X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lnet%2Flnet%2Fapi-ni.c;h=b8ba025489ee2594e0c3d107ffcc62a3bdc9699d;hb=24b1bba70c43ca64318b54aa11d7ba584ee2b4c0;hp=86541a7e90adefc4052d1bd2688d6e2f89e959c6;hpb=b4efa1b2cbfd45f85439e1bb0a4c4eb719540dcd;p=fs%2Flustre-release.git diff --git a/lnet/lnet/api-ni.c b/lnet/lnet/api-ni.c index 86541a7..8f5dc24 100644 --- a/lnet/lnet/api-ni.c +++ b/lnet/lnet/api-ni.c @@ -1,6 +1,4 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: - * +/* * GPL HEADER START * * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. @@ -17,17 +15,15 @@ * * You should have received a copy of the GNU General Public License * version 2 along with this program; If not, see - * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf - * - * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, - * CA 95054 USA or visit www.sun.com if you need additional information or - * have any questions. + * http://www.gnu.org/licenses/gpl-2.0.html * * GPL HEADER END */ /* - * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. + * + * Copyright (c) 2011, 2016, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -35,1698 +31,2697 @@ */ #define DEBUG_SUBSYSTEM S_LNET +#include +#include + #include -#ifdef __KERNEL__ #define D_LNI D_CONSOLE -#else -#define D_LNI D_CONFIG -#endif -lnet_t the_lnet; /* THE state of the network */ - -#ifdef __KERNEL__ +lnet_t the_lnet; /* THE state of the network */ +EXPORT_SYMBOL(the_lnet); static char *ip2nets = ""; -CFS_MODULE_PARM(ip2nets, "s", charp, 0444, - "LNET network <- IP table"); +module_param(ip2nets, charp, 0444); +MODULE_PARM_DESC(ip2nets, "LNET network <- IP table"); static char *networks = ""; -CFS_MODULE_PARM(networks, "s", charp, 0444, - "local networks"); +module_param(networks, charp, 0444); +MODULE_PARM_DESC(networks, "local networks"); static char *routes = ""; -CFS_MODULE_PARM(routes, "s", charp, 0444, - "routes to non-local networks"); +module_param(routes, charp, 0444); +MODULE_PARM_DESC(routes, "routes to non-local networks"); + +static int rnet_htable_size = LNET_REMOTE_NETS_HASH_DEFAULT; +module_param(rnet_htable_size, int, 0444); +MODULE_PARM_DESC(rnet_htable_size, "size of remote network hash table"); -char * +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"); + +/* + * This sequence number keeps track of how many times DLC was used to + * update the configuration. It is incremented on any DLC update and + * checked when sending a message to determine if there is a need to + * re-run the selection algorithm to handle configuration change. + * Look at lnet_select_pathway() for more details on its usage. + */ +static atomic_t lnet_dlc_seq_no = ATOMIC_INIT(0); + +static int lnet_ping(lnet_process_id_t id, signed long timeout, + lnet_process_id_t __user *ids, int n_ids); + +static char * lnet_get_routes(void) { - return routes; + return routes; } -char * +static char * lnet_get_networks(void) { - char *nets; - int rc; + char *nets; + int rc; - if (*networks != 0 && *ip2nets != 0) { - LCONSOLE_ERROR_MSG(0x101, "Please specify EITHER 'networks' or " - "'ip2nets' but not both at once\n"); - return NULL; - } + if (*networks != 0 && *ip2nets != 0) { + LCONSOLE_ERROR_MSG(0x101, "Please specify EITHER 'networks' or " + "'ip2nets' but not both at once\n"); + return NULL; + } - if (*ip2nets != 0) { - rc = lnet_parse_ip2nets(&nets, ip2nets); - return (rc == 0) ? nets : NULL; - } + if (*ip2nets != 0) { + rc = lnet_parse_ip2nets(&nets, ip2nets); + return (rc == 0) ? nets : NULL; + } - if (*networks != 0) - return networks; + if (*networks != 0) + return networks; - return "tcp"; + return "tcp"; } -void +static void lnet_init_locks(void) { - spin_lock_init (&the_lnet.ln_lock); - cfs_waitq_init (&the_lnet.ln_waitq); - init_mutex(&the_lnet.ln_lnd_mutex); - init_mutex(&the_lnet.ln_api_mutex); + spin_lock_init(&the_lnet.ln_eq_wait_lock); + init_waitqueue_head(&the_lnet.ln_eq_waitq); + init_waitqueue_head(&the_lnet.ln_rc_waitq); + mutex_init(&the_lnet.ln_lnd_mutex); + mutex_init(&the_lnet.ln_api_mutex); } -void +static void lnet_fini_locks(void) { } -#else +struct kmem_cache *lnet_mes_cachep; /* MEs kmem_cache */ +struct kmem_cache *lnet_small_mds_cachep; /* <= LNET_SMALL_MD_SIZE bytes + * MDs kmem_cache */ -char * -lnet_get_routes(void) +static int +lnet_descriptor_setup(void) { - char *str = getenv("LNET_ROUTES"); - - return (str == NULL) ? "" : str; -} - -char * -lnet_get_networks (void) -{ - static char default_networks[256]; - char *networks = getenv ("LNET_NETWORKS"); - char *ip2nets = getenv ("LNET_IP2NETS"); - char *str; - char *sep; - int len; - int nob; - int rc; - struct list_head *tmp; - -#ifdef NOT_YET - if (networks != NULL && ip2nets != NULL) { - LCONSOLE_ERROR_MSG(0x103, "Please set EITHER 'LNET_NETWORKS' or" - " 'LNET_IP2NETS' but not both at once\n"); - return NULL; - } - - if (ip2nets != NULL) { - rc = lnet_parse_ip2nets(&networks, ip2nets); - return (rc == 0) ? networks : NULL; - } -#else - ip2nets = NULL; - rc = 0; -#endif - if (networks != NULL) - return networks; + /* create specific kmem_cache for MEs and small MDs (i.e., originally + * allocated in kmem_cache). + */ + lnet_mes_cachep = kmem_cache_create("lnet_MEs", sizeof(lnet_me_t), + 0, 0, NULL); + if (!lnet_mes_cachep) + return -ENOMEM; + + lnet_small_mds_cachep = kmem_cache_create("lnet_small_MDs", + LNET_SMALL_MD_SIZE, 0, 0, + NULL); + if (!lnet_small_mds_cachep) + return -ENOMEM; + + return 0; +} - /* In userland, the default 'networks=' is the list of known net types */ +static void +lnet_descriptor_cleanup(void) +{ - len = sizeof(default_networks); - str = default_networks; - *str = 0; - sep = ""; + if (lnet_small_mds_cachep) { + kmem_cache_destroy(lnet_small_mds_cachep); + lnet_small_mds_cachep = NULL; + } - list_for_each (tmp, &the_lnet.ln_lnds) { - lnd_t *lnd = list_entry(tmp, lnd_t, lnd_list); + if (lnet_mes_cachep) { + kmem_cache_destroy(lnet_mes_cachep); + lnet_mes_cachep = NULL; + } +} - nob = snprintf(str, len, "%s%s", sep, - libcfs_lnd2str(lnd->lnd_type)); - len -= nob; - if (len < 0) { - /* overflowed the string; leave it where it was */ - *str = 0; - break; - } +static int +lnet_create_remote_nets_table(void) +{ + int i; + struct list_head *hash; + + 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)); + if (hash == NULL) { + CERROR("Failed to create remote nets hash table\n"); + return -ENOMEM; + } + + for (i = 0; i < LNET_REMOTE_NETS_HASH_SIZE; i++) + INIT_LIST_HEAD(&hash[i]); + the_lnet.ln_remote_nets_hash = hash; + return 0; +} - str += nob; - sep = ","; - } +static void +lnet_destroy_remote_nets_table(void) +{ + int i; - return default_networks; -} + if (the_lnet.ln_remote_nets_hash == NULL) + return; -# ifndef HAVE_LIBPTHREAD + for (i = 0; i < LNET_REMOTE_NETS_HASH_SIZE; i++) + LASSERT(list_empty(&the_lnet.ln_remote_nets_hash[i])); -void lnet_init_locks(void) -{ - the_lnet.ln_lock = 0; - the_lnet.ln_lnd_mutex = 0; - the_lnet.ln_api_mutex = 0; + LIBCFS_FREE(the_lnet.ln_remote_nets_hash, + LNET_REMOTE_NETS_HASH_SIZE * + sizeof(the_lnet.ln_remote_nets_hash[0])); + the_lnet.ln_remote_nets_hash = NULL; } -void lnet_fini_locks(void) +static void +lnet_destroy_locks(void) { - LASSERT (the_lnet.ln_api_mutex == 0); - LASSERT (the_lnet.ln_lnd_mutex == 0); - LASSERT (the_lnet.ln_lock == 0); -} + if (the_lnet.ln_res_lock != NULL) { + cfs_percpt_lock_free(the_lnet.ln_res_lock); + the_lnet.ln_res_lock = NULL; + } -# else + if (the_lnet.ln_net_lock != NULL) { + cfs_percpt_lock_free(the_lnet.ln_net_lock); + the_lnet.ln_net_lock = NULL; + } -void lnet_init_locks(void) + lnet_fini_locks(); +} + +static int +lnet_create_locks(void) { - pthread_cond_init(&the_lnet.ln_cond, NULL); - pthread_mutex_init(&the_lnet.ln_lock, NULL); - pthread_mutex_init(&the_lnet.ln_lnd_mutex, NULL); - pthread_mutex_init(&the_lnet.ln_api_mutex, NULL); + lnet_init_locks(); + + the_lnet.ln_res_lock = cfs_percpt_lock_alloc(lnet_cpt_table()); + if (the_lnet.ln_res_lock == NULL) + goto failed; + + the_lnet.ln_net_lock = cfs_percpt_lock_alloc(lnet_cpt_table()); + if (the_lnet.ln_net_lock == NULL) + goto failed; + + return 0; + + failed: + lnet_destroy_locks(); + return -ENOMEM; } -void lnet_fini_locks(void) +static void lnet_assert_wire_constants(void) { - pthread_mutex_destroy(&the_lnet.ln_api_mutex); - pthread_mutex_destroy(&the_lnet.ln_lnd_mutex); - pthread_mutex_destroy(&the_lnet.ln_lock); - pthread_cond_destroy(&the_lnet.ln_cond); + /* Wire protocol assertions generated by 'wirecheck' + * running on Linux robert.bartonsoftware.com 2.6.8-1.521 + * #1 Mon Aug 16 09:01:18 EDT 2004 i686 athlon i386 GNU/Linux + * with gcc version 3.3.3 20040412 (Red Hat Linux 3.3.3-7) */ + + /* Constants... */ + CLASSERT(LNET_PROTO_TCP_MAGIC == 0xeebc0ded); + CLASSERT(LNET_PROTO_TCP_VERSION_MAJOR == 1); + CLASSERT(LNET_PROTO_TCP_VERSION_MINOR == 0); + CLASSERT(LNET_MSG_ACK == 0); + CLASSERT(LNET_MSG_PUT == 1); + CLASSERT(LNET_MSG_GET == 2); + CLASSERT(LNET_MSG_REPLY == 3); + CLASSERT(LNET_MSG_HELLO == 4); + + /* Checks for struct lnet_handle_wire */ + CLASSERT((int)sizeof(struct lnet_handle_wire) == 16); + CLASSERT((int)offsetof(struct lnet_handle_wire, wh_interface_cookie) == 0); + CLASSERT((int)sizeof(((struct lnet_handle_wire *)0)->wh_interface_cookie) == 8); + CLASSERT((int)offsetof(struct lnet_handle_wire, wh_object_cookie) == 8); + CLASSERT((int)sizeof(((struct lnet_handle_wire *)0)->wh_object_cookie) == 8); + + /* Checks for struct lnet_magicversion_t */ + CLASSERT((int)sizeof(lnet_magicversion_t) == 8); + CLASSERT((int)offsetof(lnet_magicversion_t, magic) == 0); + CLASSERT((int)sizeof(((lnet_magicversion_t *)0)->magic) == 4); + CLASSERT((int)offsetof(lnet_magicversion_t, version_major) == 4); + CLASSERT((int)sizeof(((lnet_magicversion_t *)0)->version_major) == 2); + CLASSERT((int)offsetof(lnet_magicversion_t, version_minor) == 6); + CLASSERT((int)sizeof(((lnet_magicversion_t *)0)->version_minor) == 2); + + /* Checks for struct lnet_hdr_t */ + CLASSERT((int)sizeof(lnet_hdr_t) == 72); + CLASSERT((int)offsetof(lnet_hdr_t, dest_nid) == 0); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->dest_nid) == 8); + CLASSERT((int)offsetof(lnet_hdr_t, src_nid) == 8); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->src_nid) == 8); + CLASSERT((int)offsetof(lnet_hdr_t, dest_pid) == 16); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->dest_pid) == 4); + CLASSERT((int)offsetof(lnet_hdr_t, src_pid) == 20); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->src_pid) == 4); + CLASSERT((int)offsetof(lnet_hdr_t, type) == 24); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->type) == 4); + CLASSERT((int)offsetof(lnet_hdr_t, payload_length) == 28); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->payload_length) == 4); + CLASSERT((int)offsetof(lnet_hdr_t, msg) == 32); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg) == 40); + + /* Ack */ + CLASSERT((int)offsetof(lnet_hdr_t, msg.ack.dst_wmd) == 32); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.ack.dst_wmd) == 16); + CLASSERT((int)offsetof(lnet_hdr_t, msg.ack.match_bits) == 48); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.ack.match_bits) == 8); + CLASSERT((int)offsetof(lnet_hdr_t, msg.ack.mlength) == 56); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.ack.mlength) == 4); + + /* Put */ + CLASSERT((int)offsetof(lnet_hdr_t, msg.put.ack_wmd) == 32); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.put.ack_wmd) == 16); + CLASSERT((int)offsetof(lnet_hdr_t, msg.put.match_bits) == 48); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.put.match_bits) == 8); + CLASSERT((int)offsetof(lnet_hdr_t, msg.put.hdr_data) == 56); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.put.hdr_data) == 8); + CLASSERT((int)offsetof(lnet_hdr_t, msg.put.ptl_index) == 64); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.put.ptl_index) == 4); + CLASSERT((int)offsetof(lnet_hdr_t, msg.put.offset) == 68); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.put.offset) == 4); + + /* Get */ + CLASSERT((int)offsetof(lnet_hdr_t, msg.get.return_wmd) == 32); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.get.return_wmd) == 16); + CLASSERT((int)offsetof(lnet_hdr_t, msg.get.match_bits) == 48); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.get.match_bits) == 8); + CLASSERT((int)offsetof(lnet_hdr_t, msg.get.ptl_index) == 56); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.get.ptl_index) == 4); + CLASSERT((int)offsetof(lnet_hdr_t, msg.get.src_offset) == 60); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.get.src_offset) == 4); + CLASSERT((int)offsetof(lnet_hdr_t, msg.get.sink_length) == 64); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.get.sink_length) == 4); + + /* Reply */ + CLASSERT((int)offsetof(lnet_hdr_t, msg.reply.dst_wmd) == 32); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.reply.dst_wmd) == 16); + + /* Hello */ + CLASSERT((int)offsetof(lnet_hdr_t, msg.hello.incarnation) == 32); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.hello.incarnation) == 8); + CLASSERT((int)offsetof(lnet_hdr_t, msg.hello.type) == 40); + CLASSERT((int)sizeof(((lnet_hdr_t *)0)->msg.hello.type) == 4); } -# endif -#endif +static lnd_t *lnet_find_lnd_by_type(__u32 type) +{ + lnd_t *lnd; + struct list_head *tmp; + + /* holding lnd mutex */ + list_for_each(tmp, &the_lnet.ln_lnds) { + lnd = list_entry(tmp, lnd_t, lnd_list); -void lnet_assert_wire_constants (void) -{ - /* Wire protocol assertions generated by 'wirecheck' - * running on Linux robert.bartonsoftware.com 2.6.8-1.521 - * #1 Mon Aug 16 09:01:18 EDT 2004 i686 athlon i386 GNU/Linux - * with gcc version 3.3.3 20040412 (Red Hat Linux 3.3.3-7) */ - - /* Constants... */ - CLASSERT (LNET_PROTO_TCP_MAGIC == 0xeebc0ded); - CLASSERT (LNET_PROTO_TCP_VERSION_MAJOR == 1); - CLASSERT (LNET_PROTO_TCP_VERSION_MINOR == 0); - CLASSERT (LNET_MSG_ACK == 0); - CLASSERT (LNET_MSG_PUT == 1); - CLASSERT (LNET_MSG_GET == 2); - CLASSERT (LNET_MSG_REPLY == 3); - CLASSERT (LNET_MSG_HELLO == 4); - - /* Checks for struct ptl_handle_wire_t */ - CLASSERT ((int)sizeof(lnet_handle_wire_t) == 16); - CLASSERT ((int)offsetof(lnet_handle_wire_t, wh_interface_cookie) == 0); - CLASSERT ((int)sizeof(((lnet_handle_wire_t *)0)->wh_interface_cookie) == 8); - CLASSERT ((int)offsetof(lnet_handle_wire_t, wh_object_cookie) == 8); - CLASSERT ((int)sizeof(((lnet_handle_wire_t *)0)->wh_object_cookie) == 8); - - /* Checks for struct lnet_magicversion_t */ - CLASSERT ((int)sizeof(lnet_magicversion_t) == 8); - CLASSERT ((int)offsetof(lnet_magicversion_t, magic) == 0); - CLASSERT ((int)sizeof(((lnet_magicversion_t *)0)->magic) == 4); - CLASSERT ((int)offsetof(lnet_magicversion_t, version_major) == 4); - CLASSERT ((int)sizeof(((lnet_magicversion_t *)0)->version_major) == 2); - CLASSERT ((int)offsetof(lnet_magicversion_t, version_minor) == 6); - CLASSERT ((int)sizeof(((lnet_magicversion_t *)0)->version_minor) == 2); - - /* Checks for struct lnet_hdr_t */ - CLASSERT ((int)sizeof(lnet_hdr_t) == 72); - CLASSERT ((int)offsetof(lnet_hdr_t, dest_nid) == 0); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->dest_nid) == 8); - CLASSERT ((int)offsetof(lnet_hdr_t, src_nid) == 8); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->src_nid) == 8); - CLASSERT ((int)offsetof(lnet_hdr_t, dest_pid) == 16); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->dest_pid) == 4); - CLASSERT ((int)offsetof(lnet_hdr_t, src_pid) == 20); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->src_pid) == 4); - CLASSERT ((int)offsetof(lnet_hdr_t, type) == 24); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->type) == 4); - CLASSERT ((int)offsetof(lnet_hdr_t, payload_length) == 28); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->payload_length) == 4); - CLASSERT ((int)offsetof(lnet_hdr_t, msg) == 32); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg) == 40); - - /* Ack */ - CLASSERT ((int)offsetof(lnet_hdr_t, msg.ack.dst_wmd) == 32); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.ack.dst_wmd) == 16); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.ack.match_bits) == 48); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.ack.match_bits) == 8); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.ack.mlength) == 56); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.ack.mlength) == 4); - - /* Put */ - CLASSERT ((int)offsetof(lnet_hdr_t, msg.put.ack_wmd) == 32); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.put.ack_wmd) == 16); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.put.match_bits) == 48); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.put.match_bits) == 8); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.put.hdr_data) == 56); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.put.hdr_data) == 8); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.put.ptl_index) == 64); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.put.ptl_index) == 4); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.put.offset) == 68); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.put.offset) == 4); - - /* Get */ - CLASSERT ((int)offsetof(lnet_hdr_t, msg.get.return_wmd) == 32); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.get.return_wmd) == 16); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.get.match_bits) == 48); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.get.match_bits) == 8); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.get.ptl_index) == 56); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.get.ptl_index) == 4); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.get.src_offset) == 60); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.get.src_offset) == 4); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.get.sink_length) == 64); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.get.sink_length) == 4); - - /* Reply */ - CLASSERT ((int)offsetof(lnet_hdr_t, msg.reply.dst_wmd) == 32); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.reply.dst_wmd) == 16); - - /* Hello */ - CLASSERT ((int)offsetof(lnet_hdr_t, msg.hello.incarnation) == 32); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.hello.incarnation) == 8); - CLASSERT ((int)offsetof(lnet_hdr_t, msg.hello.type) == 40); - CLASSERT ((int)sizeof(((lnet_hdr_t *)0)->msg.hello.type) == 4); -} - -lnd_t * -lnet_find_lnd_by_type (int type) -{ - lnd_t *lnd; - struct list_head *tmp; - - /* holding lnd mutex */ - list_for_each (tmp, &the_lnet.ln_lnds) { - lnd = list_entry(tmp, lnd_t, lnd_list); - - if ((int)lnd->lnd_type == type) - return lnd; - } - - return NULL; + if (lnd->lnd_type == type) + return lnd; + } + return NULL; } void lnet_register_lnd (lnd_t *lnd) { - LNET_MUTEX_DOWN(&the_lnet.ln_lnd_mutex); + mutex_lock(&the_lnet.ln_lnd_mutex); - LASSERT (the_lnet.ln_init); - LASSERT (libcfs_isknown_lnd(lnd->lnd_type)); - LASSERT (lnet_find_lnd_by_type(lnd->lnd_type) == NULL); + LASSERT(libcfs_isknown_lnd(lnd->lnd_type)); + LASSERT(lnet_find_lnd_by_type(lnd->lnd_type) == NULL); - list_add_tail (&lnd->lnd_list, &the_lnet.ln_lnds); - lnd->lnd_refcount = 0; + list_add_tail(&lnd->lnd_list, &the_lnet.ln_lnds); + lnd->lnd_refcount = 0; - CDEBUG(D_NET, "%s LND registered\n", libcfs_lnd2str(lnd->lnd_type)); + CDEBUG(D_NET, "%s LND registered\n", libcfs_lnd2str(lnd->lnd_type)); - LNET_MUTEX_UP(&the_lnet.ln_lnd_mutex); + mutex_unlock(&the_lnet.ln_lnd_mutex); } +EXPORT_SYMBOL(lnet_register_lnd); void lnet_unregister_lnd (lnd_t *lnd) { - LNET_MUTEX_DOWN(&the_lnet.ln_lnd_mutex); + mutex_lock(&the_lnet.ln_lnd_mutex); - LASSERT (the_lnet.ln_init); - LASSERT (lnet_find_lnd_by_type(lnd->lnd_type) == lnd); - LASSERT (lnd->lnd_refcount == 0); + LASSERT(lnet_find_lnd_by_type(lnd->lnd_type) == lnd); + LASSERT(lnd->lnd_refcount == 0); - list_del (&lnd->lnd_list); - CDEBUG(D_NET, "%s LND unregistered\n", libcfs_lnd2str(lnd->lnd_type)); + list_del(&lnd->lnd_list); + CDEBUG(D_NET, "%s LND unregistered\n", libcfs_lnd2str(lnd->lnd_type)); - LNET_MUTEX_UP(&the_lnet.ln_lnd_mutex); + mutex_unlock(&the_lnet.ln_lnd_mutex); } +EXPORT_SYMBOL(lnet_unregister_lnd); -#ifndef LNET_USE_LIB_FREELIST - -int -lnet_descriptor_setup (void) +void +lnet_counters_get(lnet_counters_t *counters) { - return 0; + lnet_counters_t *ctr; + int i; + + memset(counters, 0, sizeof(*counters)); + + lnet_net_lock(LNET_LOCK_EX); + + cfs_percpt_for_each(ctr, i, the_lnet.ln_counters) { + counters->msgs_max += ctr->msgs_max; + counters->msgs_alloc += ctr->msgs_alloc; + counters->errors += ctr->errors; + counters->send_count += ctr->send_count; + counters->recv_count += ctr->recv_count; + counters->route_count += ctr->route_count; + counters->drop_count += ctr->drop_count; + counters->send_length += ctr->send_length; + counters->recv_length += ctr->recv_length; + counters->route_length += ctr->route_length; + counters->drop_length += ctr->drop_length; + + } + lnet_net_unlock(LNET_LOCK_EX); } +EXPORT_SYMBOL(lnet_counters_get); void -lnet_descriptor_cleanup (void) +lnet_counters_reset(void) +{ + lnet_counters_t *counters; + int i; + + lnet_net_lock(LNET_LOCK_EX); + + cfs_percpt_for_each(counters, i, the_lnet.ln_counters) + memset(counters, 0, sizeof(lnet_counters_t)); + + lnet_net_unlock(LNET_LOCK_EX); +} + +static char * +lnet_res_type2str(int type) { + switch (type) { + default: + LBUG(); + case LNET_COOKIE_TYPE_MD: + return "MD"; + case LNET_COOKIE_TYPE_ME: + return "ME"; + case LNET_COOKIE_TYPE_EQ: + return "EQ"; + } } -#else +static void +lnet_res_container_cleanup(struct lnet_res_container *rec) +{ + int count = 0; + + if (rec->rec_type == 0) /* not set yet, it's uninitialized */ + return; + + while (!list_empty(&rec->rec_active)) { + 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, lnet_eq_t, eq_list)); + + } else if (rec->rec_type == LNET_COOKIE_TYPE_MD) { + lnet_md_free(list_entry(e, lnet_libmd_t, md_list)); + + } else { /* NB: Active MEs should be attached on portals */ + LBUG(); + } + count++; + } + + if (count > 0) { + /* Found alive MD/ME/EQ, user really should unlink/free + * all of them before finalize LNet, but if someone didn't, + * we have to recycle garbage for him */ + CERROR("%d active elements on exit of %s container\n", + count, lnet_res_type2str(rec->rec_type)); + } + + if (rec->rec_lh_hash != NULL) { + LIBCFS_FREE(rec->rec_lh_hash, + LNET_LH_HASH_SIZE * sizeof(rec->rec_lh_hash[0])); + rec->rec_lh_hash = NULL; + } + + rec->rec_type = 0; /* mark it as finalized */ +} -int -lnet_freelist_init (lnet_freelist_t *fl, int n, int size) +static int +lnet_res_container_setup(struct lnet_res_container *rec, int cpt, int type) { - char *space; + int rc = 0; + int i; + + LASSERT(rec->rec_type == 0); - LASSERT (n > 0); + rec->rec_type = type; + INIT_LIST_HEAD(&rec->rec_active); - size += offsetof (lnet_freeobj_t, fo_contents); + rec->rec_lh_cookie = (cpt << LNET_COOKIE_TYPE_BITS) | type; - LIBCFS_ALLOC(space, n * size); - if (space == NULL) - return (-ENOMEM); + /* Arbitrary choice of hash table size */ + LIBCFS_CPT_ALLOC(rec->rec_lh_hash, lnet_cpt_table(), cpt, + LNET_LH_HASH_SIZE * sizeof(rec->rec_lh_hash[0])); + if (rec->rec_lh_hash == NULL) { + rc = -ENOMEM; + goto out; + } - CFS_INIT_LIST_HEAD (&fl->fl_list); - fl->fl_objs = space; - fl->fl_nobjs = n; - fl->fl_objsize = size; + for (i = 0; i < LNET_LH_HASH_SIZE; i++) + INIT_LIST_HEAD(&rec->rec_lh_hash[i]); - do - { - memset (space, 0, size); - list_add ((struct list_head *)space, &fl->fl_list); - space += size; - } while (--n != 0); + return 0; - return (0); +out: + CERROR("Failed to setup %s resource container\n", + lnet_res_type2str(type)); + lnet_res_container_cleanup(rec); + return rc; } -void -lnet_freelist_fini (lnet_freelist_t *fl) +static void +lnet_res_containers_destroy(struct lnet_res_container **recs) { - struct list_head *el; - int count; + struct lnet_res_container *rec; + int i; - if (fl->fl_nobjs == 0) - return; + cfs_percpt_for_each(rec, i, recs) + lnet_res_container_cleanup(rec); - count = 0; - for (el = fl->fl_list.next; el != &fl->fl_list; el = el->next) - count++; - - LASSERT (count == fl->fl_nobjs); - - LIBCFS_FREE(fl->fl_objs, fl->fl_nobjs * fl->fl_objsize); - memset (fl, 0, sizeof (*fl)); + cfs_percpt_free(recs); } -int -lnet_descriptor_setup (void) +static struct lnet_res_container ** +lnet_res_containers_create(int type) { - /* NB on failure caller must still call lnet_descriptor_cleanup */ - /* ****** */ - int rc; + struct lnet_res_container **recs; + struct lnet_res_container *rec; + int rc; + int i; + + recs = cfs_percpt_alloc(lnet_cpt_table(), sizeof(*rec)); + if (recs == NULL) { + CERROR("Failed to allocate %s resource containers\n", + lnet_res_type2str(type)); + return NULL; + } + + cfs_percpt_for_each(rec, i, recs) { + rc = lnet_res_container_setup(rec, i, type); + if (rc != 0) { + lnet_res_containers_destroy(recs); + return NULL; + } + } + + return recs; +} - memset (&the_lnet.ln_free_mes, 0, sizeof (the_lnet.ln_free_mes)); - memset (&the_lnet.ln_free_msgs, 0, sizeof (the_lnet.ln_free_msgs)); - memset (&the_lnet.ln_free_mds, 0, sizeof (the_lnet.ln_free_mds)); - memset (&the_lnet.ln_free_eqs, 0, sizeof (the_lnet.ln_free_eqs)); +lnet_libhandle_t * +lnet_res_lh_lookup(struct lnet_res_container *rec, __u64 cookie) +{ + /* ALWAYS called with lnet_res_lock held */ + struct list_head *head; + lnet_libhandle_t *lh; + unsigned int hash; - rc = lnet_freelist_init(&the_lnet.ln_free_mes, - MAX_MES, sizeof (lnet_me_t)); - if (rc != 0) - return (rc); + if ((cookie & LNET_COOKIE_MASK) != rec->rec_type) + return NULL; - rc = lnet_freelist_init(&the_lnet.ln_free_msgs, - MAX_MSGS, sizeof (lnet_msg_t)); - if (rc != 0) - return (rc); + hash = cookie >> (LNET_COOKIE_TYPE_BITS + LNET_CPT_BITS); + head = &rec->rec_lh_hash[hash & LNET_LH_HASH_MASK]; - rc = lnet_freelist_init(&the_lnet.ln_free_mds, - MAX_MDS, sizeof (lnet_libmd_t)); - if (rc != 0) - return (rc); + list_for_each_entry(lh, head, lh_hash_chain) { + if (lh->lh_cookie == cookie) + return lh; + } - rc = lnet_freelist_init(&the_lnet.ln_free_eqs, - MAX_EQS, sizeof (lnet_eq_t)); - return (rc); + return NULL; } void -lnet_descriptor_cleanup (void) +lnet_res_lh_initialize(struct lnet_res_container *rec, lnet_libhandle_t *lh) { - lnet_freelist_fini (&the_lnet.ln_free_mes); - lnet_freelist_fini (&the_lnet.ln_free_msgs); - lnet_freelist_fini (&the_lnet.ln_free_mds); - lnet_freelist_fini (&the_lnet.ln_free_eqs); + /* ALWAYS called with lnet_res_lock held */ + unsigned int ibits = LNET_COOKIE_TYPE_BITS + LNET_CPT_BITS; + unsigned int hash; + + lh->lh_cookie = rec->rec_lh_cookie; + rec->rec_lh_cookie += 1 << ibits; + + hash = (lh->lh_cookie >> ibits) & LNET_LH_HASH_MASK; + + list_add(&lh->lh_hash_chain, &rec->rec_lh_hash[hash]); } -#endif +static int lnet_unprepare(void); -__u64 -lnet_create_interface_cookie (void) -{ - /* NB the interface cookie in wire handles guards against delayed - * replies and ACKs appearing valid after reboot. Initialisation time, - * even if it's only implemented to millisecond resolution is probably - * easily good enough. */ - struct timeval tv; - __u64 cookie; -#ifndef __KERNEL__ - int rc = gettimeofday (&tv, NULL); - LASSERT (rc == 0); -#else - do_gettimeofday(&tv); -#endif - cookie = tv.tv_sec; - cookie *= 1000000; - cookie += tv.tv_usec; - return cookie; +static int +lnet_prepare(lnet_pid_t requested_pid) +{ + /* Prepare to bring up the network */ + struct lnet_res_container **recs; + int rc = 0; + + if (requested_pid == LNET_PID_ANY) { + /* Don't instantiate LNET just for me */ + return -ENETDOWN; + } + + LASSERT(the_lnet.ln_refcount == 0); + + the_lnet.ln_routing = 0; + + LASSERT((requested_pid & LNET_PID_USERFLAG) == 0); + the_lnet.ln_pid = requested_pid; + + INIT_LIST_HEAD(&the_lnet.ln_test_peers); + INIT_LIST_HEAD(&the_lnet.ln_peers); + INIT_LIST_HEAD(&the_lnet.ln_nets); + INIT_LIST_HEAD(&the_lnet.ln_routers); + INIT_LIST_HEAD(&the_lnet.ln_drop_rules); + INIT_LIST_HEAD(&the_lnet.ln_delay_rules); + + rc = lnet_descriptor_setup(); + if (rc != 0) + goto failed; + + rc = lnet_create_remote_nets_table(); + if (rc != 0) + goto failed; + + /* + * NB the interface cookie in wire handles guards against delayed + * replies and ACKs appearing valid after reboot. + */ + the_lnet.ln_interface_cookie = ktime_get_real_ns(); + + the_lnet.ln_counters = cfs_percpt_alloc(lnet_cpt_table(), + sizeof(lnet_counters_t)); + if (the_lnet.ln_counters == NULL) { + CERROR("Failed to allocate counters for LNet\n"); + rc = -ENOMEM; + goto failed; + } + + rc = lnet_peer_tables_create(); + if (rc != 0) + goto failed; + + rc = lnet_msg_containers_create(); + if (rc != 0) + goto failed; + + rc = lnet_res_container_setup(&the_lnet.ln_eq_container, 0, + LNET_COOKIE_TYPE_EQ); + 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; + goto failed; + } + + the_lnet.ln_md_containers = recs; + + rc = lnet_portals_create(); + if (rc != 0) { + CERROR("Failed to create portals for LNet: %d\n", rc); + goto failed; + } + + return 0; + + failed: + lnet_unprepare(); + return rc; } -int -lnet_setup_handle_hash (void) +static int +lnet_unprepare (void) { - int i; + /* 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 + * with non-zero pending count) */ - /* Arbitrary choice of hash table size */ -#ifdef __KERNEL__ - the_lnet.ln_lh_hash_size = CFS_PAGE_SIZE / sizeof (struct list_head); -#else - the_lnet.ln_lh_hash_size = (MAX_MES + MAX_MDS + MAX_EQS)/4; -#endif - LIBCFS_ALLOC(the_lnet.ln_lh_hash_table, - the_lnet.ln_lh_hash_size * sizeof (struct list_head)); - if (the_lnet.ln_lh_hash_table == NULL) - return (-ENOMEM); + lnet_fail_nid(LNET_NID_ANY, 0); + + LASSERT(the_lnet.ln_refcount == 0); + LASSERT(list_empty(&the_lnet.ln_test_peers)); + LASSERT(list_empty(&the_lnet.ln_nets)); + + lnet_portals_destroy(); + + if (the_lnet.ln_md_containers != NULL) { + lnet_res_containers_destroy(the_lnet.ln_md_containers); + the_lnet.ln_md_containers = NULL; + } - for (i = 0; i < the_lnet.ln_lh_hash_size; i++) - CFS_INIT_LIST_HEAD (&the_lnet.ln_lh_hash_table[i]); + if (the_lnet.ln_me_containers != NULL) { + lnet_res_containers_destroy(the_lnet.ln_me_containers); + the_lnet.ln_me_containers = NULL; + } - the_lnet.ln_next_object_cookie = LNET_COOKIE_TYPES; + lnet_res_container_cleanup(&the_lnet.ln_eq_container); - return (0); + lnet_msg_containers_destroy(); + lnet_peer_tables_destroy(); + lnet_rtrpools_free(0); + + if (the_lnet.ln_counters != NULL) { + cfs_percpt_free(the_lnet.ln_counters); + the_lnet.ln_counters = NULL; + } + lnet_destroy_remote_nets_table(); + lnet_descriptor_cleanup(); + + return 0; } -void -lnet_cleanup_handle_hash (void) +lnet_ni_t * +lnet_net2ni_locked(__u32 net_id, int cpt) { - if (the_lnet.ln_lh_hash_table == NULL) - return; + struct lnet_ni *ni; + struct lnet_net *net; - LIBCFS_FREE(the_lnet.ln_lh_hash_table, - the_lnet.ln_lh_hash_size * sizeof (struct list_head)); + LASSERT(cpt != LNET_LOCK_EX); + + list_for_each_entry(net, &the_lnet.ln_nets, net_list) { + if (net->net_id == net_id) { + ni = list_entry(net->net_ni_list.next, struct lnet_ni, + ni_netlist); + return ni; + } + } + + return NULL; } -lnet_libhandle_t * -lnet_lookup_cookie (__u64 cookie, int type) +lnet_ni_t * +lnet_net2ni(__u32 net) { - /* ALWAYS called with LNET_LOCK held */ - struct list_head *list; - struct list_head *el; - unsigned int hash; + lnet_ni_t *ni; - if ((cookie & (LNET_COOKIE_TYPES - 1)) != type) - return (NULL); + lnet_net_lock(0); + ni = lnet_net2ni_locked(net, 0); + lnet_net_unlock(0); - hash = ((unsigned int)cookie) % the_lnet.ln_lh_hash_size; - list = &the_lnet.ln_lh_hash_table[hash]; + return ni; +} +EXPORT_SYMBOL(lnet_net2ni); - list_for_each (el, list) { - lnet_libhandle_t *lh = list_entry (el, lnet_libhandle_t, - lh_hash_chain); +struct lnet_net * +lnet_get_net_locked(__u32 net_id) +{ + struct lnet_net *net; - if (lh->lh_cookie == cookie) - return (lh); - } + list_for_each_entry(net, &the_lnet.ln_nets, net_list) { + if (net->net_id == net_id) + return net; + } - return (NULL); + return NULL; } -void -lnet_initialise_handle (lnet_libhandle_t *lh, int type) +unsigned int +lnet_nid_cpt_hash(lnet_nid_t nid, unsigned int number) { - /* ALWAYS called with LNET_LOCK held */ - unsigned int hash; + __u64 key = nid; + unsigned int val; + + LASSERT(number >= 1 && number <= LNET_CPT_NUMBER); - LASSERT (type >= 0 && type < LNET_COOKIE_TYPES); - lh->lh_cookie = the_lnet.ln_next_object_cookie | type; - the_lnet.ln_next_object_cookie += LNET_COOKIE_TYPES; + if (number == 1) + return 0; - hash = ((unsigned int)lh->lh_cookie) % the_lnet.ln_lh_hash_size; - list_add (&lh->lh_hash_chain, &the_lnet.ln_lh_hash_table[hash]); + val = hash_long(key, LNET_CPT_BITS); + /* NB: LNET_CP_NUMBER doesn't have to be PO2 */ + if (val < number) + return val; + + return (unsigned int)(key + val + (val >> 1)) % number; } -void -lnet_invalidate_handle (lnet_libhandle_t *lh) +int +lnet_cpt_of_nid_locked(lnet_nid_t nid, struct lnet_ni *ni) { - /* ALWAYS called with LNET_LOCK held */ - list_del (&lh->lh_hash_chain); + struct lnet_net *net; + + /* must called with hold of lnet_net_lock */ + if (LNET_CPT_NUMBER == 1) + return 0; /* the only one */ + + /* + * If NI is provided then use the CPT identified in the NI cpt + * list if one exists. If one doesn't exist, then that NI is + * associated with all CPTs and it follows that the net it belongs + * to is implicitly associated with all CPTs, so just hash the nid + * and return that. + */ + if (ni != NULL) { + if (ni->ni_cpts != NULL) + return ni->ni_cpts[lnet_nid_cpt_hash(nid, + ni->ni_ncpts)]; + else + return lnet_nid_cpt_hash(nid, LNET_CPT_NUMBER); + } + + /* no NI provided so look at the net */ + net = lnet_get_net_locked(LNET_NIDNET(nid)); + + if (net != NULL && net->net_cpts != NULL) { + return net->net_cpts[lnet_nid_cpt_hash(nid, net->net_ncpts)]; + } + + return lnet_nid_cpt_hash(nid, LNET_CPT_NUMBER); } int -lnet_init_finalizers(void) +lnet_cpt_of_nid(lnet_nid_t nid, struct lnet_ni *ni) { -#ifdef __KERNEL__ - int i; + int cpt; + int cpt2; - the_lnet.ln_nfinalizers = (int) num_online_cpus(); + if (LNET_CPT_NUMBER == 1) + return 0; /* the only one */ - LIBCFS_ALLOC(the_lnet.ln_finalizers, - the_lnet.ln_nfinalizers * - sizeof(*the_lnet.ln_finalizers)); - if (the_lnet.ln_finalizers == NULL) { - CERROR("Can't allocate ln_finalizers\n"); - return -ENOMEM; - } + cpt = lnet_net_lock_current(); - for (i = 0; i < the_lnet.ln_nfinalizers; i++) - the_lnet.ln_finalizers[i] = NULL; -#else - the_lnet.ln_finalizing = 0; -#endif + cpt2 = lnet_cpt_of_nid_locked(nid, ni); + + lnet_net_unlock(cpt); - CFS_INIT_LIST_HEAD(&the_lnet.ln_finalizeq); - return 0; + return cpt2; } +EXPORT_SYMBOL(lnet_cpt_of_nid); -void -lnet_fini_finalizers(void) +int +lnet_islocalnet(__u32 net_id) { -#ifdef __KERNEL__ - int i; + struct lnet_net *net; + int cpt; + bool local; - for (i = 0; i < the_lnet.ln_nfinalizers; i++) - LASSERT (the_lnet.ln_finalizers[i] == NULL); + cpt = lnet_net_lock_current(); - LIBCFS_FREE(the_lnet.ln_finalizers, - the_lnet.ln_nfinalizers * - sizeof(*the_lnet.ln_finalizers)); -#else - LASSERT (!the_lnet.ln_finalizing); -#endif - LASSERT (list_empty(&the_lnet.ln_finalizeq)); -} + net = lnet_get_net_locked(net_id); -#ifndef __KERNEL__ -/* Temporary workaround to allow uOSS and test programs force server - * mode in userspace. See comments near ln_server_mode_flag in - * lnet/lib-types.h */ + local = net != NULL; -void -lnet_server_mode() { - the_lnet.ln_server_mode_flag = 1; + lnet_net_unlock(cpt); + + return local; } -#endif -int -lnet_prepare(lnet_pid_t requested_pid) +bool +lnet_is_ni_healthy_locked(struct lnet_ni *ni) { - /* Prepare to bring up the network */ - int rc = 0; - int i; + if (ni->ni_state == LNET_NI_STATE_ACTIVE || + ni->ni_state == LNET_NI_STATE_DEGRADED) + return true; - LASSERT (the_lnet.ln_refcount == 0); + return false; +} + +lnet_ni_t * +lnet_nid2ni_locked(lnet_nid_t nid, int cpt) +{ + struct lnet_net *net; + struct lnet_ni *ni; - the_lnet.ln_routing = 0; + LASSERT(cpt != LNET_LOCK_EX); -#ifdef __KERNEL__ - LASSERT ((requested_pid & LNET_PID_USERFLAG) == 0); - the_lnet.ln_pid = requested_pid; -#else - if (the_lnet.ln_server_mode_flag) {/* server case (uOSS) */ - LASSERT ((requested_pid & LNET_PID_USERFLAG) == 0); + list_for_each_entry(net, &the_lnet.ln_nets, net_list) { + list_for_each_entry(ni, &net->net_ni_list, ni_netlist) { + if (ni->ni_nid == nid) + return ni; + } + } - if (cfs_curproc_uid())/* Only root can run user-space server */ - return -EPERM; - the_lnet.ln_pid = requested_pid; + return NULL; +} - } else {/* client case (liblustre) */ +lnet_ni_t * +lnet_nid2ni_addref(lnet_nid_t nid) +{ + lnet_ni_t *ni; - /* My PID must be unique on this node and flag I'm userspace */ - the_lnet.ln_pid = getpid() | LNET_PID_USERFLAG; - } -#endif + lnet_net_lock(0); + ni = lnet_nid2ni_locked(nid, 0); + if (ni) + lnet_ni_addref_locked(ni, 0); + lnet_net_unlock(0); - rc = lnet_descriptor_setup(); - if (rc != 0) - goto failed0; - - memset(&the_lnet.ln_counters, 0, - sizeof(the_lnet.ln_counters)); - - CFS_INIT_LIST_HEAD (&the_lnet.ln_active_msgs); - CFS_INIT_LIST_HEAD (&the_lnet.ln_active_mds); - CFS_INIT_LIST_HEAD (&the_lnet.ln_active_eqs); - CFS_INIT_LIST_HEAD (&the_lnet.ln_test_peers); - CFS_INIT_LIST_HEAD (&the_lnet.ln_nis); - CFS_INIT_LIST_HEAD (&the_lnet.ln_zombie_nis); - CFS_INIT_LIST_HEAD (&the_lnet.ln_remote_nets); - CFS_INIT_LIST_HEAD (&the_lnet.ln_routers); - - the_lnet.ln_interface_cookie = lnet_create_interface_cookie(); - - lnet_init_rtrpools(); - - rc = lnet_setup_handle_hash (); - if (rc != 0) - goto failed0; - - rc = lnet_create_peer_table(); - if (rc != 0) - goto failed1; - - rc = lnet_init_finalizers(); - if (rc != 0) - goto failed2; - - the_lnet.ln_nportals = MAX_PORTALS; - LIBCFS_ALLOC(the_lnet.ln_portals, - the_lnet.ln_nportals * - sizeof(*the_lnet.ln_portals)); - if (the_lnet.ln_portals == NULL) { - rc = -ENOMEM; - goto failed3; - } - - for (i = 0; i < the_lnet.ln_nportals; i++) { - CFS_INIT_LIST_HEAD(&(the_lnet.ln_portals[i].ptl_ml)); - CFS_INIT_LIST_HEAD(&(the_lnet.ln_portals[i].ptl_msgq)); - the_lnet.ln_portals[i].ptl_options = 0; - } - - return 0; - - failed3: - lnet_fini_finalizers(); - failed2: - lnet_destroy_peer_table(); - failed1: - lnet_cleanup_handle_hash(); - failed0: - lnet_descriptor_cleanup(); - return rc; + return ni; } +EXPORT_SYMBOL(lnet_nid2ni_addref); int -lnet_unprepare (void) +lnet_islocalnid(lnet_nid_t nid) { - int idx; - - /* 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 - * with non-zero pending count) */ - - lnet_fail_nid(LNET_NID_ANY, 0); + struct lnet_ni *ni; + int cpt; - LASSERT (list_empty(&the_lnet.ln_test_peers)); - LASSERT (the_lnet.ln_refcount == 0); - LASSERT (list_empty(&the_lnet.ln_nis)); - LASSERT (list_empty(&the_lnet.ln_zombie_nis)); - LASSERT (the_lnet.ln_nzombie_nis == 0); + cpt = lnet_net_lock_current(); + ni = lnet_nid2ni_locked(nid, cpt); + lnet_net_unlock(cpt); - for (idx = 0; idx < the_lnet.ln_nportals; idx++) { - LASSERT (list_empty(&the_lnet.ln_portals[idx].ptl_msgq)); + return ni != NULL; +} - while (!list_empty (&the_lnet.ln_portals[idx].ptl_ml)) { - lnet_me_t *me = list_entry (the_lnet.ln_portals[idx].ptl_ml.next, - lnet_me_t, me_list); +int +lnet_count_acceptor_nets(void) +{ + /* Return the # of NIs that need the acceptor. */ + int count = 0; + struct lnet_net *net; + int cpt; + + cpt = lnet_net_lock_current(); + list_for_each_entry(net, &the_lnet.ln_nets, net_list) { + /* all socklnd type networks should have the acceptor + * thread started */ + if (net->net_lnd->lnd_accept != NULL) + count++; + } + + lnet_net_unlock(cpt); + + return count; +} - CERROR ("Active me %p on exit\n", me); - list_del (&me->me_list); - lnet_me_free (me); - } - } +static struct lnet_ping_info * +lnet_ping_info_create(int num_ni) +{ + struct lnet_ping_info *ping_info; + unsigned int infosz; + + infosz = offsetof(struct lnet_ping_info, pi_ni[num_ni]); + LIBCFS_ALLOC(ping_info, infosz); + if (ping_info == NULL) { + CERROR("Can't allocate ping info[%d]\n", num_ni); + return NULL; + } + + ping_info->pi_nnis = num_ni; + ping_info->pi_pid = the_lnet.ln_pid; + ping_info->pi_magic = LNET_PROTO_PING_MAGIC; + ping_info->pi_features = LNET_PING_FEAT_NI_STATUS; + + return ping_info; +} - while (!list_empty (&the_lnet.ln_active_mds)) { - lnet_libmd_t *md = list_entry (the_lnet.ln_active_mds.next, - lnet_libmd_t, md_list); +static inline int +lnet_get_net_ni_count_locked(struct lnet_net *net) +{ + struct lnet_ni *ni; + int count = 0; - CERROR ("Active md %p on exit\n", md); - list_del_init (&md->md_list); - lnet_md_free (md); - } + list_for_each_entry(ni, &net->net_ni_list, ni_netlist) + count++; - while (!list_empty (&the_lnet.ln_active_eqs)) { - lnet_eq_t *eq = list_entry (the_lnet.ln_active_eqs.next, - lnet_eq_t, eq_list); + return count; +} - CERROR ("Active eq %p on exit\n", eq); - list_del (&eq->eq_list); - lnet_eq_free (eq); - } +static inline int +lnet_get_ni_count(void) +{ + struct lnet_ni *ni; + struct lnet_net *net; + int count = 0; - while (!list_empty (&the_lnet.ln_active_msgs)) { - lnet_msg_t *msg = list_entry (the_lnet.ln_active_msgs.next, - lnet_msg_t, msg_activelist); + lnet_net_lock(0); - CERROR ("Active msg %p on exit\n", msg); - LASSERT (msg->msg_onactivelist); - msg->msg_onactivelist = 0; - list_del (&msg->msg_activelist); - lnet_msg_free (msg); - } + list_for_each_entry(net, &the_lnet.ln_nets, net_list) { + list_for_each_entry(ni, &net->net_ni_list, ni_netlist) + count++; + } - LIBCFS_FREE(the_lnet.ln_portals, - the_lnet.ln_nportals * sizeof(*the_lnet.ln_portals)); + lnet_net_unlock(0); - lnet_free_rtrpools(); - lnet_fini_finalizers(); - lnet_destroy_peer_table(); - lnet_cleanup_handle_hash(); - lnet_descriptor_cleanup(); + return count; +} - return (0); +static inline void +lnet_ping_info_free(struct lnet_ping_info *pinfo) +{ + LIBCFS_FREE(pinfo, + offsetof(struct lnet_ping_info, + pi_ni[pinfo->pi_nnis])); } -lnet_ni_t * -lnet_net2ni_locked (__u32 net) +static void +lnet_ping_info_destroy(void) { - struct list_head *tmp; - lnet_ni_t *ni; + struct lnet_net *net; + struct lnet_ni *ni; + + lnet_net_lock(LNET_LOCK_EX); - list_for_each (tmp, &the_lnet.ln_nis) { - ni = list_entry(tmp, lnet_ni_t, ni_list); + list_for_each_entry(net, &the_lnet.ln_nets, net_list) { + list_for_each_entry(ni, &net->net_ni_list, ni_netlist) { + lnet_ni_lock(ni); + ni->ni_status = NULL; + lnet_ni_unlock(ni); + } + } - if (LNET_NIDNET(ni->ni_nid) == net) { - lnet_ni_addref_locked(ni); - return ni; - } - } + lnet_ping_info_free(the_lnet.ln_ping_info); + the_lnet.ln_ping_info = NULL; - return NULL; + lnet_net_unlock(LNET_LOCK_EX); } -int -lnet_islocalnet (__u32 net) +static void +lnet_ping_event_handler(lnet_event_t *event) { - lnet_ni_t *ni; + struct lnet_ping_info *pinfo = event->md.user_ptr; - LNET_LOCK(); - ni = lnet_net2ni_locked(net); - if (ni != NULL) - lnet_ni_decref_locked(ni); - LNET_UNLOCK(); + if (event->unlinked) + pinfo->pi_features = LNET_PING_FEAT_INVAL; +} - return ni != NULL; +static int +lnet_ping_info_setup(struct lnet_ping_info **ppinfo, lnet_handle_md_t *md_handle, + int ni_count, bool set_eq) +{ + lnet_handle_me_t me_handle; + lnet_process_id_t id = {LNET_NID_ANY, LNET_PID_ANY}; + lnet_md_t md = {NULL}; + int rc, rc2; + + if (set_eq) { + rc = LNetEQAlloc(0, lnet_ping_event_handler, + &the_lnet.ln_ping_target_eq); + if (rc != 0) { + CERROR("Can't allocate ping EQ: %d\n", rc); + return rc; + } + } + + *ppinfo = lnet_ping_info_create(ni_count); + if (*ppinfo == NULL) { + rc = -ENOMEM; + goto failed_0; + } + + rc = LNetMEAttach(LNET_RESERVED_PORTAL, id, + LNET_PROTO_PING_MATCHBITS, 0, + LNET_UNLINK, LNET_INS_AFTER, + &me_handle); + if (rc != 0) { + CERROR("Can't create ping ME: %d\n", rc); + goto failed_1; + } + + /* initialize md content */ + md.start = *ppinfo; + md.length = offsetof(struct lnet_ping_info, + pi_ni[(*ppinfo)->pi_nnis]); + md.threshold = LNET_MD_THRESH_INF; + md.max_size = 0; + md.options = LNET_MD_OP_GET | LNET_MD_TRUNCATE | + LNET_MD_MANAGE_REMOTE; + md.user_ptr = NULL; + md.eq_handle = the_lnet.ln_ping_target_eq; + md.user_ptr = *ppinfo; + + rc = LNetMDAttach(me_handle, md, LNET_RETAIN, md_handle); + if (rc != 0) { + CERROR("Can't attach ping MD: %d\n", rc); + goto failed_2; + } + + return 0; + +failed_2: + rc2 = LNetMEUnlink(me_handle); + LASSERT(rc2 == 0); +failed_1: + lnet_ping_info_free(*ppinfo); + *ppinfo = NULL; +failed_0: + if (set_eq) + LNetEQFree(the_lnet.ln_ping_target_eq); + return rc; } -lnet_ni_t * -lnet_nid2ni_locked (lnet_nid_t nid) +static void +lnet_ping_md_unlink(struct lnet_ping_info *pinfo, lnet_handle_md_t *md_handle) { - struct list_head *tmp; - lnet_ni_t *ni; + sigset_t blocked = cfs_block_allsigs(); - list_for_each (tmp, &the_lnet.ln_nis) { - ni = list_entry(tmp, lnet_ni_t, ni_list); + LNetMDUnlink(*md_handle); + LNetInvalidateHandle(md_handle); - if (ni->ni_nid == nid) { - lnet_ni_addref_locked(ni); - return ni; - } - } + /* NB md could be busy; this just starts the unlink */ + while (pinfo->pi_features != LNET_PING_FEAT_INVAL) { + CDEBUG(D_NET, "Still waiting for ping MD to unlink\n"); + set_current_state(TASK_UNINTERRUPTIBLE); + schedule_timeout(cfs_time_seconds(1)); + } - return NULL; + cfs_restore_sigs(blocked); } -int -lnet_islocalnid (lnet_nid_t nid) +static void +lnet_ping_info_install_locked(struct lnet_ping_info *ping_info) { - lnet_ni_t *ni; + int i; + struct lnet_ni *ni; + struct lnet_net *net; + struct lnet_ni_status *ns; + + i = 0; + list_for_each_entry(net, &the_lnet.ln_nets, net_list) { + list_for_each_entry(ni, &net->net_ni_list, ni_netlist) { + LASSERT(i < ping_info->pi_nnis); + + ns = &ping_info->pi_ni[i]; + + ns->ns_nid = ni->ni_nid; - LNET_LOCK(); - ni = lnet_nid2ni_locked(nid); - if (ni != NULL) - lnet_ni_decref_locked(ni); - LNET_UNLOCK(); + lnet_ni_lock(ni); + ns->ns_status = (ni->ni_status != NULL) ? + ni->ni_status->ns_status : + LNET_NI_STATUS_UP; + ni->ni_status = ns; + lnet_ni_unlock(ni); - return ni != NULL; + i++; + } + + } } -int -lnet_count_acceptor_nis (void) +static void +lnet_ping_target_update(struct lnet_ping_info *pinfo, lnet_handle_md_t md_handle) { - /* Return the # of NIs that need the acceptor. */ - int count = 0; -#if defined(__KERNEL__) || defined(HAVE_LIBPTHREAD) - struct list_head *tmp; - lnet_ni_t *ni; + struct lnet_ping_info *old_pinfo = NULL; + lnet_handle_md_t old_md; + + /* switch the NIs to point to the new ping info created */ + lnet_net_lock(LNET_LOCK_EX); + + if (!the_lnet.ln_routing) + pinfo->pi_features |= LNET_PING_FEAT_RTE_DISABLED; + lnet_ping_info_install_locked(pinfo); + + if (the_lnet.ln_ping_info != NULL) { + old_pinfo = the_lnet.ln_ping_info; + old_md = the_lnet.ln_ping_target_md; + } + the_lnet.ln_ping_target_md = md_handle; + the_lnet.ln_ping_info = pinfo; + + lnet_net_unlock(LNET_LOCK_EX); + + if (old_pinfo != NULL) { + /* unlink the old ping info */ + lnet_ping_md_unlink(old_pinfo, &old_md); + lnet_ping_info_free(old_pinfo); + } +} - LNET_LOCK(); - list_for_each (tmp, &the_lnet.ln_nis) { - ni = list_entry(tmp, lnet_ni_t, ni_list); +static void +lnet_ping_target_fini(void) +{ + int rc; - if (ni->ni_lnd->lnd_accept != NULL) - count++; - } + lnet_ping_md_unlink(the_lnet.ln_ping_info, + &the_lnet.ln_ping_target_md); - LNET_UNLOCK(); + rc = LNetEQFree(the_lnet.ln_ping_target_eq); + LASSERT(rc == 0); -#endif /* defined(__KERNEL__) || defined(HAVE_LIBPTHREAD) */ - return count; + lnet_ping_info_destroy(); } -void -lnet_shutdown_lndnis (void) +static int +lnet_ni_tq_credits(lnet_ni_t *ni) { - int i; - int islo; - lnet_ni_t *ni; - - /* NB called holding the global mutex */ + int credits; - /* All quiet on the API front */ - LASSERT (!the_lnet.ln_shutdown); - LASSERT (the_lnet.ln_refcount == 0); - LASSERT (list_empty(&the_lnet.ln_zombie_nis)); - LASSERT (the_lnet.ln_nzombie_nis == 0); - LASSERT (list_empty(&the_lnet.ln_remote_nets)); + LASSERT(ni->ni_ncpts >= 1); - LNET_LOCK(); - the_lnet.ln_shutdown = 1; /* flag shutdown */ + if (ni->ni_ncpts == 1) + return ni->ni_net->net_tunables.lct_max_tx_credits; - /* Unlink NIs from the global table */ - while (!list_empty(&the_lnet.ln_nis)) { - ni = list_entry(the_lnet.ln_nis.next, - lnet_ni_t, ni_list); - list_del (&ni->ni_list); + credits = ni->ni_net->net_tunables.lct_max_tx_credits / ni->ni_ncpts; + credits = max(credits, 8 * ni->ni_net->net_tunables.lct_peer_tx_credits); + credits = min(credits, ni->ni_net->net_tunables.lct_max_tx_credits); - the_lnet.ln_nzombie_nis++; - lnet_ni_decref_locked(ni); /* drop apini's ref */ - } + return credits; +} - /* Drop the cached eqwait NI. */ - if (the_lnet.ln_eqwaitni != NULL) { - lnet_ni_decref_locked(the_lnet.ln_eqwaitni); - the_lnet.ln_eqwaitni = NULL; - } +static void +lnet_ni_unlink_locked(lnet_ni_t *ni) +{ + if (!list_empty(&ni->ni_cptlist)) { + list_del_init(&ni->ni_cptlist); + lnet_ni_decref_locked(ni, 0); + } + + /* move it to zombie list and nobody can find it anymore */ + LASSERT(!list_empty(&ni->ni_netlist)); + list_move(&ni->ni_netlist, &ni->ni_net->net_ni_zombie); + lnet_ni_decref_locked(ni, 0); +} - /* Drop the cached loopback NI. */ - if (the_lnet.ln_loni != NULL) { - lnet_ni_decref_locked(the_lnet.ln_loni); - the_lnet.ln_loni = NULL; - } +static void +lnet_clear_zombies_nis_locked(struct lnet_net *net) +{ + int i; + int islo; + lnet_ni_t *ni; + struct list_head *zombie_list = &net->net_ni_zombie; + + /* + * Now wait for the NIs I just nuked to show up on the zombie + * list and shut them down in guaranteed thread context + */ + i = 2; + while (!list_empty(zombie_list)) { + int *ref; + int j; + + ni = list_entry(zombie_list->next, + lnet_ni_t, ni_netlist); + list_del_init(&ni->ni_netlist); + /* the ni should be in deleting state. If it's not it's + * a bug */ + LASSERT(ni->ni_state == LNET_NI_STATE_DELETING); + cfs_percpt_for_each(ref, j, ni->ni_refs) { + if (*ref == 0) + continue; + /* still busy, add it back to zombie list */ + list_add(&ni->ni_netlist, zombie_list); + break; + } + + if (!list_empty(&ni->ni_netlist)) { + lnet_net_unlock(LNET_LOCK_EX); + ++i; + if ((i & (-i)) == i) { + CDEBUG(D_WARNING, + "Waiting for zombie LNI %s\n", + libcfs_nid2str(ni->ni_nid)); + } + set_current_state(TASK_UNINTERRUPTIBLE); + schedule_timeout(cfs_time_seconds(1)); + lnet_net_lock(LNET_LOCK_EX); + continue; + } + + lnet_net_unlock(LNET_LOCK_EX); + + islo = ni->ni_net->net_lnd->lnd_type == LOLND; + + LASSERT(!in_interrupt()); + (net->net_lnd->lnd_shutdown)(ni); + + if (!islo) + CDEBUG(D_LNI, "Removed LNI %s\n", + libcfs_nid2str(ni->ni_nid)); + + lnet_ni_free(ni); + i = 2; + lnet_net_lock(LNET_LOCK_EX); + } +} - LNET_UNLOCK(); +/* shutdown down the NI and release refcount */ +static void +lnet_shutdown_lndni(struct lnet_ni *ni) +{ + int i; + struct lnet_net *net = ni->ni_net; - /* Clear lazy portals and drop delayed messages which hold refs - * on their lnet_msg_t::msg_rxpeer */ - for (i = 0; i < the_lnet.ln_nportals; i++) - LNetClearLazyPortal(i); + lnet_net_lock(LNET_LOCK_EX); + ni->ni_state = LNET_NI_STATE_DELETING; + lnet_ni_unlink_locked(ni); + lnet_net_unlock(LNET_LOCK_EX); - /* Clear the peer table and wait for all peers to go (they hold refs on - * their NIs) */ - lnet_clear_peer_table(); + /* clear messages for this NI on the lazy portal */ + for (i = 0; i < the_lnet.ln_nportals; i++) + lnet_clear_lazy_portal(ni, i, "Shutting down NI"); - LNET_LOCK(); - /* Now wait for the NI's I just nuked to show up on apini_zombie_nis - * and shut them down in guaranteed thread context */ - i = 2; - while (the_lnet.ln_nzombie_nis != 0) { + /* Do peer table cleanup for this ni */ + lnet_peer_tables_cleanup(ni); - while (list_empty(&the_lnet.ln_zombie_nis)) { - LNET_UNLOCK(); - ++i; - if ((i & (-i)) == i) - CDEBUG(D_WARNING,"Waiting for %d zombie NIs\n", - the_lnet.ln_nzombie_nis); - cfs_pause(cfs_time_seconds(1)); - LNET_LOCK(); - } + lnet_net_lock(LNET_LOCK_EX); + lnet_clear_zombies_nis_locked(net); + lnet_net_unlock(LNET_LOCK_EX); +} - ni = list_entry(the_lnet.ln_zombie_nis.next, - lnet_ni_t, ni_list); - list_del(&ni->ni_list); - ni->ni_lnd->lnd_refcount--; +static void +lnet_shutdown_lndnet(struct lnet_net *net) +{ + struct lnet_ni *ni; - LNET_UNLOCK(); + lnet_net_lock(LNET_LOCK_EX); - islo = ni->ni_lnd->lnd_type == LOLND; + net->net_state = LNET_NET_STATE_DELETING; - LASSERT (!in_interrupt ()); - (ni->ni_lnd->lnd_shutdown)(ni); + list_del_init(&net->net_list); - /* can't deref lnd anymore now; it might have unregistered - * itself... */ + while (!list_empty(&net->net_ni_list)) { + ni = list_entry(net->net_ni_list.next, + lnet_ni_t, ni_netlist); + lnet_net_unlock(LNET_LOCK_EX); + lnet_shutdown_lndni(ni); + lnet_net_lock(LNET_LOCK_EX); + } - if (!islo) - CDEBUG(D_LNI, "Removed LNI %s\n", - libcfs_nid2str(ni->ni_nid)); + /* + * decrement ref count on lnd only when the entire network goes + * away + */ + net->net_lnd->lnd_refcount--; - LIBCFS_FREE(ni, sizeof(*ni)); + lnet_net_unlock(LNET_LOCK_EX); - LNET_LOCK(); - the_lnet.ln_nzombie_nis--; - } + lnet_net_free(net); +} - the_lnet.ln_shutdown = 0; - LNET_UNLOCK(); +static void +lnet_shutdown_lndnets(void) +{ + struct lnet_net *net; + + /* NB called holding the global mutex */ + + /* All quiet on the API front */ + LASSERT(!the_lnet.ln_shutdown); + LASSERT(the_lnet.ln_refcount == 0); + + lnet_net_lock(LNET_LOCK_EX); + the_lnet.ln_shutdown = 1; /* flag shutdown */ + + 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); + } + + /* Drop the cached loopback Net. */ + if (the_lnet.ln_loni != NULL) { + lnet_ni_decref_locked(the_lnet.ln_loni, 0); + the_lnet.ln_loni = NULL; + } + lnet_net_unlock(LNET_LOCK_EX); + + /* iterate through the net zombie list and delete each net */ + while (!list_empty(&the_lnet.ln_net_zombie)) { + net = list_entry(the_lnet.ln_net_zombie.next, + struct lnet_net, net_list); + lnet_shutdown_lndnet(net); + } + + lnet_net_lock(LNET_LOCK_EX); + the_lnet.ln_shutdown = 0; + lnet_net_unlock(LNET_LOCK_EX); +} - if (the_lnet.ln_network_tokens != NULL) { - LIBCFS_FREE(the_lnet.ln_network_tokens, - the_lnet.ln_network_tokens_nob); - the_lnet.ln_network_tokens = NULL; - } +static int +lnet_startup_lndni(struct lnet_ni *ni, struct lnet_lnd_tunables *tun) +{ + int rc = -EINVAL; + struct lnet_tx_queue *tq; + int i; + struct lnet_net *net = ni->ni_net; + + mutex_lock(&the_lnet.ln_lnd_mutex); + + if (tun) { + memcpy(&ni->ni_lnd_tunables, tun, sizeof(*tun)); + ni->ni_lnd_tunables_set = true; + } + + rc = (net->net_lnd->lnd_startup)(ni); + + mutex_unlock(&the_lnet.ln_lnd_mutex); + + if (rc != 0) { + LCONSOLE_ERROR_MSG(0x105, "Error %d starting up LNI %s\n", + rc, libcfs_lnd2str(net->net_lnd->lnd_type)); + lnet_net_lock(LNET_LOCK_EX); + net->net_lnd->lnd_refcount--; + lnet_net_unlock(LNET_LOCK_EX); + goto failed0; + } + + ni->ni_state = LNET_NI_STATE_ACTIVE; + + /* We keep a reference on the loopback net through the loopback NI */ + if (net->net_lnd->lnd_type == LOLND) { + lnet_ni_addref(ni); + LASSERT(the_lnet.ln_loni == NULL); + the_lnet.ln_loni = ni; + ni->ni_net->net_tunables.lct_peer_tx_credits = 0; + ni->ni_net->net_tunables.lct_peer_rtr_credits = 0; + ni->ni_net->net_tunables.lct_max_tx_credits = 0; + ni->ni_net->net_tunables.lct_peer_timeout = 0; + return 0; + } + + if (ni->ni_net->net_tunables.lct_peer_tx_credits == 0 || + ni->ni_net->net_tunables.lct_max_tx_credits == 0) { + LCONSOLE_ERROR_MSG(0x107, "LNI %s has no %scredits\n", + libcfs_lnd2str(net->net_lnd->lnd_type), + ni->ni_net->net_tunables.lct_peer_tx_credits == 0 ? + "" : "per-peer "); + /* shutdown the NI since if we get here then it must've already + * been started + */ + lnet_shutdown_lndni(ni); + return -EINVAL; + } + + cfs_percpt_for_each(tq, i, ni->ni_tx_queues) { + tq->tq_credits_min = + tq->tq_credits_max = + tq->tq_credits = lnet_ni_tq_credits(ni); + } + + CDEBUG(D_LNI, "Added LNI %s [%d/%d/%d/%d]\n", + libcfs_nid2str(ni->ni_nid), + ni->ni_net->net_tunables.lct_peer_tx_credits, + lnet_ni_tq_credits(ni) * LNET_CPT_NUMBER, + ni->ni_net->net_tunables.lct_peer_rtr_credits, + ni->ni_net->net_tunables.lct_peer_timeout); + + return 0; +failed0: + lnet_ni_free(ni); + return rc; } -int -lnet_startup_lndnis (void) -{ - lnd_t *lnd; - lnet_ni_t *ni; - struct list_head nilist; - int rc = 0; - int lnd_type; - int nicount = 0; - char *nets = lnet_get_networks(); - - CFS_INIT_LIST_HEAD(&nilist); - - if (nets == NULL) - goto failed; - - rc = lnet_parse_networks(&nilist, nets); - if (rc != 0) - goto failed; - - while (!list_empty(&nilist)) { - ni = list_entry(nilist.next, lnet_ni_t, ni_list); - lnd_type = LNET_NETTYP(LNET_NIDNET(ni->ni_nid)); - - LASSERT (libcfs_isknown_lnd(lnd_type)); - - LNET_MUTEX_DOWN(&the_lnet.ln_lnd_mutex); - lnd = lnet_find_lnd_by_type(lnd_type); - -#ifdef __KERNEL__ - if (lnd == NULL) { - LNET_MUTEX_UP(&the_lnet.ln_lnd_mutex); - rc = request_module("%s", libcfs_lnd2modname(lnd_type)); - LNET_MUTEX_DOWN(&the_lnet.ln_lnd_mutex); - - lnd = lnet_find_lnd_by_type(lnd_type); - if (lnd == NULL) { - LNET_MUTEX_UP(&the_lnet.ln_lnd_mutex); - CERROR("Can't load LND %s, module %s, rc=%d\n", - libcfs_lnd2str(lnd_type), - libcfs_lnd2modname(lnd_type), rc); -#ifndef CONFIG_KMOD - LCONSOLE_ERROR_MSG(0x104, "Your kernel must be " - "compiled with CONFIG_KMOD set for " - "automatic module loading."); -#endif - goto failed; - } - } -#else - if (lnd == NULL) { - LNET_MUTEX_UP(&the_lnet.ln_lnd_mutex); - CERROR("LND %s not supported\n", - libcfs_lnd2str(lnd_type)); - goto failed; - } +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; + lnd_t *lnd; + int peer_timeout = + net->net_tunables.lct_peer_timeout; + int maxtxcredits = + net->net_tunables.lct_max_tx_credits; + 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 + * 'net' is just a convenient way to pass in the list. + * if it is unique we need to find the LND and load it if + * necessary. + */ + if (lnet_net_unique(net->net_id, &the_lnet.ln_nets, &net_l)) { + lnd_type = LNET_NETTYP(net->net_id); + + LASSERT(libcfs_isknown_lnd(lnd_type)); + + if (lnd_type == CIBLND || lnd_type == OPENIBLND || + lnd_type == IIBLND || lnd_type == VIBLND) { + CERROR("LND %s obsoleted\n", libcfs_lnd2str(lnd_type)); + rc = -EINVAL; + goto failed0; + } + + mutex_lock(&the_lnet.ln_lnd_mutex); + lnd = lnet_find_lnd_by_type(lnd_type); + + if (lnd == NULL) { + mutex_unlock(&the_lnet.ln_lnd_mutex); + rc = request_module("%s", libcfs_lnd2modname(lnd_type)); + mutex_lock(&the_lnet.ln_lnd_mutex); + + lnd = lnet_find_lnd_by_type(lnd_type); + if (lnd == NULL) { + mutex_unlock(&the_lnet.ln_lnd_mutex); + CERROR("Can't load LND %s, module %s, rc=%d\n", + libcfs_lnd2str(lnd_type), + libcfs_lnd2modname(lnd_type), rc); +#ifndef HAVE_MODULE_LOADING_SUPPORT + LCONSOLE_ERROR_MSG(0x104, "Your kernel must be " + "compiled with kernel module " + "loading support."); #endif + rc = -EINVAL; + goto failed0; + } + } + + lnet_net_lock(LNET_LOCK_EX); + lnd->lnd_refcount++; + lnet_net_unlock(LNET_LOCK_EX); + + net->net_lnd = lnd; + + mutex_unlock(&the_lnet.ln_lnd_mutex); + + net_l = net; + } + + /* + * net_l: if the network being added is unique then net_l + * will point to that network + * if the network being added is not unique then + * net_l points to the existing network. + * + * When we enter the loop below, we'll pick NIs off he + * network beign added and start them up, then add them to + * a local ni list. Once we've successfully started all + * the NIs then we join the local NI list (of started up + * networks) with the net_l->net_ni_list, which should + * point to the correct network to add the new ni list to + * + * If any of the new NIs fail to start up, then we want to + * iterate through the local ni list, which should include + * any NIs which were successfully started up, and shut + * them down. + * + * After than we want to delete the network being added, + * to avoid a memory leak. + */ + + /* + * When a network uses TCP bonding then all its interfaces + * must be specified when the network is first defined: the + * TCP bonding code doesn't allow for interfaces to be added + * or removed. + */ + if (net_l != net && net_l != NULL && use_tcp_bonding && + LNET_NETTYP(net_l->net_id) == SOCKLND) { + rc = -EINVAL; + goto failed0; + } + + while (!list_empty(&net->net_ni_added)) { + ni = list_entry(net->net_ni_added.next, struct lnet_ni, + ni_netlist); + list_del_init(&ni->ni_netlist); + + /* make sure that the the NI we're about to start + * up is actually unique. if it's not fail. */ + if (!lnet_ni_unique_net(&net_l->net_ni_list, + ni->ni_interfaces[0])) { + rc = -EINVAL; + goto failed1; + } + + /* adjust the pointer the parent network, just in case it + * the net is a duplicate */ + ni->ni_net = net_l; + + rc = lnet_startup_lndni(ni, tun); + + LASSERT(ni->ni_net->net_tunables.lct_peer_timeout <= 0 || + ni->ni_net->net_lnd->lnd_query != NULL); + + if (rc < 0) + goto failed1; + + lnet_ni_addref(ni); + list_add_tail(&ni->ni_netlist, &local_ni_list); + + ni_count++; + } + + lnet_net_lock(LNET_LOCK_EX); + list_splice_tail(&local_ni_list, &net_l->net_ni_list); + lnet_incr_dlc_seq(); + lnet_net_unlock(LNET_LOCK_EX); + + /* if the network is not unique then we don't want to keep + * it around after we're done. Free it. Otherwise add that + * net to the global the_lnet.ln_nets */ + if (net_l != net && net_l != NULL) { + /* + * TODO - note. currently the tunables can not be updated + * once added + */ + lnet_net_free(net); + } else { + net->net_state = LNET_NET_STATE_ACTIVE; + /* + * restore tunables after it has been overwitten by the + * lnd + */ + if (peer_timeout != -1) + net->net_tunables.lct_peer_timeout = peer_timeout; + if (maxtxcredits != -1) + net->net_tunables.lct_max_tx_credits = maxtxcredits; + if (peerrtrcredits != -1) + net->net_tunables.lct_peer_rtr_credits = peerrtrcredits; + + lnet_net_lock(LNET_LOCK_EX); + list_add_tail(&net->net_list, &the_lnet.ln_nets); + lnet_net_unlock(LNET_LOCK_EX); + } + + return ni_count; + +failed1: + /* + * shutdown the new NIs that are being started up + * free the NET being started + */ + while (!list_empty(&local_ni_list)) { + ni = list_entry(local_ni_list.next, struct lnet_ni, + ni_netlist); + + lnet_shutdown_lndni(ni); + } + +failed0: + lnet_net_free(net); + + return rc; +} - ni->ni_refcount = 1; - - LNET_LOCK(); - lnd->lnd_refcount++; - LNET_UNLOCK(); - - ni->ni_lnd = lnd; - - rc = (lnd->lnd_startup)(ni); - - LNET_MUTEX_UP(&the_lnet.ln_lnd_mutex); - - if (rc != 0) { - LCONSOLE_ERROR_MSG(0x105, "Error %d starting up LNI %s" - "\n", - rc, libcfs_lnd2str(lnd->lnd_type)); - LNET_LOCK(); - lnd->lnd_refcount--; - LNET_UNLOCK(); - goto failed; - } - - LASSERT (ni->ni_peertimeout <= 0 || lnd->lnd_query != NULL); - - list_del(&ni->ni_list); - - LNET_LOCK(); - list_add_tail(&ni->ni_list, &the_lnet.ln_nis); - LNET_UNLOCK(); - - if (lnd->lnd_type == LOLND) { - lnet_ni_addref(ni); - LASSERT (the_lnet.ln_loni == NULL); - the_lnet.ln_loni = ni; - continue; - } - -#ifndef __KERNEL__ - if (lnd->lnd_wait != NULL) { - if (the_lnet.ln_eqwaitni == NULL) { - lnet_ni_addref(ni); - the_lnet.ln_eqwaitni = ni; - } - } else { -# ifndef HAVE_LIBPTHREAD - LCONSOLE_ERROR_MSG(0x106, "LND %s not supported in a " - "single-threaded runtime\n", - libcfs_lnd2str(lnd_type)); - goto failed; -# endif - } -#endif - if (ni->ni_peertxcredits == 0 || - ni->ni_maxtxcredits == 0) { - LCONSOLE_ERROR_MSG(0x107, "LNI %s has no %scredits\n", - libcfs_lnd2str(lnd->lnd_type), - ni->ni_peertxcredits == 0 ? - "" : "per-peer "); - goto failed; - } - - ni->ni_txcredits = ni->ni_mintxcredits = ni->ni_maxtxcredits; - - CDEBUG(D_LNI, "Added LNI %s [%d/%d/%d/%d]\n", - libcfs_nid2str(ni->ni_nid), - ni->ni_peertxcredits, ni->ni_txcredits, - ni->ni_peerrtrcredits, ni->ni_peertimeout); - - nicount++; - } - - if (the_lnet.ln_eqwaitni != NULL && nicount > 1) { - lnd_type = the_lnet.ln_eqwaitni->ni_lnd->lnd_type; - LCONSOLE_ERROR_MSG(0x109, "LND %s can only run single-network" - "\n", - libcfs_lnd2str(lnd_type)); - goto failed; - } - - return 0; +static int +lnet_startup_lndnets(struct list_head *netlist) +{ + struct lnet_net *net; + int rc; + int ni_count = 0; - failed: - lnet_shutdown_lndnis(); + while (!list_empty(netlist)) { + net = list_entry(netlist->next, struct lnet_net, net_list); + list_del_init(&net->net_list); - while (!list_empty(&nilist)) { - ni = list_entry(nilist.next, lnet_ni_t, ni_list); - list_del(&ni->ni_list); - LIBCFS_FREE(ni, sizeof(*ni)); - } + rc = lnet_startup_lndnet(net, NULL); - return -ENETDOWN; -} + if (rc < 0) + goto failed; -int -LNetInit(void) -{ - lnet_assert_wire_constants (); - LASSERT (!the_lnet.ln_init); - - memset(&the_lnet, 0, sizeof(the_lnet)); - - lnet_init_locks(); - CFS_INIT_LIST_HEAD(&the_lnet.ln_lnds); - the_lnet.ln_refcount = 0; - the_lnet.ln_init = 1; - -#ifdef __KERNEL__ - /* All LNDs apart from the LOLND are in separate modules. They - * register themselves when their module loads, and unregister - * themselves when their module is unloaded. */ -#else - /* Register LNDs - * NB the order here determines default 'networks=' order */ -# ifdef CRAY_XT3 - LNET_REGISTER_ULND(the_ptllnd); -# endif -# ifdef HAVE_LIBPTHREAD - LNET_REGISTER_ULND(the_tcplnd); -# endif -#endif - lnet_register_lnd(&the_lolnd); - return 0; + ni_count += rc; + } + + return ni_count; +failed: + lnet_shutdown_lndnets(); + + return rc; } -void -LNetFini(void) +/** + * Initialize LNet library. + * + * Automatically called at module loading time. Caller has to call + * lnet_lib_exit() after a call to lnet_lib_init(), if and only if the + * latter returned 0. It must be called exactly once. + * + * \retval 0 on success + * \retval -ve on failures. + */ +int lnet_lib_init(void) { - LASSERT (the_lnet.ln_init); - LASSERT (the_lnet.ln_refcount == 0); + int rc; + + lnet_assert_wire_constants(); + + memset(&the_lnet, 0, sizeof(the_lnet)); + + /* 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); + + LASSERT(the_lnet.ln_cpt_number > 0); + if (the_lnet.ln_cpt_number > LNET_CPT_MAX) { + /* we are under risk of consuming all lh_cookie */ + CERROR("Can't have %d CPTs for LNet (max allowed is %d), " + "please change setting of CPT-table and retry\n", + the_lnet.ln_cpt_number, LNET_CPT_MAX); + return -E2BIG; + } + + while ((1 << the_lnet.ln_cpt_bits) < the_lnet.ln_cpt_number) + the_lnet.ln_cpt_bits++; + + rc = lnet_create_locks(); + if (rc != 0) { + CERROR("Can't create LNet global locks: %d\n", rc); + return rc; + } + + the_lnet.ln_refcount = 0; + LNetInvalidateHandle(&the_lnet.ln_rc_eqh); + INIT_LIST_HEAD(&the_lnet.ln_lnds); + INIT_LIST_HEAD(&the_lnet.ln_net_zombie); + INIT_LIST_HEAD(&the_lnet.ln_rcd_zombie); + INIT_LIST_HEAD(&the_lnet.ln_rcd_deathrow); + + /* The hash table size is the number of bits it takes to express the set + * ln_num_routes, minus 1 (better to under estimate than over so we + * don't waste memory). */ + if (rnet_htable_size <= 0) + rnet_htable_size = LNET_REMOTE_NETS_HASH_DEFAULT; + else if (rnet_htable_size > LNET_REMOTE_NETS_HASH_MAX) + rnet_htable_size = LNET_REMOTE_NETS_HASH_MAX; + the_lnet.ln_remote_nets_hbits = max_t(int, 1, + order_base_2(rnet_htable_size) - 1); + + /* All LNDs apart from the LOLND are in separate modules. They + * register themselves when their module loads, and unregister + * themselves when their module is unloaded. */ + lnet_register_lnd(&the_lolnd); + return 0; +} - while (!list_empty(&the_lnet.ln_lnds)) - lnet_unregister_lnd(list_entry(the_lnet.ln_lnds.next, - lnd_t, lnd_list)); - lnet_fini_locks(); +/** + * Finalize LNet library. + * + * \pre lnet_lib_init() called with success. + * \pre All LNet users called LNetNIFini() for matching LNetNIInit() calls. + */ +void lnet_lib_exit(void) +{ + LASSERT(the_lnet.ln_refcount == 0); - the_lnet.ln_init = 0; + while (!list_empty(&the_lnet.ln_lnds)) + lnet_unregister_lnd(list_entry(the_lnet.ln_lnds.next, + lnd_t, lnd_list)); + lnet_destroy_locks(); } +/** + * Set LNet PID and start LNet interfaces, routing, and forwarding. + * + * Users must call this function at least once before any other functions. + * For each successful call there must be a corresponding call to + * LNetNIFini(). For subsequent calls to LNetNIInit(), \a requested_pid is + * ignored. + * + * The PID used by LNet may be different from the one requested. + * See LNetGetId(). + * + * \param requested_pid PID requested by the caller. + * + * \return >= 0 on success, and < 0 error code on failures. + */ int LNetNIInit(lnet_pid_t requested_pid) { - int im_a_router = 0; - int rc; - - LNET_MUTEX_DOWN(&the_lnet.ln_api_mutex); - - LASSERT (the_lnet.ln_init); - CDEBUG(D_OTHER, "refs %d\n", the_lnet.ln_refcount); - - if (the_lnet.ln_refcount > 0) { - rc = the_lnet.ln_refcount++; - goto out; - } + int im_a_router = 0; + int rc; + int ni_count; + struct lnet_ping_info *pinfo; + lnet_handle_md_t md_handle; + struct 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); + + if (the_lnet.ln_refcount > 0) { + rc = the_lnet.ln_refcount++; + mutex_unlock(&the_lnet.ln_api_mutex); + return rc; + } + + rc = lnet_prepare(requested_pid); + if (rc != 0) { + mutex_unlock(&the_lnet.ln_api_mutex); + return rc; + } + + /* create a network for Loopback network */ + net = lnet_net_alloc(LNET_MKNET(LOLND, 0), &net_head); + if (net == NULL) { + rc = -ENOMEM; + goto err_empty_list; + } + + /* Add in the loopback NI */ + if (lnet_ni_alloc(net, NULL, NULL) == NULL) { + rc = -ENOMEM; + goto err_empty_list; + } + + /* 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 + * to load networks, routes and forwarding from module parameters + * in this case. On cleanup in case of failure only clean up + * routes if it has been loaded */ + if (!the_lnet.ln_nis_from_mod_params) { + rc = lnet_parse_networks(&net_head, lnet_get_networks(), + use_tcp_bonding); + if (rc < 0) + goto err_empty_list; + } + + ni_count = lnet_startup_lndnets(&net_head); + if (ni_count < 0) { + rc = ni_count; + goto err_empty_list; + } + + if (!the_lnet.ln_nis_from_mod_params) { + rc = lnet_parse_routes(lnet_get_routes(), &im_a_router); + if (rc != 0) + goto err_shutdown_lndnis; + + rc = lnet_check_routes(); + if (rc != 0) + goto err_destroy_routes; + + rc = lnet_rtrpools_alloc(im_a_router); + if (rc != 0) + goto err_destroy_routes; + } + + rc = lnet_acceptor_start(); + if (rc != 0) + goto err_destroy_routes; + + the_lnet.ln_refcount = 1; + /* Now I may use my own API functions... */ + + rc = lnet_ping_info_setup(&pinfo, &md_handle, ni_count, true); + if (rc != 0) + goto err_acceptor_stop; + + lnet_ping_target_update(pinfo, md_handle); + + rc = lnet_router_checker_start(); + if (rc != 0) + goto err_stop_ping; + + lnet_fault_init(); + lnet_proc_init(); + + mutex_unlock(&the_lnet.ln_api_mutex); + + return 0; + +err_stop_ping: + lnet_ping_target_fini(); +err_acceptor_stop: + the_lnet.ln_refcount = 0; + lnet_acceptor_stop(); +err_destroy_routes: + if (!the_lnet.ln_nis_from_mod_params) + lnet_destroy_routes(); +err_shutdown_lndnis: + lnet_shutdown_lndnets(); +err_empty_list: + lnet_unprepare(); + LASSERT(rc < 0); + mutex_unlock(&the_lnet.ln_api_mutex); + while (!list_empty(&net_head)) { + struct lnet_net *net; + + net = list_entry(net_head.next, struct lnet_net, net_list); + list_del_init(&net->net_list); + lnet_net_free(net); + } + return rc; +} +EXPORT_SYMBOL(LNetNIInit); - lnet_get_tunables(); +/** + * Stop LNet interfaces, routing, and forwarding. + * + * Users must call this function once for each successful call to LNetNIInit(). + * Once the LNetNIFini() operation has been started, the results of pending + * API operations are undefined. + * + * \return always 0 for current implementation. + */ +int +LNetNIFini() +{ + mutex_lock(&the_lnet.ln_api_mutex); - if (requested_pid == LNET_PID_ANY) { - /* Don't instantiate LNET just for me */ - rc = -ENETDOWN; - goto failed0; - } + LASSERT(the_lnet.ln_refcount > 0); - rc = lnet_prepare(requested_pid); - if (rc != 0) - goto failed0; + if (the_lnet.ln_refcount != 1) { + the_lnet.ln_refcount--; + } else { + LASSERT(!the_lnet.ln_niinit_self); - rc = lnet_startup_lndnis(); - if (rc != 0) - goto failed1; + lnet_fault_fini(); - rc = lnet_parse_routes(lnet_get_routes(), &im_a_router); - if (rc != 0) - goto failed2; + lnet_proc_fini(); + lnet_router_checker_stop(); + lnet_ping_target_fini(); - rc = lnet_check_routes(); - if (rc != 0) - goto failed2; + /* Teardown fns that use my own API functions BEFORE here */ + the_lnet.ln_refcount = 0; - rc = lnet_alloc_rtrpools(im_a_router); - if (rc != 0) - goto failed2; + lnet_acceptor_stop(); + lnet_destroy_routes(); + lnet_shutdown_lndnets(); + lnet_unprepare(); + } - rc = lnet_acceptor_start(); - if (rc != 0) - goto failed2; + mutex_unlock(&the_lnet.ln_api_mutex); + return 0; +} +EXPORT_SYMBOL(LNetNIFini); - the_lnet.ln_refcount = 1; - /* Now I may use my own API functions... */ +/** + * Grabs the ni data from the ni structure and fills the out + * parameters + * + * \param[in] ni network interface structure + * \param[out] cpt_count the number of cpts the ni is on + * \param[out] nid Network Interface ID + * \param[out] peer_timeout NI peer timeout + * \param[out] peer_tx_crdits NI peer transmit credits + * \param[out] peer_rtr_credits NI peer router credits + * \param[out] max_tx_credits NI max transmit credit + * \param[out] net_config Network configuration + */ +static void +lnet_fill_ni_info(struct lnet_ni *ni, struct lnet_ioctl_config_data *config) +{ + struct lnet_ioctl_net_config *net_config; + struct lnet_ioctl_config_lnd_tunables *lnd_cfg = NULL; + size_t min_size, tunable_size = 0; + int i; + + if (!ni || !config) + return; + + net_config = (struct lnet_ioctl_net_config *) config->cfg_bulk; + if (!net_config) + return; + + BUILD_BUG_ON(ARRAY_SIZE(ni->ni_interfaces) != + ARRAY_SIZE(net_config->ni_interfaces)); + + for (i = 0; i < ARRAY_SIZE(ni->ni_interfaces); i++) { + if (!ni->ni_interfaces[i]) + break; + + strncpy(net_config->ni_interfaces[i], + ni->ni_interfaces[i], + sizeof(net_config->ni_interfaces[i])); + } + + config->cfg_nid = ni->ni_nid; + config->cfg_config_u.cfg_net.net_peer_timeout = + ni->ni_net->net_tunables.lct_peer_timeout; + config->cfg_config_u.cfg_net.net_max_tx_credits = + ni->ni_net->net_tunables.lct_max_tx_credits; + config->cfg_config_u.cfg_net.net_peer_tx_credits = + ni->ni_net->net_tunables.lct_peer_tx_credits; + config->cfg_config_u.cfg_net.net_peer_rtr_credits = + ni->ni_net->net_tunables.lct_peer_rtr_credits; + + net_config->ni_status = ni->ni_status->ns_status; + + if (ni->ni_cpts) { + int num_cpts = min(ni->ni_ncpts, LNET_MAX_SHOW_NUM_CPT); + + for (i = 0; i < num_cpts; i++) + net_config->ni_cpts[i] = ni->ni_cpts[i]; + + config->cfg_ncpts = num_cpts; + } + + /* + * See if user land tools sent in a newer and larger version + * of struct lnet_tunables than what the kernel uses. + */ + min_size = sizeof(*config) + sizeof(*net_config); + + if (config->cfg_hdr.ioc_len > min_size) + tunable_size = config->cfg_hdr.ioc_len - min_size; + + /* Don't copy too much data to user space */ + min_size = min(tunable_size, sizeof(ni->ni_lnd_tunables)); + lnd_cfg = (struct lnet_ioctl_config_lnd_tunables *)net_config->cfg_bulk; + + if (lnd_cfg && min_size) { + memcpy(&lnd_cfg->lt_tun, &ni->ni_lnd_tunables, min_size); + config->cfg_config_u.cfg_net.net_interface_count = 1; + + /* Tell user land that kernel side has less data */ + if (tunable_size > sizeof(ni->ni_lnd_tunables)) { + min_size = tunable_size - sizeof(ni->ni_lnd_tunables); + config->cfg_hdr.ioc_len -= min_size; + } + } +} - rc = lnet_router_checker_start(); - if (rc != 0) - goto failed3; +struct lnet_ni * +lnet_get_ni_idx_locked(int idx) +{ + struct lnet_ni *ni; + struct lnet_net *net; - rc = lnet_ping_target_init(); - if (rc != 0) - goto failed4; + list_for_each_entry(net, &the_lnet.ln_nets, net_list) { + list_for_each_entry(ni, &net->net_ni_list, ni_netlist) { + if (idx-- == 0) + return ni; + } + } - lnet_proc_init(); - goto out; + return NULL; +} - failed4: - lnet_router_checker_stop(); - failed3: - the_lnet.ln_refcount = 0; - lnet_acceptor_stop(); - failed2: - lnet_destroy_routes(); - lnet_shutdown_lndnis(); - failed1: - lnet_unprepare(); - failed0: - LASSERT (rc < 0); - out: - LNET_MUTEX_UP(&the_lnet.ln_api_mutex); - return rc; +struct lnet_ni * +lnet_get_next_ni_locked(struct lnet_net *mynet, struct lnet_ni *prev) +{ + struct lnet_ni *ni; + struct lnet_net *net = mynet; + + if (prev == NULL) { + if (net == NULL) + net = list_entry(the_lnet.ln_nets.next, struct lnet_net, + net_list); + ni = list_entry(net->net_ni_list.next, struct lnet_ni, + ni_netlist); + + return ni; + } + + if (prev->ni_netlist.next == &prev->ni_net->net_ni_list) { + /* if you reached the end of the ni list and the net is + * specified, then there are no more nis in that net */ + if (net != NULL) + return NULL; + + /* we reached the end of this net ni list. move to the + * next net */ + if (prev->ni_net->net_list.next == &the_lnet.ln_nets) + /* no more nets and no more NIs. */ + return NULL; + + /* get the next net */ + net = list_entry(prev->ni_net->net_list.next, struct lnet_net, + net_list); + /* get the ni on it */ + ni = list_entry(net->net_ni_list.next, struct lnet_ni, + ni_netlist); + + return ni; + } + + /* there are more nis left */ + ni = list_entry(prev->ni_netlist.next, struct lnet_ni, ni_netlist); + + return ni; } int -LNetNIFini() +lnet_get_net_config(struct lnet_ioctl_config_data *config) { - LNET_MUTEX_DOWN(&the_lnet.ln_api_mutex); + struct lnet_ni *ni; + int cpt; + int rc = -ENOENT; + int idx = config->cfg_count; - LASSERT (the_lnet.ln_init); - LASSERT (the_lnet.ln_refcount > 0); + cpt = lnet_net_lock_current(); - if (the_lnet.ln_refcount != 1) { - the_lnet.ln_refcount--; - } else { - LASSERT (!the_lnet.ln_niinit_self); + ni = lnet_get_ni_idx_locked(idx); - lnet_proc_fini(); - lnet_ping_target_fini(); - lnet_router_checker_stop(); + if (ni != NULL) { + rc = 0; + lnet_ni_lock(ni); + lnet_fill_ni_info(ni, config); + lnet_ni_unlock(ni); + } - /* Teardown fns that use my own API functions BEFORE here */ - the_lnet.ln_refcount = 0; - - lnet_acceptor_stop(); - lnet_destroy_routes(); - lnet_shutdown_lndnis(); - lnet_unprepare(); - } - - LNET_MUTEX_UP(&the_lnet.ln_api_mutex); - return 0; + lnet_net_unlock(cpt); + return rc; } int -LNetCtl(unsigned int cmd, void *arg) +lnet_dyn_add_ni(lnet_pid_t requested_pid, struct lnet_ioctl_config_data *conf) { - struct libcfs_ioctl_data *data = arg; - lnet_process_id_t id = {0}; - lnet_ni_t *ni; - int rc; - - LASSERT (the_lnet.ln_init); - LASSERT (the_lnet.ln_refcount > 0); - - switch (cmd) { - case IOC_LIBCFS_GET_NI: - rc = LNetGetId(data->ioc_count, &id); - data->ioc_nid = id.nid; - return rc; - - case IOC_LIBCFS_FAIL_NID: - return lnet_fail_nid(data->ioc_nid, data->ioc_count); - - case IOC_LIBCFS_ADD_ROUTE: - rc = lnet_add_route(data->ioc_net, data->ioc_count, - data->ioc_nid); - return (rc != 0) ? rc : lnet_check_routes(); - - case IOC_LIBCFS_DEL_ROUTE: - return lnet_del_route(data->ioc_net, data->ioc_nid); - - case IOC_LIBCFS_GET_ROUTE: - return lnet_get_route(data->ioc_count, - &data->ioc_net, &data->ioc_count, - &data->ioc_nid, &data->ioc_flags); - case IOC_LIBCFS_NOTIFY_ROUTER: - return lnet_notify(NULL, data->ioc_nid, data->ioc_flags, - (time_t)data->ioc_u64[0]); - - case IOC_LIBCFS_PORTALS_COMPATIBILITY: - /* This can be removed once lustre stops calling it */ - return 0; - - case IOC_LIBCFS_LNET_DIST: - rc = LNetDist(data->ioc_nid, &data->ioc_nid, &data->ioc_u32[1]); - if (rc < 0 && rc != -EHOSTUNREACH) - return rc; - - data->ioc_u32[0] = rc; - return 0; - - case IOC_LIBCFS_TESTPROTOCOMPAT: - LNET_LOCK(); - the_lnet.ln_testprotocompat = data->ioc_flags; - LNET_UNLOCK(); - return 0; - - case IOC_LIBCFS_PING: - id.nid = data->ioc_nid; - id.pid = data->ioc_u32[0]; - rc = lnet_ping(id, data->ioc_u32[1], /* timeout */ - (lnet_process_id_t *)data->ioc_pbuf1, - data->ioc_plen1/sizeof(lnet_process_id_t)); - if (rc < 0) - return rc; - data->ioc_count = rc; - return 0; - - case IOC_LIBCFS_DEBUG_PEER: { - /* CAVEAT EMPTOR: this one designed for calling directly; not - * via an ioctl */ - id = *((lnet_process_id_t *) arg); - - lnet_debug_peer(id.nid); - - ni = lnet_net2ni(LNET_NIDNET(id.nid)); - if (ni == NULL) { - CDEBUG(D_WARNING, "No NI for %s\n", libcfs_id2str(id)); - } else { - if (ni->ni_lnd->lnd_ctl == NULL) { - CDEBUG(D_WARNING, "No ctl for %s\n", - libcfs_id2str(id)); - } else { - (void)ni->ni_lnd->lnd_ctl(ni, cmd, arg); - } - - lnet_ni_decref(ni); - } - return 0; - } - - default: - ni = lnet_net2ni(data->ioc_net); - if (ni == NULL) - return -EINVAL; - - if (ni->ni_lnd->lnd_ctl == NULL) - rc = -EINVAL; - else - rc = ni->ni_lnd->lnd_ctl(ni, cmd, arg); - - lnet_ni_decref(ni); - return rc; - } - /* not reached */ + char *nets = conf->cfg_config_u.cfg_net.net_intf; + struct lnet_ping_info *pinfo; + lnet_handle_md_t md_handle; + struct lnet_net *net; + struct list_head net_head; + int rc; + lnet_remotenet_t *rnet; + int net_ni_count; + int num_acceptor_nets; + __u32 net_type; + struct lnet_ioctl_config_lnd_tunables *lnd_tunables = NULL; + + INIT_LIST_HEAD(&net_head); + + if (conf && conf->cfg_hdr.ioc_len > sizeof(*conf)) + lnd_tunables = (struct lnet_ioctl_config_lnd_tunables *)conf->cfg_bulk; + + /* Create a net/ni structures for the network string */ + rc = lnet_parse_networks(&net_head, nets, use_tcp_bonding); + if (rc <= 0) + return rc == 0 ? -EINVAL : rc; + + mutex_lock(&the_lnet.ln_api_mutex); + + if (rc > 1) { + rc = -EINVAL; /* only add one network per call */ + goto failed0; + } + + net = list_entry(net_head.next, struct lnet_net, net_list); + + lnet_net_lock(LNET_LOCK_EX); + rnet = lnet_find_rnet_locked(net->net_id); + lnet_net_unlock(LNET_LOCK_EX); + /* make sure that the net added doesn't invalidate the current + * configuration LNet is keeping */ + if (rnet != NULL) { + CERROR("Adding net %s will invalidate routing configuration\n", + nets); + rc = -EUSERS; + goto failed0; + } + + /* + * make sure you calculate the correct number of slots in the ping + * info. Since the ping info is a flattened list of all the NIs, + * we should allocate enough slots to accomodate the number of NIs + * which will be added. + * + * We can use lnet_get_net_ni_count_locked() since the net is not + * on a public list yet, so locking is not a problem + */ + net_ni_count = lnet_get_net_ni_count_locked(net); + + rc = lnet_ping_info_setup(&pinfo, &md_handle, + net_ni_count + lnet_get_ni_count(), + false); + if (rc != 0) + goto failed0; + + list_del_init(&net->net_list); + + if (lnd_tunables) + memcpy(&net->net_tunables, + &lnd_tunables->lt_cmn, sizeof(lnd_tunables->lt_cmn)); + + /* + * 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(); + + /* + * lnd_startup_lndnet() can deallocate 'net' even if it it returns + * success, because we endded up adding interfaces to an existing + * network. So grab the net_type now + */ + net_type = LNET_NETTYP(net->net_id); + + rc = lnet_startup_lndnet(net, + (lnd_tunables) ? &lnd_tunables->lt_tun : NULL); + if (rc < 0) + goto failed1; + + /* + * Start the acceptor thread if this is the first network + * being added that requires the thread. + */ + if (net_type == SOCKLND && num_acceptor_nets == 0) + { + rc = lnet_acceptor_start(); + if (rc < 0) { + /* shutdown the net that we just started */ + CERROR("Failed to start up acceptor thread\n"); + /* + * Note that if we needed to start the acceptor + * thread, then 'net' must have been the first TCP + * network, therefore was unique, and therefore + * wasn't deallocated by lnet_startup_lndnet() + */ + lnet_shutdown_lndnet(net); + goto failed1; + } + } + + lnet_ping_target_update(pinfo, md_handle); + mutex_unlock(&the_lnet.ln_api_mutex); + + return 0; + +failed1: + lnet_ping_md_unlink(pinfo, &md_handle); + lnet_ping_info_free(pinfo); +failed0: + mutex_unlock(&the_lnet.ln_api_mutex); + while (!list_empty(&net_head)) { + net = list_entry(net_head.next, struct lnet_net, net_list); + list_del_init(&net->net_list); + lnet_net_free(net); + } + return rc; } int -LNetGetId(unsigned int index, lnet_process_id_t *id) +lnet_dyn_del_ni(__u32 net_id) { - lnet_ni_t *ni; - struct list_head *tmp; - int rc = -ENOENT; + struct lnet_net *net; + struct lnet_ping_info *pinfo; + lnet_handle_md_t md_handle; + int rc; + int net_ni_count; + + /* don't allow userspace to shutdown the LOLND */ + if (LNET_NETTYP(net_id) == LOLND) + return -EINVAL; + + mutex_lock(&the_lnet.ln_api_mutex); + + lnet_net_lock(0); - LASSERT (the_lnet.ln_init); - LASSERT (the_lnet.ln_refcount > 0); + net = lnet_get_net_locked(net_id); + if (net == NULL) { + rc = -EINVAL; + goto out; + } - LNET_LOCK(); + net_ni_count = lnet_get_net_ni_count_locked(net); - list_for_each(tmp, &the_lnet.ln_nis) { - if (index-- != 0) - continue; + lnet_net_unlock(0); - ni = list_entry(tmp, lnet_ni_t, ni_list); + /* create and link a new ping info, before removing the old one */ + rc = lnet_ping_info_setup(&pinfo, &md_handle, + lnet_get_ni_count() - net_ni_count, false); + if (rc != 0) + goto out; - id->nid = ni->ni_nid; - id->pid = the_lnet.ln_pid; - rc = 0; - break; - } + lnet_shutdown_lndnet(net); - LNET_UNLOCK(); + if (lnet_count_acceptor_nets() == 0) + lnet_acceptor_stop(); - return rc; + lnet_ping_target_update(pinfo, md_handle); + +out: + mutex_unlock(&the_lnet.ln_api_mutex); + + return rc; } -void -LNetSnprintHandle(char *str, int len, lnet_handle_any_t h) +void lnet_incr_dlc_seq(void) { - snprintf(str, len, LPX64, h.cookie); + atomic_inc(&lnet_dlc_seq_no); } +__u32 lnet_get_dlc_seq_locked(void) +{ + return atomic_read(&lnet_dlc_seq_no); +} +/** + * LNet ioctl handler. + * + */ int -lnet_ping_target_init(void) -{ - lnet_handle_me_t meh; - lnet_process_id_t id; - lnet_md_t md = {0}; - int rc; - int rc2; - int n; - unsigned int infosz; - int i; - - for (n = 0; ; n++) { - rc = LNetGetId(n, &id); - if (rc == -ENOENT) - break; - - LASSERT (rc == 0); - } - - infosz = offsetof(lnet_ping_info_t, pi_nid[n]); - LIBCFS_ALLOC(the_lnet.ln_ping_info, infosz); - if (the_lnet.ln_ping_info == NULL) { - CERROR("Can't allocate ping info[%d]\n", n); - return -ENOMEM; - } - - the_lnet.ln_ping_info->pi_magic = LNET_PROTO_PING_MAGIC; - the_lnet.ln_ping_info->pi_version = LNET_PROTO_PING_VERSION; - the_lnet.ln_ping_info->pi_pid = the_lnet.ln_pid; - the_lnet.ln_ping_info->pi_nnids = n; - - for (i = 0; i < n; i++) { - rc = LNetGetId(i, &id); - LASSERT (rc == 0); - the_lnet.ln_ping_info->pi_nid[i] = id.nid; - } - - /* We can have a tiny EQ since we only need to see the unlink event on - * teardown, which by definition is the last one! */ - rc = LNetEQAlloc(2, LNET_EQ_HANDLER_NONE, &the_lnet.ln_ping_target_eq); - if (rc != 0) { - CERROR("Can't allocate ping EQ: %d\n", rc); - goto failed_0; - } - - memset(&id, 0, sizeof(lnet_process_id_t)); - id.nid = LNET_NID_ANY; - id.pid = LNET_PID_ANY; - - rc = LNetMEAttach(LNET_RESERVED_PORTAL, id, - LNET_PROTO_PING_MATCHBITS, 0, - LNET_UNLINK, LNET_INS_AFTER, - &meh); - if (rc != 0) { - CERROR("Can't create ping ME: %d\n", rc); - goto failed_1; - } - - /* initialize md content */ - md.start = the_lnet.ln_ping_info; - md.length = infosz; - md.threshold = LNET_MD_THRESH_INF; - md.max_size = 0; - md.options = LNET_MD_OP_GET | LNET_MD_TRUNCATE | - LNET_MD_MANAGE_REMOTE; - md.user_ptr = NULL; - md.eq_handle = the_lnet.ln_ping_target_eq; - - rc = LNetMDAttach(meh, md, - LNET_RETAIN, - &the_lnet.ln_ping_target_md); - if (rc != 0) { - CERROR("Can't attach ping MD: %d\n", rc); - goto failed_2; - } - - return 0; - - failed_2: - rc2 = LNetMEUnlink(meh); - LASSERT (rc2 == 0); - failed_1: - rc2 = LNetEQFree(the_lnet.ln_ping_target_eq); - LASSERT (rc2 == 0); - failed_0: - LIBCFS_FREE(the_lnet.ln_ping_info, infosz); - - return rc; +LNetCtl(unsigned int cmd, void *arg) +{ + struct libcfs_ioctl_data *data = arg; + struct lnet_ioctl_config_data *config; + lnet_process_id_t id = {0}; + lnet_ni_t *ni; + int rc; + + BUILD_BUG_ON(sizeof(struct lnet_ioctl_net_config) + + sizeof(struct lnet_ioctl_config_data) > LIBCFS_IOC_DATA_MAX); + + switch (cmd) { + case IOC_LIBCFS_GET_NI: + rc = LNetGetId(data->ioc_count, &id); + data->ioc_nid = id.nid; + return rc; + + case IOC_LIBCFS_FAIL_NID: + return lnet_fail_nid(data->ioc_nid, data->ioc_count); + + case IOC_LIBCFS_ADD_ROUTE: + config = arg; + + if (config->cfg_hdr.ioc_len < sizeof(*config)) + return -EINVAL; + + mutex_lock(&the_lnet.ln_api_mutex); + rc = lnet_add_route(config->cfg_net, + config->cfg_config_u.cfg_route.rtr_hop, + config->cfg_nid, + config->cfg_config_u.cfg_route. + rtr_priority); + if (rc == 0) { + rc = lnet_check_routes(); + if (rc != 0) + lnet_del_route(config->cfg_net, + config->cfg_nid); + } + mutex_unlock(&the_lnet.ln_api_mutex); + return rc; + + case IOC_LIBCFS_DEL_ROUTE: + config = arg; + + if (config->cfg_hdr.ioc_len < sizeof(*config)) + return -EINVAL; + + mutex_lock(&the_lnet.ln_api_mutex); + rc = lnet_del_route(config->cfg_net, config->cfg_nid); + mutex_unlock(&the_lnet.ln_api_mutex); + return rc; + + case IOC_LIBCFS_GET_ROUTE: + config = arg; + + if (config->cfg_hdr.ioc_len < sizeof(*config)) + return -EINVAL; + + return lnet_get_route(config->cfg_count, + &config->cfg_net, + &config->cfg_config_u.cfg_route.rtr_hop, + &config->cfg_nid, + &config->cfg_config_u.cfg_route.rtr_flags, + &config->cfg_config_u.cfg_route. + rtr_priority); + + case IOC_LIBCFS_GET_NET: { + size_t total = sizeof(*config) + + sizeof(struct lnet_ioctl_net_config); + config = arg; + + if (config->cfg_hdr.ioc_len < total) + return -EINVAL; + + return lnet_get_net_config(config); + } + + case IOC_LIBCFS_GET_LNET_STATS: + { + struct lnet_ioctl_lnet_stats *lnet_stats = arg; + + if (lnet_stats->st_hdr.ioc_len < sizeof(*lnet_stats)) + return -EINVAL; + + lnet_counters_get(&lnet_stats->st_cntrs); + return 0; + } + + case IOC_LIBCFS_CONFIG_RTR: + config = arg; + + if (config->cfg_hdr.ioc_len < sizeof(*config)) + return -EINVAL; + + mutex_lock(&the_lnet.ln_api_mutex); + if (config->cfg_config_u.cfg_buffers.buf_enable) { + rc = lnet_rtrpools_enable(); + mutex_unlock(&the_lnet.ln_api_mutex); + return rc; + } + lnet_rtrpools_disable(); + mutex_unlock(&the_lnet.ln_api_mutex); + return 0; + + case IOC_LIBCFS_ADD_BUF: + config = arg; + + if (config->cfg_hdr.ioc_len < sizeof(*config)) + return -EINVAL; + + mutex_lock(&the_lnet.ln_api_mutex); + rc = lnet_rtrpools_adjust(config->cfg_config_u.cfg_buffers. + buf_tiny, + config->cfg_config_u.cfg_buffers. + buf_small, + config->cfg_config_u.cfg_buffers. + buf_large); + mutex_unlock(&the_lnet.ln_api_mutex); + return rc; + + case IOC_LIBCFS_GET_BUF: { + struct lnet_ioctl_pool_cfg *pool_cfg; + size_t total = sizeof(*config) + sizeof(*pool_cfg); + + config = arg; + + if (config->cfg_hdr.ioc_len < total) + return -EINVAL; + + pool_cfg = (struct lnet_ioctl_pool_cfg *)config->cfg_bulk; + return lnet_get_rtr_pool_cfg(config->cfg_count, pool_cfg); + } + + case IOC_LIBCFS_GET_PEER_INFO: { + struct lnet_ioctl_peer *peer_info = arg; + + if (peer_info->pr_hdr.ioc_len < sizeof(*peer_info)) + return -EINVAL; + + return lnet_get_peer_info( + peer_info->pr_count, + &peer_info->pr_nid, + peer_info->pr_lnd_u.pr_peer_credits.cr_aliveness, + &peer_info->pr_lnd_u.pr_peer_credits.cr_ncpt, + &peer_info->pr_lnd_u.pr_peer_credits.cr_refcount, + &peer_info->pr_lnd_u.pr_peer_credits.cr_ni_peer_tx_credits, + &peer_info->pr_lnd_u.pr_peer_credits.cr_peer_tx_credits, + &peer_info->pr_lnd_u.pr_peer_credits.cr_peer_rtr_credits, + &peer_info->pr_lnd_u.pr_peer_credits.cr_peer_min_rtr_credits, + &peer_info->pr_lnd_u.pr_peer_credits.cr_peer_tx_qnob); + } + + case IOC_LIBCFS_NOTIFY_ROUTER: { + unsigned long jiffies_passed; + + jiffies_passed = ktime_get_real_seconds() - data->ioc_u64[0]; + jiffies_passed = cfs_time_seconds(jiffies_passed); + + return lnet_notify(NULL, data->ioc_nid, data->ioc_flags, + jiffies - jiffies_passed); + } + + case IOC_LIBCFS_LNET_DIST: + rc = LNetDist(data->ioc_nid, &data->ioc_nid, &data->ioc_u32[1]); + if (rc < 0 && rc != -EHOSTUNREACH) + return rc; + + data->ioc_u32[0] = rc; + return 0; + + case IOC_LIBCFS_TESTPROTOCOMPAT: + lnet_net_lock(LNET_LOCK_EX); + the_lnet.ln_testprotocompat = data->ioc_flags; + lnet_net_unlock(LNET_LOCK_EX); + return 0; + + case IOC_LIBCFS_LNET_FAULT: + return lnet_fault_ctl(data->ioc_flags, data); + + case IOC_LIBCFS_PING: { + signed long timeout; + + id.nid = data->ioc_nid; + id.pid = data->ioc_u32[0]; + + /* Don't block longer than 2 minutes */ + if (data->ioc_u32[1] > 120 * MSEC_PER_SEC) + return -EINVAL; + + /* If timestamp is negative then disable timeout */ + if ((s32)data->ioc_u32[1] < 0) + timeout = MAX_SCHEDULE_TIMEOUT; + else + timeout = msecs_to_jiffies(data->ioc_u32[1]); + + rc = lnet_ping(id, timeout, data->ioc_pbuf1, + data->ioc_plen1 / sizeof(lnet_process_id_t)); + if (rc < 0) + return rc; + data->ioc_count = rc; + return 0; + } + default: + ni = lnet_net2ni(data->ioc_net); + if (ni == NULL) + return -EINVAL; + + if (ni->ni_net->net_lnd->lnd_ctl == NULL) + rc = -EINVAL; + else + rc = ni->ni_net->net_lnd->lnd_ctl(ni, cmd, arg); + + return rc; + } + /* not reached */ } +EXPORT_SYMBOL(LNetCtl); -void -lnet_ping_target_fini(void) +void LNetDebugPeer(lnet_process_id_t id) { - lnet_event_t event; - int rc; - int which; - int timeout_ms = 1000; - cfs_sigset_t blocked = cfs_block_allsigs(); - - LNetMDUnlink(the_lnet.ln_ping_target_md); - /* NB md could be busy; this just starts the unlink */ + lnet_debug_peer(id.nid); +} +EXPORT_SYMBOL(LNetDebugPeer); - for (;;) { - rc = LNetEQPoll(&the_lnet.ln_ping_target_eq, 1, - timeout_ms, &event, &which); +/** + * Retrieve the lnet_process_id_t ID of LNet interface at \a index. Note that + * all interfaces share a same PID, as requested by LNetNIInit(). + * + * \param index Index of the interface to look up. + * \param id On successful return, this location will hold the + * lnet_process_id_t ID of the interface. + * + * \retval 0 If an interface exists at \a index. + * \retval -ENOENT If no interface has been found. + */ +int +LNetGetId(unsigned int index, lnet_process_id_t *id) +{ + struct lnet_ni *ni; + struct lnet_net *net; + int cpt; + int rc = -ENOENT; - /* I expect overflow... */ - LASSERT (rc >= 0 || rc == -EOVERFLOW); + LASSERT(the_lnet.ln_refcount > 0); - if (rc == 0) { - /* timed out: provide a diagnostic */ - CWARN("Still waiting for ping MD to unlink\n"); - timeout_ms *= 2; - continue; - } + cpt = lnet_net_lock_current(); - /* Got a valid event */ - if (event.unlinked) - break; - } + list_for_each_entry(net, &the_lnet.ln_nets, net_list) { + list_for_each_entry(ni, &net->net_ni_list, ni_netlist) { + if (index-- != 0) + continue; - rc = LNetEQFree(the_lnet.ln_ping_target_eq); - LASSERT (rc == 0); + id->nid = ni->ni_nid; + id->pid = the_lnet.ln_pid; + rc = 0; + break; + } + } - LIBCFS_FREE(the_lnet.ln_ping_info, - offsetof(lnet_ping_info_t, - pi_nid[the_lnet.ln_ping_info->pi_nnids])); + lnet_net_unlock(cpt); + return rc; +} +EXPORT_SYMBOL(LNetGetId); - cfs_restore_sigs(blocked); +/** + * Print a string representation of handle \a h into buffer \a str of + * \a len bytes. + */ +void +LNetSnprintHandle(char *str, int len, lnet_handle_any_t h) +{ + snprintf(str, len, "%#llx", h.cookie); } +EXPORT_SYMBOL(LNetSnprintHandle); -int -lnet_ping (lnet_process_id_t id, int timeout_ms, lnet_process_id_t *ids, int n_ids) -{ - lnet_handle_eq_t eqh; - lnet_handle_md_t mdh; - lnet_event_t event; - lnet_md_t md = {0}; - int which; - int unlinked = 0; - int replied = 0; - const int a_long_time = 60000; /* mS */ - int infosz = offsetof(lnet_ping_info_t, pi_nid[n_ids]); - lnet_ping_info_t *info; - lnet_process_id_t tmpid; - int i; - int nob; - int rc; - int rc2; - cfs_sigset_t blocked; - - if (n_ids <= 0 || - id.nid == LNET_NID_ANY || - timeout_ms > 500000 || /* arbitrary limit! */ - n_ids > 20) /* arbitrary limit! */ - return -EINVAL; - - if (id.pid == LNET_PID_ANY) - id.pid = LUSTRE_SRV_LNET_PID; - - LIBCFS_ALLOC(info, infosz); - if (info == NULL) - 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 out_0; - } - - /* initialize md content */ - md.start = info; - md.length = infosz; - md.threshold = 2; /*GET/REPLY*/ - md.max_size = 0; - md.options = LNET_MD_TRUNCATE; - md.user_ptr = NULL; - md.eq_handle = eqh; - - rc = LNetMDBind(md, LNET_UNLINK, &mdh); - if (rc != 0) { - CERROR("Can't bind MD: %d\n", rc); - goto out_1; - } - - rc = LNetGet(LNET_NID_ANY, mdh, id, - LNET_RESERVED_PORTAL, - LNET_PROTO_PING_MATCHBITS, 0); - - if (rc != 0) { - /* Don't CERROR; this could be deliberate! */ - - rc2 = LNetMDUnlink(mdh); - LASSERT (rc2 == 0); - - /* NB must wait for the UNLINK event below... */ - unlinked = 1; - timeout_ms = a_long_time; - } - - do { - /* MUST block for unlink to complete */ - if (unlinked) - blocked = cfs_block_allsigs(); - - rc2 = LNetEQPoll(&eqh, 1, timeout_ms, &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_ms = 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)); - rc = -EIO; - goto out_1; - } - - nob = rc; - LASSERT (nob >= 0 && nob <= infosz); - - rc = -EPROTO; /* if I can't parse... */ - - if (nob < 8) { - /* can't check magic/version */ - CERROR("%s: ping info too short %d\n", - libcfs_id2str(id), nob); - goto out_1; - } - - if (info->pi_magic == __swab32(LNET_PROTO_PING_MAGIC)) { - /* NB I might be swabbing garbage until I check below, but it - * doesn't matter */ - __swab32s(&info->pi_version); - __swab32s(&info->pi_pid); - __swab32s(&info->pi_nnids); - for (i = 0; i < (int)info->pi_nnids && i < (int)n_ids; i++) - __swab64s(&info->pi_nid[i]); - - } else if (info->pi_magic != LNET_PROTO_PING_MAGIC) { - CERROR("%s: Unexpected magic %08x\n", - libcfs_id2str(id), info->pi_magic); - goto out_1; - } - - if (info->pi_version != LNET_PROTO_PING_VERSION) { - CERROR("%s: Unexpected version 0x%x\n", - libcfs_id2str(id), info->pi_version); - goto out_1; - } - - if (nob < (int)offsetof(lnet_ping_info_t, pi_nid[0])) { - CERROR("%s: Short reply %d(%d min)\n", libcfs_id2str(id), - nob, (int)offsetof(lnet_ping_info_t, pi_nid[0])); - goto out_1; - } - - if ((int) info->pi_nnids < n_ids) - n_ids = info->pi_nnids; - - if (nob < (int)offsetof(lnet_ping_info_t, pi_nid[n_ids])) { - CERROR("%s: Short reply %d(%d expected)\n", libcfs_id2str(id), - nob, (int)offsetof(lnet_ping_info_t, pi_nid[n_ids])); - goto out_1; - } - - rc = -EFAULT; /* If I SEGV... */ - - for (i = 0; i < n_ids; i++) { - tmpid.pid = info->pi_pid; - tmpid.nid = info->pi_nid[i]; -#ifdef __KERNEL__ - if (copy_to_user(&ids[i], &tmpid, sizeof(tmpid))) - goto out_1; -#else - ids[i] = tmpid; -#endif - } - rc = info->pi_nnids; +static int lnet_ping(lnet_process_id_t id, signed long timeout, + lnet_process_id_t __user *ids, int n_ids) +{ + lnet_handle_eq_t eqh; + lnet_handle_md_t mdh; + lnet_event_t event; + lnet_md_t md = { NULL }; + int which; + int unlinked = 0; + int replied = 0; + const signed long a_long_time = msecs_to_jiffies(60 * MSEC_PER_SEC); + int infosz; + struct lnet_ping_info *info; + lnet_process_id_t tmpid; + int i; + int nob; + int rc; + int rc2; + sigset_t blocked; + + infosz = offsetof(struct lnet_ping_info, pi_ni[n_ids]); + + /* n_ids limit is arbitrary */ + if (n_ids <= 0 || n_ids > 20 || id.nid == LNET_NID_ANY) + return -EINVAL; + + if (id.pid == LNET_PID_ANY) + id.pid = LNET_PID_LUSTRE; + + LIBCFS_ALLOC(info, infosz); + if (info == NULL) + 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 out_0; + } + + /* initialize md content */ + md.start = info; + md.length = infosz; + md.threshold = 2; /*GET/REPLY*/ + md.max_size = 0; + md.options = LNET_MD_TRUNCATE; + md.user_ptr = NULL; + md.eq_handle = eqh; + + rc = LNetMDBind(md, LNET_UNLINK, &mdh); + if (rc != 0) { + CERROR("Can't bind MD: %d\n", rc); + goto out_1; + } + + rc = LNetGet(LNET_NID_ANY, mdh, id, + LNET_RESERVED_PORTAL, + LNET_PROTO_PING_MATCHBITS, 0); + + if (rc != 0) { + /* Don't CERROR; this could be deliberate! */ + + rc2 = LNetMDUnlink(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)); + rc = -EIO; + goto out_1; + } + + nob = rc; + LASSERT(nob >= 0 && nob <= infosz); + + rc = -EPROTO; /* if I can't parse... */ + + if (nob < 8) { + /* can't check magic/version */ + CERROR("%s: ping info too short %d\n", + libcfs_id2str(id), nob); + goto out_1; + } + + if (info->pi_magic == __swab32(LNET_PROTO_PING_MAGIC)) { + lnet_swap_pinginfo(info); + } else if (info->pi_magic != LNET_PROTO_PING_MAGIC) { + CERROR("%s: Unexpected magic %08x\n", + libcfs_id2str(id), info->pi_magic); + goto out_1; + } + + if ((info->pi_features & LNET_PING_FEAT_NI_STATUS) == 0) { + CERROR("%s: ping w/o NI status: 0x%x\n", + libcfs_id2str(id), info->pi_features); + goto out_1; + } + + if (nob < offsetof(struct lnet_ping_info, pi_ni[0])) { + CERROR("%s: Short reply %d(%d min)\n", libcfs_id2str(id), + nob, (int)offsetof(struct lnet_ping_info, pi_ni[0])); + goto out_1; + } + + if (info->pi_nnis < n_ids) + n_ids = info->pi_nnis; + + if (nob < offsetof(struct lnet_ping_info, pi_ni[n_ids])) { + CERROR("%s: Short reply %d(%d expected)\n", libcfs_id2str(id), + nob, (int)offsetof(struct lnet_ping_info, pi_ni[n_ids])); + goto out_1; + } + + rc = -EFAULT; /* If I SEGV... */ + + memset(&tmpid, 0, sizeof(tmpid)); + for (i = 0; i < n_ids; i++) { + tmpid.pid = info->pi_pid; + tmpid.nid = info->pi_ni[i].ns_nid; + if (copy_to_user(&ids[i], &tmpid, sizeof(tmpid))) + goto out_1; + } + rc = info->pi_nnis; out_1: - rc2 = LNetEQFree(eqh); - if (rc2 != 0) - CERROR("rc2 %d\n", rc2); - LASSERT (rc2 == 0); + rc2 = LNetEQFree(eqh); + if (rc2 != 0) + CERROR("rc2 %d\n", rc2); + LASSERT(rc2 == 0); out_0: - LIBCFS_FREE(info, infosz); - return rc; + LIBCFS_FREE(info, infosz); + return rc; }