf(err_str, @@ -1802,7 +1576,7 @@ int lustre_lnet_config_ni(struct lnet_dlc_network_descr *nw_descr, struct lnet_ioctl_config_lnd_tunables *tun = NULL; char buf[LNET_MAX_STR_LEN]; int rc = LUSTRE_CFG_RC_NO_ERR; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"success\""; lnet_nid_t *nids = NULL; __u32 nnids = 0; size_t len; @@ -1810,8 +1584,6 @@ int lustre_lnet_config_ni(struct lnet_dlc_network_descr *nw_descr, struct lnet_dlc_intf_descr *intf_descr, *tmp; __u32 *cpt_array; - snprintf(err_str, sizeof(err_str), "\"success\""); - if (ip2net == NULL && (nw_descr == NULL || nw_descr->nw_id == 0 || (list_empty(&nw_descr->nw_intflist) && LNET_NETTYP(nw_descr->nw_id) != GNILND))) { @@ -1949,13 +1721,11 @@ int lustre_lnet_del_ni(struct lnet_dlc_network_descr *nw_descr, { struct lnet_ioctl_config_ni data; int rc = LUSTRE_CFG_RC_NO_ERR, i; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"success\""; lnet_nid_t *nids = NULL; __u32 nnids = 0; struct lnet_dlc_intf_descr *intf_descr, *tmp; - snprintf(err_str, sizeof(err_str), "\"success\""); - if (nw_descr == NULL || nw_descr->nw_id == 0) { snprintf(err_str, sizeof(err_str), @@ -2035,9 +1805,7 @@ lustre_lnet_config_healthv(int value, bool all, lnet_nid_t nid, { 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\""); + char err_str[LNET_MAX_STR_LEN] = "\"success\""; LIBCFS_IOC_INIT_V2(data, rh_hdr); data.rh_type = type; @@ -2049,7 +1817,8 @@ lustre_lnet_config_healthv(int value, bool all, lnet_nid_t nid, if (rc != 0) { rc = -errno; snprintf(err_str, - sizeof(err_str), "Can not configure health value"); + sizeof(err_str), "Can not configure health value: %s", + strerror(errno)); } cYAML_build_error(rc, seq_no, ADD_CMD, name, err_str, err_rc); @@ -2146,13 +1915,11 @@ int lustre_lnet_show_net(char *nw, int detail, int seq_no, int str_buf_len = LNET_MAX_SHOW_NUM_CPT * 2; char str_buf[str_buf_len]; char *pos; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; bool exist = false, new_net = true; int net_num = 0; size_t buf_size = sizeof(*ni_data) + sizeof(*lnd) + sizeof(*stats); - snprintf(err_str, sizeof(err_str), "\"out of memory\""); - buf = calloc(1, buf_size); if (buf == NULL) goto out; @@ -2179,7 +1946,6 @@ int lustre_lnet_show_net(char *nw, int detail, int seq_no, goto out; for (i = 0;; i++) { - pos = str_buf; __u32 rc_net; memset(buf, 0, buf_size); @@ -2410,17 +2176,18 @@ continue_without_msg_stats: goto out; /* out put the CPTs in the format: "[x,x,x,...]" */ + pos = str_buf; limit = str_buf + str_buf_len - 3; - pos += snprintf(pos, limit - pos, "\"["); + pos += scnprintf(pos, limit - pos, "\"["); for (j = 0 ; ni_data->lic_ncpts >= 1 && j < ni_data->lic_ncpts && pos < limit; j++) { - pos += snprintf(pos, limit - pos, - "%d", ni_data->lic_cpts[j]); + pos += scnprintf(pos, limit - pos, + "%d", ni_data->lic_cpts[j]); if ((j + 1) < ni_data->lic_ncpts) - pos += snprintf(pos, limit - pos, ","); + pos += scnprintf(pos, limit - pos, ","); } - pos += snprintf(pos, 3, "]\""); + snprintf(pos, 3, "]\""); if (ni_data->lic_ncpts >= 1 && cYAML_create_string(item, "CPT", @@ -2477,9 +2244,7 @@ int lustre_lnet_enable_routing(int enable, int seq_no, struct cYAML **err_rc) { struct lnet_ioctl_config_data data; int rc = LUSTRE_CFG_RC_NO_ERR; - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"success\""); + char err_str[LNET_MAX_STR_LEN] = "\"success\""; LIBCFS_IOC_INIT_V2(data, cfg_hdr); data.cfg_config_u.cfg_buffers.buf_enable = (enable) ? 1 : 0; @@ -2507,9 +2272,7 @@ int ioctl_set_value(__u32 val, int ioc, char *name, { struct lnet_ioctl_set_value data; int rc = LUSTRE_CFG_RC_NO_ERR; - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"success\""); + char err_str[LNET_MAX_STR_LEN] = "\"success\""; LIBCFS_IOC_INIT_V2(data, sv_hdr); data.sv_value = val; @@ -2528,13 +2291,51 @@ int ioctl_set_value(__u32 val, int ioc, char *name, return rc; } -int lustre_lnet_config_hsensitivity(int sen, int seq_no, struct cYAML **err_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 err_str[LNET_MAX_STR_LEN] = "\"success\""; 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_rtr_sensitivity(int sen, int seq_no, struct cYAML **err_rc) +{ + int rc = LUSTRE_CFG_RC_NO_ERR; + char err_str[LNET_MAX_STR_LEN] = "\"success\""; + char val[LNET_MAX_STR_LEN]; + + snprintf(val, sizeof(val), "%d", sen); + + rc = write_sysfs_file(modparam_path, "router_sensitivity_percentage", val, + 1, strlen(val) + 1); + if (rc) + snprintf(err_str, sizeof(err_str), + "\"cannot configure router health sensitivity: %s\"", + strerror(errno)); + + cYAML_build_error(rc, seq_no, ADD_CMD, "router_sensitivity", 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; + char err_str[LNET_MAX_STR_LEN] = "\"success\""; + char val[LNET_MAX_STR_LEN]; snprintf(val, sizeof(val), "%d", sen); @@ -2553,11 +2354,9 @@ int lustre_lnet_config_hsensitivity(int sen, int seq_no, struct cYAML **err_rc) int lustre_lnet_config_transaction_to(int timeout, int seq_no, struct cYAML **err_rc) { int rc = LUSTRE_CFG_RC_NO_ERR; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"success\""; char val[LNET_MAX_STR_LEN]; - snprintf(err_str, sizeof(err_str), "\"success\""); - snprintf(val, sizeof(val), "%d", timeout); rc = write_sysfs_file(modparam_path, "lnet_transaction_timeout", val, @@ -2575,11 +2374,9 @@ int lustre_lnet_config_transaction_to(int timeout, int seq_no, struct cYAML **er int lustre_lnet_config_retry_count(int count, int seq_no, struct cYAML **err_rc) { int rc = LUSTRE_CFG_RC_NO_ERR; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"success\""; char val[LNET_MAX_STR_LEN]; - snprintf(err_str, sizeof(err_str), "\"success\""); - snprintf(val, sizeof(val), "%d", count); rc = write_sysfs_file(modparam_path, "lnet_retry_count", val, @@ -2594,13 +2391,41 @@ int lustre_lnet_config_retry_count(int count, int seq_no, struct cYAML **err_rc) return rc; } -int lustre_lnet_config_max_intf(int max, int seq_no, struct cYAML **err_rc) +int lustre_lnet_config_response_tracking(int val, 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]; + char val_str[LNET_MAX_STR_LEN]; - snprintf(err_str, sizeof(err_str), "\"success\""); + if (val < 0 || val > 3) { + rc = LUSTRE_CFG_RC_BAD_PARAM; + snprintf(err_str, sizeof(err_str), + "\"Valid values are: 0, 1, 2, or 3\""); + } else { + snprintf(err_str, sizeof(err_str), "\"success\""); + + snprintf(val_str, sizeof(val_str), "%d", val); + + rc = write_sysfs_file(modparam_path, "lnet_response_tracking", + val_str, 1, strlen(val_str) + 1); + if (rc) + snprintf(err_str, sizeof(err_str), + "\"cannot configure response tracking: %s\"", + strerror(errno)); + } + + cYAML_build_error(rc, seq_no, ADD_CMD, "response_tracking", err_str, + err_rc); + + return rc; +} + +int lustre_lnet_config_max_intf(int max, int seq_no, struct cYAML **err_rc) +{ + int rc = LUSTRE_CFG_RC_NO_ERR; + char err_str[LNET_MAX_STR_LEN] = "\"success\""; + char val[LNET_MAX_STR_LEN]; snprintf(val, sizeof(val), "%d", max); @@ -2619,11 +2444,9 @@ int lustre_lnet_config_max_intf(int max, int seq_no, struct cYAML **err_rc) int lustre_lnet_config_discovery(int enable, int seq_no, struct cYAML **err_rc) { int rc = LUSTRE_CFG_RC_NO_ERR; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"success\""; char val[LNET_MAX_STR_LEN]; - snprintf(err_str, sizeof(err_str), "\"success\""); - snprintf(val, sizeof(val), "%u", (enable) ? 0 : 1); rc = write_sysfs_file(modparam_path, "lnet_peer_discovery_disabled", val, @@ -2639,6 +2462,29 @@ int lustre_lnet_config_discovery(int enable, int seq_no, struct cYAML **err_rc) } +int lustre_lnet_config_drop_asym_route(int drop, int seq_no, + struct cYAML **err_rc) +{ + int rc = LUSTRE_CFG_RC_NO_ERR; + char err_str[LNET_MAX_STR_LEN] = "\"success\""; + char val[LNET_MAX_STR_LEN]; + + snprintf(val, sizeof(val), "%u", (drop) ? 1 : 0); + + rc = write_sysfs_file(modparam_path, "lnet_drop_asym_route", val, + 1, strlen(val) + 1); + if (rc) + snprintf(err_str, sizeof(err_str), + "\"cannot configure drop asym route: %s\"", + strerror(errno)); + + cYAML_build_error(rc, seq_no, ADD_CMD, "drop_asym_route", + err_str, err_rc); + + return rc; + +} + int lustre_lnet_config_numa_range(int range, int seq_no, struct cYAML **err_rc) { return ioctl_set_value(range, IOC_LIBCFS_SET_NUMA_RANGE, @@ -2650,9 +2496,7 @@ int lustre_lnet_config_buffers(int tiny, int small, int large, int seq_no, { struct lnet_ioctl_config_data data; int rc = LUSTRE_CFG_RC_NO_ERR; - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"success\""); + char err_str[LNET_MAX_STR_LEN] = "\"success\""; /* -1 indicates to ignore changes to this field */ if (tiny < -1 || small < -1 || large < -1) { @@ -2697,12 +2541,10 @@ int lustre_lnet_show_routing(int seq_no, struct cYAML **show_rc, *type_node = NULL, *item = NULL, *cpt = NULL, *first_seq = NULL, *buffers = NULL; int i, j; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; char node_name[LNET_MAX_STR_LEN]; bool exist = false; - snprintf(err_str, sizeof(err_str), "\"out of memory\""); - buf = calloc(1, sizeof(*data) + sizeof(*pool_cfg)); if (buf == NULL) goto out; @@ -2882,13 +2724,11 @@ int lustre_lnet_show_peer(char *knid, int detail, int seq_no, *first_seq = NULL, *peer_root = NULL, *tmp = NULL, *msg_statistics = NULL, *statistics = NULL, *yhstats; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; struct lnet_process_id *list = NULL; void *data = NULL; void *lpni_data; - - snprintf(err_str, sizeof(err_str), - "\"out of memory\""); + bool exist = false; /* create struct cYAML root object */ root = cYAML_create_object(NULL, NULL); @@ -2948,6 +2788,7 @@ int lustre_lnet_show_peer(char *knid, int detail, int seq_no, l_errno = ENOMEM; goto out; } + for (i = 0; i < count; i++) { for (;;) { memset(&peer_info, 0, sizeof(peer_info)); @@ -2979,6 +2820,7 @@ int lustre_lnet_show_peer(char *knid, int detail, int seq_no, goto out; } } + exist = true; peer = cYAML_create_seq_item(peer_root); if (peer == NULL) @@ -3150,7 +2992,7 @@ int lustre_lnet_show_peer(char *knid, int detail, int seq_no, 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; @@ -3190,12 +3032,9 @@ int lustre_lnet_list_peer(int seq_no, int i = 0; int l_errno = 0; struct cYAML *root = NULL, *list_root = NULL, *first_seq = NULL; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; struct lnet_process_id *list = NULL; - snprintf(err_str, sizeof(err_str), - "\"out of memory\""); - memset(&peer_info, 0, sizeof(peer_info)); /* create struct cYAML root object */ @@ -3307,7 +3146,7 @@ static void add_to_global(struct cYAML *show_rc, struct cYAML *node, } static int build_global_yaml_entry(char *err_str, int err_len, int seq_no, - char *name, __u32 value, + char *name, __u64 value, struct cYAML **show_rc, struct cYAML **err_rc, int err) { @@ -3359,9 +3198,7 @@ static int ioctl_show_global_values(int ioc, int seq_no, char *name, struct lnet_ioctl_set_value data; int rc; int l_errno = 0; - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"out of memory\""); + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; LIBCFS_IOC_INIT_V2(data, sv_hdr); @@ -3374,32 +3211,127 @@ static int ioctl_show_global_values(int ioc, int seq_no, char *name, name, strerror(l_errno)); } - return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, name, - data.sv_value, show_rc, err_rc, l_errno); + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, name, + 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] = "\"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) +{ + int rc = LUSTRE_CFG_RC_OUT_OF_MEM; + char val[LNET_MAX_STR_LEN]; + int sen = -1, l_errno = 0; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; + + rc = read_sysfs_file(modparam_path, "lnet_health_sensitivity", val, + 1, sizeof(val)); + if (rc) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"cannot get health sensitivity: %d\"", rc); + } else { + sen = atoi(val); + } + + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "health_sensitivity", sen, show_rc, + err_rc, l_errno); } -int lustre_lnet_show_hsensitivity(int seq_no, struct cYAML **show_rc, - struct cYAML **err_rc) +int lustre_lnet_show_rtr_sensitivity(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 sen = -1, l_errno = 0; - char err_str[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; - snprintf(err_str, sizeof(err_str), "\"out of memory\""); - - rc = read_sysfs_file(modparam_path, "lnet_health_sensitivity", val, + rc = read_sysfs_file(modparam_path, "router_sensitivity_percentage", val, 1, sizeof(val)); if (rc) { l_errno = -errno; snprintf(err_str, sizeof(err_str), - "\"cannot get health sensitivity: %d\"", rc); + "\"cannot get router sensitivity percentage: %d\"", rc); } else { sen = atoi(val); } return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, - "health_sensitivity", sen, show_rc, + "router_sensitivity", sen, show_rc, + err_rc, l_errno); +} + +int lustre_lnet_show_lnd_timeout(int seq_no, struct cYAML **show_rc, + struct cYAML **err_rc) +{ + char val[LNET_MAX_STR_LEN]; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; + int lnd_to = -1; + int l_errno = 0; + int rc; + int fd; + glob_t path; + + rc = cfs_get_param_paths(&path, "lnet_lnd_timeout"); + if (rc < 0) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"cannot get LND timeout: %d\"", rc); + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "lnd_timeout", lnd_to, show_rc, + err_rc, l_errno); + } + + fd = open(path.gl_pathv[0], O_RDONLY); + if (fd < 0) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"error opening %s\"", path.gl_pathv[0]); + goto failed; + } + + rc = read(fd, val, sizeof(val)); + if (rc < 0) + l_errno = -errno; + + close(fd); + + if (rc < 0) { + snprintf(err_str, sizeof(err_str), + "\"error reading %s\"", path.gl_pathv[0]); + goto failed; + } + + lnd_to = atoi(val); + +failed: + cfs_free_param_data(&path); + + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "lnd_timeout", lnd_to, show_rc, err_rc, l_errno); } @@ -3409,9 +3341,7 @@ int lustre_lnet_show_transaction_to(int seq_no, struct cYAML **show_rc, int rc = LUSTRE_CFG_RC_OUT_OF_MEM; char val[LNET_MAX_STR_LEN]; int tto = -1, l_errno = 0; - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"out of memory\""); + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; rc = read_sysfs_file(modparam_path, "lnet_transaction_timeout", val, 1, sizeof(val)); @@ -3434,9 +3364,7 @@ int lustre_lnet_show_retry_count(int seq_no, struct cYAML **show_rc, int rc = LUSTRE_CFG_RC_OUT_OF_MEM; char val[LNET_MAX_STR_LEN]; int retry_count = -1, l_errno = 0; - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"out of memory\""); + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; rc = read_sysfs_file(modparam_path, "lnet_retry_count", val, 1, sizeof(val)); @@ -3453,15 +3381,143 @@ int lustre_lnet_show_retry_count(int seq_no, struct cYAML **show_rc, err_rc, l_errno); } +int lustre_lnet_calc_service_id(__u64 *service_id) +{ + int rc = LUSTRE_CFG_RC_OUT_OF_MEM; + char val[LNET_MAX_STR_LEN]; + int service_port = -1, l_errno = 0; + + rc = read_sysfs_file(o2ib_modparam_path, "service", val, + 1, sizeof(val)); + if (rc) { + l_errno = errno; + fprintf(stderr, "error:\n msg: \"cannot get service port: %s (%d)\"\n", + strerror(l_errno), -l_errno); + return rc; + } else { + service_port = atoi(val); + } + + *service_id = htobe64(((__u64)RDMA_PS_TCP << 16) + service_port); + + return LUSTRE_CFG_RC_NO_ERR; +} + +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] = "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_response_tracking(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 rsp_tracking = -1, l_errno = 0; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; + + rc = read_sysfs_file(modparam_path, "lnet_response_tracking", val, + 1, sizeof(val)); + if (rc) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"cannot get lnet_response_tracking value: %d\"", rc); + } else { + rsp_tracking = atoi(val); + } + + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "response_tracking", rsp_tracking, + show_rc, err_rc, l_errno); +} + int lustre_lnet_show_max_intf(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 max_intf = -1, l_errno = 0; - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"out of memory\""); + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; rc = read_sysfs_file(modparam_path, "lnet_interfaces_max", val, 1, sizeof(val)); @@ -3484,9 +3540,7 @@ int lustre_lnet_show_discovery(int seq_no, struct cYAML **show_rc, int rc = LUSTRE_CFG_RC_OUT_OF_MEM; char val[LNET_MAX_STR_LEN]; int discovery = -1, l_errno = 0; - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"out of memory\""); + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; rc = read_sysfs_file(modparam_path, "lnet_peer_discovery_disabled", val, 1, sizeof(val)); @@ -3508,6 +3562,29 @@ int lustre_lnet_show_discovery(int seq_no, struct cYAML **show_rc, err_rc, l_errno); } +int lustre_lnet_show_drop_asym_route(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 drop_asym_route = -1, l_errno = 0; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; + + rc = read_sysfs_file(modparam_path, "lnet_drop_asym_route", val, + 1, sizeof(val)); + if (rc) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"cannot get drop asym route setting: %d\"", rc); + } else { + drop_asym_route = atoi(val); + } + + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "drop_asym_route", drop_asym_route, + show_rc, err_rc, l_errno); +} + int lustre_lnet_show_numa_range(int seq_no, struct cYAML **show_rc, struct cYAML **err_rc) { @@ -3519,17 +3596,16 @@ int lustre_lnet_show_stats(int seq_no, struct cYAML **show_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]; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; struct cYAML *root = NULL, *stats = NULL; - snprintf(err_str, sizeof(err_str), "\"out of memory\""); - 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), @@ -3541,111 +3617,113 @@ int lustre_lnet_show_stats(int seq_no, struct cYAML **show_rc, 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, "rst_alloc", - data.st_cntrs.rst_alloc) == NULL) + if (!cYAML_create_number(stats, "rst_alloc", + cntrs->lct_health.lch_rst_alloc)) goto out; - if (cYAML_create_number(stats, "errors", - data.st_cntrs.errors) == NULL) + if (!cYAML_create_number(stats, "errors", + cntrs->lct_common.lcc_errors)) goto out; - if (cYAML_create_number(stats, "send_count", - data.st_cntrs.send_count) == NULL) + if (!cYAML_create_number(stats, "send_count", + cntrs->lct_common.lcc_send_count)) goto out; - if (cYAML_create_number(stats, "resend_count", - data.st_cntrs.resend_count) == NULL) + if (!cYAML_create_number(stats, "resend_count", + cntrs->lct_health.lch_resend_count)) goto out; - if (cYAML_create_number(stats, "response_timeout_count", - data.st_cntrs.response_timeout_count) == NULL) + if (!cYAML_create_number(stats, "response_timeout_count", + cntrs->lct_health.lch_response_timeout_count)) goto out; - if (cYAML_create_number(stats, "local_interrupt_count", - data.st_cntrs.local_interrupt_count) == NULL) + if (!cYAML_create_number(stats, "local_interrupt_count", + cntrs->lct_health.lch_local_interrupt_count)) goto out; - if (cYAML_create_number(stats, "local_dropped_count", - data.st_cntrs.local_dropped_count) == NULL) + if (!cYAML_create_number(stats, "local_dropped_count", + cntrs->lct_health.lch_local_dropped_count)) goto out; - if (cYAML_create_number(stats, "local_aborted_count", - data.st_cntrs.local_aborted_count) == NULL) + if (!cYAML_create_number(stats, "local_aborted_count", + cntrs->lct_health.lch_local_aborted_count)) goto out; - if (cYAML_create_number(stats, "local_no_route_count", - data.st_cntrs.local_no_route_count) == NULL) + if (!cYAML_create_number(stats, "local_no_route_count", + cntrs->lct_health.lch_local_no_route_count)) goto out; - if (cYAML_create_number(stats, "local_timeout_count", - data.st_cntrs.local_timeout_count) == 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", - data.st_cntrs.local_error_count) == NULL) + 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", - data.st_cntrs.remote_dropped_count) == NULL) + 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", - data.st_cntrs.remote_error_count) == NULL) + 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", - data.st_cntrs.remote_timeout_count) == NULL) + 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", - data.st_cntrs.network_timeout_count) == NULL) + if (!cYAML_create_number(stats, "network_timeout_count", + cntrs->lct_health.lch_network_timeout_count)) goto out; - if (cYAML_create_number(stats, "recv_count", - data.st_cntrs.recv_count) == NULL) + if (!cYAML_create_number(stats, "recv_count", + cntrs->lct_common.lcc_recv_count)) goto out; - if (cYAML_create_number(stats, "route_count", - data.st_cntrs.route_count) == NULL) + if (!cYAML_create_number(stats, "route_count", + cntrs->lct_common.lcc_route_count)) goto out; - if (cYAML_create_number(stats, "drop_count", - data.st_cntrs.drop_count) == NULL) + if (!cYAML_create_number(stats, "drop_count", + cntrs->lct_common.lcc_drop_count)) goto out; - if (cYAML_create_number(stats, "send_length", - data.st_cntrs.send_length) == NULL) + if (!cYAML_create_number(stats, "send_length", + cntrs->lct_common.lcc_send_length)) goto out; - if (cYAML_create_number(stats, "recv_length", - data.st_cntrs.recv_length) == NULL) + if (!cYAML_create_number(stats, "recv_length", + cntrs->lct_common.lcc_recv_length)) goto out; - if (cYAML_create_number(stats, "route_length", - data.st_cntrs.route_length) == NULL) + if (!cYAML_create_number(stats, "route_length", + cntrs->lct_common.lcc_route_length)) goto out; - if (cYAML_create_number(stats, "drop_length", - data.st_cntrs.drop_length) == NULL) + if (!cYAML_create_number(stats, "drop_length", + cntrs->lct_common.lcc_drop_length)) goto out; - if (show_rc == NULL) + if (!show_rc) cYAML_print_tree(root); snprintf(err_str, sizeof(err_str), "\"success\""); @@ -3673,36 +3751,24 @@ typedef int (*cmd_handler_t)(struct cYAML *tree, static int handle_yaml_config_route(struct cYAML *tree, struct cYAML **show_rc, struct cYAML **err_rc) { - struct cYAML *net, *gw, *hop, *prio, *seq_no; + struct cYAML *net, *gw, *hop, *prio, *sen, *seq_no; net = cYAML_get_object_item(tree, "net"); gw = cYAML_get_object_item(tree, "gateway"); hop = cYAML_get_object_item(tree, "hop"); prio = cYAML_get_object_item(tree, "priority"); + sen = cYAML_get_object_item(tree, "health_sensitivity"); seq_no = cYAML_get_object_item(tree, "seq_no"); return lustre_lnet_config_route((net) ? net->cy_valuestring : NULL, (gw) ? gw->cy_valuestring : NULL, (hop) ? hop->cy_valueint : -1, (prio) ? prio->cy_valueint : -1, + (sen) ? sen->cy_valueint : -1, (seq_no) ? seq_no->cy_valueint : -1, err_rc); } -static void yaml_free_string_array(char **array, int num) -{ - int i; - char **sub_array = array; - - for (i = 0; i < num; i++) { - if (*sub_array != NULL) - free(*sub_array); - sub_array++; - } - if (array) - free(array); -} - /* * interfaces: * 0: ['['']'] @@ -4058,12 +4124,17 @@ static int handle_yaml_del_ni(struct cYAML *tree, struct cYAML **show_rc, return rc; } -static int yaml_copy_peer_nids(struct cYAML *nids_entry, char ***nidsppp, - char *prim_nid, bool del) +/* Create a nidstring parseable by the nidstrings library from the nid + * information encoded in the CYAML structure. + * NOTE: Caller must free memory allocated to nidstr + */ +static int yaml_nids2nidstr(struct cYAML *nids_entry, char **nidstr, + char *prim_nid, int cmd) { + int num_strs = 0, rc; + size_t buf_size, buf_pos, nidstr_len = 0; + char *buffer; struct cYAML *child = NULL, *entry = NULL; - char **nids = NULL; - int num = 0, rc = LUSTRE_CFG_RC_NO_ERR; if (cYAML_is_sequence(nids_entry)) { while (cYAML_get_next_seq_item(nids_entry, &child)) { @@ -4073,145 +4144,170 @@ static int yaml_copy_peer_nids(struct cYAML *nids_entry, char ***nidsppp, continue; if (prim_nid && - (strcmp(entry->cy_valuestring, prim_nid) - == 0) && del) { - /* - * primary nid is present in the list of - * nids so that means we want to delete - * the entire peer, so no need to go - * further. Just delete the entire peer. - */ - return 0; + (strcmp(entry->cy_valuestring, prim_nid) == 0)) { + if (cmd == LNETCTL_DEL_CMD) { + /* + * primary nid is present in the list of + * nids so that means we want to delete + * the entire peer, so no need to go + * further. Just delete the entire peer. + */ + return LUSTRE_CFG_RC_NO_ERR; + } else { + continue; + } } - num++; + /* + * + 1 for the space separating each string, and + * accounts for the terminating null char + */ + nidstr_len += strlen(entry->cy_valuestring) + 1; + num_strs++; } } - if (num == 0) + if (num_strs == 0 && !prim_nid) return LUSTRE_CFG_RC_MISSING_PARAM; + else if (num_strs == 0) /* Only the primary nid was given to add/del */ + return LUSTRE_CFG_RC_NO_ERR; - nids = calloc(sizeof(*nids), num); - if (!nids) + buffer = malloc(nidstr_len); + if (!buffer) return LUSTRE_CFG_RC_OUT_OF_MEM; /* now grab all the nids */ - num = 0; + rc = 0; + buf_pos = 0; + buf_size = nidstr_len; child = NULL; while (cYAML_get_next_seq_item(nids_entry, &child)) { entry = cYAML_get_object_item(child, "nid"); if (!entry || !entry->cy_valuestring) continue; - nids[num] = calloc(strlen(entry->cy_valuestring) + 1, 1); - if (!nids[num]) { - rc = LUSTRE_CFG_RC_OUT_OF_MEM; - goto failed; + if (prim_nid && + (strcmp(entry->cy_valuestring, prim_nid) == 0)) + continue; + + if (buf_pos) { + rc = snprintf(buffer + buf_pos, buf_size, " "); + buf_pos += (rc < buf_size) ? rc : buf_size; + buf_size = nidstr_len - buf_pos; } - strncpy(nids[num], entry->cy_valuestring, - strlen(entry->cy_valuestring)); - num++; + + rc = snprintf(buffer + buf_pos, buf_size, "%s", + entry->cy_valuestring); + buf_pos += (rc < buf_size) ? rc : buf_size; + buf_size = nidstr_len - buf_pos; } - rc = num; - *nidsppp = nids; - return rc; + *nidstr = buffer; -failed: - if (nids != NULL) - yaml_free_string_array(nids, num); - *nidsppp = NULL; - return rc; + return LUSTRE_CFG_RC_NO_ERR; } -static int handle_yaml_config_peer(struct cYAML *tree, struct cYAML **show_rc, - struct cYAML **err_rc) +static int handle_yaml_peer_common(struct cYAML *tree, struct cYAML **show_rc, + struct cYAML **err_rc, int cmd) { - char **nids = NULL; - int num, rc; - struct cYAML *seq_no, *prim_nid, *non_mr, *ip2nets, *peer_nis; + int rc, num_nids = 0, seqn; + bool mr_value = false; + char *nidstr = NULL, *prim_nidstr; char err_str[LNET_MAX_STR_LEN]; + struct cYAML *seq_no, *prim_nid, *mr, *peer_nis; + lnet_nid_t lnet_nidlist[LNET_MAX_NIDS_PER_PEER]; + lnet_nid_t pnid = LNET_NID_ANY; seq_no = cYAML_get_object_item(tree, "seq_no"); + seqn = seq_no ? seq_no->cy_valueint : -1; + prim_nid = cYAML_get_object_item(tree, "primary nid"); - non_mr = cYAML_get_object_item(tree, "non_mr"); - ip2nets = cYAML_get_object_item(tree, "ip2nets"); peer_nis = cYAML_get_object_item(tree, "peer ni"); + if (!prim_nid) { + rc = LUSTRE_CFG_RC_BAD_PARAM; + snprintf(err_str, LNET_MAX_STR_LEN, + "\"primary nid\" must be specified"); + goto failed; + } + + prim_nidstr = prim_nid->cy_valuestring; - if (ip2nets && (prim_nid || peer_nis)) { + /* if the provided primary NID is bad, no need to go any further */ + pnid = libcfs_str2nid(prim_nidstr); + if (pnid == LNET_NID_ANY) { rc = LUSTRE_CFG_RC_BAD_PARAM; - snprintf(err_str, sizeof(err_str), - "ip2nets can not be specified along side prim_nid" - " or peer ni fields"); - cYAML_build_error(rc, (seq_no) ? seq_no->cy_valueint : -1, - ADD_CMD, "peer", err_str, err_rc); - return rc; + snprintf(err_str, LNET_MAX_STR_LEN, + "badly formatted primary NID: %s", prim_nidstr); + goto failed; } - num = yaml_copy_peer_nids((ip2nets) ? ip2nets : peer_nis, &nids, - (prim_nid) ? prim_nid->cy_valuestring : NULL, - false); + rc = yaml_nids2nidstr(peer_nis, &nidstr, prim_nidstr, cmd); + if (rc == LUSTRE_CFG_RC_MISSING_PARAM) { + snprintf(err_str, LNET_MAX_STR_LEN, + "No nids defined in YAML block"); + goto failed; + } else if (rc == LUSTRE_CFG_RC_OUT_OF_MEM) { + snprintf(err_str, LNET_MAX_STR_LEN, "out of memory"); + goto failed; + } else if (rc != LUSTRE_CFG_RC_NO_ERR) { + snprintf(err_str, LNET_MAX_STR_LEN, + "Unrecognized error %d", rc); + goto failed; + } - if (num < 0) { - snprintf(err_str, sizeof(err_str), - "error copying nids from YAML block"); - cYAML_build_error(num, (seq_no) ? seq_no->cy_valueint : -1, - ADD_CMD, "peer", err_str, err_rc); - return num; + num_nids = 0; + if (nidstr) { + num_nids = lustre_lnet_parse_nidstr(nidstr, lnet_nidlist, + LNET_MAX_NIDS_PER_PEER, + err_str); + if (num_nids < 0) { + rc = num_nids; + goto failed; + } + } + + if (cmd == LNETCTL_ADD_CMD) { + mr = cYAML_get_object_item(tree, "Multi-Rail"); + mr_value = true; + if (mr && mr->cy_valuestring) { + if (strcmp(mr->cy_valuestring, "False") == 0) + mr_value = false; + else if (strcmp(mr->cy_valuestring, "True") != 0) { + rc = LUSTRE_CFG_RC_BAD_PARAM; + snprintf(err_str, LNET_MAX_STR_LEN, + "Multi-Rail must be set to \"True\" or \"False\" found \"%s\"", + mr->cy_valuestring); + goto failed; + } + } } - rc = lustre_lnet_config_peer_nid((prim_nid) ? prim_nid->cy_valuestring : NULL, - nids, num, - (non_mr) ? false : true, - (ip2nets) ? true : false, - (seq_no) ? seq_no->cy_valueint : -1, - err_rc); + rc = lustre_lnet_mod_peer_nidlist(pnid, lnet_nidlist, cmd, + num_nids, mr_value, seqn, + err_rc); + +failed: + if (nidstr) + free(nidstr); + + if (rc != LUSTRE_CFG_RC_NO_ERR) + cYAML_build_error(rc, seqn, "peer", + cmd == LNETCTL_ADD_CMD ? ADD_CMD : DEL_CMD, + err_str, err_rc); - yaml_free_string_array(nids, num); return rc; } +static int handle_yaml_config_peer(struct cYAML *tree, struct cYAML **show_rc, + struct cYAML **err_rc) +{ + return handle_yaml_peer_common(tree, show_rc, err_rc, LNETCTL_ADD_CMD); +} + static int handle_yaml_del_peer(struct cYAML *tree, struct cYAML **show_rc, struct cYAML **err_rc) { - char **nids = NULL; - int num, rc; - struct cYAML *seq_no, *prim_nid, *ip2nets, *peer_nis; - char err_str[LNET_MAX_STR_LEN]; - - seq_no = cYAML_get_object_item(tree, "seq_no"); - prim_nid = cYAML_get_object_item(tree, "primary nid"); - ip2nets = cYAML_get_object_item(tree, "ip2nets"); - peer_nis = cYAML_get_object_item(tree, "peer ni"); - - if (ip2nets && (prim_nid || peer_nis)) { - rc = LUSTRE_CFG_RC_BAD_PARAM; - snprintf(err_str, sizeof(err_str), - "ip2nets can not be specified along side prim_nid" - " or peer ni fields"); - cYAML_build_error(rc, (seq_no) ? seq_no->cy_valueint : -1, - DEL_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, - true); - if (num < 0) { - snprintf(err_str, sizeof(err_str), - "error copying nids from YAML block"); - cYAML_build_error(num, (seq_no) ? seq_no->cy_valueint : -1, - ADD_CMD, "peer", err_str, err_rc); - return num; - } - - rc = lustre_lnet_del_peer_nid((prim_nid) ? prim_nid->cy_valuestring : NULL, - nids, num, (ip2nets) ? true : false, - (seq_no) ? seq_no->cy_valueint : -1, - err_rc); - - yaml_free_string_array(nids, num); - return rc; + return handle_yaml_peer_common(tree, show_rc, err_rc, LNETCTL_DEL_CMD); } static int handle_yaml_config_buffers(struct cYAML *tree, @@ -4402,7 +4498,7 @@ static int handle_yaml_config_global_settings(struct cYAML *tree, struct cYAML **err_rc) { struct cYAML *max_intf, *numa, *discovery, *retry, *tto, *seq_no, - *sen; + *sen, *recov, *rsen, *drop_asym_route, *rsp_tracking; int rc = 0; seq_no = cYAML_get_object_item(tree, "seq_no"); @@ -4427,6 +4523,13 @@ static int handle_yaml_config_global_settings(struct cYAML *tree, : -1, err_rc); + drop_asym_route = cYAML_get_object_item(tree, "drop_asym_route"); + if (drop_asym_route) + rc = lustre_lnet_config_drop_asym_route( + drop_asym_route->cy_valueint, + seq_no ? seq_no->cy_valueint : -1, + err_rc); + retry = cYAML_get_object_item(tree, "retry_count"); if (retry) rc = lustre_lnet_config_retry_count(retry->cy_valueint, @@ -4448,6 +4551,27 @@ static int handle_yaml_config_global_settings(struct cYAML *tree, : -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); + + rsen = cYAML_get_object_item(tree, "router_sensitivity"); + if (rsen) + rc = lustre_lnet_config_rtr_sensitivity(rsen->cy_valueint, + seq_no ? seq_no->cy_valueint + : -1, + err_rc); + + rsp_tracking = cYAML_get_object_item(tree, "response_tracking"); + if (rsp_tracking) + rc = lustre_lnet_config_response_tracking(rsp_tracking->cy_valueint, + seq_no ? seq_no->cy_valueint + : -1, + err_rc); + return rc; } @@ -4455,7 +4579,7 @@ static int handle_yaml_del_global_settings(struct cYAML *tree, struct cYAML **show_rc, struct cYAML **err_rc) { - struct cYAML *max_intf, *numa, *discovery, *seq_no; + struct cYAML *max_intf, *numa, *discovery, *seq_no, *drop_asym_route; int rc = 0; seq_no = cYAML_get_object_item(tree, "seq_no"); @@ -4481,6 +4605,12 @@ static int handle_yaml_del_global_settings(struct cYAML *tree, : -1, err_rc); + /* asymmetrical route messages are accepted by default */ + drop_asym_route = cYAML_get_object_item(tree, "drop_asym_route"); + if (drop_asym_route) + rc = lustre_lnet_config_drop_asym_route( + 0, seq_no ? seq_no->cy_valueint : -1, err_rc); + return rc; } @@ -4489,7 +4619,7 @@ static int handle_yaml_show_global_settings(struct cYAML *tree, struct cYAML **err_rc) { struct cYAML *max_intf, *numa, *discovery, *retry, *tto, *seq_no, - *sen; + *sen, *recov, *rsen, *drop_asym_route, *rsp_tracking; int rc = 0; seq_no = cYAML_get_object_item(tree, "seq_no"); @@ -4511,6 +4641,12 @@ static int handle_yaml_show_global_settings(struct cYAML *tree, : -1, show_rc, err_rc); + drop_asym_route = cYAML_get_object_item(tree, "drop_asym_route"); + if (drop_asym_route) + rc = lustre_lnet_show_drop_asym_route( + seq_no ? seq_no->cy_valueint : -1, + show_rc, err_rc); + retry = cYAML_get_object_item(tree, "retry_count"); if (retry) rc = lustre_lnet_show_retry_count(seq_no ? seq_no->cy_valueint @@ -4529,6 +4665,25 @@ static int handle_yaml_show_global_settings(struct cYAML *tree, : -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); + + rsen = cYAML_get_object_item(tree, "router_sensitivity"); + if (rsen) + rc = lustre_lnet_show_hsensitivity(seq_no ? seq_no->cy_valueint + : -1, + show_rc, err_rc); + + rsp_tracking = cYAML_get_object_item(tree, "response_tracking"); + if (rsp_tracking) + rc = lustre_lnet_show_response_tracking(seq_no ? + seq_no->cy_valueint : + -1, + show_rc, err_rc); + return rc; }