X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lnet%2Flnet%2Fapi-ni.c;h=05c0cb34fdba53bfac7b21de4e9eaf3f0de3c9c2;hp=966f38b3c0fb80fda93de8b5ca5cc7f4239a3ae1;hb=45722de60a8fb5fee86a07ce5e3861b31dcedf59;hpb=181c53563dc835cb96c3bdc9698d514bade47901 diff --git a/lnet/lnet/api-ni.c b/lnet/lnet/api-ni.c index 966f38b..05c0cb3 100644 --- a/lnet/lnet/api-ni.c +++ b/lnet/lnet/api-ni.c @@ -37,6 +37,9 @@ #include #include #include +#ifdef HAVE_SCHED_HEADERS +#include +#endif #include @@ -180,7 +183,7 @@ 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 10 +#define LNET_TRANSACTION_TIMEOUT_HEALTH_DEFAULT 50 unsigned lnet_transaction_timeout = LNET_TRANSACTION_TIMEOUT_HEALTH_DEFAULT; static int transaction_to_set(const char *val, cfs_kernel_param_arg_t *kp); @@ -200,7 +203,7 @@ 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 3 +#define LNET_RETRY_COUNT_HEALTH_DEFAULT 2 unsigned lnet_retry_count = LNET_RETRY_COUNT_HEALTH_DEFAULT; static int retry_count_set(const char *val, cfs_kernel_param_arg_t *kp); #ifdef HAVE_KERNEL_PARAM_OPS @@ -346,9 +349,15 @@ discovery_set(const char *val, cfs_kernel_param_arg_t *kp) 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 = value; mutex_unlock(&the_lnet.ln_api_mutex); return 0; } @@ -362,7 +371,23 @@ 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); + /* + * Always update the peers. This will result in a push to the + * peers with the updated capabilities feature mask. The peer can + * then take appropriate action to update its representation of + * the node. + * + * If discovery is already off, turn it on first before pushing + * the update. The discovery flag must be on before pushing. + * otherwise if the flag is on and we're turning it off then push + * first before turning the flag off. In the former case the flag + * is being set twice, but I find it's better to do that rather + * than have duplicate code in an if/else statement. + */ + if (*discovery > 0 && value == 0) + *discovery = value; lnet_push_update_to_peers(1); + *discovery = value; mutex_unlock(&the_lnet.ln_api_mutex); @@ -511,17 +536,17 @@ intf_max_set(const char *val, cfs_kernel_param_arg_t *kp) return 0; } -static char * +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 " @@ -545,7 +570,6 @@ lnet_init_locks(void) { spin_lock_init(&the_lnet.ln_eq_wait_lock); spin_lock_init(&the_lnet.ln_msg_resend_lock); - init_waitqueue_head(&the_lnet.ln_eq_waitq); init_completion(&the_lnet.ln_mt_wait_complete); mutex_init(&the_lnet.ln_lnd_mutex); } @@ -553,9 +577,11 @@ lnet_init_locks(void) struct kmem_cache *lnet_mes_cachep; /* MEs kmem_cache */ struct kmem_cache *lnet_small_mds_cachep; /* <= LNET_SMALL_MD_SIZE bytes * MDs kmem_cache */ +struct kmem_cache *lnet_rspt_cachep; /* response tracker cache */ +struct kmem_cache *lnet_msg_cachep; static int -lnet_descriptor_setup(void) +lnet_slab_setup(void) { /* create specific kmem_cache for MEs and small MDs (i.e., originally * allocated in kmem_cache). @@ -571,12 +597,32 @@ lnet_descriptor_setup(void) if (!lnet_small_mds_cachep) return -ENOMEM; + lnet_rspt_cachep = kmem_cache_create("lnet_rspt", sizeof(struct lnet_rsp_tracker), + 0, 0, NULL); + if (!lnet_rspt_cachep) + return -ENOMEM; + + lnet_msg_cachep = kmem_cache_create("lnet_msg", sizeof(struct lnet_msg), + 0, 0, NULL); + if (!lnet_msg_cachep) + return -ENOMEM; + return 0; } static void -lnet_descriptor_cleanup(void) +lnet_slab_cleanup(void) { + if (lnet_msg_cachep) { + kmem_cache_destroy(lnet_msg_cachep); + lnet_msg_cachep = NULL; + } + + + if (lnet_rspt_cachep) { + kmem_cache_destroy(lnet_rspt_cachep); + lnet_rspt_cachep = NULL; + } if (lnet_small_mds_cachep) { kmem_cache_destroy(lnet_small_mds_cachep); @@ -597,7 +643,7 @@ lnet_create_remote_nets_table(void) LASSERT(the_lnet.ln_remote_nets_hash == NULL); LASSERT(the_lnet.ln_remote_nets_hbits > 0); - LIBCFS_ALLOC(hash, LNET_REMOTE_NETS_HASH_SIZE * sizeof(*hash)); + CFS_ALLOC_PTR_ARRAY(hash, LNET_REMOTE_NETS_HASH_SIZE); if (hash == NULL) { CERROR("Failed to create remote nets hash table\n"); return -ENOMEM; @@ -620,9 +666,8 @@ lnet_destroy_remote_nets_table(void) for (i = 0; i < LNET_REMOTE_NETS_HASH_SIZE; i++) LASSERT(list_empty(&the_lnet.ln_remote_nets_hash[i])); - LIBCFS_FREE(the_lnet.ln_remote_nets_hash, - LNET_REMOTE_NETS_HASH_SIZE * - sizeof(the_lnet.ln_remote_nets_hash[0])); + CFS_FREE_PTR_ARRAY(the_lnet.ln_remote_nets_hash, + LNET_REMOTE_NETS_HASH_SIZE); the_lnet.ln_remote_nets_hash = NULL; } @@ -954,10 +999,7 @@ lnet_res_container_cleanup(struct lnet_res_container *rec) struct list_head *e = rec->rec_active.next; list_del_init(e); - if (rec->rec_type == LNET_COOKIE_TYPE_EQ) { - lnet_eq_free(list_entry(e, struct lnet_eq, eq_list)); - - } else if (rec->rec_type == LNET_COOKIE_TYPE_MD) { + if (rec->rec_type == LNET_COOKIE_TYPE_MD) { lnet_md_free(list_entry(e, struct lnet_libmd, md_list)); } else { /* NB: Active MEs should be attached on portals */ @@ -975,8 +1017,7 @@ lnet_res_container_cleanup(struct lnet_res_container *rec) } if (rec->rec_lh_hash != NULL) { - LIBCFS_FREE(rec->rec_lh_hash, - LNET_LH_HASH_SIZE * sizeof(rec->rec_lh_hash[0])); + CFS_FREE_PTR_ARRAY(rec->rec_lh_hash, LNET_LH_HASH_SIZE); rec->rec_lh_hash = NULL; } @@ -1145,10 +1186,10 @@ 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); - LNetInvalidateEQHandle(&the_lnet.ln_mt_eqh); + the_lnet.ln_mt_handler = NULL; init_completion(&the_lnet.ln_started); - rc = lnet_descriptor_setup(); + rc = lnet_slab_setup(); if (rc != 0) goto failed; @@ -1183,14 +1224,6 @@ lnet_prepare(lnet_pid_t requested_pid) if (rc != 0) goto failed; - recs = lnet_res_containers_create(LNET_COOKIE_TYPE_ME); - if (recs == NULL) { - rc = -ENOMEM; - goto failed; - } - - the_lnet.ln_me_containers = recs; - recs = lnet_res_containers_create(LNET_COOKIE_TYPE_MD); if (recs == NULL) { rc = -ENOMEM; @@ -1221,8 +1254,6 @@ 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 @@ -1239,11 +1270,8 @@ lnet_unprepare (void) the_lnet.ln_mt_zombie_rstqs = NULL; } - if (!LNetEQHandleIsInvalid(the_lnet.ln_mt_eqh)) { - rc = LNetEQFree(the_lnet.ln_mt_eqh); - LNetInvalidateEQHandle(&the_lnet.ln_mt_eqh); - LASSERT(rc == 0); - } + lnet_assert_handler_unused(the_lnet.ln_mt_handler); + the_lnet.ln_mt_handler = NULL; lnet_portals_destroy(); @@ -1252,11 +1280,6 @@ lnet_unprepare (void) the_lnet.ln_md_containers = NULL; } - if (the_lnet.ln_me_containers != NULL) { - lnet_res_containers_destroy(the_lnet.ln_me_containers); - the_lnet.ln_me_containers = NULL; - } - lnet_res_container_cleanup(&the_lnet.ln_eq_container); lnet_msg_containers_destroy(); @@ -1268,7 +1291,7 @@ lnet_unprepare (void) the_lnet.ln_counters = NULL; } lnet_destroy_remote_nets_table(); - lnet_descriptor_cleanup(); + lnet_slab_cleanup(); return 0; } @@ -1495,6 +1518,7 @@ 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); } @@ -1504,7 +1528,7 @@ lnet_ping_buffer_alloc(int nnis, gfp_t gfp) void lnet_ping_buffer_free(struct lnet_ping_buffer *pbuf) { - LASSERT(lnet_ping_buffer_numref(pbuf) == 0); + LASSERT(atomic_read(&pbuf->pb_refcnt) == 0); LIBCFS_FREE(pbuf, LNET_PING_BUFFER_SIZE(pbuf->pb_nnis)); } @@ -1621,7 +1645,7 @@ 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; } @@ -1651,7 +1675,7 @@ 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); @@ -1666,18 +1690,13 @@ lnet_ping_target_setup(struct lnet_ping_buffer **ppbuf, .nid = LNET_NID_ANY, .pid = LNET_PID_ANY }; - struct lnet_handle_me me_handle; + struct lnet_me *me; struct lnet_md md = { NULL }; - int rc, rc2; + int rc; - if (set_eq) { - rc = LNetEQAlloc(0, lnet_ping_target_event_handler, - &the_lnet.ln_ping_target_eq); - if (rc != 0) { - CERROR("Can't allocate ping buffer EQ: %d\n", rc); - return 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) { @@ -1686,11 +1705,11 @@ lnet_ping_target_setup(struct lnet_ping_buffer **ppbuf, } /* Ping target ME/MD */ - rc = LNetMEAttach(LNET_RESERVED_PORTAL, id, + me = LNetMEAttach(LNET_RESERVED_PORTAL, id, LNET_PROTO_PING_MATCHBITS, 0, - LNET_UNLINK, LNET_INS_AFTER, - &me_handle); - if (rc != 0) { + LNET_UNLINK, LNET_INS_AFTER); + if (IS_ERR(me)) { + rc = PTR_ERR(me); CERROR("Can't create ping target ME: %d\n", rc); goto fail_decref_ping_buffer; } @@ -1702,10 +1721,10 @@ 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_handle, 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; @@ -1715,17 +1734,12 @@ lnet_ping_target_setup(struct lnet_ping_buffer **ppbuf, return 0; fail_unlink_ping_me: - rc2 = LNetMEUnlink(me_handle); - LASSERT(rc2 == 0); + LNetMEUnlink(me); fail_decref_ping_buffer: - LASSERT(lnet_ping_buffer_numref(*ppbuf) == 1); + 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; } @@ -1733,19 +1747,13 @@ static void lnet_ping_md_unlink(struct lnet_ping_buffer *pbuf, struct lnet_handle_md *ping_mdh) { - sigset_t blocked = cfs_block_allsigs(); - LNetMDUnlink(*ping_mdh); LNetInvalidateMDHandle(ping_mdh); /* NB the MD could be busy; this just starts the unlink */ - while (lnet_ping_buffer_numref(pbuf) > 1) { - CDEBUG(D_NET, "Still waiting for ping data MD to unlink\n"); - set_current_state(TASK_UNINTERRUPTIBLE); - schedule_timeout(cfs_time_seconds(1)); - } - - cfs_restore_sigs(blocked); + wait_var_event_warning(&pbuf->pb_refcnt, + atomic_read(&pbuf->pb_refcnt) <= 1, + "Still waiting for ping data MD to unlink\n"); } static void @@ -1832,66 +1840,46 @@ 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_handle_me meh; 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; - } - - rc = LNetMEAttach(LNET_RESERVED_PORTAL, id, - LNET_PROTO_PING_MATCHBITS, 0, - LNET_UNLINK, LNET_INS_AFTER, - &meh); - if (rc) { - 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(meh, 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_meh; + 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; @@ -1902,34 +1890,82 @@ 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_meh: - LNetMEUnlink(meh); -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); + LNetMEUnlink(me); + 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); } @@ -1941,12 +1977,11 @@ static int lnet_push_target_init(void) if (the_lnet.ln_push_target) return -EALREADY; - rc = LNetEQAlloc(0, lnet_push_target_event_handler, - &the_lnet.ln_push_target_eq); - if (rc) { - 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; @@ -1954,8 +1989,8 @@ static int lnet_push_target_init(void) rc = lnet_push_target_resize(); if (rc) { - LNetEQFree(the_lnet.ln_push_target_eq); - LNetInvalidateEQHandle(&the_lnet.ln_push_target_eq); + LNetClearLazyPortal(LNET_RESERVED_PORTAL); + the_lnet.ln_push_target_handler = NULL; } return rc; @@ -1972,18 +2007,18 @@ static void lnet_push_target_fini(void) LNetInvalidateMDHandle(&the_lnet.ln_push_target_md); /* Wait for the unlink to complete. */ - while (lnet_ping_buffer_numref(the_lnet.ln_push_target) > 1) { - CDEBUG(D_NET, "Still waiting for ping data MD to unlink\n"); - set_current_state(TASK_UNINTERRUPTIBLE); - schedule_timeout(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); - LNetInvalidateEQHandle(&the_lnet.ln_push_target_eq); + LNetClearLazyPortal(LNET_RESERVED_PORTAL); + lnet_assert_handler_unused(the_lnet.ln_push_target_handler); + the_lnet.ln_push_target_handler = NULL; } static int @@ -2051,8 +2086,7 @@ lnet_clear_zombies_nis_locked(struct lnet_net *net) "Waiting for zombie LNI %s\n", libcfs_nid2str(ni->ni_nid)); } - set_current_state(TASK_UNINTERRUPTIBLE); - schedule_timeout(cfs_time_seconds(1)); + schedule_timeout_uninterruptible(cfs_time_seconds(1)); lnet_net_lock(LNET_LOCK_EX); continue; } @@ -2062,7 +2096,14 @@ lnet_clear_zombies_nis_locked(struct lnet_net *net) islo = ni->ni_net->net_lnd->lnd_type == LOLND; LASSERT(!in_interrupt()); + /* Holding the mutex makes it safe for lnd_shutdown + * to call module_put(). Module unload cannot finish + * until lnet_unregister_lnd() completes, and that + * requires the mutex. + */ + mutex_lock(&the_lnet.ln_lnd_mutex); (net->net_lnd->lnd_shutdown)(ni); + mutex_unlock(&the_lnet.ln_lnd_mutex); if (!islo) CDEBUG(D_LNI, "Removed LNI %s\n", @@ -2127,11 +2168,9 @@ static void lnet_shutdown_lndnets(void) { struct lnet_net *net; - struct list_head resend; + LIST_HEAD(resend); struct lnet_msg *msg, *tmp; - INIT_LIST_HEAD(&resend); - /* NB called holding the global mutex */ /* All quiet on the API front */ @@ -2141,16 +2180,12 @@ lnet_shutdown_lndnets(void) lnet_net_lock(LNET_LOCK_EX); the_lnet.ln_state = LNET_STATE_STOPPING; - while (!list_empty(&the_lnet.ln_nets)) { - /* - * move the nets to the zombie list to avoid them being - * picked up for new work. LONET is also included in the - * Nets that will be moved to the zombie list - */ - net = list_entry(the_lnet.ln_nets.next, - struct lnet_net, net_list); - list_move(&net->net_list, &the_lnet.ln_net_zombie); - } + /* + * move the nets to the zombie list to avoid them being + * picked up for new work. LONET is also included in the + * Nets that will be moved to the zombie list + */ + list_splice_init(&the_lnet.ln_nets, &the_lnet.ln_net_zombie); /* Drop the cached loopback Net. */ if (the_lnet.ln_loni != NULL) { @@ -2262,21 +2297,19 @@ static int lnet_startup_lndnet(struct lnet_net *net, struct lnet_lnd_tunables *tun) { struct lnet_ni *ni; - struct lnet_net *net_l = NULL; - struct list_head local_ni_list; - int rc; - int ni_count = 0; - __u32 lnd_type; + struct lnet_net *net_l = NULL; + LIST_HEAD(local_ni_list); + int rc; + int ni_count = 0; + __u32 lnd_type; const struct lnet_lnd *lnd; - int peer_timeout = + int peer_timeout = net->net_tunables.lct_peer_timeout; - int maxtxcredits = + int maxtxcredits = net->net_tunables.lct_max_tx_credits; - int peerrtrcredits = + int peerrtrcredits = net->net_tunables.lct_peer_rtr_credits; - INIT_LIST_HEAD(&local_ni_list); - /* * make sure that this net is unique. If it isn't then * we are adding interfaces to an already existing network, and @@ -2374,9 +2407,6 @@ lnet_startup_lndnet(struct lnet_net *net, struct lnet_lnd_tunables *tun) if (rc < 0) goto failed1; - LASSERT(ni->ni_net->net_tunables.lct_peer_timeout <= 0 || - ni->ni_net->net_lnd->lnd_query != NULL); - lnet_ni_addref(ni); list_add_tail(&ni->ni_netlist, &local_ni_list); @@ -2489,8 +2519,8 @@ int lnet_lib_init(void) lnet_assert_wire_constants(); /* refer to global cfs_cpt_table for now */ - the_lnet.ln_cpt_table = cfs_cpt_table; - the_lnet.ln_cpt_number = cfs_cpt_number(cfs_cpt_table); + the_lnet.ln_cpt_table = cfs_cpt_tab; + the_lnet.ln_cpt_number = cfs_cpt_number(cfs_cpt_tab); LASSERT(the_lnet.ln_cpt_number > 0); if (the_lnet.ln_cpt_number > LNET_CPT_MAX) { @@ -2574,11 +2604,9 @@ LNetNIInit(lnet_pid_t requested_pid) int ni_count; struct lnet_ping_buffer *pbuf; struct lnet_handle_md ping_mdh; - struct list_head net_head; + LIST_HEAD(net_head); struct lnet_net *net; - INIT_LIST_HEAD(&net_head); - mutex_lock(&the_lnet.ln_api_mutex); CDEBUG(D_OTHER, "refs %d\n", the_lnet.ln_refcount); @@ -2650,11 +2678,7 @@ LNetNIInit(lnet_pid_t requested_pid) lnet_ping_target_update(pbuf, ping_mdh); - rc = LNetEQAlloc(0, lnet_mt_event_handler, &the_lnet.ln_mt_eqh); - if (rc != 0) { - 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) @@ -2783,7 +2807,7 @@ 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; @@ -2875,7 +2899,7 @@ 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; @@ -3069,7 +3093,6 @@ 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); @@ -3110,14 +3133,6 @@ 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, @@ -3135,7 +3150,7 @@ 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 */ @@ -3163,11 +3178,9 @@ 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; - struct list_head net_head; - - INIT_LIST_HEAD(&net_head); + LIST_HEAD(net_head); rc = lnet_parse_ip2nets(&nets, ip2nets); if (rc < 0) @@ -3290,8 +3303,7 @@ 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); @@ -3318,8 +3330,7 @@ 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); @@ -3346,13 +3357,11 @@ unlock_api_mutex: int lnet_dyn_add_net(struct lnet_ioctl_config_data *conf) { - struct lnet_net *net; - struct list_head net_head; - int rc; + struct lnet_net *net; + LIST_HEAD(net_head); + int rc; struct lnet_ioctl_config_lnd_tunables tun; - char *nets = conf->cfg_config_u.cfg_net.net_intf; - - INIT_LIST_HEAD(&net_head); + 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); @@ -3431,8 +3440,7 @@ 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); @@ -4084,24 +4092,44 @@ LNetGetId(unsigned int index, struct lnet_process_id *id) } EXPORT_SYMBOL(LNetGetId); +struct ping_data { + int rc; + int replied; + struct lnet_handle_md mdh; + struct completion completion; +}; + +static void +lnet_ping_event_handler(struct lnet_event *event) +{ + struct ping_data *pd = event->md_user_ptr; + + CDEBUG(D_NET, "ping event (%d %d)%s\n", + event->type, event->status, + event->unlinked ? " unlinked" : ""); + + if (event->status) { + if (!pd->rc) + pd->rc = event->status; + } else if (event->type == LNET_EVENT_REPLY) { + pd->replied = 1; + pd->rc = event->mlength; + } + if (event->unlinked) + complete(&pd->completion); +} + static int lnet_ping(struct lnet_process_id id, signed long timeout, struct lnet_process_id __user *ids, int n_ids) { - struct lnet_handle_eq eqh; - struct lnet_handle_md mdh; - struct lnet_event event; struct lnet_md md = { NULL }; - int which; - int unlinked = 0; - int replied = 0; - const signed long a_long_time = cfs_time_seconds(60); + struct ping_data pd = { 0 }; struct lnet_ping_buffer *pbuf; struct lnet_process_id tmpid; int i; int nob; int rc; int rc2; - sigset_t blocked; /* n_ids limit is arbitrary */ if (n_ids <= 0 || id.nid == LNET_NID_ANY) @@ -4121,92 +4149,46 @@ static int lnet_ping(struct lnet_process_id id, signed long timeout, if (!pbuf) return -ENOMEM; - /* NB 2 events max (including any unlink event) */ - rc = LNetEQAlloc(2, LNET_EQ_HANDLER_NONE, &eqh); - if (rc != 0) { - 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); md.threshold = 2; /* GET/REPLY */ md.max_size = 0; md.options = LNET_MD_TRUNCATE; - md.user_ptr = NULL; - md.eq_handle = eqh; + md.user_ptr = &pd; + md.handler = lnet_ping_event_handler; - rc = LNetMDBind(md, LNET_UNLINK, &mdh); + init_completion(&pd.completion); + + 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, mdh, id, + rc = LNetGet(LNET_NID_ANY, pd.mdh, id, LNET_RESERVED_PORTAL, LNET_PROTO_PING_MATCHBITS, 0, false); if (rc != 0) { /* Don't CERROR; this could be deliberate! */ - rc2 = LNetMDUnlink(mdh); + rc2 = LNetMDUnlink(pd.mdh); LASSERT(rc2 == 0); /* NB must wait for the UNLINK event below... */ - unlinked = 1; - timeout = a_long_time; - } - - do { - /* MUST block for unlink to complete */ - if (unlinked) - blocked = cfs_block_allsigs(); - - rc2 = LNetEQPoll(&eqh, 1, timeout, &event, &which); - - if (unlinked) - cfs_restore_sigs(blocked); - - CDEBUG(D_NET, "poll %d(%d %d)%s\n", rc2, - (rc2 <= 0) ? -1 : event.type, - (rc2 <= 0) ? -1 : event.status, - (rc2 > 0 && event.unlinked) ? " unlinked" : ""); - - LASSERT(rc2 != -EOVERFLOW); /* can't miss anything */ - - if (rc2 <= 0 || event.status != 0) { - /* timeout or error */ - if (!replied && rc == 0) - rc = (rc2 < 0) ? rc2 : - (rc2 == 0) ? -ETIMEDOUT : - event.status; - - if (!unlinked) { - /* Ensure completion in finite time... */ - LNetMDUnlink(mdh); - /* No assertion (racing with network) */ - unlinked = 1; - timeout = a_long_time; - } else if (rc2 == 0) { - /* timed out waiting for unlink */ - CWARN("ping %s: late network completion\n", - libcfs_id2str(id)); - } - } else if (event.type == LNET_EVENT_REPLY) { - replied = 1; - rc = event.mlength; - } - } while (rc2 <= 0 || !event.unlinked); + } - if (!replied) { - if (rc >= 0) - CWARN("%s: Unexpected rc >= 0 but no reply!\n", - libcfs_id2str(id)); + if (wait_for_completion_timeout(&pd.completion, timeout) == 0) { + /* Ensure completion in finite time... */ + LNetMDUnlink(pd.mdh); + wait_for_completion(&pd.completion); + } + if (!pd.replied) { rc = -EIO; - goto fail_free_eq; + goto fail_ping_buffer_decref; } - nob = rc; + nob = pd.rc; LASSERT(nob >= 0 && nob <= LNET_PING_INFO_SIZE(n_ids)); rc = -EPROTO; /* if I can't parse... */ @@ -4214,7 +4196,7 @@ 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)) { @@ -4222,20 +4204,20 @@ 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) @@ -4245,7 +4227,7 @@ 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()... */ @@ -4255,16 +4237,10 @@ 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(eqh); - if (rc2 != 0) - CERROR("rc2 %d\n", rc2); - LASSERT(rc2 == 0); - fail_ping_buffer_decref: lnet_ping_buffer_decref(pbuf); return rc; @@ -4281,8 +4257,6 @@ lnet_discover(struct lnet_process_id id, __u32 force, int cpt; int i; int rc; - int max_intf = lnet_interfaces_max; - size_t buf_size; if (n_ids <= 0 || id.nid == LNET_NID_ANY) @@ -4292,15 +4266,13 @@ 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; - - buf_size = n_ids * sizeof(*buf); + if (n_ids > lnet_interfaces_max) + n_ids = lnet_interfaces_max; - LIBCFS_ALLOC(buf, buf_size); + CFS_ALLOC_PTR_ARRAY(buf, n_ids); if (!buf) return -ENOMEM; @@ -4326,11 +4298,6 @@ 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) { @@ -4339,21 +4306,17 @@ 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); - LIBCFS_FREE(buf, buf_size); + if (rc >= 0) + if (copy_to_user(ids, buf, rc * sizeof(*buf))) + rc = -EFAULT; + CFS_FREE_PTR_ARRAY(buf, n_ids); return rc; }