X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lnet%2Futils%2Flnetconfig%2Fliblnetconfig.c;h=a5a3eebb6f85e27b291311e5e143488a9d637ded;hb=3e5c6620fd0b0511498d14d38e8610d08f6da7b3;hp=885ef06d3ffa46f3a81904f2e108fc62a65ecc39;hpb=756abb9cf00b936b3d9149f50925b0e02ca0e4fd;p=fs%2Flustre-release.git diff --git a/lnet/utils/lnetconfig/liblnetconfig.c b/lnet/utils/lnetconfig/liblnetconfig.c index 885ef06..a5a3eeb 100644 --- a/lnet/utils/lnetconfig/liblnetconfig.c +++ b/lnet/utils/lnetconfig/liblnetconfig.c @@ -18,7 +18,7 @@ * * LGPL HEADER END * - * Copyright (c) 2014, 2016, Intel Corporation. + * Copyright (c) 2014, 2017, Intel Corporation. * * Author: * Amir Shehata @@ -47,9 +47,13 @@ #include #include "liblnd.h" #include +#include +#include #include +#include #include "liblnetconfig.h" -#include "cyaml.h" +#include +#include #define CONFIG_CMD "configure" #define UNCONFIG_CMD "unconfigure" @@ -57,6 +61,20 @@ #define DEL_CMD "del" #define SHOW_CMD "show" #define DBG_CMD "dbg" +#define MANAGE_CMD "manage" + +#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"}; /* * lustre_lnet_ip_range_descr @@ -77,6 +95,62 @@ struct lustre_lnet_ip2nets { struct list_head ip2nets_ip_ranges; }; +int open_sysfs_file(const char *path, const char *attr, const int mode) +{ + int fd; + char filename[LNET_MAX_STR_LEN]; + + if (strlen(path) + strlen(attr) >= LNET_MAX_STR_LEN) + return -1; + + snprintf(filename, sizeof(filename), "%s%s", + path, attr); + + fd = open(filename, mode); + + return fd; +} + +static int read_sysfs_file(const char *path, const char *attr, + void *val, const size_t size, const int nelem) +{ + int fd; + int rc = LUSTRE_CFG_RC_GENERIC_ERR; + + fd = open_sysfs_file(path, attr, O_RDONLY); + if (fd == -1) + return LUSTRE_CFG_RC_NO_MATCH; + + if (read(fd, val, size * nelem) == -1) + goto close_fd; + + rc = LUSTRE_CFG_RC_NO_ERR; + +close_fd: + close(fd); + return rc; +} + +static int write_sysfs_file(const char *path, const char *attr, + void *val, const size_t size, const int nelem) +{ + int fd; + int rc = LUSTRE_CFG_RC_GENERIC_ERR; + + fd = open_sysfs_file(path, attr, O_WRONLY | O_TRUNC); + if (fd == -1) + return LUSTRE_CFG_RC_NO_MATCH; + + if (write(fd, val, size * nelem) == -1) + goto close_fd; + + rc = LUSTRE_CFG_RC_NO_ERR; + +close_fd: + close(fd); + return rc; +} + /* * free_intf_descr * frees the memory allocated for an intf descriptor. @@ -176,93 +250,66 @@ int lustre_lnet_add_intf_descr(struct list_head *list, char *intf, int len) void lustre_lnet_init_nw_descr(struct lnet_dlc_network_descr *nw_descr) { if (nw_descr != NULL) { + nw_descr->nw_id = 0; INIT_LIST_HEAD(&nw_descr->network_on_rule); INIT_LIST_HEAD(&nw_descr->nw_intflist); } } -int lustre_lnet_parse_nids(char *nids, char **array, int size, - char ***out_array) +int lustre_lnet_parse_nidstr(char *nidstr, lnet_nid_t *lnet_nidlist, + int max_nids, char *err_str) { - 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; + int rc, num_nids = 0; + struct list_head nidlist; - /* count the number or new nids, by counting the number of commas */ - while (comma) { - comma = strchr(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 = strchr(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; } /* @@ -332,8 +379,7 @@ failed: int lustre_lnet_config_lib_init(void) { - return register_ioc_dev(LNET_DEV_ID, LNET_DEV_PATH, - LNET_DEV_MAJOR, LNET_DEV_MINOR); + return register_ioc_dev(LNET_DEV_ID, LNET_DEV_PATH); } void lustre_lnet_config_lib_uninit(void) @@ -347,9 +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); @@ -374,232 +418,458 @@ int lustre_lnet_config_ni_system(bool up, bool load_ni_from_mod, return rc; } -static lnet_nid_t *allocate_create_nid_array(char **nids, __u32 num_nids, - char *err_str) +static int dispatch_peer_ni_cmd(__u32 cmd, struct lnet_ioctl_peer_cfg *data, + char *err_str, char *cmd_str) { - lnet_nid_t *array = NULL; - __u32 i; + int rc; - if (!nids || num_nids == 0) { - snprintf(err_str, LNET_MAX_STR_LEN, "no NIDs to add"); - return NULL; + rc = l_ioctl(LNET_DEV_ID, cmd, data); + if (rc) { + rc = -errno; + snprintf(err_str, LNET_MAX_STR_LEN, + "\"%s peer ni operation failed: %s\"", + cmd_str, strerror(errno)); } - array = calloc(sizeof(*array) * num_nids, 1); - if (array == NULL) { - snprintf(err_str, LNET_MAX_STR_LEN, "out of memory"); - return NULL; - } + return rc; +} - for (i = 0; i < num_nids; i++) { - array[i] = libcfs_str2nid(nids[i]); - if (array[i] == LNET_NID_ANY) { - free(array); - snprintf(err_str, LNET_MAX_STR_LEN, - "bad NID: '%s'", - nids[i]); - return NULL; +static int infra_ping_nid(char *ping_nids, char *oper, int param, int ioc_call, + int seq_no, struct cYAML **show_rc, + struct cYAML **err_rc) +{ + void *data = NULL; + struct lnet_ioctl_ping_data ping; + struct cYAML *root = NULL, *ping_node = NULL, *item = NULL, + *first_seq = NULL, *tmp = NULL, *peer_ni = NULL; + struct lnet_process_id id; + char err_str[LNET_MAX_STR_LEN] = {0}; + char *sep, *token, *end; + char buf[6]; + size_t len; + int rc = LUSTRE_CFG_RC_OUT_OF_MEM; + int i; + bool flag = false; + + len = (sizeof(struct lnet_process_id) * LNET_INTERFACES_MAX_DEFAULT); + + data = calloc(1, len); + if (data == NULL) + goto out; + + /* create struct cYAML root object */ + root = cYAML_create_object(NULL, NULL); + if (root == NULL) + goto out; + + ping_node = cYAML_create_seq(root, oper); + if (ping_node == NULL) + goto out; + + /* tokenise each nid in string ping_nids */ + token = strtok(ping_nids, ","); + + do { + item = cYAML_create_seq_item(ping_node); + if (item == NULL) + goto out; + + if (first_seq == NULL) + first_seq = item; + + /* check if '-' is a part of NID, token */ + sep = strchr(token, '-'); + if (sep == NULL) { + id.pid = LNET_PID_ANY; + /* if no net is specified, libcfs_str2nid() will assume tcp */ + id.nid = libcfs_str2nid(token); + if (id.nid == LNET_NID_ANY) { + snprintf(err_str, sizeof(err_str), + "\"cannot parse NID '%s'\"", + token); + rc = LUSTRE_CFG_RC_BAD_PARAM; + cYAML_build_error(rc, seq_no, MANAGE_CMD, + oper, err_str, err_rc); + continue; + } + } else { + if (token[0] == 'u' || token[0] == 'U') + id.pid = (strtoul(&token[1], &end, 0) | + (LNET_PID_USERFLAG)); + else + id.pid = strtoul(token, &end, 0); + + /* assuming '-' is part of hostname */ + if (end != sep) { + id.pid = LNET_PID_ANY; + id.nid = libcfs_str2nid(token); + if (id.nid == LNET_NID_ANY) { + snprintf(err_str, sizeof(err_str), + "\"cannot parse NID '%s'\"", + token); + rc = LUSTRE_CFG_RC_BAD_PARAM; + cYAML_build_error(rc, seq_no, MANAGE_CMD, + oper, err_str, + err_rc); + continue; + } + } else { + id.nid = libcfs_str2nid(sep + 1); + if (id.nid == LNET_NID_ANY) { + snprintf(err_str, sizeof(err_str), + "\"cannot parse NID '%s'\"", + token); + rc = LUSTRE_CFG_RC_BAD_PARAM; + cYAML_build_error(rc, seq_no, MANAGE_CMD, + oper, err_str, + err_rc); + continue; + } + } + } + LIBCFS_IOC_INIT_V2(ping, ping_hdr); + ping.ping_hdr.ioc_len = sizeof(ping); + ping.ping_id = id; + ping.op_param = param; + ping.ping_count = LNET_INTERFACES_MAX_DEFAULT; + ping.ping_buf = data; + + rc = l_ioctl(LNET_DEV_ID, ioc_call, &ping); + if (rc != 0) { + snprintf(err_str, + sizeof(err_str), "failed to %s %s: %s\n", oper, + id.pid == LNET_PID_ANY ? + libcfs_nid2str(id.nid) : + libcfs_id2str(id), strerror(errno)); + rc = LUSTRE_CFG_RC_BAD_PARAM; + cYAML_build_error(rc, seq_no, MANAGE_CMD, + oper, err_str, err_rc); + continue; + } + + if (cYAML_create_string(item, "primary nid", + libcfs_nid2str(ping.ping_id.nid)) == NULL) + goto out; + + if (cYAML_create_string(item, "Multi-Rail", ping.mr_info ? + "True" : "False") == NULL) + goto out; + + tmp = cYAML_create_seq(item, "peer ni"); + if (tmp == NULL) + goto out; + + for (i = 0; i < ping.ping_count; i++) { + if (ping.ping_buf[i].nid == LNET_NID_LO_0) + continue; + peer_ni = cYAML_create_seq_item(tmp); + if (peer_ni == NULL) + goto out; + memset(buf, 0, sizeof buf); + snprintf(buf, sizeof buf, "nid"); + if (cYAML_create_string(peer_ni, buf, + libcfs_nid2str(ping.ping_buf[i].nid)) == NULL) + goto out; + } + + flag = true; + + } while ((token = strtok(NULL, ",")) != NULL); + + if (flag) + rc = LUSTRE_CFG_RC_NO_ERR; + +out: + if (data) + free(data); + 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; + show_node = cYAML_get_object_item(*show_rc, oper); + if (show_node != NULL && cYAML_is_sequence(show_node)) { + cYAML_insert_child(show_node, first_seq); + free(ping_node); + free(root); + } else if (show_node == NULL) { + cYAML_insert_sibling((*show_rc)->cy_child, + ping_node); + free(root); + } else { + cYAML_free_tree(root); } + } else { + *show_rc = root; } - return array; + return rc; } -static int dispatch_peer_ni_cmd(lnet_nid_t pnid, lnet_nid_t nid, __u32 cmd, - struct lnet_ioctl_peer_cfg *data, - char *err_str, char *cmd_str) +int lustre_lnet_ping_nid(char *ping_nids, int timeout, int seq_no, + struct cYAML **show_rc, struct cYAML **err_rc) { int rc; - data->prcfg_prim_nid = pnid; - data->prcfg_cfg_nid = nid; + rc = infra_ping_nid(ping_nids, "ping", timeout, IOC_LIBCFS_PING_PEER, + seq_no, show_rc, err_rc); + return rc; +} - rc = l_ioctl(LNET_DEV_ID, cmd, data); - if (rc != 0) { - rc = -errno; - snprintf(err_str, - LNET_MAX_STR_LEN, - "\"cannot %s peer ni: %s\"", - (cmd_str) ? cmd_str : "add", strerror(errno)); - } +int lustre_lnet_discover_nid(char *ping_nids, int force, int seq_no, + struct cYAML **show_rc, struct cYAML **err_rc) +{ + int rc; + rc = infra_ping_nid(ping_nids, "discover", force, IOC_LIBCFS_DISCOVER, + seq_no, show_rc, err_rc); return rc; } -int lustre_lnet_config_peer_nid(char *pnid, char **nid, int num_nids, - bool mr, int seq_no, struct cYAML **err_rc) +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 lnet_ioctl_peer_cfg data; - lnet_nid_t prim_nid = LNET_NID_ANY; - int rc = LUSTRE_CFG_RC_NO_ERR; - int idx = 0; - bool nid0_used = false; - char err_str[LNET_MAX_STR_LEN] = {0}; - lnet_nid_t *nids = allocate_create_nid_array(nid, num_nids, err_str); + int rc, nid_idx; - if (pnid) { - prim_nid = libcfs_str2nid(pnid); - if (prim_nid == LNET_NID_ANY) { - snprintf(err_str, sizeof(err_str), - "bad key NID: '%s'", - pnid); - rc = LUSTRE_CFG_RC_MISSING_PARAM; - goto out; - } - } else if (!nids || nids[0] == LNET_NID_ANY) { - snprintf(err_str, sizeof(err_str), - "no NIDs provided for configuration"); - rc = LUSTRE_CFG_RC_MISSING_PARAM; - goto out; - } else { - prim_nid = LNET_NID_ANY; - } + 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; - snprintf(err_str, sizeof(err_str), "\"Success\""); + rc = dispatch_peer_ni_cmd(cmd, &data, err_str, cmd_type); - LIBCFS_IOC_INIT_V2(data, prcfg_hdr); - data.prcfg_mr = mr; + if (rc) + return rc; + } - /* - * if prim_nid is not specified use the first nid in the list of - * nids provided as the prim_nid. NOTE: on entering 'if' we must - * have at least 1 NID + /* Add or delete any specified NIs associated with the specified + * (or implied) primary nid */ - if (prim_nid == LNET_NID_ANY) { - nid0_used = true; - prim_nid = nids[0]; - } + 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]; - /* Create the prim_nid first */ - rc = dispatch_peer_ni_cmd(prim_nid, LNET_NID_ANY, - IOC_LIBCFS_ADD_PEER_NI, - &data, err_str, "add"); + rc = dispatch_peer_ni_cmd(cmd, &data, err_str, cmd_type); - if (rc != 0) - goto out; + if (rc) + return rc; + } - /* add the rest of the nids to the key nid if any are available */ - for (idx = nid0_used ? 1 : 0 ; nids && idx < num_nids; idx++) { - /* - * 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 + 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) */ - rc = dispatch_peer_ni_cmd(prim_nid, nids[idx], - IOC_LIBCFS_ADD_PEER_NI, &data, - err_str, "add"); + LIBCFS_IOC_INIT_V2(data, prcfg_hdr); + data.prcfg_prim_nid = nidlist[0]; + data.prcfg_cfg_nid = LNET_NID_ANY; - if (rc != 0) - goto out; + rc = dispatch_peer_ni_cmd(cmd, &data, err_str, cmd_type); } -out: - if (nids != NULL) - free(nids); - cYAML_build_error(rc, seq_no, ADD_CMD, "peer_ni", err_str, err_rc); return rc; } -int lustre_lnet_del_peer_nid(char *pnid, char **nid, int num_nids, - 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) { - struct lnet_ioctl_peer_cfg data; - lnet_nid_t prim_nid; int rc = LUSTRE_CFG_RC_NO_ERR; - int idx = 0; - char err_str[LNET_MAX_STR_LEN] = {0}; - lnet_nid_t *nids = allocate_create_nid_array(nid, num_nids, err_str); - - if (pnid == NULL) { - snprintf(err_str, sizeof(err_str), - "\"Primary nid is not provided\""); - rc = LUSTRE_CFG_RC_MISSING_PARAM; + 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; + + 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; - } else { - prim_nid = libcfs_str2nid(pnid); - if (prim_nid == LNET_NID_ANY) { - rc = LUSTRE_CFG_RC_BAD_PARAM; - snprintf(err_str, sizeof(err_str), - "bad key NID: '%s'", - pnid); - goto out; - } } + lnet_nidlist2[0] = pnid; + memcpy(&lnet_nidlist2[1], lnet_nidlist, sizeof(*lnet_nidlist) * + (num_nids - 1)); + + rc = lustre_lnet_handle_peer_nidlist(lnet_nidlist2, + num_nids, is_mr, ioc_cmd, + cmd_str, err_str); +out: + if (lnet_nidlist2) + free(lnet_nidlist2); + + cYAML_build_error(rc, seq_no, cmd_str, "peer_ni", err_str, err_rc); + return rc; +} + +static void +replace_sep(char *str, char sep, char newsep) +{ + 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; - LIBCFS_IOC_INIT_V2(data, prcfg_hdr); - if (!nids || nids[0] == LNET_NID_ANY) { - rc = dispatch_peer_ni_cmd(prim_nid, LNET_NID_ANY, - IOC_LIBCFS_DEL_PEER_NI, - &data, err_str, "del"); + if (!prim_nid) { + rc = LUSTRE_CFG_RC_BAD_PARAM; + snprintf(err_str, LNET_MAX_STR_LEN, + "--prim_nid must be specified"); goto out; } - for (idx = 0; nids && idx < num_nids; idx++) { - rc = dispatch_peer_ni_cmd(prim_nid, nids[idx], - IOC_LIBCFS_DEL_PEER_NI, &data, - err_str, "del"); + 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; + } - if (rc != 0) + 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; } + rc = lustre_lnet_mod_peer_nidlist(pnid, lnet_nidlist, + cmd, num_nids, is_mr, + -1, err_rc); + out: - if (nids != NULL) - free(nids); - cYAML_build_error(rc, seq_no, DEL_CMD, "peer_ni", err_str, err_rc); + if (rc != LUSTRE_CFG_RC_NO_ERR) + cYAML_build_error(rc, -1, "peer", + cmd == LNETCTL_ADD_CMD ? "add" : "del", + 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; - __u32 net = LNET_NIDNET(LNET_NID_ANY); - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"Success\""); + lnet_nid_t lnet_nidlist[LNET_MAX_NIDS_PER_PEER]; - if (nw == NULL || gw == NULL) { - snprintf(err_str, - sizeof(err_str), - "\"missing mandatory parameter(s): '%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; } - net = libcfs_str2net(nw); - if (net == LNET_NIDNET(LNET_NID_ANY)) { - snprintf(err_str, - sizeof(err_str), - "\"cannot parse net %s\"", nw); + rnet = libcfs_str2net(nw); + if (rnet == LNET_NET_ANY) { + snprintf(err_str, LNET_MAX_STR_LEN, + "\"cannot parse remote net %s\"", nw); rc = LUSTRE_CFG_RC_BAD_PARAM; goto out; } - gateway_nid = libcfs_str2nid(gw); - if (gateway_nid == LNET_NID_ANY) { - snprintf(err_str, - sizeof(err_str), - "\"cannot parse gateway NID '%s'\"", gw); - rc = LUSTRE_CFG_RC_BAD_PARAM; + replace_sep(nidstr, ',', ' '); + rc = lustre_lnet_parse_nidstr(nidstr, lnet_nidlist, + LNET_MAX_NIDS_PER_PEER, err_str); + if (rc < 0) goto out; + + num_nids = 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; + } + + data.cfg_nid = lnet_nidlist[idx]; + + 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); + + if (rc != 0 && errno != EEXIST && + errno != EHOSTUNREACH) { + rc = -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, + cmd == LNETCTL_ADD_CMD ? ADD_CMD : DEL_CMD, "route", + err_str, err_rc); + + return rc; +} + +int lustre_lnet_config_route(char *nw, char *nidstr, int hops, int prio, + int sen, int seq_no, struct cYAML **err_rc) +{ + int rc; + char err_str[LNET_MAX_STR_LEN] = "\"generic error\""; + 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); + 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; } @@ -607,114 +877,57 @@ int lustre_lnet_config_route(char *nw, char *gw, int hops, int prio, if (prio == -1) { prio = 0; } else if (prio < 0) { - snprintf(err_str, - sizeof(err_str), - "\"invalid priority %d, must be greater than 0\"", - prio); + 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; } - LIBCFS_IOC_INIT_V2(data, cfg_hdr); - data.cfg_net = net; - data.cfg_config_u.cfg_route.rtr_hop = hops; - data.cfg_config_u.cfg_route.rtr_priority = prio; - data.cfg_nid = gateway_nid; - - rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_ADD_ROUTE, &data); - if (rc != 0) { - rc = -errno; - snprintf(err_str, - sizeof(err_str), - "\"cannot add route: %s\"", strerror(errno)); + 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; } + 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, ADD_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_del_route(char *nw, char *nidstr, 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 net = LNET_NIDNET(LNET_NID_ANY); - char err_str[LNET_MAX_STR_LEN]; - - snprintf(err_str, sizeof(err_str), "\"Success\""); - - if (nw == NULL || gw == NULL) { - snprintf(err_str, - sizeof(err_str), - "\"missing mandatory parameter(s): '%s'\"", - (nw == NULL && gw == NULL) ? "network, gateway" : - (nw == NULL) ? "network" : "gateway"); - rc = LUSTRE_CFG_RC_MISSING_PARAM; - goto out; - } - - net = libcfs_str2net(nw); - if (net == LNET_NIDNET(LNET_NID_ANY)) { - snprintf(err_str, - sizeof(err_str), - "\"cannot parse net '%s'\"", nw); - rc = LUSTRE_CFG_RC_BAD_PARAM; - goto out; - } - - gateway_nid = libcfs_str2nid(gw); - if (gateway_nid == LNET_NID_ANY) { - snprintf(err_str, - sizeof(err_str), - "\"cannot parse gateway NID '%s'\"", gw); - rc = LUSTRE_CFG_RC_BAD_PARAM; - goto out; - } - - LIBCFS_IOC_INIT_V2(data, cfg_hdr); - data.cfg_net = net; - 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; - } - -out: - cYAML_build_error(rc, seq_no, DEL_CMD, "route", err_str, err_rc); - - return 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) + struct cYAML **err_rc, bool backup) { struct lnet_ioctl_config_data data; lnet_nid_t gateway_nid; int rc = LUSTRE_CFG_RC_OUT_OF_MEM; int l_errno = 0; - __u32 net = LNET_NIDNET(LNET_NID_ANY); + __u32 net = LNET_NET_ANY; 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)) { + if (net == LNET_NET_ANY) { snprintf(err_str, sizeof(err_str), "\"cannot parse net '%s'\"", nw); @@ -724,7 +937,7 @@ int lustre_lnet_show_route(char *nw, char *gw, int hops, int prio, int detail, } else { /* show all routes without filtering on net */ - net = LNET_NIDNET(LNET_NID_ANY); + net = LNET_NET_ANY; } if (gw != NULL) { @@ -759,6 +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; @@ -769,7 +985,7 @@ int lustre_lnet_show_route(char *nw, char *gw, int hops, int prio, int detail, } /* filter on provided data */ - if (net != LNET_NIDNET(LNET_NID_ANY) && + if (net != LNET_NET_ANY && net != data.cfg_net) continue; @@ -816,11 +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_string(item, "state", - data.cfg_config_u.cfg_route. - rtr_flags ? + 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", + rt_alive ? "up" : "down") == NULL) goto out; + + if (!backup && + cYAML_create_string(item, "type", + rt_multi_hop? + "multi-hop" : "single-hop") == NULL) + goto out; } } @@ -891,51 +1123,116 @@ static int socket_intf_query(int request, char *intf, return rc; } +static int lustre_lnet_queryip(struct lnet_dlc_intf_descr *intf, __u32 *ip) +{ + struct ifreq ifr; + int rc; + + memset(&ifr, 0, sizeof(ifr)); + rc = socket_intf_query(SIOCGIFFLAGS, intf->intf_name, &ifr); + if (rc != 0) + return LUSTRE_CFG_RC_BAD_PARAM; + + if ((ifr.ifr_flags & IFF_UP) == 0) + return LUSTRE_CFG_RC_BAD_PARAM; + + memset(&ifr, 0, sizeof(ifr)); + rc = socket_intf_query(SIOCGIFADDR, intf->intf_name, &ifr); + if (rc != 0) + return LUSTRE_CFG_RC_BAD_PARAM; + + *ip = ((struct sockaddr_in *)&ifr.ifr_addr)->sin_addr.s_addr; + *ip = bswap_32(*ip); + + return LUSTRE_CFG_RC_NO_ERR; +} + /* * for each interface in the array of interfaces find the IP address of * that interface, create its nid and add it to an array of NIDs. * Stop if any of the interfaces is down */ static int lustre_lnet_intf2nids(struct lnet_dlc_network_descr *nw, - lnet_nid_t **nids, __u32 *nnids) + lnet_nid_t **nids, __u32 *nnids, + char *err_str, size_t str_len) { int i = 0, count = 0, rc; - struct ifreq ifr; - __u32 ip; struct lnet_dlc_intf_descr *intf; + char val[LNET_MAX_STR_LEN]; + __u32 ip; + int gni_num; + char *endp; + unsigned int num; + - if (nw == NULL || nids == NULL) + if (nw == NULL || nids == NULL) { + snprintf(err_str, str_len, + "\"unexpected parameters to lustre_lnet_intf2nids()\""); return LUSTRE_CFG_RC_BAD_PARAM; + } - list_for_each_entry(intf, &nw->nw_intflist, intf_on_network) - count++; + if (LNET_NETTYP(nw->nw_id) == GNILND) { + count = 1; + } else { + list_for_each_entry(intf, &nw->nw_intflist, intf_on_network) + count++; + } *nids = calloc(count, sizeof(lnet_nid_t)); - if (*nids == NULL) + if (*nids == NULL) { + snprintf(err_str, str_len, + "\"out of memory\""); return LUSTRE_CFG_RC_OUT_OF_MEM; - - list_for_each_entry(intf, &nw->nw_intflist, intf_on_network) { - memset(&ifr, 0, sizeof(ifr)); - rc = socket_intf_query(SIOCGIFFLAGS, intf->intf_name, &ifr); - if (rc != 0) - goto failed; - - if ((ifr.ifr_flags & IFF_UP) == 0) { - rc = LUSTRE_CFG_RC_BAD_PARAM; + } + /* + * special case the GNI interface since it doesn't have an IP + * address. The assumption is that there can only be one GNI + * interface in the system. No interface name is provided. + */ + if (LNET_NETTYP(nw->nw_id) == GNILND) { + rc = read_sysfs_file(gni_nid_path, "nid", val, + 1, sizeof(val)); + if (rc) { + snprintf(err_str, str_len, + "\"cannot read gni nid\""); goto failed; } + gni_num = atoi(val); - memset(&ifr, 0, sizeof(ifr)); - rc = socket_intf_query(SIOCGIFADDR, intf->intf_name, &ifr); - if (rc != 0) - goto failed; + (*nids)[i] = LNET_MKNID(nw->nw_id, gni_num); - ip = ((struct sockaddr_in *)&ifr.ifr_addr)->sin_addr.s_addr; - ip = bswap_32(ip); - (*nids)[i] = LNET_MKNID(nw->nw_id, ip); - i++; + goto out; + } + + /* look at the other interfaces */ + list_for_each_entry(intf, &nw->nw_intflist, intf_on_network) { + 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++; + } } +out: *nnids = count; return 0; @@ -1104,25 +1401,32 @@ int lustre_lnet_match_ip_to_intf(struct ifaddrs *ifa, return LUSTRE_CFG_RC_MATCH; } -int lustre_lnet_resolve_ip2nets_rule(struct lustre_lnet_ip2nets *ip2nets, - lnet_nid_t **nids, __u32 *nnids) +static int lustre_lnet_resolve_ip2nets_rule(struct lustre_lnet_ip2nets *ip2nets, + lnet_nid_t **nids, __u32 *nnids, + char *err_str, size_t str_len) { struct ifaddrs *ifa; int rc = LUSTRE_CFG_RC_NO_ERR; rc = getifaddrs(&ifa); - if (rc < 0) + if (rc < 0) { + snprintf(err_str, str_len, + "\"failed to get interface addresses: %d\"", -errno); return -errno; + } rc = lustre_lnet_match_ip_to_intf(ifa, &ip2nets->ip2nets_net.nw_intflist, &ip2nets->ip2nets_ip_ranges); if (rc != LUSTRE_CFG_RC_MATCH) { + snprintf(err_str, str_len, + "\"couldn't match ip to existing interfaces\""); freeifaddrs(ifa); return rc; } - rc = lustre_lnet_intf2nids(&ip2nets->ip2nets_net, nids, nnids); + rc = lustre_lnet_intf2nids(&ip2nets->ip2nets_net, nids, nnids, + err_str, str_len); if (rc != LUSTRE_CFG_RC_NO_ERR) { *nids = NULL; *nnids = 0; @@ -1223,9 +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, @@ -1240,13 +1542,10 @@ lustre_lnet_config_ip2nets(struct lustre_lnet_ip2nets *ip2nets, * The memory is allocated in that function then freed here when * it's no longer needed. */ - rc = lustre_lnet_resolve_ip2nets_rule(ip2nets, &nids, &nnids); - if (rc != LUSTRE_CFG_RC_NO_ERR && rc != LUSTRE_CFG_RC_MATCH) { - snprintf(err_str, - sizeof(err_str), - "\"cannot resolve ip2nets rule\""); + rc = lustre_lnet_resolve_ip2nets_rule(ip2nets, &nids, &nnids, err_str, + sizeof(err_str)); + if (rc != LUSTRE_CFG_RC_NO_ERR && rc != LUSTRE_CFG_RC_MATCH) goto out; - } if (list_empty(&ip2nets->ip2nets_net.nw_intflist)) { snprintf(err_str, sizeof(err_str), @@ -1277,7 +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; @@ -1285,12 +1584,14 @@ 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) { + if (ip2net == NULL && (nw_descr == NULL || nw_descr->nw_id == 0 || + (list_empty(&nw_descr->nw_intflist) && + LNET_NETTYP(nw_descr->nw_id) != GNILND))) { snprintf(err_str, sizeof(err_str), - "\"mandatory parameters not specified.\""); + "\"missing mandatory parameters in NI config: '%s'\"", + (nw_descr == NULL) ? "network , interface" : + (nw_descr->nw_id == 0) ? "network" : "interface"); rc = LUSTRE_CFG_RC_MISSING_PARAM; goto out; } @@ -1362,7 +1663,7 @@ int lustre_lnet_config_ni(struct lnet_dlc_network_descr *nw_descr, goto out; } - if (nw_descr->nw_id == LNET_NIDNET(LNET_NID_ANY)) { + if (nw_descr->nw_id == LNET_NET_ANY) { snprintf(err_str, sizeof(err_str), "\"cannot parse net '%s'\"", @@ -1371,7 +1672,11 @@ int lustre_lnet_config_ni(struct lnet_dlc_network_descr *nw_descr, goto out; } - if (list_empty(&nw_descr->nw_intflist)) { + /* + * special case the GNI since no interface name is expected + */ + if (list_empty(&nw_descr->nw_intflist) && + (LNET_NETTYP(nw_descr->nw_id) != GNILND)) { snprintf(err_str, sizeof(err_str), "\"no interface name provided\""); @@ -1379,10 +1684,9 @@ int lustre_lnet_config_ni(struct lnet_dlc_network_descr *nw_descr, goto out; } - rc = lustre_lnet_intf2nids(nw_descr, &nids, &nnids); + rc = lustre_lnet_intf2nids(nw_descr, &nids, &nnids, + err_str, sizeof(err_str)); if (rc != 0) { - snprintf(err_str, sizeof(err_str), - "\"bad parameter\""); rc = LUSTRE_CFG_RC_BAD_PARAM; goto out; } @@ -1417,17 +1721,17 @@ 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) { + if (nw_descr == NULL || nw_descr->nw_id == 0) { snprintf(err_str, sizeof(err_str), - "\"missing mandatory parameter\""); + "\"missing mandatory parameter in deleting NI: '%s'\"", + (nw_descr == NULL) ? "network , interface" : + (nw_descr->nw_id == 0) ? "network" : "interface"); rc = LUSTRE_CFG_RC_MISSING_PARAM; goto out; } @@ -1435,7 +1739,7 @@ int lustre_lnet_del_ni(struct lnet_dlc_network_descr *nw_descr, if (LNET_NETTYP(nw_descr->nw_id) == LOLND) return LUSTRE_CFG_RC_NO_ERR; - if (nw_descr->nw_id == LNET_NIDNET(LNET_NID_ANY)) { + if (nw_descr->nw_id == LNET_NET_ANY) { snprintf(err_str, sizeof(err_str), "\"cannot parse net '%s'\"", @@ -1444,10 +1748,9 @@ int lustre_lnet_del_ni(struct lnet_dlc_network_descr *nw_descr, goto out; } - rc = lustre_lnet_intf2nids(nw_descr, &nids, &nnids); + rc = lustre_lnet_intf2nids(nw_descr, &nids, &nnids, + err_str, sizeof(err_str)); if (rc != 0) { - snprintf(err_str, sizeof(err_str), - "\"bad parameter\""); rc = LUSTRE_CFG_RC_BAD_PARAM; goto out; } @@ -1495,31 +1798,128 @@ out: 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] = "\"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) +{ + if (cYAML_create_number(yaml, "put", + counts->ico_put_count) + == NULL) + return false; + if (cYAML_create_number(yaml, "get", + counts->ico_get_count) + == NULL) + return false; + if (cYAML_create_number(yaml, "reply", + counts->ico_reply_count) + == NULL) + return false; + if (cYAML_create_number(yaml, "ack", + counts->ico_ack_count) + == NULL) + return false; + if (cYAML_create_number(yaml, "hello", + counts->ico_hello_count) + == NULL) + return false; + + return true; +} + +static struct lnet_ioctl_comm_count * +get_counts(struct lnet_ioctl_element_msg_stats *msg_stats, int idx) +{ + if (idx == 0) + return &msg_stats->im_send_stats; + if (idx == 1) + return &msg_stats->im_recv_stats; + if (idx == 2) + return &msg_stats->im_drop_stats; + + return NULL; +} + int lustre_lnet_show_net(char *nw, int detail, int seq_no, - struct cYAML **show_rc, struct cYAML **err_rc) + struct cYAML **show_rc, struct cYAML **err_rc, + bool backup) { char *buf; struct lnet_ioctl_config_ni *ni_data; struct lnet_ioctl_config_lnd_tunables *lnd; struct lnet_ioctl_element_stats *stats; - __u32 net = LNET_NIDNET(LNET_NID_ANY); - __u32 prev_net = LNET_NIDNET(LNET_NID_ANY); + struct lnet_ioctl_element_msg_stats msg_stats; + struct lnet_ioctl_local_ni_hstats hstats; + __u32 net = LNET_NET_ANY; + __u32 prev_net = LNET_NET_ANY; int rc = LUSTRE_CFG_RC_OUT_OF_MEM, i, j; int l_errno = 0; struct cYAML *root = NULL, *tunables = NULL, *net_node = NULL, *interfaces = NULL, *item = NULL, *first_seq = NULL, - *tmp = NULL, *statistics = NULL; + *tmp = NULL, *statistics = NULL, + *yhstats = NULL; 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; @@ -1528,7 +1928,7 @@ int lustre_lnet_show_net(char *nw, int detail, int seq_no, if (nw != NULL) { net = libcfs_str2net(nw); - if (net == LNET_NIDNET(LNET_NID_ANY)) { + if (net == LNET_NET_ANY) { snprintf(err_str, sizeof(err_str), "\"cannot parse net '%s'\"", nw); @@ -1546,7 +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); @@ -1568,10 +1967,14 @@ int lustre_lnet_show_net(char *nw, int detail, int seq_no, rc_net = LNET_NIDNET(ni_data->lic_nid); /* filter on provided data */ - if (net != LNET_NIDNET(LNET_NID_ANY) && + if (net != LNET_NET_ANY && net != rc_net) continue; + /* if we're backing up don't store lo */ + if (backup && LNET_NETTYP(rc_net) == LOLND) + continue; + /* default rc to -1 in case we hit the goto */ rc = -1; exist = true; @@ -1605,11 +2008,13 @@ int lustre_lnet_show_net(char *nw, int detail, int seq_no, if (first_seq == NULL) first_seq = item; - if (cYAML_create_string(item, "nid", + if (!backup && + cYAML_create_string(item, "nid", libcfs_nid2str(ni_data->lic_nid)) == NULL) goto out; - if (cYAML_create_string(item, + if (!backup && + cYAML_create_string(item, "status", (ni_data->lic_status == LNET_NI_STATUS_UP) ? @@ -1638,6 +2043,10 @@ int lustre_lnet_show_net(char *nw, int detail, int seq_no, if (detail) { char *limit; + int k; + + if (backup) + goto continue_without_msg_stats; statistics = cYAML_create_object(item, "statistics"); if (statistics == NULL) @@ -1658,6 +2067,82 @@ int lustre_lnet_show_net(char *nw, int detail, int seq_no, == NULL) goto out; + if (detail < 2) + goto continue_without_msg_stats; + + LIBCFS_IOC_INIT_V2(msg_stats, im_hdr); + msg_stats.im_hdr.ioc_len = sizeof(msg_stats); + msg_stats.im_idx = i; + + rc = l_ioctl(LNET_DEV_ID, + IOC_LIBCFS_GET_LOCAL_NI_MSG_STATS, + &msg_stats); + if (rc != 0) { + l_errno = errno; + goto continue_without_msg_stats; + } + + for (k = 0; k < 3; k++) { + struct lnet_ioctl_comm_count *counts; + struct cYAML *msg_statistics = NULL; + + msg_statistics = cYAML_create_object(item, + (char *)gmsg_stat_names[k]); + if (msg_statistics == NULL) + goto out; + + counts = get_counts(&msg_stats, k); + if (counts == NULL) + goto out; + + if (!add_msg_stats_to_yaml_blk(msg_statistics, + counts)) + goto out; + } + + LIBCFS_IOC_INIT_V2(hstats, hlni_hdr); + hstats.hlni_nid = ni_data->lic_nid; + /* grab health stats */ + rc = l_ioctl(LNET_DEV_ID, + IOC_LIBCFS_GET_LOCAL_HSTATS, + &hstats); + if (rc != 0) { + l_errno = errno; + goto continue_without_msg_stats; + } + yhstats = cYAML_create_object(item, "health stats"); + if (!yhstats) + goto out; + if (cYAML_create_number(yhstats, "health value", + hstats.hlni_health_value) + == NULL) + goto out; + if (cYAML_create_number(yhstats, "interrupts", + hstats.hlni_local_interrupt) + == NULL) + goto out; + if (cYAML_create_number(yhstats, "dropped", + hstats.hlni_local_dropped) + == NULL) + goto out; + if (cYAML_create_number(yhstats, "aborted", + hstats.hlni_local_aborted) + == NULL) + goto out; + if (cYAML_create_number(yhstats, "no route", + hstats.hlni_local_no_route) + == NULL) + goto out; + if (cYAML_create_number(yhstats, "timeouts", + hstats.hlni_local_timeout) + == NULL) + goto out; + if (cYAML_create_number(yhstats, "error", + hstats.hlni_local_error) + == NULL) + goto out; + +continue_without_msg_stats: tunables = cYAML_create_object(item, "tunables"); if (!tunables) goto out; @@ -1666,36 +2151,43 @@ int lustre_lnet_show_net(char *nw, int detail, int seq_no, if (rc != LUSTRE_CFG_RC_NO_ERR) goto out; - tunables = cYAML_create_object(item, "lnd tunables"); - if (tunables == NULL) - goto out; - rc = lustre_ni_show_tunables(tunables, LNET_NETTYP(rc_net), &lnd->lt_tun); - if (rc != LUSTRE_CFG_RC_NO_ERR) + if (rc != LUSTRE_CFG_RC_NO_ERR && + rc != LUSTRE_CFG_RC_NO_MATCH) goto out; - if (cYAML_create_number(item, "tcp bonding", - ni_data->lic_tcp_bonding) - == NULL) - goto out; + if (rc != LUSTRE_CFG_RC_NO_MATCH) { + tunables = cYAML_create_object(item, + "lnd tunables"); + if (tunables == NULL) + goto out; + } - if (cYAML_create_number(item, "dev cpt", + if (!backup && + cYAML_create_number(item, "dev cpt", ni_data->lic_dev_cpt) == NULL) goto out; + if (!backup && + cYAML_create_number(item, "tcp bonding", + ni_data->lic_tcp_bonding) + == NULL) + 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", @@ -1752,9 +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; @@ -1777,80 +2267,298 @@ out: return rc; } -int lustre_lnet_config_numa_range(int range, int seq_no, struct cYAML **err_rc) +int ioctl_set_value(__u32 val, int ioc, char *name, + int seq_no, struct cYAML **err_rc) { - struct lnet_ioctl_numa_range data; + 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\""); - - if (range < 0) { - snprintf(err_str, - sizeof(err_str), - "\"range must be >= 0\""); - rc = LUSTRE_CFG_RC_OUT_OF_RANGE_PARAM; - goto out; - } + char err_str[LNET_MAX_STR_LEN] = "\"success\""; - LIBCFS_IOC_INIT_V2(data, nr_hdr); - data.nr_range = range; + LIBCFS_IOC_INIT_V2(data, sv_hdr); + data.sv_value = val; - rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_SET_NUMA_RANGE, &data); + rc = l_ioctl(LNET_DEV_ID, ioc , &data); if (rc != 0) { rc = -errno; snprintf(err_str, sizeof(err_str), - "\"cannot configure buffers: %s\"", strerror(errno)); - goto out; + "\"cannot configure %s to %d: %s\"", name, + val, strerror(errno)); } -out: - cYAML_build_error(rc, seq_no, ADD_CMD, "numa_range", err_str, err_rc); + cYAML_build_error(rc, seq_no, ADD_CMD, name, err_str, err_rc); return rc; } -int lustre_lnet_config_buffers(int tiny, int small, int large, int seq_no, - struct cYAML **err_rc) +int lustre_lnet_config_recov_intrv(int intrv, 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\""); - - /* -1 indicates to ignore changes to this field */ - if (tiny < -1 || small < -1 || large < -1) { - snprintf(err_str, - sizeof(err_str), - "\"tiny, small and large must be >= 0\""); - rc = LUSTRE_CFG_RC_OUT_OF_RANGE_PARAM; - goto out; - } + char err_str[LNET_MAX_STR_LEN] = "\"success\""; + char val[LNET_MAX_STR_LEN]; - LIBCFS_IOC_INIT_V2(data, cfg_hdr); - data.cfg_config_u.cfg_buffers.buf_tiny = tiny; - data.cfg_config_u.cfg_buffers.buf_small = small; - data.cfg_config_u.cfg_buffers.buf_large = large; + snprintf(val, sizeof(val), "%d", intrv); - rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_ADD_BUF, &data); - if (rc != 0) { - rc = -errno; - snprintf(err_str, - sizeof(err_str), - "\"cannot configure buffers: %s\"", strerror(errno)); - goto out; - } + 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)); -out: - cYAML_build_error(rc, seq_no, ADD_CMD, "buf", err_str, err_rc); + 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); + + rc = write_sysfs_file(modparam_path, "lnet_health_sensitivity", val, + 1, strlen(val) + 1); + if (rc) + snprintf(err_str, sizeof(err_str), + "\"cannot configure health sensitivity: %s\"", + strerror(errno)); + + cYAML_build_error(rc, seq_no, ADD_CMD, "health_sensitivity", err_str, err_rc); + + return 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] = "\"success\""; + char val[LNET_MAX_STR_LEN]; + + snprintf(val, sizeof(val), "%d", timeout); + + rc = write_sysfs_file(modparam_path, "lnet_transaction_timeout", val, + 1, strlen(val) + 1); + if (rc) + snprintf(err_str, sizeof(err_str), + "\"cannot configure transaction timeout: %s\"", + strerror(errno)); + + cYAML_build_error(rc, seq_no, ADD_CMD, "transaction_timeout", err_str, err_rc); + + return rc; +} + +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] = "\"success\""; + char val[LNET_MAX_STR_LEN]; + + snprintf(val, sizeof(val), "%d", count); + + rc = write_sysfs_file(modparam_path, "lnet_retry_count", val, + 1, strlen(val) + 1); + if (rc) + snprintf(err_str, sizeof(err_str), + "\"cannot configure retry count: %s\"", + strerror(errno)); + + cYAML_build_error(rc, seq_no, ADD_CMD, "retry_count", err_str, err_rc); + + return 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_str[LNET_MAX_STR_LEN]; + + 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_recovery_limit(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_str[LNET_MAX_STR_LEN]; + + if (val < 0) { + rc = LUSTRE_CFG_RC_BAD_PARAM; + snprintf(err_str, sizeof(err_str), + "\"Must be greater than or equal to 0\""); + } else { + snprintf(err_str, sizeof(err_str), "\"success\""); + + snprintf(val_str, sizeof(val_str), "%d", val); + + rc = write_sysfs_file(modparam_path, "lnet_recovery_limit", + val_str, 1, strlen(val_str) + 1); + if (rc) + snprintf(err_str, sizeof(err_str), + "\"cannot configure recovery limit: %s\"", + strerror(errno)); + } + + cYAML_build_error(rc, seq_no, ADD_CMD, "recovery_limit", 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); + + rc = write_sysfs_file(modparam_path, "lnet_interfaces_max", val, + 1, strlen(val) + 1); + if (rc) + snprintf(err_str, sizeof(err_str), + "\"cannot configure max interfaces: %s\"", + strerror(errno)); + + cYAML_build_error(rc, seq_no, ADD_CMD, "max_interfaces", err_str, err_rc); + + return 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] = "\"success\""; + char val[LNET_MAX_STR_LEN]; + + snprintf(val, sizeof(val), "%u", (enable) ? 0 : 1); + + rc = write_sysfs_file(modparam_path, "lnet_peer_discovery_disabled", val, + 1, strlen(val) + 1); + if (rc) + snprintf(err_str, sizeof(err_str), + "\"cannot configure discovery: %s\"", + strerror(errno)); + + cYAML_build_error(rc, seq_no, ADD_CMD, "discovery", err_str, err_rc); + + return 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, + "numa_range", seq_no, err_rc); +} + +int lustre_lnet_config_buffers(int tiny, int small, int large, 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] = "\"success\""; + + /* -1 indicates to ignore changes to this field */ + if (tiny < -1 || small < -1 || large < -1) { + snprintf(err_str, + sizeof(err_str), + "\"tiny, small and large must be >= 0\""); + rc = LUSTRE_CFG_RC_OUT_OF_RANGE_PARAM; + goto out; + } + + LIBCFS_IOC_INIT_V2(data, cfg_hdr); + data.cfg_config_u.cfg_buffers.buf_tiny = tiny; + data.cfg_config_u.cfg_buffers.buf_small = small; + data.cfg_config_u.cfg_buffers.buf_large = large; + + rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_ADD_BUF, &data); + if (rc != 0) { + rc = -errno; + snprintf(err_str, + sizeof(err_str), + "\"cannot configure buffers: %s\"", strerror(errno)); + goto out; + } + +out: + cYAML_build_error(rc, seq_no, ADD_CMD, "buf", err_str, err_rc); return rc; } int lustre_lnet_show_routing(int seq_no, struct cYAML **show_rc, - struct cYAML **err_rc) + struct cYAML **err_rc, bool backup) { struct lnet_ioctl_config_data *data; struct lnet_ioctl_pool_cfg *pool_cfg = NULL; @@ -1863,12 +2571,10 @@ 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; @@ -1879,7 +2585,10 @@ int lustre_lnet_show_routing(int seq_no, struct cYAML **show_rc, if (root == NULL) goto out; - pools_node = cYAML_create_seq(root, "routing"); + if (backup) + pools_node = cYAML_create_object(root, "routing"); + else + pools_node = cYAML_create_seq(root, "routing"); if (pools_node == NULL) goto out; @@ -1899,6 +2608,9 @@ int lustre_lnet_show_routing(int seq_no, struct cYAML **show_rc, pool_cfg = (struct lnet_ioctl_pool_cfg *)data->cfg_bulk; + if (backup) + goto calculate_buffers; + snprintf(node_name, sizeof(node_name), "cpt[%d]", i); item = cYAML_create_seq_item(pools_node); if (item == NULL) @@ -1911,24 +2623,31 @@ int lustre_lnet_show_routing(int seq_no, struct cYAML **show_rc, if (cpt == NULL) goto out; +calculate_buffers: /* create the tree and print */ for (j = 0; j < LNET_NRBPOOLS; j++) { - type_node = cYAML_create_object(cpt, pools[j]); - if (type_node == NULL) - goto out; - if (cYAML_create_number(type_node, "npages", + if (!backup) { + type_node = cYAML_create_object(cpt, pools[j]); + if (type_node == NULL) + goto out; + } + if (!backup && + cYAML_create_number(type_node, "npages", pool_cfg->pl_pools[j].pl_npages) == NULL) goto out; - if (cYAML_create_number(type_node, "nbuffers", + if (!backup && + cYAML_create_number(type_node, "nbuffers", pool_cfg->pl_pools[j]. pl_nbuffers) == NULL) goto out; - if (cYAML_create_number(type_node, "credits", + if (!backup && + cYAML_create_number(type_node, "credits", pool_cfg->pl_pools[j]. pl_credits) == NULL) goto out; - if (cYAML_create_number(type_node, "mincredits", + if (!backup && + cYAML_create_number(type_node, "mincredits", pool_cfg->pl_pools[j]. pl_mincredits) == NULL) goto out; @@ -1939,6 +2658,15 @@ int lustre_lnet_show_routing(int seq_no, struct cYAML **show_rc, } if (pool_cfg != NULL) { + if (backup) { + if (cYAML_create_number(pools_node, "enable", + pool_cfg->pl_routing) == + NULL) + goto out; + + goto add_buffer_section; + } + item = cYAML_create_seq_item(pools_node); if (item == NULL) goto out; @@ -1948,6 +2676,7 @@ int lustre_lnet_show_routing(int seq_no, struct cYAML **show_rc, goto out; } +add_buffer_section: /* create a buffers entry in the show. This is necessary so that * if the YAML output is used to configure a node, the buffer * configuration takes hold */ @@ -2003,7 +2732,8 @@ out: } int lustre_lnet_show_peer(char *knid, int detail, int seq_no, - struct cYAML **show_rc, struct cYAML **err_rc) + struct cYAML **show_rc, struct cYAML **err_rc, + bool backup) { /* * TODO: This function is changing in a future patch to accommodate @@ -2012,21 +2742,23 @@ int lustre_lnet_show_peer(char *knid, int detail, int seq_no, struct lnet_ioctl_peer_cfg peer_info; struct lnet_peer_ni_credit_info *lpni_cri; struct lnet_ioctl_element_stats *lpni_stats; - int rc = LUSTRE_CFG_RC_OUT_OF_MEM, ncpt = 0, i = 0, j = 0; + 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; int l_errno = 0; + __u32 count; + __u32 size; struct cYAML *root = NULL, *peer = NULL, *peer_ni = NULL, - *first_seq = NULL, *peer_root = NULL, *tmp = NULL; - char err_str[LNET_MAX_STR_LEN]; - lnet_nid_t prev_primary_nid = LNET_NID_ANY, primary_nid = LNET_NID_ANY; - int data_size = sizeof(*lpni_cri) + sizeof(*lpni_stats); - char *data = malloc(data_size); - bool new_peer = true; - - snprintf(err_str, sizeof(err_str), - "\"out of memory\""); - - if (data == NULL) - goto out; + *first_seq = NULL, *peer_root = NULL, *tmp = NULL, + *msg_statistics = NULL, *statistics = NULL, + *yhstats; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; + struct lnet_process_id *list = NULL; + void *data = NULL; + void *lpni_data; + bool exist = false; /* create struct cYAML root object */ root = cYAML_create_object(NULL, NULL); @@ -2037,72 +2769,142 @@ int lustre_lnet_show_peer(char *knid, int detail, int seq_no, if (peer_root == NULL) goto out; - if (knid != NULL) - primary_nid = libcfs_str2nid(knid); - - do { - for (i = 0;; i++) { - memset(data, 0, data_size); + count = 1000; + size = count * sizeof(struct lnet_process_id); + list = malloc(size); + if (list == NULL) { + l_errno = ENOMEM; + goto out; + } + if (knid != NULL) { + list[0].nid = libcfs_str2nid(knid); + count = 1; + } else { + for (;;) { memset(&peer_info, 0, sizeof(peer_info)); LIBCFS_IOC_INIT_V2(peer_info, prcfg_hdr); peer_info.prcfg_hdr.ioc_len = sizeof(peer_info); - peer_info.prcfg_count = i; - peer_info.prcfg_bulk = (void *)data; - peer_info.prcfg_size = data_size; + peer_info.prcfg_size = size; + peer_info.prcfg_bulk = list; + + l_errno = 0; + rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_GET_PEER_LIST, + &peer_info); + count = peer_info.prcfg_count; + if (rc == 0) + break; + l_errno = errno; + if (l_errno != E2BIG) { + snprintf(err_str, + sizeof(err_str), + "\"cannot get peer list: %s\"", + strerror(l_errno)); + rc = -l_errno; + goto out; + } + free(list); + size = peer_info.prcfg_size; + list = malloc(size); + if (list == NULL) { + l_errno = ENOMEM; + goto out; + } + } + } - rc = l_ioctl(LNET_DEV_ID, - IOC_LIBCFS_GET_PEER_NI, &peer_info); - if (rc != 0) { - l_errno = errno; + size = 4096; + data = malloc(size); + if (data == NULL) { + l_errno = ENOMEM; + goto out; + } + + for (i = 0; i < count; i++) { + for (;;) { + memset(&peer_info, 0, sizeof(peer_info)); + LIBCFS_IOC_INIT_V2(peer_info, prcfg_hdr); + peer_info.prcfg_hdr.ioc_len = sizeof(peer_info); + peer_info.prcfg_prim_nid = list[i].nid; + peer_info.prcfg_size = size; + peer_info.prcfg_bulk = data; + + l_errno = 0; + rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_GET_PEER_NI, + &peer_info); + if (rc == 0) break; + l_errno = errno; + if (l_errno != E2BIG) { + snprintf(err_str, + sizeof(err_str), + "\"cannot get peer information: %s\"", + strerror(l_errno)); + rc = -l_errno; + goto out; } + free(data); + size = peer_info.prcfg_size; + data = malloc(size); + if (data == NULL) { + l_errno = ENOMEM; + goto out; + } + } + exist = true; - if (primary_nid != LNET_NID_ANY && - primary_nid != peer_info.prcfg_prim_nid) - continue; + peer = cYAML_create_seq_item(peer_root); + if (peer == NULL) + goto out; - lpni_cri = peer_info.prcfg_bulk; - lpni_stats = peer_info.prcfg_bulk + sizeof(*lpni_cri); + if (first_seq == NULL) + first_seq = peer; - peer = cYAML_create_seq_item(peer_root); - if (peer == NULL) + lnet_nid_t pnid = peer_info.prcfg_prim_nid; + if (cYAML_create_string(peer, "primary nid", + libcfs_nid2str(pnid)) + == NULL) + goto out; + if (cYAML_create_string(peer, "Multi-Rail", + peer_info.prcfg_mr ? "True" : "False") + == NULL) + goto out; + /* + * print out the state of the peer only if details are + * requested + */ + if (detail >= 3) { + if (!backup && + cYAML_create_number(peer, "peer state", + peer_info.prcfg_state) + == NULL) goto out; + } - if (peer_info.prcfg_prim_nid != prev_primary_nid) { - prev_primary_nid = peer_info.prcfg_prim_nid; - new_peer = true; - } - - if (new_peer) { - lnet_nid_t pnid = peer_info.prcfg_prim_nid; - if (cYAML_create_string(peer, "primary nid", - libcfs_nid2str(pnid)) - == NULL) - goto out; - if (cYAML_create_string(peer, "Multi-Rail", - peer_info.prcfg_mr ? - "True" : "False") - == NULL) - goto out; - tmp = cYAML_create_seq(peer, "peer ni"); - if (tmp == NULL) - goto out; - new_peer = false; - } + tmp = cYAML_create_seq(peer, "peer ni"); + if (tmp == NULL) + goto out; - if (first_seq == NULL) - first_seq = peer; + lpni_data = data; + for (j = 0; j < peer_info.prcfg_count; j++) { + nidp = lpni_data; + lpni_cri = (void*)nidp + sizeof(nidp); + lpni_stats = (void *)lpni_cri + sizeof(*lpni_cri); + msg_stats = (void *)lpni_stats + sizeof(*lpni_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; if (cYAML_create_string(peer_ni, "nid", - libcfs_nid2str - (peer_info.prcfg_cfg_nid)) + libcfs_nid2str(*nidp)) == NULL) goto out; + if (backup) + continue; + if (cYAML_create_string(peer_ni, "state", lpni_cri->cr_aliveness) == NULL) @@ -2141,62 +2943,554 @@ int lustre_lnet_show_peer(char *knid, int detail, int seq_no, == NULL) goto out; - if (cYAML_create_number(peer_ni, "send_count", + if (cYAML_create_number(peer_ni, "refcount", + lpni_cri->cr_refcount) == NULL) + goto out; + + statistics = cYAML_create_object(peer_ni, "statistics"); + if (statistics == NULL) + goto out; + + if (cYAML_create_number(statistics, "send_count", lpni_stats->iel_send_count) == NULL) goto out; - if (cYAML_create_number(peer_ni, "recv_count", + if (cYAML_create_number(statistics, "recv_count", lpni_stats->iel_recv_count) == NULL) goto out; - if (cYAML_create_number(peer_ni, "drop_count", + if (cYAML_create_number(statistics, "drop_count", lpni_stats->iel_drop_count) == NULL) goto out; - if (cYAML_create_number(peer_ni, "refcount", - lpni_cri->cr_refcount) == NULL) + if (detail < 2) + continue; + + for (k = 0; k < 3; k++) { + struct lnet_ioctl_comm_count *counts; + + msg_statistics = cYAML_create_object(peer_ni, + (char *) gmsg_stat_names[k]); + if (msg_statistics == NULL) + goto out; + + counts = get_counts(msg_stats, k); + if (counts == NULL) + goto out; + + if (!add_msg_stats_to_yaml_blk(msg_statistics, + counts)) + 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; } + } + + /* print output iff show_rc is not provided */ + if (show_rc == NULL) + cYAML_print_tree(root); + + snprintf(err_str, sizeof(err_str), "\"success\""); + rc = LUSTRE_CFG_RC_NO_ERR; - if (l_errno != ENOENT) { +out: + free(list); + free(data); + 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; + /* find the peer node, if one doesn't exist then + * insert one. Otherwise add to the one there + */ + show_node = cYAML_get_object_item(*show_rc, + "peer"); + if (show_node != NULL && cYAML_is_sequence(show_node)) { + cYAML_insert_child(show_node, first_seq); + free(peer_root); + free(root); + } else if (show_node == NULL) { + cYAML_insert_sibling((*show_rc)->cy_child, + peer_root); + free(root); + } else { + cYAML_free_tree(root); + } + } else { + *show_rc = root; + } + + cYAML_build_error(rc, seq_no, SHOW_CMD, "peer", err_str, + err_rc); + + return rc; +} + +int lustre_lnet_list_peer(int seq_no, + struct cYAML **show_rc, struct cYAML **err_rc) +{ + struct lnet_ioctl_peer_cfg peer_info; + int rc = LUSTRE_CFG_RC_OUT_OF_MEM; + __u32 count; + __u32 size; + int i = 0; + int l_errno = 0; + struct cYAML *root = NULL, *list_root = NULL, *first_seq = NULL; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; + struct lnet_process_id *list = NULL; + + memset(&peer_info, 0, sizeof(peer_info)); + + /* create struct cYAML root object */ + root = cYAML_create_object(NULL, NULL); + if (root == NULL) + goto out; + + list_root = cYAML_create_seq(root, "peer list"); + if (list_root == NULL) + goto out; + + count = 1000; + size = count * sizeof(struct lnet_process_id); + list = malloc(size); + if (list == NULL) { + l_errno = ENOMEM; + goto out; + } + for (;;) { + LIBCFS_IOC_INIT_V2(peer_info, prcfg_hdr); + peer_info.prcfg_hdr.ioc_len = sizeof(peer_info); + peer_info.prcfg_size = size; + peer_info.prcfg_bulk = list; + + l_errno = 0; + rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_GET_PEER_LIST, &peer_info); + count = peer_info.prcfg_count; + if (rc == 0) + break; + l_errno = errno; + if (l_errno != E2BIG) { snprintf(err_str, sizeof(err_str), - "\"cannot get peer information: %s\"", + "\"cannot get peer list: %s\"", strerror(l_errno)); rc = -l_errno; goto out; } + free(list); + size = peer_info.prcfg_size; + list = malloc(size); + if (list == NULL) { + l_errno = ENOMEM; + goto out; + } + } + + /* count is now the actual number of ids in the list. */ + for (i = 0; i < count; i++) { + if (cYAML_create_string(list_root, "nid", + libcfs_nid2str(list[i].nid)) + == NULL) + goto out; + } + + /* print output iff show_rc is not provided */ + if (show_rc == NULL) + cYAML_print_tree(root); + + snprintf(err_str, sizeof(err_str), "\"success\""); + rc = LUSTRE_CFG_RC_NO_ERR; + +out: + if (list != NULL) + free(list); + 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 peer node, if one doesn't exist then + * insert one. Otherwise add to the one there + */ + show_node = cYAML_get_object_item(*show_rc, + "peer"); + if (show_node != NULL && cYAML_is_sequence(show_node)) { + cYAML_insert_child(show_node, first_seq); + free(list_root); + free(root); + } else if (show_node == NULL) { + cYAML_insert_sibling((*show_rc)->cy_child, + list_root); + free(root); + } else { + cYAML_free_tree(root); + } + } else { + *show_rc = root; + } + + cYAML_build_error(rc, seq_no, SHOW_CMD, "peer", err_str, + err_rc); + + return rc; +} + +static void add_to_global(struct cYAML *show_rc, struct cYAML *node, + struct cYAML *root) +{ + struct cYAML *show_node; + + show_node = cYAML_get_object_item(show_rc, "global"); + if (show_node != NULL) + cYAML_insert_sibling(show_node->cy_child, + node->cy_child); + else + cYAML_insert_sibling(show_rc->cy_child, + node); + free(root); +} + +static int build_global_yaml_entry(char *err_str, int err_len, int seq_no, + char *name, __u64 value, + struct cYAML **show_rc, + struct cYAML **err_rc, int err) +{ + int rc = LUSTRE_CFG_RC_OUT_OF_MEM; + struct cYAML *root = NULL, *global = NULL; + + if (err) { + rc = err; + goto out; + } + + root = cYAML_create_object(NULL, NULL); + if (root == NULL) + goto out; + + global = cYAML_create_object(root, "global"); + if (global == NULL) + goto out; + + if (cYAML_create_number(global, name, + value) == NULL) + goto out; + + if (show_rc == NULL) + cYAML_print_tree(root); + + snprintf(err_str, err_len, "\"success\""); + + rc = LUSTRE_CFG_RC_NO_ERR; + +out: + if (show_rc == NULL || rc != LUSTRE_CFG_RC_NO_ERR) { + cYAML_free_tree(root); + } else if (show_rc != NULL && *show_rc != NULL) { + add_to_global(*show_rc, global, root); + } else { + *show_rc = root; + } + + cYAML_build_error(rc, seq_no, SHOW_CMD, "global", err_str, err_rc); + + return rc; +} + +static int ioctl_show_global_values(int ioc, int seq_no, char *name, + struct cYAML **show_rc, + struct cYAML **err_rc) +{ + struct lnet_ioctl_set_value data; + int rc; + int l_errno = 0; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; + + LIBCFS_IOC_INIT_V2(data, sv_hdr); + + rc = l_ioctl(LNET_DEV_ID, ioc, &data); + if (rc != 0) { + l_errno = -errno; + snprintf(err_str, + sizeof(err_str), + "\"cannot get %s: %s\"", + 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); +} + +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_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] = "\"out of memory\""; + + 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 router sensitivity percentage: %d\"", rc); + } else { + sen = atoi(val); + } + + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "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); +} + +int lustre_lnet_show_transaction_to(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 tto = -1, l_errno = 0; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; + + rc = read_sysfs_file(modparam_path, "lnet_transaction_timeout", val, + 1, sizeof(val)); + if (rc) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"cannot get transaction timeout: %d\"", rc); + } else { + tto = atoi(val); + } + + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "transaction_timeout", tto, show_rc, + err_rc, l_errno); +} + +int lustre_lnet_show_retry_count(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 retry_count = -1, l_errno = 0; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; + + rc = read_sysfs_file(modparam_path, "lnet_retry_count", val, + 1, sizeof(val)); + if (rc) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"cannot get retry count: %d\"", rc); + } else { + retry_count = atoi(val); + } + + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "retry_count", retry_count, 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; - j++; - } while (j < ncpt); + rc = -EINVAL; - /* print output iff show_rc is not provided */ - if (show_rc == NULL) - cYAML_print_tree(root); + 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\""); - rc = LUSTRE_CFG_RC_NO_ERR; + 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 peer node, if one doesn't exist then - * insert one. Otherwise add to the one there + /* 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, - "peer"); + show_node = cYAML_get_object_item(*show_rc, name); if (show_node != NULL && cYAML_is_sequence(show_node)) { - cYAML_insert_child(show_node, first_seq); - free(peer_root); + cYAML_insert_child(show_node, nids); + free(nids); free(root); } else if (show_node == NULL) { cYAML_insert_sibling((*show_rc)->cy_child, - peer_root); + nids); free(root); } else { cYAML_free_tree(root); @@ -2205,86 +3499,168 @@ out: *show_rc = root; } - cYAML_build_error(rc, seq_no, SHOW_CMD, "peer", err_str, - err_rc); + cYAML_build_error(rc, seq_no, SHOW_CMD, name, err_str, err_rc); return rc; } -int lustre_lnet_show_numa_range(int seq_no, struct cYAML **show_rc, - struct cYAML **err_rc) +int lustre_lnet_show_local_ni_recovq(int seq_no, struct cYAML **show_rc, + struct cYAML **err_rc) { - struct lnet_ioctl_numa_range data; - int rc; - int l_errno; + 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_recovery_limit(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 recov_limit = -1, l_errno = 0; char err_str[LNET_MAX_STR_LEN]; - struct cYAML *root = NULL, *range = NULL; snprintf(err_str, sizeof(err_str), "\"out of memory\""); - LIBCFS_IOC_INIT_V2(data, nr_hdr); - - rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_GET_NUMA_RANGE, &data); - if (rc != 0) { - l_errno = errno; - snprintf(err_str, - sizeof(err_str), - "\"cannot get numa range: %s\"", - strerror(l_errno)); - rc = -l_errno; - goto out; + rc = read_sysfs_file(modparam_path, "lnet_recovery_limit", val, + 1, sizeof(val)); + if (rc) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"cannot get lnet_recovery_limit value: %d\"", rc); + } else { + recov_limit = atoi(val); } - rc = LUSTRE_CFG_RC_OUT_OF_MEM; + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "recovery_limit", recov_limit, + show_rc, err_rc, l_errno); +} - root = cYAML_create_object(NULL, NULL); - if (root == NULL) - goto out; +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] = "\"out of memory\""; + + rc = read_sysfs_file(modparam_path, "lnet_interfaces_max", val, + 1, sizeof(val)); + if (rc) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"cannot get max interfaces: %d\"", rc); + } else { + max_intf = atoi(val); + } - range = cYAML_create_object(root, "numa"); - if (range == NULL) - goto out; + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "max_intf", max_intf, show_rc, + err_rc, l_errno); +} - if (cYAML_create_number(range, "range", - data.nr_range) == NULL) - goto out; +int lustre_lnet_show_discovery(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 discovery = -1, l_errno = 0; + char err_str[LNET_MAX_STR_LEN] = "\"out of memory\""; + + rc = read_sysfs_file(modparam_path, "lnet_peer_discovery_disabled", val, + 1, sizeof(val)); + if (rc) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"cannot get discovery setting: %d\"", rc); + } else { + /* + * The kernel stores a discovery disabled value. User space + * shows whether discovery is enabled. So the value must be + * inverted. + */ + discovery = !atoi(val); + } - if (show_rc == NULL) - cYAML_print_tree(root); + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "discovery", discovery, show_rc, + err_rc, l_errno); +} - snprintf(err_str, sizeof(err_str), "\"success\""); - rc = LUSTRE_CFG_RC_NO_ERR; -out: - if (show_rc == NULL || rc != LUSTRE_CFG_RC_NO_ERR) { - cYAML_free_tree(root); - } else if (show_rc != NULL && *show_rc != NULL) { - cYAML_insert_sibling((*show_rc)->cy_child, - root->cy_child); - free(root); +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 { - *show_rc = root; + drop_asym_route = atoi(val); } - cYAML_build_error(rc, seq_no, SHOW_CMD, "numa", err_str, err_rc); + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "drop_asym_route", drop_asym_route, + show_rc, err_rc, l_errno); +} - return rc; +int lustre_lnet_show_numa_range(int seq_no, struct cYAML **show_rc, + struct cYAML **err_rc) +{ + return ioctl_show_global_values(IOC_LIBCFS_GET_NUMA_RANGE, seq_no, + "numa_range", show_rc, err_rc); } 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), @@ -2296,59 +3672,113 @@ 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, "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\""); @@ -2376,36 +3806,24 @@ 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: ['['']'] @@ -2761,18 +4179,18 @@ static int handle_yaml_del_ni(struct cYAML *tree, struct cYAML **show_rc, return rc; } -static int yaml_copy_peer_nids(struct cYAML *tree, char ***nidsppp, 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) { - struct cYAML *nids_entry = NULL, *child = NULL, *entry = NULL, - *prim_nid = NULL; - char **nids = NULL; - int num = 0, rc = LUSTRE_CFG_RC_NO_ERR; - - prim_nid = cYAML_get_object_item(tree, "primary nid"); - if (!prim_nid || !prim_nid->cy_valuestring) - return LUSTRE_CFG_RC_MISSING_PARAM; + int num_strs = 0, rc; + size_t buf_size, buf_pos, nidstr_len = 0; + char *buffer; + struct cYAML *child = NULL, *entry = NULL; - nids_entry = cYAML_get_object_item(tree, "peer ni"); if (cYAML_is_sequence(nids_entry)) { while (cYAML_get_next_seq_item(nids_entry, &child)) { entry = cYAML_get_object_item(child, "nid"); @@ -2780,105 +4198,173 @@ static int yaml_copy_peer_nids(struct cYAML *tree, char ***nidsppp, bool del) if (!entry || !entry->cy_valuestring) continue; - if ((strcmp(entry->cy_valuestring, prim_nid->cy_valuestring) - == 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; + if (prim_nid && + (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, 1); - if (nids == NULL) + 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; - - num = yaml_copy_peer_nids(tree, &nids, false); - if (num < 0) - return num; + 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"); + 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; + } - rc = lustre_lnet_config_peer_nid((prim_nid) ? prim_nid->cy_valuestring : NULL, - nids, num, - (non_mr) ? false : true, - (seq_no) ? seq_no->cy_valueint : -1, - err_rc); + prim_nidstr = prim_nid->cy_valuestring; - yaml_free_string_array(nids, num); - return rc; -} + /* 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, LNET_MAX_STR_LEN, + "badly formatted primary NID: %s", prim_nidstr); + goto failed; + } -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; + 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; + } + + 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; + } + } + } - num = yaml_copy_peer_nids(tree, &nids, true); - if (num < 0) - return num; + rc = lustre_lnet_mod_peer_nidlist(pnid, lnet_nidlist, cmd, + num_nids, mr_value, seqn, + err_rc); - seq_no = cYAML_get_object_item(tree, "seq_no"); - prim_nid = cYAML_get_object_item(tree, "primary nid"); +failed: + if (nidstr) + free(nidstr); - rc = lustre_lnet_del_peer_nid((prim_nid) ? prim_nid->cy_valuestring : NULL, - nids, num, - (seq_no) ? seq_no->cy_valueint : -1, - err_rc); + 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) +{ + return handle_yaml_peer_common(tree, show_rc, err_rc, LNETCTL_DEL_CMD); +} + static int handle_yaml_config_buffers(struct cYAML *tree, struct cYAML **show_rc, struct cYAML **err_rc) @@ -2972,8 +4458,7 @@ static int handle_yaml_show_route(struct cYAML *tree, struct cYAML **show_rc, (prio) ? prio->cy_valueint : -1, (detail) ? detail->cy_valueint : 0, (seq_no) ? seq_no->cy_valueint : -1, - show_rc, - err_rc); + show_rc, err_rc, false); } static int handle_yaml_show_net(struct cYAML *tree, struct cYAML **show_rc, @@ -2988,8 +4473,7 @@ static int handle_yaml_show_net(struct cYAML *tree, struct cYAML **show_rc, return lustre_lnet_show_net((net) ? net->cy_valuestring : NULL, (detail) ? detail->cy_valueint : 0, (seq_no) ? seq_no->cy_valueint : -1, - show_rc, - err_rc); + show_rc, err_rc, false); } static int handle_yaml_show_routing(struct cYAML *tree, struct cYAML **show_rc, @@ -3000,7 +4484,7 @@ static int handle_yaml_show_routing(struct cYAML *tree, struct cYAML **show_rc, seq_no = cYAML_get_object_item(tree, "seq_no"); return lustre_lnet_show_routing((seq_no) ? seq_no->cy_valueint : -1, - show_rc, err_rc); + show_rc, err_rc, false); } static int handle_yaml_show_peers(struct cYAML *tree, struct cYAML **show_rc, @@ -3015,7 +4499,7 @@ static int handle_yaml_show_peers(struct cYAML *tree, struct cYAML **show_rc, return lustre_lnet_show_peer((nid) ? nid->cy_valuestring : NULL, (detail) ? detail->cy_valueint : 0, (seq_no) ? seq_no->cy_valueint : -1, - show_rc, err_rc); + show_rc, err_rc, false); } static int handle_yaml_show_stats(struct cYAML *tree, struct cYAML **show_rc, @@ -3064,6 +4548,251 @@ static int handle_yaml_show_numa(struct cYAML *tree, struct cYAML **show_rc, show_rc, err_rc); } +static int handle_yaml_config_global_settings(struct cYAML *tree, + struct cYAML **show_rc, + struct cYAML **err_rc) +{ + struct cYAML *max_intf, *numa, *discovery, *retry, *tto, *seq_no, + *sen, *recov, *rsen, *drop_asym_route, *rsp_tracking, + *recov_limit; + int rc = 0; + + seq_no = cYAML_get_object_item(tree, "seq_no"); + max_intf = cYAML_get_object_item(tree, "max_intf"); + if (max_intf) + rc = lustre_lnet_config_max_intf(max_intf->cy_valueint, + seq_no ? seq_no->cy_valueint + : -1, + err_rc); + + numa = cYAML_get_object_item(tree, "numa_range"); + if (numa) + rc = lustre_lnet_config_numa_range(numa->cy_valueint, + seq_no ? seq_no->cy_valueint + : -1, + err_rc); + + discovery = cYAML_get_object_item(tree, "discovery"); + if (discovery) + rc = lustre_lnet_config_discovery(discovery->cy_valueint, + seq_no ? seq_no->cy_valueint + : -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, + seq_no ? seq_no->cy_valueint + : -1, + err_rc); + + tto = cYAML_get_object_item(tree, "transaction_timeout"); + if (tto) + rc = lustre_lnet_config_transaction_to(tto->cy_valueint, + seq_no ? seq_no->cy_valueint + : -1, + err_rc); + + sen = cYAML_get_object_item(tree, "health_sensitivity"); + if (sen) + rc = lustre_lnet_config_hsensitivity(sen->cy_valueint, + seq_no ? seq_no->cy_valueint + : -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); + + recov_limit = cYAML_get_object_item(tree, "recovery_limit"); + if (recov_limit) + rc = lustre_lnet_config_recovery_limit(recov_limit->cy_valueint, + seq_no ? seq_no->cy_valueint + : -1, + err_rc); + + return rc; +} + +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, *drop_asym_route; + int rc = 0; + + seq_no = cYAML_get_object_item(tree, "seq_no"); + max_intf = cYAML_get_object_item(tree, "max_intf"); + if (max_intf) + rc = lustre_lnet_config_max_intf(LNET_INTERFACES_MAX_DEFAULT, + seq_no ? seq_no->cy_valueint + : -1, + err_rc); + + numa = cYAML_get_object_item(tree, "numa_range"); + if (numa) + rc = lustre_lnet_config_numa_range(0, + seq_no ? seq_no->cy_valueint + : -1, + err_rc); + + /* peer discovery is enabled by default */ + discovery = cYAML_get_object_item(tree, "discovery"); + if (discovery) + rc = lustre_lnet_config_discovery(1, + seq_no ? seq_no->cy_valueint + : -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; +} + +static int handle_yaml_show_global_settings(struct cYAML *tree, + struct cYAML **show_rc, + struct cYAML **err_rc) +{ + struct cYAML *max_intf, *numa, *discovery, *retry, *tto, *seq_no, + *sen, *recov, *rsen, *drop_asym_route, *rsp_tracking, + *recov_limit; + int rc = 0; + + seq_no = cYAML_get_object_item(tree, "seq_no"); + max_intf = cYAML_get_object_item(tree, "max_intf"); + if (max_intf) + rc = lustre_lnet_show_max_intf(seq_no ? seq_no->cy_valueint + : -1, + show_rc, err_rc); + + numa = cYAML_get_object_item(tree, "numa_range"); + if (numa) + rc = lustre_lnet_show_numa_range(seq_no ? seq_no->cy_valueint + : -1, + show_rc, err_rc); + + discovery = cYAML_get_object_item(tree, "discovery"); + if (discovery) + rc = lustre_lnet_show_discovery(seq_no ? seq_no->cy_valueint + : -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 + : -1, + show_rc, err_rc); + + tto = cYAML_get_object_item(tree, "transaction_timeout"); + if (tto) + rc = lustre_lnet_show_transaction_to(seq_no ? seq_no->cy_valueint + : -1, + show_rc, err_rc); + + sen = cYAML_get_object_item(tree, "health_sensitivity"); + if (sen) + rc = lustre_lnet_show_hsensitivity(seq_no ? seq_no->cy_valueint + : -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); + + recov_limit = cYAML_get_object_item(tree, "recovery_limit"); + if (recov_limit) + rc = lustre_lnet_show_recovery_limit(seq_no ? + seq_no->cy_valueint : + -1, + show_rc, err_rc); + + return rc; +} + +static int handle_yaml_ping(struct cYAML *tree, struct cYAML **show_rc, + struct cYAML **err_rc) +{ + struct cYAML *seq_no, *nid, *timeout; + + seq_no = cYAML_get_object_item(tree, "seq_no"); + nid = cYAML_get_object_item(tree, "primary nid"); + timeout = cYAML_get_object_item(tree, "timeout"); + + return lustre_lnet_ping_nid((nid) ? nid->cy_valuestring : NULL, + (timeout) ? timeout->cy_valueint : 1000, + (seq_no) ? seq_no->cy_valueint : -1, + show_rc, err_rc); +} + +static int handle_yaml_discover(struct cYAML *tree, struct cYAML **show_rc, + struct cYAML **err_rc) +{ + struct cYAML *seq_no, *nid, *force; + + seq_no = cYAML_get_object_item(tree, "seq_no"); + nid = cYAML_get_object_item(tree, "primary nid"); + force = cYAML_get_object_item(tree, "force"); + + return lustre_lnet_discover_nid((nid) ? nid->cy_valuestring : NULL, + (force) ? force->cy_valueint : 0, + (seq_no) ? seq_no->cy_valueint : -1, + show_rc, err_rc); +} + +static int handle_yaml_no_op() +{ + return LUSTRE_CFG_RC_NO_ERR; +} + struct lookup_cmd_hdlr_tbl { char *name; cmd_handler_t cb; @@ -3076,25 +4805,53 @@ static struct lookup_cmd_hdlr_tbl lookup_config_tbl[] = { { .name = "peer", .cb = handle_yaml_config_peer }, { .name = "routing", .cb = handle_yaml_config_routing }, { .name = "buffers", .cb = handle_yaml_config_buffers }, + { .name = "statistics", .cb = handle_yaml_no_op }, + { .name = "global", .cb = handle_yaml_config_global_settings}, { .name = "numa", .cb = handle_yaml_config_numa }, + { .name = "ping", .cb = handle_yaml_no_op }, + { .name = "discover", .cb = handle_yaml_no_op }, { .name = NULL } }; static struct lookup_cmd_hdlr_tbl lookup_del_tbl[] = { { .name = "route", .cb = handle_yaml_del_route }, { .name = "net", .cb = handle_yaml_del_ni }, + { .name = "ip2nets", .cb = handle_yaml_no_op }, { .name = "peer", .cb = handle_yaml_del_peer }, { .name = "routing", .cb = handle_yaml_del_routing }, + { .name = "buffers", .cb = handle_yaml_no_op }, + { .name = "statistics", .cb = handle_yaml_no_op }, + { .name = "global", .cb = handle_yaml_del_global_settings}, { .name = "numa", .cb = handle_yaml_del_numa }, + { .name = "ping", .cb = handle_yaml_no_op }, + { .name = "discover", .cb = handle_yaml_no_op }, { .name = NULL } }; static struct lookup_cmd_hdlr_tbl lookup_show_tbl[] = { { .name = "route", .cb = handle_yaml_show_route }, { .name = "net", .cb = handle_yaml_show_net }, - { .name = "buffers", .cb = handle_yaml_show_routing }, - { .name = "routing", .cb = handle_yaml_show_routing }, { .name = "peer", .cb = handle_yaml_show_peers }, + { .name = "ip2nets", .cb = handle_yaml_no_op }, + { .name = "routing", .cb = handle_yaml_show_routing }, + { .name = "buffers", .cb = handle_yaml_show_routing }, { .name = "statistics", .cb = handle_yaml_show_stats }, + { .name = "global", .cb = handle_yaml_show_global_settings}, { .name = "numa", .cb = handle_yaml_show_numa }, + { .name = "ping", .cb = handle_yaml_no_op }, + { .name = "discover", .cb = handle_yaml_no_op }, + { .name = NULL } }; + +static struct lookup_cmd_hdlr_tbl lookup_exec_tbl[] = { + { .name = "route", .cb = handle_yaml_no_op }, + { .name = "net", .cb = handle_yaml_no_op }, + { .name = "peer", .cb = handle_yaml_no_op }, + { .name = "ip2nets", .cb = handle_yaml_no_op }, + { .name = "routing", .cb = handle_yaml_no_op }, + { .name = "buffers", .cb = handle_yaml_no_op }, + { .name = "statistics", .cb = handle_yaml_no_op }, + { .name = "global", .cb = handle_yaml_no_op }, + { .name = "numa", .cb = handle_yaml_no_op }, + { .name = "ping", .cb = handle_yaml_ping }, + { .name = "discover", .cb = handle_yaml_discover }, { .name = NULL } }; static cmd_handler_t lookup_fn(char *key, @@ -3176,3 +4933,8 @@ int lustre_yaml_show(char *f, struct cYAML **show_rc, struct cYAML **err_rc) show_rc, err_rc); } +int lustre_yaml_exec(char *f, struct cYAML **show_rc, struct cYAML **err_rc) +{ + return lustre_yaml_cb_helper(f, lookup_exec_tbl, + show_rc, err_rc); +}