X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lnet%2Flnet%2Fapi-ni.c;h=2297772a2d9d8613f327cc69c408f8934d92f7bf;hp=9105eb108ad7250a09aab8d9db75f1044a32e413;hb=1a2bf911b9793648de3abbc88b9f77eb2237dc50;hpb=d277f2ae95d0d0580059561706face3accfe3618 diff --git a/lnet/lnet/api-ni.c b/lnet/lnet/api-ni.c index 9105eb1..2297772 100644 --- a/lnet/lnet/api-ni.c +++ b/lnet/lnet/api-ni.c @@ -74,7 +74,7 @@ MODULE_PARM_DESC(rnet_htable_size, "size of remote network hash table"); static int use_tcp_bonding = false; module_param(use_tcp_bonding, int, 0444); MODULE_PARM_DESC(use_tcp_bonding, - "Set to 1 to use socklnd bonding. 0 to use Multi-Rail"); + "use_tcp_bonding parameter has been deprecated"); unsigned int lnet_numa_range = 0; module_param(lnet_numa_range, uint, 0444); @@ -124,6 +124,11 @@ module_param_call(lnet_recovery_interval, recovery_interval_set, param_get_int, MODULE_PARM_DESC(lnet_recovery_interval, "Interval to recover unhealthy interfaces in seconds"); +unsigned int lnet_recovery_limit; +module_param(lnet_recovery_limit, uint, 0644); +MODULE_PARM_DESC(lnet_recovery_limit, + "How long to attempt recovery of unhealthy peer interfaces in seconds. Set to 0 to allow indefinite recovery"); + static int lnet_interfaces_max = LNET_INTERFACES_MAX_DEFAULT; static int intf_max_set(const char *val, cfs_kernel_param_arg_t *kp); @@ -182,10 +187,8 @@ module_param_call(lnet_drop_asym_route, drop_asym_route_set, param_get_int, MODULE_PARM_DESC(lnet_drop_asym_route, "Set to 1 to drop asymmetrical route messages."); -#define LNET_TRANSACTION_TIMEOUT_NO_HEALTH_DEFAULT 50 -#define LNET_TRANSACTION_TIMEOUT_HEALTH_DEFAULT 50 - -unsigned lnet_transaction_timeout = LNET_TRANSACTION_TIMEOUT_HEALTH_DEFAULT; +#define LNET_TRANSACTION_TIMEOUT_DEFAULT 50 +unsigned int lnet_transaction_timeout = LNET_TRANSACTION_TIMEOUT_DEFAULT; static int transaction_to_set(const char *val, cfs_kernel_param_arg_t *kp); #ifdef HAVE_KERNEL_PARAM_OPS static struct kernel_param_ops param_ops_transaction_timeout = { @@ -203,8 +206,8 @@ module_param_call(lnet_transaction_timeout, transaction_to_set, param_get_int, MODULE_PARM_DESC(lnet_transaction_timeout, "Maximum number of seconds to wait for a peer response."); -#define LNET_RETRY_COUNT_HEALTH_DEFAULT 2 -unsigned lnet_retry_count = LNET_RETRY_COUNT_HEALTH_DEFAULT; +#define LNET_RETRY_COUNT_DEFAULT 2 +unsigned int lnet_retry_count = LNET_RETRY_COUNT_DEFAULT; static int retry_count_set(const char *val, cfs_kernel_param_arg_t *kp); #ifdef HAVE_KERNEL_PARAM_OPS static struct kernel_param_ops param_ops_retry_count = { @@ -222,8 +225,34 @@ module_param_call(lnet_retry_count, retry_count_set, param_get_int, MODULE_PARM_DESC(lnet_retry_count, "Maximum number of times to retry transmitting a message"); +unsigned int lnet_response_tracking = 3; +static int response_tracking_set(const char *val, cfs_kernel_param_arg_t *kp); + +#ifdef HAVE_KERNEL_PARAM_OPS +static struct kernel_param_ops param_ops_response_tracking = { + .set = response_tracking_set, + .get = param_get_int, +}; + +#define param_check_response_tracking(name, p) \ + __param_check(name, p, int) +module_param(lnet_response_tracking, response_tracking, 0644); +#else +module_param_call(lnet_response_tracking, response_tracking_set, param_get_int, + &lnet_response_tracking, 0644); +#endif +MODULE_PARM_DESC(lnet_response_tracking, + "(0|1|2|3) LNet Internal Only|GET Reply only|PUT ACK only|Full Tracking (default)"); + +#define LNET_LND_TIMEOUT_DEFAULT ((LNET_TRANSACTION_TIMEOUT_DEFAULT - 1) / \ + (LNET_RETRY_COUNT_DEFAULT + 1)) +unsigned int lnet_lnd_timeout = LNET_LND_TIMEOUT_DEFAULT; +static void lnet_set_lnd_timeout(void) +{ + lnet_lnd_timeout = (lnet_transaction_timeout - 1) / + (lnet_retry_count + 1); +} -unsigned lnet_lnd_timeout = LNET_LND_DEFAULT_TIMEOUT; unsigned int lnet_current_net_count; /* @@ -267,21 +296,9 @@ sensitivity_set(const char *val, cfs_kernel_param_arg_t *kp) return -EINVAL; } - /* - * if we're turning on health then use the health timeout - * defaults. - */ - if (*sensitivity == 0 && value != 0) { - lnet_transaction_timeout = LNET_TRANSACTION_TIMEOUT_HEALTH_DEFAULT; - lnet_retry_count = LNET_RETRY_COUNT_HEALTH_DEFAULT; - /* - * if we're turning off health then use the no health timeout - * default. - */ - } else if (*sensitivity != 0 && value == 0) { - lnet_transaction_timeout = - LNET_TRANSACTION_TIMEOUT_NO_HEALTH_DEFAULT; + if (*sensitivity != 0 && value == 0 && lnet_retry_count != 0) { lnet_retry_count = 0; + lnet_set_lnd_timeout(); } *sensitivity = value; @@ -326,7 +343,7 @@ static int discovery_set(const char *val, cfs_kernel_param_arg_t *kp) { int rc; - unsigned *discovery = (unsigned *)kp->arg; + unsigned *discovery_off = (unsigned *)kp->arg; unsigned long value; struct lnet_ping_buffer *pbuf; @@ -344,14 +361,20 @@ discovery_set(const char *val, cfs_kernel_param_arg_t *kp) */ mutex_lock(&the_lnet.ln_api_mutex); - if (value == *discovery) { + if (value == *discovery_off) { mutex_unlock(&the_lnet.ln_api_mutex); return 0; } - *discovery = value; - + /* + * We still want to set the discovery value even when LNet is not + * running. This is the case when LNet is being loaded and we want + * the module parameters to take effect. Otherwise if we're + * changing the value dynamically, we want to set it after + * updating the peers + */ if (the_lnet.ln_state != LNET_STATE_RUNNING) { + *discovery_off = value; mutex_unlock(&the_lnet.ln_api_mutex); return 0; } @@ -365,7 +388,10 @@ discovery_set(const char *val, cfs_kernel_param_arg_t *kp) pbuf->pb_info.pi_features |= LNET_PING_FEAT_DISCOVERY; lnet_net_unlock(LNET_LOCK_EX); - lnet_push_update_to_peers(1); + /* only send a push when we're turning off discovery */ + if (*discovery_off <= 0 && value > 0) + lnet_push_update_to_peers(1); + *discovery_off = value; mutex_unlock(&the_lnet.ln_api_mutex); @@ -423,7 +449,7 @@ transaction_to_set(const char *val, cfs_kernel_param_arg_t *kp) */ mutex_lock(&the_lnet.ln_api_mutex); - if (value < lnet_retry_count || value == 0) { + if (value <= lnet_retry_count || value == 0) { mutex_unlock(&the_lnet.ln_api_mutex); CERROR("Invalid value for lnet_transaction_timeout (%lu). " "Has to be greater than lnet_retry_count (%u)\n", @@ -437,10 +463,10 @@ transaction_to_set(const char *val, cfs_kernel_param_arg_t *kp) } *transaction_to = value; - if (lnet_retry_count == 0) - lnet_lnd_timeout = value; - else - lnet_lnd_timeout = value / lnet_retry_count; + /* Update the lnet_lnd_timeout now that we've modified the + * transaction timeout + */ + lnet_set_lnd_timeout(); mutex_unlock(&the_lnet.ln_api_mutex); @@ -466,9 +492,9 @@ retry_count_set(const char *val, cfs_kernel_param_arg_t *kp) */ mutex_lock(&the_lnet.ln_api_mutex); - if (lnet_health_sensitivity == 0) { + if (lnet_health_sensitivity == 0 && value > 0) { mutex_unlock(&the_lnet.ln_api_mutex); - CERROR("Can not set retry_count when health feature is turned off\n"); + CERROR("Can not set lnet_retry_count when health feature is turned off\n"); return -EINVAL; } @@ -482,10 +508,10 @@ retry_count_set(const char *val, cfs_kernel_param_arg_t *kp) *retry_count = value; - if (value == 0) - lnet_lnd_timeout = lnet_transaction_timeout; - else - lnet_lnd_timeout = lnet_transaction_timeout / value; + /* Update the lnet_lnd_timeout now that we've modified the + * retry count + */ + lnet_set_lnd_timeout(); mutex_unlock(&the_lnet.ln_api_mutex); @@ -514,17 +540,40 @@ intf_max_set(const char *val, cfs_kernel_param_arg_t *kp) return 0; } -static char * +static int +response_tracking_set(const char *val, cfs_kernel_param_arg_t *kp) +{ + int rc; + unsigned long new_value; + + rc = kstrtoul(val, 0, &new_value); + if (rc) { + CERROR("Invalid value for 'lnet_response_tracking'\n"); + return -EINVAL; + } + + if (new_value < 0 || new_value > 3) { + CWARN("Invalid value (%lu) for 'lnet_response_tracking'\n", + new_value); + return -EINVAL; + } + + lnet_response_tracking = new_value; + + return 0; +} + +static const char * lnet_get_routes(void) { return routes; } -static char * +static const char * lnet_get_networks(void) { - char *nets; - int rc; + const char *nets; + int rc; if (*networks != 0 && *ip2nets != 0) { LCONSOLE_ERROR_MSG(0x101, "Please specify EITHER 'networks' or " @@ -866,16 +915,17 @@ lnet_unregister_lnd(const struct lnet_lnd *lnd) } EXPORT_SYMBOL(lnet_unregister_lnd); -void -lnet_counters_get_common(struct lnet_counters_common *common) +static void +lnet_counters_get_common_locked(struct lnet_counters_common *common) { struct lnet_counters *ctr; int i; + /* FIXME !!! Their is no assert_lnet_net_locked() to ensure this + * actually called under the protection of the lnet_net_lock. + */ memset(common, 0, sizeof(*common)); - lnet_net_lock(LNET_LOCK_EX); - cfs_percpt_for_each(ctr, i, the_lnet.ln_counters) { common->lcc_msgs_max += ctr->lct_common.lcc_msgs_max; common->lcc_msgs_alloc += ctr->lct_common.lcc_msgs_alloc; @@ -889,23 +939,33 @@ lnet_counters_get_common(struct lnet_counters_common *common) common->lcc_route_length += ctr->lct_common.lcc_route_length; common->lcc_drop_length += ctr->lct_common.lcc_drop_length; } +} + +void +lnet_counters_get_common(struct lnet_counters_common *common) +{ + lnet_net_lock(LNET_LOCK_EX); + lnet_counters_get_common_locked(common); lnet_net_unlock(LNET_LOCK_EX); } EXPORT_SYMBOL(lnet_counters_get_common); -void +int lnet_counters_get(struct lnet_counters *counters) { struct lnet_counters *ctr; struct lnet_counters_health *health = &counters->lct_health; - int i; + int i, rc = 0; memset(counters, 0, sizeof(*counters)); - lnet_counters_get_common(&counters->lct_common); - lnet_net_lock(LNET_LOCK_EX); + if (the_lnet.ln_state != LNET_STATE_RUNNING) + GOTO(out_unlock, rc = -ENODEV); + + lnet_counters_get_common_locked(&counters->lct_common); + cfs_percpt_for_each(ctr, i, the_lnet.ln_counters) { health->lch_rst_alloc += ctr->lct_health.lch_rst_alloc; health->lch_resend_count += ctr->lct_health.lch_resend_count; @@ -932,7 +992,9 @@ lnet_counters_get(struct lnet_counters *counters) health->lch_network_timeout_count += ctr->lct_health.lch_network_timeout_count; } +out_unlock: lnet_net_unlock(LNET_LOCK_EX); + return rc; } EXPORT_SYMBOL(lnet_counters_get); @@ -944,9 +1006,12 @@ lnet_counters_reset(void) lnet_net_lock(LNET_LOCK_EX); + if (the_lnet.ln_state != LNET_STATE_RUNNING) + goto avoid_reset; + cfs_percpt_for_each(counters, i, the_lnet.ln_counters) memset(counters, 0, sizeof(struct lnet_counters)); - +avoid_reset: lnet_net_unlock(LNET_LOCK_EX); } @@ -1164,7 +1229,7 @@ lnet_prepare(lnet_pid_t requested_pid) INIT_LIST_HEAD(&the_lnet.ln_mt_localNIRecovq); INIT_LIST_HEAD(&the_lnet.ln_mt_peerNIRecovq); init_waitqueue_head(&the_lnet.ln_dc_waitq); - the_lnet.ln_mt_eq = NULL; + the_lnet.ln_mt_handler = NULL; init_completion(&the_lnet.ln_started); rc = lnet_slab_setup(); @@ -1232,8 +1297,6 @@ lnet_prepare(lnet_pid_t requested_pid) static int lnet_unprepare (void) { - int rc; - /* NB no LNET_LOCK since this is the last reference. All LND instances * have shut down already, so it is safe to unlink and free all * descriptors, even those that appear committed to a network op (eg MD @@ -1250,11 +1313,8 @@ lnet_unprepare (void) the_lnet.ln_mt_zombie_rstqs = NULL; } - if (the_lnet.ln_mt_eq) { - rc = LNetEQFree(the_lnet.ln_mt_eq); - the_lnet.ln_mt_eq = NULL; - LASSERT(rc == 0); - } + lnet_assert_handler_unused(the_lnet.ln_mt_handler); + the_lnet.ln_mt_handler = NULL; lnet_portals_destroy(); @@ -1501,6 +1561,7 @@ lnet_ping_buffer_alloc(int nnis, gfp_t gfp) LIBCFS_ALLOC_GFP(pbuf, LNET_PING_BUFFER_SIZE(nnis), gfp); if (pbuf) { pbuf->pb_nnis = nnis; + pbuf->pb_needs_post = false; atomic_set(&pbuf->pb_refcnt, 1); } @@ -1627,7 +1688,7 @@ lnet_ping_info_validate(struct lnet_ping_info *pinfo) /* Loopback is guaranteed to be present */ if (pinfo->pi_nnis < 1 || pinfo->pi_nnis > lnet_interfaces_max) return -ERANGE; - if (LNET_NETTYP(LNET_NIDNET(LNET_PING_INFO_LONI(pinfo))) != LOLND) + if (LNET_PING_INFO_LONI(pinfo) != LNET_NID_LO_0) return -EPROTO; return 0; } @@ -1657,7 +1718,7 @@ lnet_ping_target_destroy(void) static void lnet_ping_target_event_handler(struct lnet_event *event) { - struct lnet_ping_buffer *pbuf = event->md.user_ptr; + struct lnet_ping_buffer *pbuf = event->md_user_ptr; if (event->unlinked) lnet_ping_buffer_decref(pbuf); @@ -1674,17 +1735,11 @@ lnet_ping_target_setup(struct lnet_ping_buffer **ppbuf, }; struct lnet_me *me; struct lnet_md md = { NULL }; - int rc, rc2; - - if (set_eq) { - the_lnet.ln_ping_target_eq = - LNetEQAlloc(lnet_ping_target_event_handler); - if (IS_ERR(the_lnet.ln_ping_target_eq)) { - rc = PTR_ERR(the_lnet.ln_ping_target_eq); - CERROR("Can't allocate ping buffer EQ: %d\n", rc); - return rc; - } - } + int rc; + + if (set_eq) + the_lnet.ln_ping_target_handler = + lnet_ping_target_event_handler; *ppbuf = lnet_ping_target_create(ni_count); if (*ppbuf == NULL) { @@ -1709,29 +1764,23 @@ lnet_ping_target_setup(struct lnet_ping_buffer **ppbuf, md.max_size = 0; md.options = LNET_MD_OP_GET | LNET_MD_TRUNCATE | LNET_MD_MANAGE_REMOTE; - md.eq_handle = the_lnet.ln_ping_target_eq; + md.handler = the_lnet.ln_ping_target_handler; md.user_ptr = *ppbuf; - rc = LNetMDAttach(me, md, LNET_RETAIN, ping_mdh); + rc = LNetMDAttach(me, &md, LNET_RETAIN, ping_mdh); if (rc != 0) { CERROR("Can't attach ping target MD: %d\n", rc); - goto fail_unlink_ping_me; + goto fail_decref_ping_buffer; } lnet_ping_buffer_addref(*ppbuf); return 0; -fail_unlink_ping_me: - LNetMEUnlink(me); fail_decref_ping_buffer: LASSERT(atomic_read(&(*ppbuf)->pb_refcnt) == 1); lnet_ping_buffer_decref(*ppbuf); *ppbuf = NULL; fail_free_eq: - if (set_eq) { - rc2 = LNetEQFree(the_lnet.ln_ping_target_eq); - LASSERT(rc2 == 0); - } return rc; } @@ -1743,10 +1792,9 @@ lnet_ping_md_unlink(struct lnet_ping_buffer *pbuf, LNetInvalidateMDHandle(ping_mdh); /* NB the MD could be busy; this just starts the unlink */ - while (atomic_read(&pbuf->pb_refcnt) > 1) { - CDEBUG(D_NET, "Still waiting for ping data MD to unlink\n"); - schedule_timeout_uninterruptible(cfs_time_seconds(1)); - } + wait_var_event_warning(&pbuf->pb_refcnt, + atomic_read(&pbuf->pb_refcnt) <= 1, + "Still waiting for ping data MD to unlink\n"); } static void @@ -1833,67 +1881,46 @@ lnet_ping_target_update(struct lnet_ping_buffer *pbuf, static void lnet_ping_target_fini(void) { - int rc; - lnet_ping_md_unlink(the_lnet.ln_ping_target, &the_lnet.ln_ping_target_md); - rc = LNetEQFree(the_lnet.ln_ping_target_eq); - LASSERT(rc == 0); - + lnet_assert_handler_unused(the_lnet.ln_ping_target_handler); lnet_ping_target_destroy(); } /* Resize the push target. */ int lnet_push_target_resize(void) { - struct lnet_process_id id = { LNET_NID_ANY, LNET_PID_ANY }; - struct lnet_md md = { NULL }; - struct lnet_me *me; struct lnet_handle_md mdh; struct lnet_handle_md old_mdh; struct lnet_ping_buffer *pbuf; struct lnet_ping_buffer *old_pbuf; - int nnis = the_lnet.ln_push_target_nnis; + int nnis; int rc; +again: + nnis = the_lnet.ln_push_target_nnis; if (nnis <= 0) { - rc = -EINVAL; - goto fail_return; + CDEBUG(D_NET, "Invalid nnis %d\n", nnis); + return -EINVAL; } -again: + + /* NB: lnet_ping_buffer_alloc() sets pbuf refcount to 1. That ref is + * dropped when we need to resize again (see "old_pbuf" below) or when + * LNet is shutdown (see lnet_push_target_fini()) + */ pbuf = lnet_ping_buffer_alloc(nnis, GFP_NOFS); if (!pbuf) { - rc = -ENOMEM; - goto fail_return; - } - - me = LNetMEAttach(LNET_RESERVED_PORTAL, id, - LNET_PROTO_PING_MATCHBITS, 0, - LNET_UNLINK, LNET_INS_AFTER); - - if (IS_ERR(me)) { - rc = PTR_ERR(me); - CERROR("Can't create push target ME: %d\n", rc); - goto fail_decref_pbuf; + CDEBUG(D_NET, "Can't allocate pbuf for nnis %d\n", nnis); + return -ENOMEM; } - /* initialize md content */ - md.start = &pbuf->pb_info; - md.length = LNET_PING_INFO_SIZE(nnis); - md.threshold = LNET_MD_THRESH_INF; - md.max_size = 0; - md.options = LNET_MD_OP_PUT | LNET_MD_TRUNCATE | - LNET_MD_MANAGE_REMOTE; - md.user_ptr = pbuf; - md.eq_handle = the_lnet.ln_push_target_eq; - - rc = LNetMDAttach(me, md, LNET_RETAIN, &mdh); + rc = lnet_push_target_post(pbuf, &mdh); if (rc) { - CERROR("Can't attach push MD: %d\n", rc); - goto fail_unlink_me; + CDEBUG(D_NET, "Failed to post push target: %d\n", rc); + lnet_ping_buffer_decref(pbuf); + return rc; } - lnet_ping_buffer_addref(pbuf); lnet_net_lock(LNET_LOCK_EX); old_pbuf = the_lnet.ln_push_target; @@ -1904,34 +1931,81 @@ again: if (old_pbuf) { LNetMDUnlink(old_mdh); + /* Drop ref set by lnet_ping_buffer_alloc() */ lnet_ping_buffer_decref(old_pbuf); } + /* Received another push or reply that requires a larger buffer */ if (nnis < the_lnet.ln_push_target_nnis) goto again; CDEBUG(D_NET, "nnis %d success\n", nnis); - return 0; +} -fail_unlink_me: - LNetMEUnlink(me); -fail_decref_pbuf: - lnet_ping_buffer_decref(pbuf); -fail_return: - CDEBUG(D_NET, "nnis %d error %d\n", nnis, rc); - return rc; +int lnet_push_target_post(struct lnet_ping_buffer *pbuf, + struct lnet_handle_md *mdhp) +{ + struct lnet_process_id id = { LNET_NID_ANY, LNET_PID_ANY }; + struct lnet_md md = { NULL }; + struct lnet_me *me; + int rc; + + me = LNetMEAttach(LNET_RESERVED_PORTAL, id, + LNET_PROTO_PING_MATCHBITS, 0, + LNET_UNLINK, LNET_INS_AFTER); + if (IS_ERR(me)) { + rc = PTR_ERR(me); + CERROR("Can't create push target ME: %d\n", rc); + return rc; + } + + pbuf->pb_needs_post = false; + + /* This reference is dropped by lnet_push_target_event_handler() */ + lnet_ping_buffer_addref(pbuf); + + /* initialize md content */ + md.start = &pbuf->pb_info; + md.length = LNET_PING_INFO_SIZE(pbuf->pb_nnis); + md.threshold = 1; + md.max_size = 0; + md.options = LNET_MD_OP_PUT | LNET_MD_TRUNCATE; + md.user_ptr = pbuf; + md.handler = the_lnet.ln_push_target_handler; + + rc = LNetMDAttach(me, &md, LNET_UNLINK, mdhp); + if (rc) { + CERROR("Can't attach push MD: %d\n", rc); + lnet_ping_buffer_decref(pbuf); + pbuf->pb_needs_post = true; + return rc; + } + + CDEBUG(D_NET, "posted push target %p\n", pbuf); + + return 0; } static void lnet_push_target_event_handler(struct lnet_event *ev) { - struct lnet_ping_buffer *pbuf = ev->md.user_ptr; + struct lnet_ping_buffer *pbuf = ev->md_user_ptr; + + CDEBUG(D_NET, "type %d status %d unlinked %d\n", ev->type, ev->status, + ev->unlinked); if (pbuf->pb_info.pi_magic == __swab32(LNET_PROTO_PING_MAGIC)) lnet_swap_pinginfo(pbuf); + if (ev->type == LNET_EVENT_UNLINK) { + /* Drop ref added by lnet_push_target_post() */ + lnet_ping_buffer_decref(pbuf); + return; + } + lnet_peer_push_event(ev); if (ev->unlinked) + /* Drop ref added by lnet_push_target_post */ lnet_ping_buffer_decref(pbuf); } @@ -1943,13 +2017,11 @@ static int lnet_push_target_init(void) if (the_lnet.ln_push_target) return -EALREADY; - the_lnet.ln_push_target_eq = - LNetEQAlloc(lnet_push_target_event_handler); - if (IS_ERR(the_lnet.ln_push_target_eq)) { - rc = PTR_ERR(the_lnet.ln_push_target_eq); - CERROR("Can't allocated push target EQ: %d\n", rc); - return rc; - } + the_lnet.ln_push_target_handler = + lnet_push_target_event_handler; + + rc = LNetSetLazyPortal(LNET_RESERVED_PORTAL); + LASSERT(rc == 0); /* Start at the required minimum, we'll enlarge if required. */ the_lnet.ln_push_target_nnis = LNET_INTERFACES_MIN; @@ -1957,8 +2029,8 @@ static int lnet_push_target_init(void) rc = lnet_push_target_resize(); if (rc) { - LNetEQFree(the_lnet.ln_push_target_eq); - the_lnet.ln_push_target_eq = NULL; + LNetClearLazyPortal(LNET_RESERVED_PORTAL); + the_lnet.ln_push_target_handler = NULL; } return rc; @@ -1975,17 +2047,18 @@ static void lnet_push_target_fini(void) LNetInvalidateMDHandle(&the_lnet.ln_push_target_md); /* Wait for the unlink to complete. */ - while (atomic_read(&the_lnet.ln_push_target->pb_refcnt) > 1) { - CDEBUG(D_NET, "Still waiting for ping data MD to unlink\n"); - schedule_timeout_uninterruptible(cfs_time_seconds(1)); - } + wait_var_event_warning(&the_lnet.ln_push_target->pb_refcnt, + atomic_read(&the_lnet.ln_push_target->pb_refcnt) <= 1, + "Still waiting for ping data MD to unlink\n"); + /* Drop ref set by lnet_ping_buffer_alloc() */ lnet_ping_buffer_decref(the_lnet.ln_push_target); the_lnet.ln_push_target = NULL; the_lnet.ln_push_target_nnis = 0; - LNetEQFree(the_lnet.ln_push_target_eq); - the_lnet.ln_push_target_eq = NULL; + LNetClearLazyPortal(LNET_RESERVED_PORTAL); + lnet_assert_handler_unused(the_lnet.ln_push_target_handler); + the_lnet.ln_push_target_handler = NULL; } static int @@ -2046,7 +2119,13 @@ lnet_clear_zombies_nis_locked(struct lnet_net *net) } if (!list_empty(&ni->ni_netlist)) { + /* Unlock mutex while waiting to allow other + * threads to read the LNet state and fall through + * to avoid deadlock + */ lnet_net_unlock(LNET_LOCK_EX); + mutex_unlock(&the_lnet.ln_api_mutex); + ++i; if ((i & (-i)) == i) { CDEBUG(D_WARNING, @@ -2054,6 +2133,8 @@ lnet_clear_zombies_nis_locked(struct lnet_net *net) libcfs_nid2str(ni->ni_nid)); } schedule_timeout_uninterruptible(cfs_time_seconds(1)); + + mutex_lock(&the_lnet.ln_api_mutex); lnet_net_lock(LNET_LOCK_EX); continue; } @@ -2603,6 +2684,9 @@ LNetNIInit(lnet_pid_t requested_pid) goto err_empty_list; } + if (use_tcp_bonding) + CWARN("'use_tcp_bonding' option has been deprecated. See LU-13641\n"); + /* If LNet is being initialized via DLC it is possible * that the user requests not to load module parameters (ones which * are supported by DLC) on initialization. Therefore, make sure not @@ -2645,12 +2729,7 @@ LNetNIInit(lnet_pid_t requested_pid) lnet_ping_target_update(pbuf, ping_mdh); - the_lnet.ln_mt_eq = LNetEQAlloc(lnet_mt_event_handler); - if (IS_ERR(the_lnet.ln_mt_eq)) { - rc = PTR_ERR(the_lnet.ln_mt_eq); - CERROR("Can't allocate monitor thread EQ: %d\n", rc); - goto err_stop_ping; - } + the_lnet.ln_mt_handler = lnet_mt_event_handler; rc = lnet_push_target_init(); if (rc != 0) @@ -2779,7 +2858,7 @@ lnet_fill_ni_info(struct lnet_ni *ni, struct lnet_ioctl_config_ni *cfg_ni, } cfg_ni->lic_nid = ni->ni_nid; - if (LNET_NETTYP(LNET_NIDNET(ni->ni_nid)) == LOLND) + if (ni->ni_nid == LNET_NID_LO_0) cfg_ni->lic_status = LNET_NI_STATUS_UP; else cfg_ni->lic_status = ni->ni_status->ns_status; @@ -2871,7 +2950,7 @@ lnet_fill_ni_info_legacy(struct lnet_ni *ni, config->cfg_config_u.cfg_net.net_peer_rtr_credits = ni->ni_net->net_tunables.lct_peer_rtr_credits; - if (LNET_NETTYP(LNET_NIDNET(ni->ni_nid)) == LOLND) + if (ni->ni_nid == LNET_NID_LO_0) net_config->ni_status = LNET_NI_STATUS_UP; else net_config->ni_status = ni->ni_status->ns_status; @@ -3065,7 +3144,6 @@ static int lnet_add_net_common(struct lnet_net *net, int rc; struct lnet_remotenet *rnet; int net_ni_count; - int num_acceptor_nets; lnet_net_lock(LNET_LOCK_EX); rnet = lnet_find_rnet_locked(net->net_id); @@ -3106,14 +3184,6 @@ static int lnet_add_net_common(struct lnet_net *net, else memset(&net->net_tunables, -1, sizeof(net->net_tunables)); - /* - * before starting this network get a count of the current TCP - * networks which require the acceptor thread running. If that - * count is == 0 before we start up this network, then we'd want to - * start up the acceptor thread after starting up this network - */ - num_acceptor_nets = lnet_count_acceptor_nets(); - net_id = net->net_id; rc = lnet_startup_lndnet(net, @@ -3131,7 +3201,7 @@ static int lnet_add_net_common(struct lnet_net *net, * Start the acceptor thread if this is the first network * being added that requires the thread. */ - if (net->net_lnd->lnd_accept && num_acceptor_nets == 0) { + if (net->net_lnd->lnd_accept) { rc = lnet_acceptor_start(); if (rc < 0) { /* shutdown the net that we just started */ @@ -3155,11 +3225,24 @@ failed: return rc; } +static void +lnet_set_tune_defaults(struct lnet_ioctl_config_lnd_tunables *tun) +{ + if (tun) { + if (!tun->lt_cmn.lct_peer_timeout) + tun->lt_cmn.lct_peer_timeout = DEFAULT_PEER_TIMEOUT; + if (!tun->lt_cmn.lct_peer_tx_credits) + tun->lt_cmn.lct_peer_tx_credits = DEFAULT_PEER_CREDITS; + if (!tun->lt_cmn.lct_max_tx_credits) + tun->lt_cmn.lct_max_tx_credits = DEFAULT_CREDITS; + } +} + static int lnet_handle_legacy_ip2nets(char *ip2nets, struct lnet_ioctl_config_lnd_tunables *tun) { struct lnet_net *net; - char *nets; + const char *nets; int rc; LIST_HEAD(net_head); @@ -3171,6 +3254,8 @@ static int lnet_handle_legacy_ip2nets(char *ip2nets, if (rc < 0) return rc; + lnet_set_tune_defaults(tun); + mutex_lock(&the_lnet.ln_api_mutex); while (!list_empty(&net_head)) { net = list_entry(net_head.next, struct lnet_net, net_list); @@ -3232,6 +3317,8 @@ int lnet_dyn_add_ni(struct lnet_ioctl_config_ni *conf) if (!ni) return -ENOMEM; + lnet_set_tune_defaults(tun); + mutex_lock(&the_lnet.ln_api_mutex); rc = lnet_add_net_common(net, tun); @@ -3284,8 +3371,7 @@ int lnet_dyn_del_ni(struct lnet_ioctl_config_ni *conf) lnet_shutdown_lndnet(net); - if (lnet_count_acceptor_nets() == 0) - lnet_acceptor_stop(); + lnet_acceptor_stop(); lnet_ping_target_update(pbuf, ping_mdh); @@ -3312,8 +3398,7 @@ int lnet_dyn_del_ni(struct lnet_ioctl_config_ni *conf) lnet_shutdown_lndni(ni); - if (lnet_count_acceptor_nets() == 0) - lnet_acceptor_stop(); + lnet_acceptor_stop(); lnet_ping_target_update(pbuf, ping_mdh); @@ -3344,7 +3429,7 @@ lnet_dyn_add_net(struct lnet_ioctl_config_data *conf) LIST_HEAD(net_head); int rc; struct lnet_ioctl_config_lnd_tunables tun; - char *nets = conf->cfg_config_u.cfg_net.net_intf; + const char *nets = conf->cfg_config_u.cfg_net.net_intf; /* Create a net/ni structures for the network string */ rc = lnet_parse_networks(&net_head, nets, use_tcp_bonding); @@ -3366,13 +3451,16 @@ lnet_dyn_add_net(struct lnet_ioctl_config_data *conf) memset(&tun, 0, sizeof(tun)); tun.lt_cmn.lct_peer_timeout = - conf->cfg_config_u.cfg_net.net_peer_timeout; + (!conf->cfg_config_u.cfg_net.net_peer_timeout) ? DEFAULT_PEER_TIMEOUT : + conf->cfg_config_u.cfg_net.net_peer_timeout; tun.lt_cmn.lct_peer_tx_credits = - conf->cfg_config_u.cfg_net.net_peer_tx_credits; + (!conf->cfg_config_u.cfg_net.net_peer_tx_credits) ? DEFAULT_PEER_CREDITS : + conf->cfg_config_u.cfg_net.net_peer_tx_credits; tun.lt_cmn.lct_peer_rtr_credits = conf->cfg_config_u.cfg_net.net_peer_rtr_credits; tun.lt_cmn.lct_max_tx_credits = - conf->cfg_config_u.cfg_net.net_max_tx_credits; + (!conf->cfg_config_u.cfg_net.net_max_tx_credits) ? DEFAULT_CREDITS : + conf->cfg_config_u.cfg_net.net_max_tx_credits; rc = lnet_add_net_common(net, &tun); @@ -3423,8 +3511,7 @@ lnet_dyn_del_net(__u32 net_id) lnet_shutdown_lndnet(net); - if (lnet_count_acceptor_nets() == 0) - lnet_acceptor_stop(); + lnet_acceptor_stop(); lnet_ping_target_update(pbuf, ping_mdh); @@ -3680,9 +3767,9 @@ LNetCtl(unsigned int cmd, void *arg) return -EINVAL; mutex_lock(&the_lnet.ln_api_mutex); - lnet_counters_get(&lnet_stats->st_cntrs); + rc = lnet_counters_get(&lnet_stats->st_cntrs); mutex_unlock(&the_lnet.ln_api_mutex); - return 0; + return rc; } case IOC_LIBCFS_CONFIG_RTR: @@ -4086,7 +4173,7 @@ struct ping_data { static void lnet_ping_event_handler(struct lnet_event *event) { - struct ping_data *pd = event->md.user_ptr; + struct ping_data *pd = event->md_user_ptr; CDEBUG(D_NET, "ping event (%d %d)%s\n", event->type, event->status, @@ -4106,7 +4193,6 @@ lnet_ping_event_handler(struct lnet_event *event) static int lnet_ping(struct lnet_process_id id, signed long timeout, struct lnet_process_id __user *ids, int n_ids) { - struct lnet_eq *eq; struct lnet_md md = { NULL }; struct ping_data pd = { 0 }; struct lnet_ping_buffer *pbuf; @@ -4134,13 +4220,6 @@ static int lnet_ping(struct lnet_process_id id, signed long timeout, if (!pbuf) return -ENOMEM; - eq = LNetEQAlloc(lnet_ping_event_handler); - if (IS_ERR(eq)) { - rc = PTR_ERR(eq); - CERROR("Can't allocate EQ: %d\n", rc); - goto fail_ping_buffer_decref; - } - /* initialize md content */ md.start = &pbuf->pb_info; md.length = LNET_PING_INFO_SIZE(n_ids); @@ -4148,14 +4227,14 @@ static int lnet_ping(struct lnet_process_id id, signed long timeout, md.max_size = 0; md.options = LNET_MD_TRUNCATE; md.user_ptr = &pd; - md.eq_handle = eq; + md.handler = lnet_ping_event_handler; init_completion(&pd.completion); - rc = LNetMDBind(md, LNET_UNLINK, &pd.mdh); + rc = LNetMDBind(&md, LNET_UNLINK, &pd.mdh); if (rc != 0) { CERROR("Can't bind MD: %d\n", rc); - goto fail_free_eq; + goto fail_ping_buffer_decref; } rc = LNetGet(LNET_NID_ANY, pd.mdh, id, @@ -4177,7 +4256,7 @@ static int lnet_ping(struct lnet_process_id id, signed long timeout, } if (!pd.replied) { rc = -EIO; - goto fail_free_eq; + goto fail_ping_buffer_decref; } nob = pd.rc; @@ -4188,7 +4267,7 @@ static int lnet_ping(struct lnet_process_id id, signed long timeout, if (nob < 8) { CERROR("%s: ping info too short %d\n", libcfs_id2str(id), nob); - goto fail_free_eq; + goto fail_ping_buffer_decref; } if (pbuf->pb_info.pi_magic == __swab32(LNET_PROTO_PING_MAGIC)) { @@ -4196,20 +4275,20 @@ static int lnet_ping(struct lnet_process_id id, signed long timeout, } else if (pbuf->pb_info.pi_magic != LNET_PROTO_PING_MAGIC) { CERROR("%s: Unexpected magic %08x\n", libcfs_id2str(id), pbuf->pb_info.pi_magic); - goto fail_free_eq; + goto fail_ping_buffer_decref; } if ((pbuf->pb_info.pi_features & LNET_PING_FEAT_NI_STATUS) == 0) { CERROR("%s: ping w/o NI status: 0x%x\n", libcfs_id2str(id), pbuf->pb_info.pi_features); - goto fail_free_eq; + goto fail_ping_buffer_decref; } if (nob < LNET_PING_INFO_SIZE(0)) { CERROR("%s: Short reply %d(%d min)\n", libcfs_id2str(id), nob, (int)LNET_PING_INFO_SIZE(0)); - goto fail_free_eq; + goto fail_ping_buffer_decref; } if (pbuf->pb_info.pi_nnis < n_ids) @@ -4219,7 +4298,7 @@ static int lnet_ping(struct lnet_process_id id, signed long timeout, CERROR("%s: Short reply %d(%d expected)\n", libcfs_id2str(id), nob, (int)LNET_PING_INFO_SIZE(n_ids)); - goto fail_free_eq; + goto fail_ping_buffer_decref; } rc = -EFAULT; /* if I segv in copy_to_user()... */ @@ -4229,16 +4308,10 @@ static int lnet_ping(struct lnet_process_id id, signed long timeout, tmpid.pid = pbuf->pb_info.pi_pid; tmpid.nid = pbuf->pb_info.pi_ni[i].ns_nid; if (copy_to_user(&ids[i], &tmpid, sizeof(tmpid))) - goto fail_free_eq; + goto fail_ping_buffer_decref; } rc = pbuf->pb_info.pi_nnis; - fail_free_eq: - rc2 = LNetEQFree(eq); - if (rc2 != 0) - CERROR("rc2 %d\n", rc2); - LASSERT(rc2 == 0); - fail_ping_buffer_decref: lnet_ping_buffer_decref(pbuf); return rc; @@ -4255,7 +4328,6 @@ lnet_discover(struct lnet_process_id id, __u32 force, int cpt; int i; int rc; - int max_intf = lnet_interfaces_max; if (n_ids <= 0 || id.nid == LNET_NID_ANY) @@ -4265,11 +4337,11 @@ lnet_discover(struct lnet_process_id id, __u32 force, id.pid = LNET_PID_LUSTRE; /* - * if the user buffer has more space than the max_intf - * then only fill it up to max_intf + * If the user buffer has more space than the lnet_interfaces_max, + * then only fill it up to lnet_interfaces_max. */ - if (n_ids > max_intf) - n_ids = max_intf; + if (n_ids > lnet_interfaces_max) + n_ids = lnet_interfaces_max; CFS_ALLOC_PTR_ARRAY(buf, n_ids); if (!buf) @@ -4297,11 +4369,6 @@ lnet_discover(struct lnet_process_id id, __u32 force, if (rc) goto out_decref; - /* Peer may have changed. */ - lp = lpni->lpni_peer_net->lpn_peer; - if (lp->lp_nnis < n_ids) - n_ids = lp->lp_nnis; - i = 0; p = NULL; while ((p = lnet_get_next_peer_ni_locked(lp, NULL, p)) != NULL) { @@ -4310,20 +4377,16 @@ lnet_discover(struct lnet_process_id id, __u32 force, if (++i >= n_ids) break; } + rc = i; - lnet_net_unlock(cpt); - - rc = -EFAULT; - if (copy_to_user(ids, buf, n_ids * sizeof(*buf))) - goto out_relock; - rc = n_ids; -out_relock: - lnet_net_lock(cpt); out_decref: lnet_peer_ni_decref_locked(lpni); out: lnet_net_unlock(cpt); + if (rc >= 0) + if (copy_to_user(ids, buf, rc * sizeof(*buf))) + rc = -EFAULT; CFS_FREE_PTR_ARRAY(buf, n_ids); return rc;