4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
31 * lnet/klnds/socklnd/socklnd.c
33 * Author: Zach Brown <zab@zabbo.net>
34 * Author: Peter J. Braam <braam@clusterfs.com>
35 * Author: Phil Schwan <phil@clusterfs.com>
36 * Author: Eric Barton <eric@bartonsoftware.com>
39 #include <linux/ethtool.h>
40 #include <linux/inetdevice.h>
42 #include <linux/sunrpc/addr.h>
43 #include <net/addrconf.h>
45 static const struct lnet_lnd the_ksocklnd;
46 struct ksock_nal_data ksocknal_data;
48 static struct ksock_interface *
49 ksocknal_ip2iface(struct lnet_ni *ni, struct sockaddr *addr)
51 struct ksock_net *net = ni->ni_data;
52 struct ksock_interface *iface;
54 iface = &net->ksnn_interface;
56 if (rpc_cmp_addr((struct sockaddr *)&iface->ksni_addr, addr))
62 static struct ksock_interface *
63 ksocknal_index2iface(struct lnet_ni *ni, int index)
65 struct ksock_net *net = ni->ni_data;
66 struct ksock_interface *iface;
68 iface = &net->ksnn_interface;
70 if (iface->ksni_index == index)
76 static int ksocknal_ip2index(struct sockaddr *addr, struct lnet_ni *ni)
78 struct net_device *dev;
80 DECLARE_CONST_IN_IFADDR(ifa);
82 if (addr->sa_family != AF_INET && addr->sa_family != AF_INET6)
86 for_each_netdev(ni->ni_net_ns, dev) {
87 int flags = dev_get_flags(dev);
88 struct in_device *in_dev;
90 if (flags & IFF_LOOPBACK) /* skip the loopback IF */
93 if (!(flags & IFF_UP))
96 switch (addr->sa_family) {
98 in_dev = __in_dev_get_rcu(dev);
102 in_dev_for_each_ifa_rcu(ifa, in_dev) {
103 if (ifa->ifa_local ==
104 ((struct sockaddr_in *)addr)->sin_addr.s_addr)
109 #if IS_ENABLED(CONFIG_IPV6)
111 struct inet6_dev *in6_dev;
112 const struct inet6_ifaddr *ifa6;
113 struct sockaddr_in6 *addr6 = (struct sockaddr_in6*)addr;
115 in6_dev = __in6_dev_get(dev);
119 list_for_each_entry_rcu(ifa6, &in6_dev->addr_list, if_list) {
120 if (ipv6_addr_cmp(&ifa6->addr,
121 &addr6->sin6_addr) == 0)
126 #endif /* IS_ENABLED(CONFIG_IPV6) */
136 static struct ksock_conn_cb *
137 ksocknal_create_conn_cb(struct sockaddr *addr)
139 struct ksock_conn_cb *conn_cb;
141 LIBCFS_ALLOC(conn_cb, sizeof(*conn_cb));
145 refcount_set(&conn_cb->ksnr_refcount, 1);
146 conn_cb->ksnr_peer = NULL;
147 conn_cb->ksnr_retry_interval = 0; /* OK to connect at any time */
148 rpc_copy_addr((struct sockaddr *)&conn_cb->ksnr_addr, addr);
149 rpc_set_port((struct sockaddr *)&conn_cb->ksnr_addr,
151 conn_cb->ksnr_myiface = -1;
152 conn_cb->ksnr_scheduled = 0;
153 conn_cb->ksnr_connecting = 0;
154 conn_cb->ksnr_connected = 0;
155 conn_cb->ksnr_deleted = 0;
156 conn_cb->ksnr_conn_count = 0;
157 conn_cb->ksnr_ctrl_conn_count = 0;
158 conn_cb->ksnr_blki_conn_count = 0;
159 conn_cb->ksnr_blko_conn_count = 0;
160 conn_cb->ksnr_max_conns = 0;
166 ksocknal_destroy_conn_cb(struct ksock_conn_cb *conn_cb)
168 LASSERT(refcount_read(&conn_cb->ksnr_refcount) == 0);
170 if (conn_cb->ksnr_peer)
171 ksocknal_peer_decref(conn_cb->ksnr_peer);
173 LIBCFS_FREE(conn_cb, sizeof(*conn_cb));
176 static struct ksock_peer_ni *
177 ksocknal_create_peer(struct lnet_ni *ni, struct lnet_processid *id)
179 int cpt = lnet_nid2cpt(&id->nid, ni);
180 struct ksock_net *net = ni->ni_data;
181 struct ksock_peer_ni *peer_ni;
183 LASSERT(!LNET_NID_IS_ANY(&id->nid));
184 LASSERT(id->pid != LNET_PID_ANY);
185 LASSERT(!in_interrupt());
187 if (!atomic_inc_unless_negative(&net->ksnn_npeers)) {
188 CERROR("Can't create peer_ni: network shutdown\n");
189 return ERR_PTR(-ESHUTDOWN);
192 LIBCFS_CPT_ALLOC(peer_ni, lnet_cpt_table(), cpt, sizeof(*peer_ni));
194 atomic_dec(&net->ksnn_npeers);
195 return ERR_PTR(-ENOMEM);
198 peer_ni->ksnp_ni = ni;
199 peer_ni->ksnp_id = *id;
200 refcount_set(&peer_ni->ksnp_refcount, 1); /* 1 ref for caller */
201 peer_ni->ksnp_closing = 0;
202 peer_ni->ksnp_accepting = 0;
203 peer_ni->ksnp_proto = NULL;
204 peer_ni->ksnp_last_alive = 0;
205 peer_ni->ksnp_zc_next_cookie = SOCKNAL_KEEPALIVE_PING + 1;
206 peer_ni->ksnp_conn_cb = NULL;
208 INIT_LIST_HEAD(&peer_ni->ksnp_conns);
209 INIT_LIST_HEAD(&peer_ni->ksnp_tx_queue);
210 INIT_LIST_HEAD(&peer_ni->ksnp_zc_req_list);
211 spin_lock_init(&peer_ni->ksnp_lock);
217 ksocknal_destroy_peer(struct ksock_peer_ni *peer_ni)
219 struct ksock_net *net = peer_ni->ksnp_ni->ni_data;
221 CDEBUG (D_NET, "peer_ni %s %p deleted\n",
222 libcfs_idstr(&peer_ni->ksnp_id), peer_ni);
224 LASSERT(refcount_read(&peer_ni->ksnp_refcount) == 0);
225 LASSERT(peer_ni->ksnp_accepting == 0);
226 LASSERT(list_empty(&peer_ni->ksnp_conns));
227 LASSERT(peer_ni->ksnp_conn_cb == NULL);
228 LASSERT(list_empty(&peer_ni->ksnp_tx_queue));
229 LASSERT(list_empty(&peer_ni->ksnp_zc_req_list));
231 LIBCFS_FREE(peer_ni, sizeof(*peer_ni));
233 /* NB a peer_ni's connections and conn_cb keep a reference on their
234 * peer_ni until they are destroyed, so we can be assured that _all_
235 * state to do with this peer_ni has been cleaned up when its refcount
238 if (atomic_dec_and_test(&net->ksnn_npeers))
239 wake_up_var(&net->ksnn_npeers);
242 struct ksock_peer_ni *
243 ksocknal_find_peer_locked(struct lnet_ni *ni, struct lnet_processid *id)
245 struct ksock_peer_ni *peer_ni;
246 unsigned long hash = nidhash(&id->nid);
248 hash_for_each_possible(ksocknal_data.ksnd_peers, peer_ni,
250 LASSERT(!peer_ni->ksnp_closing);
252 if (peer_ni->ksnp_ni != ni)
255 if (!nid_same(&peer_ni->ksnp_id.nid, &id->nid) ||
256 peer_ni->ksnp_id.pid != id->pid)
259 CDEBUG(D_NET, "got peer_ni [%p] -> %s (%d)\n",
260 peer_ni, libcfs_idstr(id),
261 refcount_read(&peer_ni->ksnp_refcount));
267 struct ksock_peer_ni *
268 ksocknal_find_peer(struct lnet_ni *ni, struct lnet_processid *id)
270 struct ksock_peer_ni *peer_ni;
272 read_lock(&ksocknal_data.ksnd_global_lock);
273 peer_ni = ksocknal_find_peer_locked(ni, id);
274 if (peer_ni != NULL) /* +1 ref for caller? */
275 ksocknal_peer_addref(peer_ni);
276 read_unlock(&ksocknal_data.ksnd_global_lock);
282 ksocknal_unlink_peer_locked(struct ksock_peer_ni *peer_ni)
285 struct ksock_interface *iface;
287 for (i = 0; i < peer_ni->ksnp_n_passive_ips; i++) {
288 struct sockaddr_in sa = { .sin_family = AF_INET };
289 LASSERT(i < LNET_INTERFACES_NUM);
290 sa.sin_addr.s_addr = htonl(peer_ni->ksnp_passive_ips[i]);
292 iface = ksocknal_ip2iface(peer_ni->ksnp_ni,
293 (struct sockaddr *)&sa);
295 * All IPs in peer_ni->ksnp_passive_ips[] come from the
296 * interface list, therefore the call must succeed.
298 LASSERT(iface != NULL);
300 CDEBUG(D_NET, "peer_ni=%p iface=%p ksni_nroutes=%d\n",
301 peer_ni, iface, iface->ksni_nroutes);
302 iface->ksni_npeers--;
305 LASSERT(list_empty(&peer_ni->ksnp_conns));
306 LASSERT(peer_ni->ksnp_conn_cb == NULL);
307 LASSERT(!peer_ni->ksnp_closing);
308 peer_ni->ksnp_closing = 1;
309 hlist_del(&peer_ni->ksnp_list);
310 /* lose peerlist's ref */
311 ksocknal_peer_decref(peer_ni);
315 ksocknal_get_peer_info(struct lnet_ni *ni, int index,
316 struct lnet_processid *id, __u32 *myip, __u32 *peer_ip,
317 int *port, int *conn_count, int *share_count)
319 struct ksock_peer_ni *peer_ni;
320 struct ksock_conn_cb *conn_cb;
325 read_lock(&ksocknal_data.ksnd_global_lock);
327 hash_for_each(ksocknal_data.ksnd_peers, i, peer_ni, ksnp_list) {
329 if (peer_ni->ksnp_ni != ni)
332 if (peer_ni->ksnp_n_passive_ips == 0 &&
333 peer_ni->ksnp_conn_cb == NULL) {
337 *id = peer_ni->ksnp_id;
347 for (j = 0; j < peer_ni->ksnp_n_passive_ips; j++) {
351 *id = peer_ni->ksnp_id;
352 *myip = peer_ni->ksnp_passive_ips[j];
361 if (peer_ni->ksnp_conn_cb) {
365 conn_cb = peer_ni->ksnp_conn_cb;
367 *id = peer_ni->ksnp_id;
368 if (conn_cb->ksnr_addr.ss_family == AF_INET) {
369 struct sockaddr_in *sa =
370 (void *)&conn_cb->ksnr_addr;
372 rc = choose_ipv4_src(myip,
373 conn_cb->ksnr_myiface,
374 ntohl(sa->sin_addr.s_addr),
376 *peer_ip = ntohl(sa->sin_addr.s_addr);
377 *port = ntohs(sa->sin_port);
380 *peer_ip = 0xFFFFFFFF;
384 *conn_count = conn_cb->ksnr_conn_count;
390 read_unlock(&ksocknal_data.ksnd_global_lock);
395 ksocknal_get_conn_count_by_type(struct ksock_conn_cb *conn_cb,
398 unsigned int count = 0;
401 case SOCKLND_CONN_CONTROL:
402 count = conn_cb->ksnr_ctrl_conn_count;
404 case SOCKLND_CONN_BULK_IN:
405 count = conn_cb->ksnr_blki_conn_count;
407 case SOCKLND_CONN_BULK_OUT:
408 count = conn_cb->ksnr_blko_conn_count;
410 case SOCKLND_CONN_ANY:
411 count = conn_cb->ksnr_conn_count;
422 ksocknal_get_conns_per_peer(struct ksock_peer_ni *peer_ni)
424 struct lnet_ni *ni = peer_ni->ksnp_ni;
425 struct lnet_ioctl_config_socklnd_tunables *tunables;
429 tunables = &ni->ni_lnd_tunables.lnd_tun_u.lnd_sock;
431 return tunables->lnd_conns_per_peer;
435 ksocknal_incr_conn_count(struct ksock_conn_cb *conn_cb,
438 conn_cb->ksnr_conn_count++;
440 /* check if all connections of the given type got created */
442 case SOCKLND_CONN_CONTROL:
443 conn_cb->ksnr_ctrl_conn_count++;
444 /* there's a single control connection per peer,
445 * two in case of loopback
447 conn_cb->ksnr_connected |= BIT(type);
449 case SOCKLND_CONN_BULK_IN:
450 conn_cb->ksnr_blki_conn_count++;
451 if (conn_cb->ksnr_blki_conn_count >= conn_cb->ksnr_max_conns)
452 conn_cb->ksnr_connected |= BIT(type);
454 case SOCKLND_CONN_BULK_OUT:
455 conn_cb->ksnr_blko_conn_count++;
456 if (conn_cb->ksnr_blko_conn_count >= conn_cb->ksnr_max_conns)
457 conn_cb->ksnr_connected |= BIT(type);
459 case SOCKLND_CONN_ANY:
460 if (conn_cb->ksnr_conn_count >= conn_cb->ksnr_max_conns)
461 conn_cb->ksnr_connected |= BIT(type);
468 CDEBUG(D_NET, "Add conn type %d, ksnr_connected %x ksnr_max_conns %d\n",
469 type, conn_cb->ksnr_connected, conn_cb->ksnr_max_conns);
474 ksocknal_decr_conn_count(struct ksock_conn_cb *conn_cb,
477 conn_cb->ksnr_conn_count--;
479 /* check if all connections of the given type got created */
481 case SOCKLND_CONN_CONTROL:
482 conn_cb->ksnr_ctrl_conn_count--;
483 /* there's a single control connection per peer,
484 * two in case of loopback
486 if (conn_cb->ksnr_ctrl_conn_count == 0)
487 conn_cb->ksnr_connected &= ~BIT(type);
489 case SOCKLND_CONN_BULK_IN:
490 conn_cb->ksnr_blki_conn_count--;
491 if (conn_cb->ksnr_blki_conn_count < conn_cb->ksnr_max_conns)
492 conn_cb->ksnr_connected &= ~BIT(type);
494 case SOCKLND_CONN_BULK_OUT:
495 conn_cb->ksnr_blko_conn_count--;
496 if (conn_cb->ksnr_blko_conn_count < conn_cb->ksnr_max_conns)
497 conn_cb->ksnr_connected &= ~BIT(type);
499 case SOCKLND_CONN_ANY:
500 if (conn_cb->ksnr_conn_count < conn_cb->ksnr_max_conns)
501 conn_cb->ksnr_connected &= ~BIT(type);
508 CDEBUG(D_NET, "Del conn type %d, ksnr_connected %x ksnr_max_conns %d\n",
509 type, conn_cb->ksnr_connected, conn_cb->ksnr_max_conns);
513 ksocknal_associate_cb_conn_locked(struct ksock_conn_cb *conn_cb,
514 struct ksock_conn *conn)
516 struct ksock_peer_ni *peer_ni = conn_cb->ksnr_peer;
517 int type = conn->ksnc_type;
518 struct ksock_interface *iface;
521 conn_iface = ksocknal_ip2index((struct sockaddr *)&conn->ksnc_myaddr,
523 conn->ksnc_conn_cb = conn_cb;
524 ksocknal_conn_cb_addref(conn_cb);
526 if (conn_cb->ksnr_myiface != conn_iface) {
527 if (conn_cb->ksnr_myiface < 0) {
528 /* route wasn't bound locally yet (the initial route) */
529 CDEBUG(D_NET, "Binding %s %pIS to interface %d\n",
530 libcfs_idstr(&peer_ni->ksnp_id),
535 "Rebinding %s %pIS from interface %d to %d\n",
536 libcfs_idstr(&peer_ni->ksnp_id),
538 conn_cb->ksnr_myiface,
541 iface = ksocknal_index2iface(peer_ni->ksnp_ni,
542 conn_cb->ksnr_myiface);
544 iface->ksni_nroutes--;
546 conn_cb->ksnr_myiface = conn_iface;
547 iface = ksocknal_index2iface(peer_ni->ksnp_ni,
548 conn_cb->ksnr_myiface);
550 iface->ksni_nroutes++;
553 ksocknal_incr_conn_count(conn_cb, type);
555 /* Successful connection => further attempts can
556 * proceed immediately
558 conn_cb->ksnr_retry_interval = 0;
562 ksocknal_add_conn_cb_locked(struct ksock_peer_ni *peer_ni,
563 struct ksock_conn_cb *conn_cb)
565 struct ksock_conn *conn;
566 struct ksock_net *net = peer_ni->ksnp_ni->ni_data;
568 LASSERT(!peer_ni->ksnp_closing);
569 LASSERT(!conn_cb->ksnr_peer);
570 LASSERT(!conn_cb->ksnr_scheduled);
571 LASSERT(!conn_cb->ksnr_connecting);
572 LASSERT(conn_cb->ksnr_connected == 0);
574 conn_cb->ksnr_peer = peer_ni;
575 ksocknal_peer_addref(peer_ni);
577 /* set the conn_cb's interface to the current net's interface */
578 conn_cb->ksnr_myiface = net->ksnn_interface.ksni_index;
579 net->ksnn_interface.ksni_nroutes++;
581 /* peer_ni's route list takes over my ref on 'route' */
582 peer_ni->ksnp_conn_cb = conn_cb;
584 list_for_each_entry(conn, &peer_ni->ksnp_conns, ksnc_list) {
585 if (!rpc_cmp_addr((struct sockaddr *)&conn->ksnc_peeraddr,
586 (struct sockaddr *)&conn_cb->ksnr_addr))
589 ksocknal_associate_cb_conn_locked(conn_cb, conn);
590 /* keep going (typed conns) */
595 ksocknal_del_conn_cb_locked(struct ksock_conn_cb *conn_cb)
597 struct ksock_peer_ni *peer_ni = conn_cb->ksnr_peer;
598 struct ksock_interface *iface;
599 struct ksock_conn *conn;
600 struct ksock_conn *cnxt;
602 LASSERT(!conn_cb->ksnr_deleted);
604 /* Close associated conns */
605 list_for_each_entry_safe(conn, cnxt, &peer_ni->ksnp_conns, ksnc_list) {
606 if (conn->ksnc_conn_cb != conn_cb)
609 ksocknal_close_conn_locked(conn, 0);
612 if (conn_cb->ksnr_myiface >= 0) {
613 iface = ksocknal_index2iface(peer_ni->ksnp_ni,
614 conn_cb->ksnr_myiface);
616 iface->ksni_nroutes--;
619 conn_cb->ksnr_deleted = 1;
620 ksocknal_conn_cb_decref(conn_cb); /* drop peer_ni's ref */
621 peer_ni->ksnp_conn_cb = NULL;
623 if (list_empty(&peer_ni->ksnp_conns)) {
624 /* I've just removed the last route to a peer_ni with no active
627 ksocknal_unlink_peer_locked(peer_ni);
632 ksocknal_add_peer(struct lnet_ni *ni, struct lnet_processid *id,
633 struct sockaddr *addr)
635 struct ksock_peer_ni *peer_ni;
636 struct ksock_peer_ni *peer2;
637 struct ksock_conn_cb *conn_cb;
639 if (LNET_NID_IS_ANY(&id->nid) ||
640 id->pid == LNET_PID_ANY)
643 /* Have a brand new peer_ni ready... */
644 peer_ni = ksocknal_create_peer(ni, id);
646 return PTR_ERR(peer_ni);
648 conn_cb = ksocknal_create_conn_cb(addr);
650 ksocknal_peer_decref(peer_ni);
654 write_lock_bh(&ksocknal_data.ksnd_global_lock);
656 /* always called with a ref on ni, so shutdown can't have started */
657 LASSERT(atomic_read(&((struct ksock_net *)ni->ni_data)->ksnn_npeers)
660 peer2 = ksocknal_find_peer_locked(ni, id);
662 ksocknal_peer_decref(peer_ni);
665 /* peer_ni table takes my ref on peer_ni */
666 hash_add(ksocknal_data.ksnd_peers, &peer_ni->ksnp_list,
670 if (peer_ni->ksnp_conn_cb) {
671 ksocknal_conn_cb_decref(conn_cb);
673 ksocknal_add_conn_cb_locked(peer_ni, conn_cb);
674 /* Remember conns_per_peer setting at the time
675 * of connection initiation. It will define the
676 * max number of conns per type for this conn_cb
679 conn_cb->ksnr_max_conns = ksocknal_get_conns_per_peer(peer_ni);
682 write_unlock_bh(&ksocknal_data.ksnd_global_lock);
688 ksocknal_del_peer_locked(struct ksock_peer_ni *peer_ni)
690 struct ksock_conn *conn;
691 struct ksock_conn *cnxt;
692 struct ksock_conn_cb *conn_cb;
694 LASSERT(!peer_ni->ksnp_closing);
696 /* Extra ref prevents peer_ni disappearing until I'm done with it */
697 ksocknal_peer_addref(peer_ni);
698 conn_cb = peer_ni->ksnp_conn_cb;
700 ksocknal_del_conn_cb_locked(conn_cb);
702 list_for_each_entry_safe(conn, cnxt, &peer_ni->ksnp_conns,
704 ksocknal_close_conn_locked(conn, 0);
706 ksocknal_peer_decref(peer_ni);
707 /* NB peer_ni unlinks itself when last conn/conn_cb is removed */
711 ksocknal_del_peer(struct lnet_ni *ni, struct lnet_processid *id)
714 struct hlist_node *pnxt;
715 struct ksock_peer_ni *peer_ni;
721 write_lock_bh(&ksocknal_data.ksnd_global_lock);
723 if (id && !LNET_NID_IS_ANY(&id->nid)) {
724 lo = hash_min(nidhash(&id->nid),
725 HASH_BITS(ksocknal_data.ksnd_peers));
729 hi = HASH_SIZE(ksocknal_data.ksnd_peers) - 1;
732 for (i = lo; i <= hi; i++) {
733 hlist_for_each_entry_safe(peer_ni, pnxt,
734 &ksocknal_data.ksnd_peers[i],
736 if (peer_ni->ksnp_ni != ni)
739 if (!((!id || LNET_NID_IS_ANY(&id->nid) ||
740 nid_same(&peer_ni->ksnp_id.nid, &id->nid)) &&
741 (!id || id->pid == LNET_PID_ANY ||
742 peer_ni->ksnp_id.pid == id->pid)))
745 ksocknal_peer_addref(peer_ni); /* a ref for me... */
747 ksocknal_del_peer_locked(peer_ni);
749 if (peer_ni->ksnp_closing &&
750 !list_empty(&peer_ni->ksnp_tx_queue)) {
751 LASSERT(list_empty(&peer_ni->ksnp_conns));
752 LASSERT(peer_ni->ksnp_conn_cb == NULL);
754 list_splice_init(&peer_ni->ksnp_tx_queue,
758 ksocknal_peer_decref(peer_ni); /* ...till here */
760 rc = 0; /* matched! */
764 write_unlock_bh(&ksocknal_data.ksnd_global_lock);
766 ksocknal_txlist_done(ni, &zombies, -ENETDOWN);
771 static struct ksock_conn *
772 ksocknal_get_conn_by_idx(struct lnet_ni *ni, int index)
774 struct ksock_peer_ni *peer_ni;
775 struct ksock_conn *conn;
778 read_lock(&ksocknal_data.ksnd_global_lock);
780 hash_for_each(ksocknal_data.ksnd_peers, i, peer_ni, ksnp_list) {
781 LASSERT(!peer_ni->ksnp_closing);
783 if (peer_ni->ksnp_ni != ni)
786 list_for_each_entry(conn, &peer_ni->ksnp_conns,
791 ksocknal_conn_addref(conn);
792 read_unlock(&ksocknal_data.ksnd_global_lock);
797 read_unlock(&ksocknal_data.ksnd_global_lock);
801 static struct ksock_sched *
802 ksocknal_choose_scheduler_locked(unsigned int cpt)
804 struct ksock_sched *sched = ksocknal_data.ksnd_schedulers[cpt];
807 if (sched->kss_nthreads == 0) {
808 cfs_percpt_for_each(sched, i, ksocknal_data.ksnd_schedulers) {
809 if (sched->kss_nthreads > 0) {
810 CDEBUG(D_NET, "scheduler[%d] has no threads. selected scheduler[%d]\n",
811 cpt, sched->kss_cpt);
822 ksocknal_accept(struct lnet_ni *ni, struct socket *sock)
824 struct ksock_connreq *cr;
826 struct sockaddr_storage peer;
828 rc = lnet_sock_getaddr(sock, true, &peer);
830 CERROR("Can't determine new connection's address\n");
834 LIBCFS_ALLOC(cr, sizeof(*cr));
836 LCONSOLE_ERROR_MSG(0x12f,
837 "Dropping connection request from %pIS: memory exhausted\n",
844 cr->ksncr_sock = sock;
846 spin_lock_bh(&ksocknal_data.ksnd_connd_lock);
848 list_add_tail(&cr->ksncr_list, &ksocknal_data.ksnd_connd_connreqs);
849 wake_up(&ksocknal_data.ksnd_connd_waitq);
851 spin_unlock_bh(&ksocknal_data.ksnd_connd_lock);
856 ksocknal_connecting(struct ksock_conn_cb *conn_cb, struct sockaddr *sa)
859 rpc_cmp_addr((struct sockaddr *)&conn_cb->ksnr_addr, sa))
860 return conn_cb->ksnr_connecting;
865 ksocknal_create_conn(struct lnet_ni *ni, struct ksock_conn_cb *conn_cb,
866 struct socket *sock, int type)
868 rwlock_t *global_lock = &ksocknal_data.ksnd_global_lock;
870 struct lnet_processid peerid;
872 struct ksock_conn *conn;
873 struct ksock_conn *conn2;
874 struct ksock_peer_ni *peer_ni = NULL;
875 struct ksock_peer_ni *peer2;
876 struct ksock_sched *sched;
877 struct ksock_hello_msg *hello;
880 struct ksock_tx *txtmp;
887 active = (conn_cb != NULL);
889 LASSERT(active == (type != SOCKLND_CONN_NONE));
891 LIBCFS_ALLOC(conn, sizeof(*conn));
897 conn->ksnc_peer = NULL;
898 conn->ksnc_conn_cb = NULL;
899 conn->ksnc_sock = sock;
900 /* 2 ref, 1 for conn, another extra ref prevents socket
901 * being closed before establishment of connection */
902 refcount_set(&conn->ksnc_sock_refcount, 2);
903 conn->ksnc_type = type;
904 ksocknal_lib_save_callback(sock, conn);
905 refcount_set(&conn->ksnc_conn_refcount, 1); /* 1 ref for me */
907 conn->ksnc_rx_ready = 0;
908 conn->ksnc_rx_scheduled = 0;
910 INIT_LIST_HEAD(&conn->ksnc_tx_queue);
911 conn->ksnc_tx_ready = 0;
912 conn->ksnc_tx_scheduled = 0;
913 conn->ksnc_tx_carrier = NULL;
914 atomic_set (&conn->ksnc_tx_nob, 0);
916 LIBCFS_ALLOC(hello, offsetof(struct ksock_hello_msg,
917 kshm_ips[LNET_INTERFACES_NUM]));
923 /* stash conn's local and remote addrs */
924 rc = ksocknal_lib_get_conn_addrs(conn);
928 /* Find out/confirm peer_ni's NID and connection type and get the
929 * vector of interfaces she's willing to let me connect to.
930 * Passive connections use the listener timeout since the peer_ni sends
935 peer_ni = conn_cb->ksnr_peer;
936 LASSERT(ni == peer_ni->ksnp_ni);
938 /* Active connection sends HELLO eagerly */
939 hello->kshm_nips = 0;
940 peerid = peer_ni->ksnp_id;
942 write_lock_bh(global_lock);
943 conn->ksnc_proto = peer_ni->ksnp_proto;
944 write_unlock_bh(global_lock);
946 if (conn->ksnc_proto == NULL) {
947 conn->ksnc_proto = &ksocknal_protocol_v3x;
948 #if SOCKNAL_VERSION_DEBUG
949 if (*ksocknal_tunables.ksnd_protocol == 2)
950 conn->ksnc_proto = &ksocknal_protocol_v2x;
951 else if (*ksocknal_tunables.ksnd_protocol == 1)
952 conn->ksnc_proto = &ksocknal_protocol_v1x;
956 rc = ksocknal_send_hello(ni, conn, &peerid.nid, hello);
960 peerid.nid = LNET_ANY_NID;
961 peerid.pid = LNET_PID_ANY;
963 /* Passive, get protocol from peer_ni */
964 conn->ksnc_proto = NULL;
967 rc = ksocknal_recv_hello(ni, conn, hello, &peerid, &incarnation);
971 LASSERT(rc == 0 || active);
972 LASSERT(conn->ksnc_proto != NULL);
973 LASSERT(!LNET_NID_IS_ANY(&peerid.nid));
975 cpt = lnet_nid2cpt(&peerid.nid, ni);
978 ksocknal_peer_addref(peer_ni);
979 write_lock_bh(global_lock);
981 peer_ni = ksocknal_create_peer(ni, &peerid);
982 if (IS_ERR(peer_ni)) {
983 rc = PTR_ERR(peer_ni);
987 write_lock_bh(global_lock);
989 /* called with a ref on ni, so shutdown can't have started */
990 LASSERT(atomic_read(&((struct ksock_net *)ni->ni_data)->ksnn_npeers) >= 0);
992 peer2 = ksocknal_find_peer_locked(ni, &peerid);
994 /* NB this puts an "empty" peer_ni in the peer_ni
995 * table (which takes my ref) */
996 hash_add(ksocknal_data.ksnd_peers,
997 &peer_ni->ksnp_list, nidhash(&peerid.nid));
999 ksocknal_peer_decref(peer_ni);
1004 ksocknal_peer_addref(peer_ni);
1005 peer_ni->ksnp_accepting++;
1007 /* Am I already connecting to this guy? Resolve in
1008 * favour of higher NID...
1010 if (memcmp(&peerid.nid, &ni->ni_nid, sizeof(peerid.nid)) < 0 &&
1011 ksocknal_connecting(peer_ni->ksnp_conn_cb,
1012 ((struct sockaddr *) &conn->ksnc_peeraddr))) {
1014 warn = "connection race resolution";
1019 if (peer_ni->ksnp_closing ||
1020 (active && conn_cb->ksnr_deleted)) {
1021 /* peer_ni/conn_cb got closed under me */
1023 warn = "peer_ni/conn_cb removed";
1027 if (peer_ni->ksnp_proto == NULL) {
1028 /* Never connected before.
1029 * NB recv_hello may have returned EPROTO to signal my peer_ni
1030 * wants a different protocol than the one I asked for.
1032 LASSERT(list_empty(&peer_ni->ksnp_conns));
1034 peer_ni->ksnp_proto = conn->ksnc_proto;
1035 peer_ni->ksnp_incarnation = incarnation;
1038 if (peer_ni->ksnp_proto != conn->ksnc_proto ||
1039 peer_ni->ksnp_incarnation != incarnation) {
1040 /* peer_ni rebooted or I've got the wrong protocol version */
1041 ksocknal_close_peer_conns_locked(peer_ni, NULL, 0);
1043 peer_ni->ksnp_proto = NULL;
1045 warn = peer_ni->ksnp_incarnation != incarnation ?
1046 "peer_ni rebooted" :
1047 "wrong proto version";
1057 warn = "lost conn race";
1060 warn = "retry with different protocol version";
1064 /* Refuse to duplicate an existing connection, unless this is a
1065 * loopback connection */
1066 if (!rpc_cmp_addr((struct sockaddr *)&conn->ksnc_peeraddr,
1067 (struct sockaddr *)&conn->ksnc_myaddr)) {
1068 list_for_each_entry(conn2, &peer_ni->ksnp_conns, ksnc_list) {
1070 (struct sockaddr *)&conn2->ksnc_peeraddr,
1071 (struct sockaddr *)&conn->ksnc_peeraddr) ||
1073 (struct sockaddr *)&conn2->ksnc_myaddr,
1074 (struct sockaddr *)&conn->ksnc_myaddr) ||
1075 conn2->ksnc_type != conn->ksnc_type)
1079 /* If max conns per type is not registered in conn_cb
1080 * as ksnr_max_conns, use ni's conns_per_peer
1082 if ((peer_ni->ksnp_conn_cb &&
1083 num_dup < peer_ni->ksnp_conn_cb->ksnr_max_conns) ||
1084 (!peer_ni->ksnp_conn_cb &&
1085 num_dup < ksocknal_get_conns_per_peer(peer_ni)))
1088 /* Reply on a passive connection attempt so the peer_ni
1089 * realises we're connected.
1099 /* If the connection created by this route didn't bind to the IP
1100 * address the route connected to, the connection/route matching
1101 * code below probably isn't going to work.
1104 !rpc_cmp_addr((struct sockaddr *)&conn_cb->ksnr_addr,
1105 (struct sockaddr *)&conn->ksnc_peeraddr)) {
1106 CERROR("Route %s %pIS connected to %pIS\n",
1107 libcfs_idstr(&peer_ni->ksnp_id),
1108 &conn_cb->ksnr_addr,
1109 &conn->ksnc_peeraddr);
1112 /* Search for a conn_cb corresponding to the new connection and
1113 * create an association. This allows incoming connections created
1114 * by conn_cbs in my peer_ni to match my own conn_cb entries so I don't
1115 * continually create duplicate conn_cbs.
1117 conn_cb = peer_ni->ksnp_conn_cb;
1119 if (conn_cb && rpc_cmp_addr((struct sockaddr *)&conn->ksnc_peeraddr,
1120 (struct sockaddr *)&conn_cb->ksnr_addr))
1121 ksocknal_associate_cb_conn_locked(conn_cb, conn);
1123 conn->ksnc_peer = peer_ni; /* conn takes my ref on peer_ni */
1124 peer_ni->ksnp_last_alive = ktime_get_seconds();
1125 peer_ni->ksnp_send_keepalive = 0;
1126 peer_ni->ksnp_error = 0;
1128 sched = ksocknal_choose_scheduler_locked(cpt);
1130 CERROR("no schedulers available. node is unhealthy\n");
1134 * The cpt might have changed if we ended up selecting a non cpt
1135 * native scheduler. So use the scheduler's cpt instead.
1137 cpt = sched->kss_cpt;
1138 sched->kss_nconns++;
1139 conn->ksnc_scheduler = sched;
1141 conn->ksnc_tx_last_post = ktime_get_seconds();
1142 /* Set the deadline for the outgoing HELLO to drain */
1143 conn->ksnc_tx_bufnob = sock->sk->sk_wmem_queued;
1144 conn->ksnc_tx_deadline = ktime_get_seconds() +
1146 smp_mb(); /* order with adding to peer_ni's conn list */
1148 list_add(&conn->ksnc_list, &peer_ni->ksnp_conns);
1149 ksocknal_conn_addref(conn);
1151 ksocknal_new_packet(conn, 0);
1153 conn->ksnc_zc_capable = ksocknal_lib_zc_capable(conn);
1155 /* Take packets blocking for this connection. */
1156 list_for_each_entry_safe(tx, txtmp, &peer_ni->ksnp_tx_queue, tx_list) {
1157 if (conn->ksnc_proto->pro_match_tx(conn, tx, tx->tx_nonblk) ==
1161 list_del(&tx->tx_list);
1162 ksocknal_queue_tx_locked(tx, conn);
1165 write_unlock_bh(global_lock);
1166 /* We've now got a new connection. Any errors from here on are just
1167 * like "normal" comms errors and we close the connection normally.
1168 * NB (a) we still have to send the reply HELLO for passive
1170 * (b) normal I/O on the conn is blocked until I setup and call the
1174 CDEBUG(D_NET, "New conn %s p %d.x %pIS -> %pISp"
1175 " incarnation:%lld sched[%d]\n",
1176 libcfs_idstr(&peerid), conn->ksnc_proto->pro_version,
1177 &conn->ksnc_myaddr, &conn->ksnc_peeraddr,
1181 hello->kshm_nips = 0;
1182 rc = ksocknal_send_hello(ni, conn, &peerid.nid, hello);
1185 LIBCFS_FREE(hello, offsetof(struct ksock_hello_msg,
1186 kshm_ips[LNET_INTERFACES_NUM]));
1188 /* setup the socket AFTER I've received hello (it disables
1189 * SO_LINGER). I might call back to the acceptor who may want
1190 * to send a protocol version response and then close the
1191 * socket; this ensures the socket only tears down after the
1192 * response has been sent.
1195 rc = ksocknal_lib_setup_sock(sock);
1197 write_lock_bh(global_lock);
1199 /* NB my callbacks block while I hold ksnd_global_lock */
1200 ksocknal_lib_set_callback(sock, conn);
1203 peer_ni->ksnp_accepting--;
1205 write_unlock_bh(global_lock);
1208 write_lock_bh(global_lock);
1209 if (!conn->ksnc_closing) {
1210 /* could be closed by another thread */
1211 ksocknal_close_conn_locked(conn, rc);
1213 write_unlock_bh(global_lock);
1214 } else if (ksocknal_connsock_addref(conn) == 0) {
1215 /* Allow I/O to proceed. */
1216 ksocknal_read_callback(conn);
1217 ksocknal_write_callback(conn);
1218 ksocknal_connsock_decref(conn);
1221 ksocknal_connsock_decref(conn);
1222 ksocknal_conn_decref(conn);
1227 if (!peer_ni->ksnp_closing &&
1228 list_empty(&peer_ni->ksnp_conns) &&
1229 peer_ni->ksnp_conn_cb == NULL) {
1230 list_splice_init(&peer_ni->ksnp_tx_queue, &zombies);
1231 ksocknal_unlink_peer_locked(peer_ni);
1234 write_unlock_bh(global_lock);
1238 CERROR("Not creating conn %s type %d: %s\n",
1239 libcfs_idstr(&peerid), conn->ksnc_type, warn);
1241 CDEBUG(D_NET, "Not creating conn %s type %d: %s\n",
1242 libcfs_idstr(&peerid), conn->ksnc_type, warn);
1247 /* Request retry by replying with CONN_NONE
1248 * ksnc_proto has been set already
1250 conn->ksnc_type = SOCKLND_CONN_NONE;
1251 hello->kshm_nips = 0;
1252 ksocknal_send_hello(ni, conn, &peerid.nid, hello);
1255 write_lock_bh(global_lock);
1256 peer_ni->ksnp_accepting--;
1257 write_unlock_bh(global_lock);
1261 * If we get here without an error code, just use -EALREADY.
1262 * Depending on how we got here, the error may be positive
1263 * or negative. Normalize the value for ksocknal_txlist_done().
1265 rc2 = (rc == 0 ? -EALREADY : (rc > 0 ? -rc : rc));
1266 ksocknal_txlist_done(ni, &zombies, rc2);
1267 ksocknal_peer_decref(peer_ni);
1271 LIBCFS_FREE(hello, offsetof(struct ksock_hello_msg,
1272 kshm_ips[LNET_INTERFACES_NUM]));
1274 LIBCFS_FREE(conn, sizeof(*conn));
1283 ksocknal_close_conn_locked(struct ksock_conn *conn, int error)
1285 /* This just does the immmediate housekeeping, and queues the
1286 * connection for the reaper to terminate.
1287 * Caller holds ksnd_global_lock exclusively in irq context */
1288 struct ksock_peer_ni *peer_ni = conn->ksnc_peer;
1289 struct ksock_conn_cb *conn_cb;
1290 struct ksock_conn *conn2;
1292 int duplicate_count = 0;
1294 LASSERT(peer_ni->ksnp_error == 0);
1295 LASSERT(!conn->ksnc_closing);
1296 conn->ksnc_closing = 1;
1298 /* ksnd_deathrow_conns takes over peer_ni's ref */
1299 list_del(&conn->ksnc_list);
1301 conn_cb = conn->ksnc_conn_cb;
1302 if (conn_cb != NULL) {
1303 /* dissociate conn from cb... */
1304 LASSERT(!conn_cb->ksnr_deleted);
1306 conn_count = ksocknal_get_conn_count_by_type(conn_cb,
1308 /* connected bit is set only if all connections
1309 * of the given type got created
1311 if (conn_count == conn_cb->ksnr_max_conns)
1312 LASSERT((conn_cb->ksnr_connected &
1313 BIT(conn->ksnc_type)) != 0);
1315 if (conn_count == 1) {
1316 list_for_each_entry(conn2, &peer_ni->ksnp_conns,
1318 if (conn2->ksnc_conn_cb == conn_cb &&
1319 conn2->ksnc_type == conn->ksnc_type)
1320 duplicate_count += 1;
1322 if (duplicate_count > 0)
1323 CERROR("Found %d duplicate conns type %d\n",
1327 ksocknal_decr_conn_count(conn_cb, conn->ksnc_type);
1329 conn->ksnc_conn_cb = NULL;
1331 /* drop conn's ref on conn_cb */
1332 ksocknal_conn_cb_decref(conn_cb);
1335 if (list_empty(&peer_ni->ksnp_conns)) {
1336 /* No more connections to this peer_ni */
1338 if (!list_empty(&peer_ni->ksnp_tx_queue)) {
1339 struct ksock_tx *tx;
1341 LASSERT(conn->ksnc_proto == &ksocknal_protocol_v3x);
1343 /* throw them to the last connection...,
1344 * these TXs will be send to /dev/null by scheduler */
1345 list_for_each_entry(tx, &peer_ni->ksnp_tx_queue,
1347 ksocknal_tx_prep(conn, tx);
1349 spin_lock_bh(&conn->ksnc_scheduler->kss_lock);
1350 list_splice_init(&peer_ni->ksnp_tx_queue,
1351 &conn->ksnc_tx_queue);
1352 spin_unlock_bh(&conn->ksnc_scheduler->kss_lock);
1355 /* renegotiate protocol version */
1356 peer_ni->ksnp_proto = NULL;
1357 /* stash last conn close reason */
1358 peer_ni->ksnp_error = error;
1360 if (peer_ni->ksnp_conn_cb == NULL) {
1361 /* I've just closed last conn belonging to a
1362 * peer_ni with no connections to it
1364 ksocknal_unlink_peer_locked(peer_ni);
1368 spin_lock_bh(&ksocknal_data.ksnd_reaper_lock);
1370 list_add_tail(&conn->ksnc_list, &ksocknal_data.ksnd_deathrow_conns);
1371 wake_up(&ksocknal_data.ksnd_reaper_waitq);
1373 spin_unlock_bh(&ksocknal_data.ksnd_reaper_lock);
1377 ksocknal_peer_failed(struct ksock_peer_ni *peer_ni)
1379 bool notify = false;
1380 time64_t last_alive = 0;
1382 /* There has been a connection failure or comms error; but I'll only
1383 * tell LNET I think the peer_ni is dead if it's to another kernel and
1384 * there are no connections or connection attempts in existence. */
1386 read_lock(&ksocknal_data.ksnd_global_lock);
1388 if ((peer_ni->ksnp_id.pid & LNET_PID_USERFLAG) == 0 &&
1389 list_empty(&peer_ni->ksnp_conns) &&
1390 peer_ni->ksnp_accepting == 0 &&
1391 !ksocknal_find_connecting_conn_cb_locked(peer_ni)) {
1393 last_alive = peer_ni->ksnp_last_alive;
1396 read_unlock(&ksocknal_data.ksnd_global_lock);
1399 lnet_notify(peer_ni->ksnp_ni,
1400 lnet_nid_to_nid4(&peer_ni->ksnp_id.nid),
1401 false, false, last_alive);
1405 ksocknal_finalize_zcreq(struct ksock_conn *conn)
1407 struct ksock_peer_ni *peer_ni = conn->ksnc_peer;
1408 struct ksock_tx *tx;
1409 struct ksock_tx *tmp;
1412 /* NB safe to finalize TXs because closing of socket will
1413 * abort all buffered data */
1414 LASSERT(conn->ksnc_sock == NULL);
1416 spin_lock(&peer_ni->ksnp_lock);
1418 list_for_each_entry_safe(tx, tmp, &peer_ni->ksnp_zc_req_list,
1420 if (tx->tx_conn != conn)
1423 LASSERT(tx->tx_msg.ksm_zc_cookies[0] != 0);
1425 tx->tx_msg.ksm_zc_cookies[0] = 0;
1426 tx->tx_zc_aborted = 1; /* mark it as not-acked */
1427 list_move(&tx->tx_zc_list, &zlist);
1430 spin_unlock(&peer_ni->ksnp_lock);
1432 while ((tx = list_first_entry_or_null(&zlist, struct ksock_tx,
1433 tx_zc_list)) != NULL) {
1434 list_del(&tx->tx_zc_list);
1435 ksocknal_tx_decref(tx);
1440 ksocknal_terminate_conn(struct ksock_conn *conn)
1442 /* This gets called by the reaper (guaranteed thread context) to
1443 * disengage the socket from its callbacks and close it.
1444 * ksnc_refcount will eventually hit zero, and then the reaper will
1447 struct ksock_peer_ni *peer_ni = conn->ksnc_peer;
1448 struct ksock_sched *sched = conn->ksnc_scheduler;
1449 bool failed = false;
1451 LASSERT(conn->ksnc_closing);
1453 /* wake up the scheduler to "send" all remaining packets to /dev/null */
1454 spin_lock_bh(&sched->kss_lock);
1456 /* a closing conn is always ready to tx */
1457 conn->ksnc_tx_ready = 1;
1459 if (!conn->ksnc_tx_scheduled &&
1460 !list_empty(&conn->ksnc_tx_queue)) {
1461 list_add_tail(&conn->ksnc_tx_list,
1462 &sched->kss_tx_conns);
1463 conn->ksnc_tx_scheduled = 1;
1464 /* extra ref for scheduler */
1465 ksocknal_conn_addref(conn);
1467 wake_up(&sched->kss_waitq);
1470 spin_unlock_bh(&sched->kss_lock);
1472 /* serialise with callbacks */
1473 write_lock_bh(&ksocknal_data.ksnd_global_lock);
1475 ksocknal_lib_reset_callback(conn->ksnc_sock, conn);
1477 /* OK, so this conn may not be completely disengaged from its
1478 * scheduler yet, but it _has_ committed to terminate...
1480 conn->ksnc_scheduler->kss_nconns--;
1482 if (peer_ni->ksnp_error != 0) {
1483 /* peer_ni's last conn closed in error */
1484 LASSERT(list_empty(&peer_ni->ksnp_conns));
1486 peer_ni->ksnp_error = 0; /* avoid multiple notifications */
1489 write_unlock_bh(&ksocknal_data.ksnd_global_lock);
1492 ksocknal_peer_failed(peer_ni);
1494 /* The socket is closed on the final put; either here, or in
1495 * ksocknal_{send,recv}msg(). Since we set up the linger2 option
1496 * when the connection was established, this will close the socket
1497 * immediately, aborting anything buffered in it. Any hung
1498 * zero-copy transmits will therefore complete in finite time.
1500 ksocknal_connsock_decref(conn);
1504 ksocknal_queue_zombie_conn(struct ksock_conn *conn)
1506 /* Queue the conn for the reaper to destroy */
1507 LASSERT(refcount_read(&conn->ksnc_conn_refcount) == 0);
1508 spin_lock_bh(&ksocknal_data.ksnd_reaper_lock);
1510 list_add_tail(&conn->ksnc_list, &ksocknal_data.ksnd_zombie_conns);
1511 wake_up(&ksocknal_data.ksnd_reaper_waitq);
1513 spin_unlock_bh(&ksocknal_data.ksnd_reaper_lock);
1517 ksocknal_destroy_conn(struct ksock_conn *conn)
1521 /* Final coup-de-grace of the reaper */
1522 CDEBUG(D_NET, "connection %p\n", conn);
1524 LASSERT(refcount_read(&conn->ksnc_conn_refcount) == 0);
1525 LASSERT(refcount_read(&conn->ksnc_sock_refcount) == 0);
1526 LASSERT(conn->ksnc_sock == NULL);
1527 LASSERT(conn->ksnc_conn_cb == NULL);
1528 LASSERT(!conn->ksnc_tx_scheduled);
1529 LASSERT(!conn->ksnc_rx_scheduled);
1530 LASSERT(list_empty(&conn->ksnc_tx_queue));
1532 /* complete current receive if any */
1533 switch (conn->ksnc_rx_state) {
1534 case SOCKNAL_RX_LNET_PAYLOAD:
1535 last_rcv = conn->ksnc_rx_deadline -
1537 CERROR("Completing partial receive from %s[%d], ip %pISp, with error, wanted: %d, left: %d, last alive is %lld secs ago\n",
1538 libcfs_idstr(&conn->ksnc_peer->ksnp_id),
1540 &conn->ksnc_peeraddr,
1541 conn->ksnc_rx_nob_wanted, conn->ksnc_rx_nob_left,
1542 ktime_get_seconds() - last_rcv);
1543 if (conn->ksnc_lnet_msg)
1544 conn->ksnc_lnet_msg->msg_health_status =
1545 LNET_MSG_STATUS_REMOTE_ERROR;
1546 lnet_finalize(conn->ksnc_lnet_msg, -EIO);
1548 case SOCKNAL_RX_LNET_HEADER:
1549 if (conn->ksnc_rx_started)
1550 CERROR("Incomplete receive of lnet header from %s, ip %pISp, with error, protocol: %d.x.\n",
1551 libcfs_idstr(&conn->ksnc_peer->ksnp_id),
1552 &conn->ksnc_peeraddr,
1553 conn->ksnc_proto->pro_version);
1555 case SOCKNAL_RX_KSM_HEADER:
1556 if (conn->ksnc_rx_started)
1557 CERROR("Incomplete receive of ksock message from %s, ip %pISp, with error, protocol: %d.x.\n",
1558 libcfs_idstr(&conn->ksnc_peer->ksnp_id),
1559 &conn->ksnc_peeraddr,
1560 conn->ksnc_proto->pro_version);
1562 case SOCKNAL_RX_SLOP:
1563 if (conn->ksnc_rx_started)
1564 CERROR("Incomplete receive of slops from %s, ip %pISp, with error\n",
1565 libcfs_idstr(&conn->ksnc_peer->ksnp_id),
1566 &conn->ksnc_peeraddr);
1573 ksocknal_peer_decref(conn->ksnc_peer);
1575 LIBCFS_FREE(conn, sizeof(*conn));
1579 ksocknal_close_peer_conns_locked(struct ksock_peer_ni *peer_ni,
1580 struct sockaddr *addr, int why)
1582 struct ksock_conn *conn;
1583 struct ksock_conn *cnxt;
1586 list_for_each_entry_safe(conn, cnxt, &peer_ni->ksnp_conns, ksnc_list) {
1589 (struct sockaddr *)&conn->ksnc_peeraddr)) {
1591 ksocknal_close_conn_locked(conn, why);
1599 ksocknal_close_conn_and_siblings(struct ksock_conn *conn, int why)
1601 struct ksock_peer_ni *peer_ni = conn->ksnc_peer;
1604 write_lock_bh(&ksocknal_data.ksnd_global_lock);
1606 count = ksocknal_close_peer_conns_locked(
1607 peer_ni, (struct sockaddr *)&conn->ksnc_peeraddr, why);
1609 write_unlock_bh(&ksocknal_data.ksnd_global_lock);
1615 ksocknal_close_matching_conns(struct lnet_processid *id, __u32 ipaddr)
1617 struct ksock_peer_ni *peer_ni;
1618 struct hlist_node *pnxt;
1623 struct sockaddr_in sa = {.sin_family = AF_INET};
1625 write_lock_bh(&ksocknal_data.ksnd_global_lock);
1627 if (!LNET_NID_IS_ANY(&id->nid)) {
1628 lo = hash_min(nidhash(&id->nid),
1629 HASH_BITS(ksocknal_data.ksnd_peers));
1633 hi = HASH_SIZE(ksocknal_data.ksnd_peers) - 1;
1636 sa.sin_addr.s_addr = htonl(ipaddr);
1637 for (i = lo; i <= hi; i++) {
1638 hlist_for_each_entry_safe(peer_ni, pnxt,
1639 &ksocknal_data.ksnd_peers[i],
1642 if (!((LNET_NID_IS_ANY(&id->nid) ||
1643 nid_same(&id->nid, &peer_ni->ksnp_id.nid)) &&
1644 (id->pid == LNET_PID_ANY ||
1645 id->pid == peer_ni->ksnp_id.pid)))
1648 count += ksocknal_close_peer_conns_locked(
1650 ipaddr ? (struct sockaddr *)&sa : NULL, 0);
1654 write_unlock_bh(&ksocknal_data.ksnd_global_lock);
1656 /* wildcards always succeed */
1657 if (LNET_NID_IS_ANY(&id->nid) || id->pid == LNET_PID_ANY ||
1661 return (count == 0 ? -ENOENT : 0);
1665 ksocknal_notify_gw_down(struct lnet_nid *gw_nid)
1667 /* The router is telling me she's been notified of a change in
1670 struct lnet_processid id = {
1671 .pid = LNET_PID_ANY,
1675 CDEBUG(D_NET, "gw %s down\n", libcfs_nidstr(gw_nid));
1677 /* If the gateway crashed, close all open connections... */
1678 ksocknal_close_matching_conns(&id, 0);
1681 /* We can only establish new connections
1682 * if we have autroutes, and these connect on demand.
1687 ksocknal_push_peer(struct ksock_peer_ni *peer_ni)
1691 struct ksock_conn *conn;
1693 for (index = 0; ; index++) {
1694 read_lock(&ksocknal_data.ksnd_global_lock);
1699 list_for_each_entry(conn, &peer_ni->ksnp_conns, ksnc_list) {
1701 ksocknal_conn_addref(conn);
1706 read_unlock(&ksocknal_data.ksnd_global_lock);
1711 ksocknal_lib_push_conn (conn);
1712 ksocknal_conn_decref(conn);
1717 ksocknal_push(struct lnet_ni *ni, struct lnet_processid *id)
1724 if (!LNET_NID_IS_ANY(&id->nid)) {
1725 lo = hash_min(nidhash(&id->nid),
1726 HASH_BITS(ksocknal_data.ksnd_peers));
1730 hi = HASH_SIZE(ksocknal_data.ksnd_peers) - 1;
1733 for (bkt = lo; bkt <= hi; bkt++) {
1734 int peer_off; /* searching offset in peer_ni hash table */
1736 for (peer_off = 0; ; peer_off++) {
1737 struct ksock_peer_ni *peer_ni;
1740 read_lock(&ksocknal_data.ksnd_global_lock);
1741 hlist_for_each_entry(peer_ni,
1742 &ksocknal_data.ksnd_peers[bkt],
1744 if (!((LNET_NID_IS_ANY(&id->nid) ||
1746 &peer_ni->ksnp_id.nid)) &&
1747 (id->pid == LNET_PID_ANY ||
1748 id->pid == peer_ni->ksnp_id.pid)))
1751 if (i++ == peer_off) {
1752 ksocknal_peer_addref(peer_ni);
1756 read_unlock(&ksocknal_data.ksnd_global_lock);
1758 if (i <= peer_off) /* no match */
1762 ksocknal_push_peer(peer_ni);
1763 ksocknal_peer_decref(peer_ni);
1770 ksocknal_ctl(struct lnet_ni *ni, unsigned int cmd, void *arg)
1772 struct lnet_processid id = {};
1773 struct libcfs_ioctl_data *data = arg;
1777 case IOC_LIBCFS_GET_INTERFACE: {
1778 struct ksock_net *net = ni->ni_data;
1779 struct ksock_interface *iface;
1780 struct sockaddr_in *sa;
1782 read_lock(&ksocknal_data.ksnd_global_lock);
1784 if (data->ioc_count >= 1) {
1788 iface = &net->ksnn_interface;
1790 sa = (void *)&iface->ksni_addr;
1791 if (sa->sin_family == AF_INET)
1792 data->ioc_u32[0] = ntohl(sa->sin_addr.s_addr);
1794 data->ioc_u32[0] = 0xFFFFFFFF;
1795 data->ioc_u32[1] = iface->ksni_netmask;
1796 data->ioc_u32[2] = iface->ksni_npeers;
1797 data->ioc_u32[3] = iface->ksni_nroutes;
1800 read_unlock(&ksocknal_data.ksnd_global_lock);
1804 case IOC_LIBCFS_GET_PEER: {
1809 int share_count = 0;
1811 rc = ksocknal_get_peer_info(ni, data->ioc_count,
1812 &id, &myip, &ip, &port,
1813 &conn_count, &share_count);
1816 if (!nid_is_nid4(&id.nid))
1818 data->ioc_nid = lnet_nid_to_nid4(&id.nid);
1819 data->ioc_count = share_count;
1820 data->ioc_u32[0] = ip;
1821 data->ioc_u32[1] = port;
1822 data->ioc_u32[2] = myip;
1823 data->ioc_u32[3] = conn_count;
1824 data->ioc_u32[4] = id.pid;
1828 case IOC_LIBCFS_ADD_PEER: {
1829 struct sockaddr_in sa = {.sin_family = AF_INET};
1831 id.pid = LNET_PID_LUSTRE;
1832 lnet_nid4_to_nid(data->ioc_nid, &id.nid);
1833 sa.sin_addr.s_addr = htonl(data->ioc_u32[0]);
1834 sa.sin_port = htons(data->ioc_u32[1]);
1835 return ksocknal_add_peer(ni, &id, (struct sockaddr *)&sa);
1837 case IOC_LIBCFS_DEL_PEER:
1838 lnet_nid4_to_nid(data->ioc_nid, &id.nid);
1839 id.pid = LNET_PID_ANY;
1840 return ksocknal_del_peer(ni, &id);
1842 case IOC_LIBCFS_GET_CONN: {
1846 struct ksock_conn *conn = ksocknal_get_conn_by_idx(ni, data->ioc_count);
1847 struct sockaddr_in *psa = (void *)&conn->ksnc_peeraddr;
1848 struct sockaddr_in *mysa = (void *)&conn->ksnc_myaddr;
1853 ksocknal_lib_get_conn_tunables(conn, &txmem, &rxmem, &nagle);
1855 data->ioc_count = txmem;
1856 data->ioc_nid = lnet_nid_to_nid4(&conn->ksnc_peer->ksnp_id.nid);
1857 data->ioc_flags = nagle;
1858 if (psa->sin_family == AF_INET)
1859 data->ioc_u32[0] = ntohl(psa->sin_addr.s_addr);
1861 data->ioc_u32[0] = 0xFFFFFFFF;
1862 data->ioc_u32[1] = rpc_get_port((struct sockaddr *)
1863 &conn->ksnc_peeraddr);
1864 if (mysa->sin_family == AF_INET)
1865 data->ioc_u32[2] = ntohl(mysa->sin_addr.s_addr);
1867 data->ioc_u32[2] = 0xFFFFFFFF;
1868 data->ioc_u32[3] = conn->ksnc_type;
1869 data->ioc_u32[4] = conn->ksnc_scheduler->kss_cpt;
1870 data->ioc_u32[5] = rxmem;
1871 data->ioc_u32[6] = conn->ksnc_peer->ksnp_id.pid;
1872 ksocknal_conn_decref(conn);
1876 case IOC_LIBCFS_CLOSE_CONNECTION:
1877 lnet_nid4_to_nid(data->ioc_nid, &id.nid);
1878 id.pid = LNET_PID_ANY;
1879 return ksocknal_close_matching_conns(&id,
1882 case IOC_LIBCFS_REGISTER_MYNID:
1883 /* Ignore if this is a noop */
1884 if (nid_is_nid4(&ni->ni_nid) &&
1885 data->ioc_nid == lnet_nid_to_nid4(&ni->ni_nid))
1888 CERROR("obsolete IOC_LIBCFS_REGISTER_MYNID: %s(%s)\n",
1889 libcfs_nid2str(data->ioc_nid),
1890 libcfs_nidstr(&ni->ni_nid));
1893 case IOC_LIBCFS_PUSH_CONNECTION:
1894 lnet_nid4_to_nid(data->ioc_nid, &id.nid);
1895 id.pid = LNET_PID_ANY;
1896 return ksocknal_push(ni, &id);
1905 ksocknal_free_buffers (void)
1907 LASSERT (atomic_read(&ksocknal_data.ksnd_nactive_txs) == 0);
1909 if (ksocknal_data.ksnd_schedulers != NULL)
1910 cfs_percpt_free(ksocknal_data.ksnd_schedulers);
1912 spin_lock(&ksocknal_data.ksnd_tx_lock);
1914 if (!list_empty(&ksocknal_data.ksnd_idle_noop_txs)) {
1916 struct ksock_tx *tx;
1918 list_splice_init(&ksocknal_data.ksnd_idle_noop_txs, &zlist);
1919 spin_unlock(&ksocknal_data.ksnd_tx_lock);
1921 while ((tx = list_first_entry_or_null(&zlist, struct ksock_tx,
1922 tx_list)) != NULL) {
1923 list_del(&tx->tx_list);
1924 LIBCFS_FREE(tx, tx->tx_desc_size);
1927 spin_unlock(&ksocknal_data.ksnd_tx_lock);
1931 static int ksocknal_get_link_status(struct net_device *dev)
1937 if (!netif_running(dev)) {
1939 CDEBUG(D_NET, "device not running\n");
1941 /* Some devices may not be providing link settings */
1942 else if (dev->ethtool_ops->get_link) {
1943 ret = dev->ethtool_ops->get_link(dev);
1944 CDEBUG(D_NET, "get_link returns %u\n", ret);
1951 ksocknal_handle_link_state_change(struct net_device *dev,
1952 unsigned char operstate)
1954 struct lnet_ni *ni = NULL;
1955 struct ksock_net *net;
1956 struct ksock_net *cnxt;
1958 unsigned char link_down = !(operstate == IF_OPER_UP);
1959 struct in_device *in_dev;
1960 bool found_ip = false;
1961 struct ksock_interface *ksi = NULL;
1962 struct sockaddr_in *sa;
1963 DECLARE_CONST_IN_IFADDR(ifa);
1965 ifindex = dev->ifindex;
1967 if (!ksocknal_data.ksnd_nnets)
1970 list_for_each_entry_safe(net, cnxt, &ksocknal_data.ksnd_nets,
1973 ksi = &net->ksnn_interface;
1974 sa = (void *)&ksi->ksni_addr;
1977 if (ksi->ksni_index != ifindex ||
1978 strcmp(ksi->ksni_name, dev->name))
1983 in_dev = __in_dev_get_rtnl(dev);
1985 CDEBUG(D_NET, "Interface %s has no IPv4 status.\n",
1987 CDEBUG(D_NET, "set link fatal state to 1\n");
1988 atomic_set(&ni->ni_fatal_error_on, 1);
1991 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1992 if (sa->sin_addr.s_addr == ifa->ifa_local)
1998 CDEBUG(D_NET, "Interface %s has no matching ip\n",
2000 CDEBUG(D_NET, "set link fatal state to 1\n");
2001 atomic_set(&ni->ni_fatal_error_on, 1);
2006 CDEBUG(D_NET, "set link fatal state to 1\n");
2007 atomic_set(&ni->ni_fatal_error_on, link_down);
2009 CDEBUG(D_NET, "set link fatal state to %u\n",
2010 (ksocknal_get_link_status(dev) == 0));
2011 atomic_set(&ni->ni_fatal_error_on,
2012 (ksocknal_get_link_status(dev) == 0));
2021 ksocknal_handle_inetaddr_change(struct in_ifaddr *ifa, unsigned long event)
2024 struct ksock_net *net;
2025 struct ksock_net *cnxt;
2026 struct net_device *event_netdev = ifa->ifa_dev->dev;
2028 struct ksock_interface *ksi = NULL;
2029 struct sockaddr_in *sa;
2031 if (!ksocknal_data.ksnd_nnets)
2034 ifindex = event_netdev->ifindex;
2036 list_for_each_entry_safe(net, cnxt, &ksocknal_data.ksnd_nets,
2039 ksi = &net->ksnn_interface;
2040 sa = (void *)&ksi->ksni_addr;
2042 if (ksi->ksni_index != ifindex ||
2043 strcmp(ksi->ksni_name, event_netdev->name))
2046 if (sa->sin_addr.s_addr == ifa->ifa_local) {
2047 CDEBUG(D_NET, "set link fatal state to %u\n",
2048 (event == NETDEV_DOWN));
2050 atomic_set(&ni->ni_fatal_error_on,
2051 (event == NETDEV_DOWN));
2058 /************************************
2059 * Net device notifier event handler
2060 ************************************/
2061 static int ksocknal_device_event(struct notifier_block *unused,
2062 unsigned long event, void *ptr)
2064 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2065 unsigned char operstate;
2067 operstate = dev->operstate;
2069 CDEBUG(D_NET, "devevent: status=%ld, iface=%s ifindex %d state %u\n",
2070 event, dev->name, dev->ifindex, operstate);
2076 ksocknal_handle_link_state_change(dev, operstate);
2083 /************************************
2084 * Inetaddr notifier event handler
2085 ************************************/
2086 static int ksocknal_inetaddr_event(struct notifier_block *unused,
2087 unsigned long event, void *ptr)
2089 struct in_ifaddr *ifa = ptr;
2091 CDEBUG(D_NET, "addrevent: status %ld ip addr %pI4, netmask %pI4.\n",
2092 event, &ifa->ifa_address, &ifa->ifa_mask);
2098 ksocknal_handle_inetaddr_change(ifa, event);
2105 static struct notifier_block ksocknal_dev_notifier_block = {
2106 .notifier_call = ksocknal_device_event,
2109 static struct notifier_block ksocknal_inetaddr_notifier_block = {
2110 .notifier_call = ksocknal_inetaddr_event,
2114 ksocknal_base_shutdown(void)
2116 struct ksock_sched *sched;
2117 struct ksock_peer_ni *peer_ni;
2120 CDEBUG(D_MALLOC, "before NAL cleanup: kmem %lld\n",
2121 libcfs_kmem_read());
2122 LASSERT (ksocknal_data.ksnd_nnets == 0);
2124 if (ksocknal_data.ksnd_init == SOCKNAL_INIT_ALL) {
2125 unregister_netdevice_notifier(&ksocknal_dev_notifier_block);
2126 unregister_inetaddr_notifier(&ksocknal_inetaddr_notifier_block);
2129 switch (ksocknal_data.ksnd_init) {
2134 case SOCKNAL_INIT_ALL:
2135 case SOCKNAL_INIT_DATA:
2136 hash_for_each(ksocknal_data.ksnd_peers, i, peer_ni, ksnp_list)
2139 LASSERT(list_empty(&ksocknal_data.ksnd_nets));
2140 LASSERT(list_empty(&ksocknal_data.ksnd_enomem_conns));
2141 LASSERT(list_empty(&ksocknal_data.ksnd_zombie_conns));
2142 LASSERT(list_empty(&ksocknal_data.ksnd_connd_connreqs));
2143 LASSERT(list_empty(&ksocknal_data.ksnd_connd_routes));
2145 if (ksocknal_data.ksnd_schedulers != NULL) {
2146 cfs_percpt_for_each(sched, i,
2147 ksocknal_data.ksnd_schedulers) {
2149 LASSERT(list_empty(&sched->kss_tx_conns));
2150 LASSERT(list_empty(&sched->kss_rx_conns));
2151 LASSERT(list_empty(&sched->kss_zombie_noop_txs));
2152 LASSERT(sched->kss_nconns == 0);
2156 /* flag threads to terminate; wake and wait for them to die */
2157 ksocknal_data.ksnd_shuttingdown = 1;
2158 wake_up_all(&ksocknal_data.ksnd_connd_waitq);
2159 wake_up(&ksocknal_data.ksnd_reaper_waitq);
2161 if (ksocknal_data.ksnd_schedulers != NULL) {
2162 cfs_percpt_for_each(sched, i,
2163 ksocknal_data.ksnd_schedulers)
2164 wake_up_all(&sched->kss_waitq);
2167 wait_var_event_warning(&ksocknal_data.ksnd_nthreads,
2168 atomic_read(&ksocknal_data.ksnd_nthreads) == 0,
2169 "waiting for %d threads to terminate\n",
2170 atomic_read(&ksocknal_data.ksnd_nthreads));
2172 ksocknal_free_buffers();
2174 ksocknal_data.ksnd_init = SOCKNAL_INIT_NOTHING;
2178 CDEBUG(D_MALLOC, "after NAL cleanup: kmem %lld\n",
2179 libcfs_kmem_read());
2181 module_put(THIS_MODULE);
2185 ksocknal_base_startup(void)
2187 struct ksock_sched *sched;
2191 LASSERT(ksocknal_data.ksnd_init == SOCKNAL_INIT_NOTHING);
2192 LASSERT(ksocknal_data.ksnd_nnets == 0);
2194 memset(&ksocknal_data, 0, sizeof(ksocknal_data)); /* zero pointers */
2196 hash_init(ksocknal_data.ksnd_peers);
2198 rwlock_init(&ksocknal_data.ksnd_global_lock);
2199 INIT_LIST_HEAD(&ksocknal_data.ksnd_nets);
2201 spin_lock_init(&ksocknal_data.ksnd_reaper_lock);
2202 INIT_LIST_HEAD(&ksocknal_data.ksnd_enomem_conns);
2203 INIT_LIST_HEAD(&ksocknal_data.ksnd_zombie_conns);
2204 INIT_LIST_HEAD(&ksocknal_data.ksnd_deathrow_conns);
2205 init_waitqueue_head(&ksocknal_data.ksnd_reaper_waitq);
2207 spin_lock_init(&ksocknal_data.ksnd_connd_lock);
2208 INIT_LIST_HEAD(&ksocknal_data.ksnd_connd_connreqs);
2209 INIT_LIST_HEAD(&ksocknal_data.ksnd_connd_routes);
2210 init_waitqueue_head(&ksocknal_data.ksnd_connd_waitq);
2212 spin_lock_init(&ksocknal_data.ksnd_tx_lock);
2213 INIT_LIST_HEAD(&ksocknal_data.ksnd_idle_noop_txs);
2215 /* NB memset above zeros whole of ksocknal_data */
2217 /* flag lists/ptrs/locks initialised */
2218 ksocknal_data.ksnd_init = SOCKNAL_INIT_DATA;
2219 if (!try_module_get(THIS_MODULE))
2222 /* Create a scheduler block per available CPT */
2223 ksocknal_data.ksnd_schedulers = cfs_percpt_alloc(lnet_cpt_table(),
2225 if (ksocknal_data.ksnd_schedulers == NULL)
2228 cfs_percpt_for_each(sched, i, ksocknal_data.ksnd_schedulers) {
2232 * make sure not to allocate more threads than there are
2233 * cores/CPUs in teh CPT
2235 nthrs = cfs_cpt_weight(lnet_cpt_table(), i);
2236 if (*ksocknal_tunables.ksnd_nscheds > 0) {
2237 nthrs = min(nthrs, *ksocknal_tunables.ksnd_nscheds);
2240 * max to half of CPUs, assume another half should be
2241 * reserved for upper layer modules
2243 nthrs = min(max(SOCKNAL_NSCHEDS, nthrs >> 1), nthrs);
2246 sched->kss_nthreads_max = nthrs;
2249 spin_lock_init(&sched->kss_lock);
2250 INIT_LIST_HEAD(&sched->kss_rx_conns);
2251 INIT_LIST_HEAD(&sched->kss_tx_conns);
2252 INIT_LIST_HEAD(&sched->kss_zombie_noop_txs);
2253 init_waitqueue_head(&sched->kss_waitq);
2256 ksocknal_data.ksnd_connd_starting = 0;
2257 ksocknal_data.ksnd_connd_failed_stamp = 0;
2258 ksocknal_data.ksnd_connd_starting_stamp = ktime_get_real_seconds();
2259 /* must have at least 2 connds to remain responsive to accepts while
2261 if (*ksocknal_tunables.ksnd_nconnds < SOCKNAL_CONND_RESV + 1)
2262 *ksocknal_tunables.ksnd_nconnds = SOCKNAL_CONND_RESV + 1;
2264 if (*ksocknal_tunables.ksnd_nconnds_max <
2265 *ksocknal_tunables.ksnd_nconnds) {
2266 ksocknal_tunables.ksnd_nconnds_max =
2267 ksocknal_tunables.ksnd_nconnds;
2270 for (i = 0; i < *ksocknal_tunables.ksnd_nconnds; i++) {
2271 spin_lock_bh(&ksocknal_data.ksnd_connd_lock);
2272 ksocknal_data.ksnd_connd_starting++;
2273 spin_unlock_bh(&ksocknal_data.ksnd_connd_lock);
2275 rc = ksocknal_thread_start(ksocknal_connd,
2276 (void *)((uintptr_t)i),
2277 "socknal_cd%02d", i);
2279 spin_lock_bh(&ksocknal_data.ksnd_connd_lock);
2280 ksocknal_data.ksnd_connd_starting--;
2281 spin_unlock_bh(&ksocknal_data.ksnd_connd_lock);
2282 CERROR("Can't spawn socknal connd: %d\n", rc);
2287 rc = ksocknal_thread_start(ksocknal_reaper, NULL, "socknal_reaper");
2289 CERROR ("Can't spawn socknal reaper: %d\n", rc);
2293 register_netdevice_notifier(&ksocknal_dev_notifier_block);
2294 register_inetaddr_notifier(&ksocknal_inetaddr_notifier_block);
2296 /* flag everything initialised */
2297 ksocknal_data.ksnd_init = SOCKNAL_INIT_ALL;
2302 ksocknal_base_shutdown();
2307 ksocknal_debug_peerhash(struct lnet_ni *ni)
2309 struct ksock_peer_ni *peer_ni;
2312 read_lock(&ksocknal_data.ksnd_global_lock);
2314 hash_for_each(ksocknal_data.ksnd_peers, i, peer_ni, ksnp_list) {
2315 struct ksock_conn_cb *conn_cb;
2316 struct ksock_conn *conn;
2318 if (peer_ni->ksnp_ni != ni)
2321 CWARN("Active peer_ni on shutdown: %s, ref %d, closing %d, accepting %d, err %d, zcookie %llu, txq %d, zc_req %d\n",
2322 libcfs_idstr(&peer_ni->ksnp_id),
2323 refcount_read(&peer_ni->ksnp_refcount),
2324 peer_ni->ksnp_closing,
2325 peer_ni->ksnp_accepting, peer_ni->ksnp_error,
2326 peer_ni->ksnp_zc_next_cookie,
2327 !list_empty(&peer_ni->ksnp_tx_queue),
2328 !list_empty(&peer_ni->ksnp_zc_req_list));
2330 conn_cb = peer_ni->ksnp_conn_cb;
2332 CWARN("ConnCB: ref %d, schd %d, conn %d, cnted %d, del %d\n",
2333 refcount_read(&conn_cb->ksnr_refcount),
2334 conn_cb->ksnr_scheduled, conn_cb->ksnr_connecting,
2335 conn_cb->ksnr_connected, conn_cb->ksnr_deleted);
2338 list_for_each_entry(conn, &peer_ni->ksnp_conns, ksnc_list) {
2339 CWARN("Conn: ref %d, sref %d, t %d, c %d\n",
2340 refcount_read(&conn->ksnc_conn_refcount),
2341 refcount_read(&conn->ksnc_sock_refcount),
2342 conn->ksnc_type, conn->ksnc_closing);
2347 read_unlock(&ksocknal_data.ksnd_global_lock);
2352 ksocknal_shutdown(struct lnet_ni *ni)
2354 struct ksock_net *net = ni->ni_data;
2356 LASSERT(ksocknal_data.ksnd_init == SOCKNAL_INIT_ALL);
2357 LASSERT(ksocknal_data.ksnd_nnets > 0);
2359 /* prevent new peers */
2360 atomic_add(SOCKNAL_SHUTDOWN_BIAS, &net->ksnn_npeers);
2362 /* Delete all peers */
2363 ksocknal_del_peer(ni, NULL);
2365 /* Wait for all peer_ni state to clean up */
2366 wait_var_event_warning(&net->ksnn_npeers,
2367 atomic_read(&net->ksnn_npeers) ==
2368 SOCKNAL_SHUTDOWN_BIAS,
2369 "waiting for %d peers to disconnect\n",
2370 ksocknal_debug_peerhash(ni) +
2371 atomic_read(&net->ksnn_npeers) -
2372 SOCKNAL_SHUTDOWN_BIAS);
2374 LASSERT(net->ksnn_interface.ksni_npeers == 0);
2375 LASSERT(net->ksnn_interface.ksni_nroutes == 0);
2377 list_del(&net->ksnn_list);
2378 LIBCFS_FREE(net, sizeof(*net));
2380 ksocknal_data.ksnd_nnets--;
2381 if (ksocknal_data.ksnd_nnets == 0)
2382 ksocknal_base_shutdown();
2386 ksocknal_search_new_ipif(struct ksock_net *net)
2389 char *ifnam = &net->ksnn_interface.ksni_name[0];
2390 char *colon = strchr(ifnam, ':');
2392 struct ksock_net *tmp;
2397 list_for_each_entry(tmp, &ksocknal_data.ksnd_nets, ksnn_list) {
2398 char *ifnam2 = &tmp->ksnn_interface.ksni_name[0];
2399 char *colon2 = strchr(ifnam2, ':');
2404 found = strcmp(ifnam, ifnam2) == 0;
2417 ksocknal_start_schedulers(struct ksock_sched *sched)
2423 if (sched->kss_nthreads == 0) {
2424 if (*ksocknal_tunables.ksnd_nscheds > 0) {
2425 nthrs = sched->kss_nthreads_max;
2427 nthrs = cfs_cpt_weight(lnet_cpt_table(),
2429 nthrs = min(max(SOCKNAL_NSCHEDS, nthrs >> 1), nthrs);
2430 nthrs = min(SOCKNAL_NSCHEDS_HIGH, nthrs);
2432 nthrs = min(nthrs, sched->kss_nthreads_max);
2434 LASSERT(sched->kss_nthreads <= sched->kss_nthreads_max);
2435 /* increase two threads if there is new interface */
2436 nthrs = min(2, sched->kss_nthreads_max - sched->kss_nthreads);
2439 for (i = 0; i < nthrs; i++) {
2442 id = KSOCK_THREAD_ID(sched->kss_cpt, sched->kss_nthreads + i);
2443 rc = ksocknal_thread_start(ksocknal_scheduler, (void *)id,
2444 "socknal_sd%02d_%02d",
2446 (int)KSOCK_THREAD_SID(id));
2450 CERROR("Can't spawn thread %d for scheduler[%d]: %d\n",
2451 sched->kss_cpt, (int) KSOCK_THREAD_SID(id), rc);
2455 sched->kss_nthreads += i;
2460 ksocknal_net_start_threads(struct ksock_net *net, __u32 *cpts, int ncpts)
2462 int newif = ksocknal_search_new_ipif(net);
2466 if (ncpts > 0 && ncpts > cfs_cpt_number(lnet_cpt_table()))
2469 for (i = 0; i < ncpts; i++) {
2470 struct ksock_sched *sched;
2471 int cpt = (cpts == NULL) ? i : cpts[i];
2473 LASSERT(cpt < cfs_cpt_number(lnet_cpt_table()));
2474 sched = ksocknal_data.ksnd_schedulers[cpt];
2476 if (!newif && sched->kss_nthreads > 0)
2479 rc = ksocknal_start_schedulers(sched);
2487 ksocknal_startup(struct lnet_ni *ni)
2489 struct ksock_net *net;
2490 struct ksock_interface *ksi = NULL;
2491 struct lnet_inetdev *ifaces = NULL;
2492 struct sockaddr_in *sa;
2496 LASSERT (ni->ni_net->net_lnd == &the_ksocklnd);
2497 if (ksocknal_data.ksnd_init == SOCKNAL_INIT_NOTHING) {
2498 rc = ksocknal_base_startup();
2502 LIBCFS_ALLOC(net, sizeof(*net));
2505 net->ksnn_incarnation = ktime_get_real_ns();
2508 ksocknal_tunables_setup(ni);
2510 rc = lnet_inet_enumerate(&ifaces, ni->ni_net_ns);
2514 ksi = &net->ksnn_interface;
2516 /* Use the first discovered interface or look in the list */
2517 if (ni->ni_interface) {
2518 for (i = 0; i < rc; i++)
2519 if (strcmp(ifaces[i].li_name, ni->ni_interface) == 0)
2522 /* ni_interfaces doesn't contain the interface we want */
2524 CERROR("ksocklnd: failed to find interface %s\n",
2530 ni->ni_dev_cpt = ifaces[i].li_cpt;
2531 ksi->ksni_index = ifaces[i].li_index;
2532 sa = (void *)&ksi->ksni_addr;
2533 memset(sa, 0, sizeof(*sa));
2534 sa->sin_family = AF_INET;
2535 sa->sin_addr.s_addr = htonl(ifaces[i].li_ipaddr);
2536 ksi->ksni_netmask = ifaces[i].li_netmask;
2537 strlcpy(ksi->ksni_name, ifaces[i].li_name, sizeof(ksi->ksni_name));
2539 /* call it before add it to ksocknal_data.ksnd_nets */
2540 rc = ksocknal_net_start_threads(net, ni->ni_cpts, ni->ni_ncpts);
2545 LASSERT(ksi->ksni_addr.ss_family == AF_INET);
2546 ni->ni_nid.nid_addr[0] =
2547 ((struct sockaddr_in *)&ksi->ksni_addr)->sin_addr.s_addr;
2548 list_add(&net->ksnn_list, &ksocknal_data.ksnd_nets);
2550 ksocknal_data.ksnd_nnets++;
2555 LIBCFS_FREE(net, sizeof(*net));
2557 if (ksocknal_data.ksnd_nnets == 0)
2558 ksocknal_base_shutdown();
2563 static void __exit ksocklnd_exit(void)
2565 lnet_unregister_lnd(&the_ksocklnd);
2568 static const struct lnet_lnd the_ksocklnd = {
2569 .lnd_type = SOCKLND,
2570 .lnd_startup = ksocknal_startup,
2571 .lnd_shutdown = ksocknal_shutdown,
2572 .lnd_ctl = ksocknal_ctl,
2573 .lnd_send = ksocknal_send,
2574 .lnd_recv = ksocknal_recv,
2575 .lnd_notify_peer_down = ksocknal_notify_gw_down,
2576 .lnd_accept = ksocknal_accept,
2579 static int __init ksocklnd_init(void)
2583 /* check ksnr_connected/connecting field large enough */
2584 BUILD_BUG_ON(SOCKLND_CONN_NTYPES > 4);
2585 BUILD_BUG_ON(SOCKLND_CONN_ACK != SOCKLND_CONN_BULK_IN);
2587 rc = ksocknal_tunables_init();
2591 lnet_register_lnd(&the_ksocklnd);
2596 MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
2597 MODULE_DESCRIPTION("TCP Socket LNet Network Driver");
2598 MODULE_VERSION("2.8.0");
2599 MODULE_LICENSE("GPL");
2601 module_init(ksocklnd_init);
2602 module_exit(ksocklnd_exit);