-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
+/*
+ * GPL HEADER START
*
- * Copyright (C) 2001, 2002 Cluster File Systems, Inc.
- * Author: Zach Brown <zab@zabbo.net>
- * Author: Peter J. Braam <braam@clusterfs.com>
- * Author: Phil Schwan <phil@clusterfs.com>
- * Author: Eric Barton <eric@bartonsoftware.com>
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
- * This file is part of Portals, http://www.sf.net/projects/sandiaportals/
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 only,
+ * as published by the Free Software Foundation.
*
- * Portals is free software; you can redistribute it and/or
- * modify it under the terms of version 2 of the GNU General Public
- * License as published by the Free Software Foundation.
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License version 2 for more details (a copy is included
+ * in the LICENSE file that accompanied this code).
*
- * Portals is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * GNU General Public License for more details.
+ * You should have received a copy of the GNU General Public License
+ * version 2 along with this program; If not, see
+ * http://www.gnu.org/licenses/gpl-2.0.html
*
- * You should have received a copy of the GNU General Public License
- * along with Portals; if not, write to the Free Software
- * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ * GPL HEADER END
+ */
+/*
+ * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Use is subject to license terms.
+ *
+ * Copyright (c) 2011, 2017, Intel Corporation.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
+ *
+ * lnet/klnds/socklnd/socklnd.c
+ *
+ * Author: Zach Brown <zab@zabbo.net>
+ * Author: Peter J. Braam <braam@clusterfs.com>
+ * Author: Phil Schwan <phil@clusterfs.com>
+ * Author: Eric Barton <eric@bartonsoftware.com>
*/
-#include "socknal.h"
-
-nal_t ksocknal_api;
-ksock_nal_data_t ksocknal_data;
-ptl_handle_ni_t ksocknal_ni;
-ksock_tunables_t ksocknal_tunables;
-
-kpr_nal_interface_t ksocknal_router_interface = {
- kprni_nalid: SOCKNAL,
- kprni_arg: &ksocknal_data,
- kprni_fwd: ksocknal_fwd_packet,
- kprni_notify: ksocknal_notify,
-};
-
-int
-ksocknal_set_mynid(ptl_nid_t nid)
-{
- lib_ni_t *ni = &ksocknal_lib.libnal_ni;
-
- /* FIXME: we have to do this because we call lib_init() at module
- * insertion time, which is before we have 'mynid' available. lib_init
- * sets the NAL's nid, which it uses to tell other nodes where packets
- * are coming from. This is not a very graceful solution to this
- * problem. */
-
- CDEBUG(D_IOCTL, "setting mynid to "LPX64" (old nid="LPX64")\n",
- nid, ni->ni_pid.nid);
+#include "socklnd.h"
+#include <linux/inetdevice.h>
- ni->ni_pid.nid = nid;
- return (0);
-}
+static struct lnet_lnd the_ksocklnd;
+struct ksock_nal_data ksocknal_data;
-ksock_interface_t *
-ksocknal_ip2iface(__u32 ip)
+static struct ksock_interface *
+ksocknal_ip2iface(struct lnet_ni *ni, __u32 ip)
{
- int i;
- ksock_interface_t *iface;
+ struct ksock_net *net = ni->ni_data;
+ int i;
+ struct ksock_interface *iface;
- for (i = 0; i < ksocknal_data.ksnd_ninterfaces; i++) {
- LASSERT(i < SOCKNAL_MAX_INTERFACES);
- iface = &ksocknal_data.ksnd_interfaces[i];
+ for (i = 0; i < net->ksnn_ninterfaces; i++) {
+ LASSERT(i < LNET_INTERFACES_NUM);
+ iface = &net->ksnn_interfaces[i];
- if (iface->ksni_ipaddr == ip)
- return (iface);
- }
+ if (iface->ksni_ipaddr == ip)
+ return iface;
+ }
- return (NULL);
+ return NULL;
}
-ksock_route_t *
-ksocknal_create_route (__u32 ipaddr, int port)
+static struct ksock_route *
+ksocknal_create_route(__u32 ipaddr, int port)
{
- ksock_route_t *route;
+ struct ksock_route *route;
- PORTAL_ALLOC (route, sizeof (*route));
- if (route == NULL)
- return (NULL);
+ LIBCFS_ALLOC (route, sizeof (*route));
+ if (route == NULL)
+ return (NULL);
- atomic_set (&route->ksnr_refcount, 1);
- route->ksnr_peer = NULL;
- route->ksnr_timeout = cfs_time_current();
- route->ksnr_retry_interval = SOCKNAL_MIN_RECONNECT_INTERVAL;
- route->ksnr_ipaddr = ipaddr;
+ atomic_set (&route->ksnr_refcount, 1);
+ route->ksnr_peer = NULL;
+ route->ksnr_retry_interval = 0; /* OK to connect at any time */
+ route->ksnr_ipaddr = ipaddr;
route->ksnr_port = port;
+ route->ksnr_scheduled = 0;
route->ksnr_connecting = 0;
route->ksnr_connected = 0;
route->ksnr_deleted = 0;
}
void
-ksocknal_destroy_route (ksock_route_t *route)
+ksocknal_destroy_route(struct ksock_route *route)
{
- if (route->ksnr_peer != NULL)
- ksocknal_put_peer (route->ksnr_peer);
+ LASSERT (atomic_read(&route->ksnr_refcount) == 0);
- PORTAL_FREE (route, sizeof (*route));
-}
+ if (route->ksnr_peer != NULL)
+ ksocknal_peer_decref(route->ksnr_peer);
-void
-ksocknal_put_route (ksock_route_t *route)
-{
- CDEBUG (D_OTHER, "putting route[%p] (%d)\n",
- route, atomic_read (&route->ksnr_refcount));
-
- LASSERT (atomic_read (&route->ksnr_refcount) > 0);
- if (!atomic_dec_and_test (&route->ksnr_refcount))
- return;
-
- ksocknal_destroy_route (route);
+ LIBCFS_FREE (route, sizeof (*route));
}
-ksock_peer_t *
-ksocknal_create_peer (ptl_nid_t nid)
+static int
+ksocknal_create_peer(struct ksock_peer_ni **peerp, struct lnet_ni *ni,
+ struct lnet_process_id id)
{
- ksock_peer_t *peer;
+ int cpt = lnet_cpt_of_nid(id.nid, ni);
+ struct ksock_net *net = ni->ni_data;
+ struct ksock_peer_ni *peer_ni;
- LASSERT (nid != PTL_NID_ANY);
+ LASSERT(id.nid != LNET_NID_ANY);
+ LASSERT(id.pid != LNET_PID_ANY);
+ LASSERT(!in_interrupt());
- PORTAL_ALLOC (peer, sizeof (*peer));
- if (peer == NULL)
- return (NULL);
+ LIBCFS_CPT_ALLOC(peer_ni, lnet_cpt_table(), cpt, sizeof(*peer_ni));
+ if (peer_ni == NULL)
+ return -ENOMEM;
- memset (peer, 0, sizeof (*peer)); /* NULL pointers/clear flags etc */
+ peer_ni->ksnp_ni = ni;
+ peer_ni->ksnp_id = id;
+ atomic_set(&peer_ni->ksnp_refcount, 1); /* 1 ref for caller */
+ peer_ni->ksnp_closing = 0;
+ peer_ni->ksnp_accepting = 0;
+ peer_ni->ksnp_proto = NULL;
+ peer_ni->ksnp_last_alive = 0;
+ peer_ni->ksnp_zc_next_cookie = SOCKNAL_KEEPALIVE_PING + 1;
- peer->ksnp_nid = nid;
- atomic_set (&peer->ksnp_refcount, 1); /* 1 ref for caller */
- peer->ksnp_closing = 0;
- CFS_INIT_LIST_HEAD (&peer->ksnp_conns);
- CFS_INIT_LIST_HEAD (&peer->ksnp_routes);
- CFS_INIT_LIST_HEAD (&peer->ksnp_tx_queue);
+ INIT_LIST_HEAD(&peer_ni->ksnp_conns);
+ INIT_LIST_HEAD(&peer_ni->ksnp_routes);
+ INIT_LIST_HEAD(&peer_ni->ksnp_tx_queue);
+ INIT_LIST_HEAD(&peer_ni->ksnp_zc_req_list);
+ spin_lock_init(&peer_ni->ksnp_lock);
- atomic_inc (&ksocknal_data.ksnd_npeers);
- return (peer);
-}
+ spin_lock_bh(&net->ksnn_lock);
-void
-ksocknal_destroy_peer (ksock_peer_t *peer)
-{
- CDEBUG (D_NET, "peer "LPX64" %p deleted\n", peer->ksnp_nid, peer);
+ if (net->ksnn_shutdown) {
+ spin_unlock_bh(&net->ksnn_lock);
- LASSERT (atomic_read (&peer->ksnp_refcount) == 0);
- LASSERT (list_empty (&peer->ksnp_conns));
- LASSERT (list_empty (&peer->ksnp_routes));
- LASSERT (list_empty (&peer->ksnp_tx_queue));
+ LIBCFS_FREE(peer_ni, sizeof(*peer_ni));
+ CERROR("Can't create peer_ni: network shutdown\n");
+ return -ESHUTDOWN;
+ }
- PORTAL_FREE (peer, sizeof (*peer));
+ net->ksnn_npeers++;
- /* NB a peer's connections and autoconnect routes keep a reference
- * on their peer until they are destroyed, so we can be assured
- * that _all_ state to do with this peer has been cleaned up when
- * its refcount drops to zero. */
- atomic_dec (&ksocknal_data.ksnd_npeers);
+ spin_unlock_bh(&net->ksnn_lock);
+
+ *peerp = peer_ni;
+ return 0;
}
void
-ksocknal_put_peer (ksock_peer_t *peer)
+ksocknal_destroy_peer(struct ksock_peer_ni *peer_ni)
{
- CDEBUG (D_OTHER, "putting peer[%p] -> "LPX64" (%d)\n",
- peer, peer->ksnp_nid,
- atomic_read (&peer->ksnp_refcount));
-
- LASSERT (atomic_read (&peer->ksnp_refcount) > 0);
- if (!atomic_dec_and_test (&peer->ksnp_refcount))
- return;
-
- ksocknal_destroy_peer (peer);
+ struct ksock_net *net = peer_ni->ksnp_ni->ni_data;
+
+ CDEBUG (D_NET, "peer_ni %s %p deleted\n",
+ libcfs_id2str(peer_ni->ksnp_id), peer_ni);
+
+ LASSERT(atomic_read(&peer_ni->ksnp_refcount) == 0);
+ LASSERT(peer_ni->ksnp_accepting == 0);
+ LASSERT(list_empty(&peer_ni->ksnp_conns));
+ LASSERT(list_empty(&peer_ni->ksnp_routes));
+ LASSERT(list_empty(&peer_ni->ksnp_tx_queue));
+ LASSERT(list_empty(&peer_ni->ksnp_zc_req_list));
+
+ LIBCFS_FREE(peer_ni, sizeof(*peer_ni));
+
+ /* NB a peer_ni's connections and routes keep a reference on their peer_ni
+ * until they are destroyed, so we can be assured that _all_ state to
+ * do with this peer_ni has been cleaned up when its refcount drops to
+ * zero. */
+ spin_lock_bh(&net->ksnn_lock);
+ net->ksnn_npeers--;
+ spin_unlock_bh(&net->ksnn_lock);
}
-ksock_peer_t *
-ksocknal_find_peer_locked (ptl_nid_t nid)
+struct ksock_peer_ni *
+ksocknal_find_peer_locked(struct lnet_ni *ni, struct lnet_process_id id)
{
- struct list_head *peer_list = ksocknal_nid2peerlist (nid);
- struct list_head *tmp;
- ksock_peer_t *peer;
+ struct list_head *peer_list = ksocknal_nid2peerlist(id.nid);
+ struct list_head *tmp;
+ struct ksock_peer_ni *peer_ni;
- list_for_each (tmp, peer_list) {
+ list_for_each(tmp, peer_list) {
+ peer_ni = list_entry(tmp, struct ksock_peer_ni, ksnp_list);
- peer = list_entry (tmp, ksock_peer_t, ksnp_list);
+ LASSERT(!peer_ni->ksnp_closing);
- LASSERT (!peer->ksnp_closing);
+ if (peer_ni->ksnp_ni != ni)
+ continue;
- if (peer->ksnp_nid != nid)
- continue;
+ if (peer_ni->ksnp_id.nid != id.nid ||
+ peer_ni->ksnp_id.pid != id.pid)
+ continue;
- CDEBUG(D_NET, "got peer [%p] -> "LPX64" (%d)\n",
- peer, nid, atomic_read (&peer->ksnp_refcount));
- return (peer);
- }
- return (NULL);
+ CDEBUG(D_NET, "got peer_ni [%p] -> %s (%d)\n",
+ peer_ni, libcfs_id2str(id),
+ atomic_read(&peer_ni->ksnp_refcount));
+ return peer_ni;
+ }
+ return NULL;
}
-ksock_peer_t *
-ksocknal_get_peer (ptl_nid_t nid)
+struct ksock_peer_ni *
+ksocknal_find_peer(struct lnet_ni *ni, struct lnet_process_id id)
{
- ksock_peer_t *peer;
+ struct ksock_peer_ni *peer_ni;
- read_lock (&ksocknal_data.ksnd_global_lock);
- peer = ksocknal_find_peer_locked (nid);
- if (peer != NULL) /* +1 ref for caller? */
- atomic_inc (&peer->ksnp_refcount);
- read_unlock (&ksocknal_data.ksnd_global_lock);
+ read_lock(&ksocknal_data.ksnd_global_lock);
+ peer_ni = ksocknal_find_peer_locked(ni, id);
+ if (peer_ni != NULL) /* +1 ref for caller? */
+ ksocknal_peer_addref(peer_ni);
+ read_unlock(&ksocknal_data.ksnd_global_lock);
- return (peer);
+ return (peer_ni);
}
-void
-ksocknal_unlink_peer_locked (ksock_peer_t *peer)
+static void
+ksocknal_unlink_peer_locked(struct ksock_peer_ni *peer_ni)
{
- int i;
- __u32 ip;
-
- for (i = 0; i < peer->ksnp_n_passive_ips; i++) {
- LASSERT (i < SOCKNAL_MAX_INTERFACES);
- ip = peer->ksnp_passive_ips[i];
-
- ksocknal_ip2iface(ip)->ksni_npeers--;
- }
-
- LASSERT (list_empty(&peer->ksnp_conns));
- LASSERT (list_empty(&peer->ksnp_routes));
- LASSERT (!peer->ksnp_closing);
- peer->ksnp_closing = 1;
- list_del (&peer->ksnp_list);
- /* lose peerlist's ref */
- ksocknal_put_peer (peer);
+ int i;
+ __u32 ip;
+ struct ksock_interface *iface;
+
+ for (i = 0; i < peer_ni->ksnp_n_passive_ips; i++) {
+ LASSERT(i < LNET_INTERFACES_NUM);
+ ip = peer_ni->ksnp_passive_ips[i];
+
+ iface = ksocknal_ip2iface(peer_ni->ksnp_ni, ip);
+ /*
+ * All IPs in peer_ni->ksnp_passive_ips[] come from the
+ * interface list, therefore the call must succeed.
+ */
+ LASSERT(iface != NULL);
+
+ CDEBUG(D_NET, "peer_ni=%p iface=%p ksni_nroutes=%d\n",
+ peer_ni, iface, iface->ksni_nroutes);
+ iface->ksni_npeers--;
+ }
+
+ LASSERT(list_empty(&peer_ni->ksnp_conns));
+ LASSERT(list_empty(&peer_ni->ksnp_routes));
+ LASSERT(!peer_ni->ksnp_closing);
+ peer_ni->ksnp_closing = 1;
+ list_del(&peer_ni->ksnp_list);
+ /* lose peerlist's ref */
+ ksocknal_peer_decref(peer_ni);
}
-int
-ksocknal_get_peer_info (int index, ptl_nid_t *nid,
- __u32 *myip, __u32 *peer_ip, int *port,
- int *conn_count, int *share_count)
+static int
+ksocknal_get_peer_info(struct lnet_ni *ni, int index,
+ struct lnet_process_id *id, __u32 *myip, __u32 *peer_ip,
+ int *port, int *conn_count, int *share_count)
{
- ksock_peer_t *peer;
- struct list_head *ptmp;
- ksock_route_t *route;
- struct list_head *rtmp;
- int i;
- int j;
- int rc = -ENOENT;
+ struct ksock_peer_ni *peer_ni;
+ struct list_head *ptmp;
+ struct ksock_route *route;
+ struct list_head *rtmp;
+ int i;
+ int j;
+ int rc = -ENOENT;
- read_lock (&ksocknal_data.ksnd_global_lock);
+ read_lock(&ksocknal_data.ksnd_global_lock);
- for (i = 0; i < ksocknal_data.ksnd_peer_hash_size; i++) {
+ for (i = 0; i < ksocknal_data.ksnd_peer_hash_size; i++) {
+ list_for_each(ptmp, &ksocknal_data.ksnd_peers[i]) {
+ peer_ni = list_entry(ptmp, struct ksock_peer_ni, ksnp_list);
- list_for_each (ptmp, &ksocknal_data.ksnd_peers[i]) {
- peer = list_entry (ptmp, ksock_peer_t, ksnp_list);
+ if (peer_ni->ksnp_ni != ni)
+ continue;
- if (peer->ksnp_n_passive_ips == 0 &&
- list_empty(&peer->ksnp_routes)) {
- if (index-- > 0)
- continue;
+ if (peer_ni->ksnp_n_passive_ips == 0 &&
+ list_empty(&peer_ni->ksnp_routes)) {
+ if (index-- > 0)
+ continue;
- *nid = peer->ksnp_nid;
+ *id = peer_ni->ksnp_id;
*myip = 0;
*peer_ip = 0;
*port = 0;
goto out;
}
- for (j = 0; j < peer->ksnp_n_passive_ips; j++) {
- if (index-- > 0)
- continue;
+ for (j = 0; j < peer_ni->ksnp_n_passive_ips; j++) {
+ if (index-- > 0)
+ continue;
- *nid = peer->ksnp_nid;
- *myip = peer->ksnp_passive_ips[j];
+ *id = peer_ni->ksnp_id;
+ *myip = peer_ni->ksnp_passive_ips[j];
*peer_ip = 0;
*port = 0;
*conn_count = 0;
goto out;
}
- list_for_each (rtmp, &peer->ksnp_routes) {
- if (index-- > 0)
- continue;
-
- route = list_entry(rtmp, ksock_route_t,
- ksnr_list);
-
- *nid = peer->ksnp_nid;
- *myip = route->ksnr_myipaddr;
- *peer_ip = route->ksnr_ipaddr;
- *port = route->ksnr_port;
- *conn_count = route->ksnr_conn_count;
- *share_count = route->ksnr_share_count;
- rc = 0;
- goto out;
- }
- }
- }
- out:
- read_unlock (&ksocknal_data.ksnd_global_lock);
- return (rc);
+ list_for_each(rtmp, &peer_ni->ksnp_routes) {
+ if (index-- > 0)
+ continue;
+
+ route = list_entry(rtmp, struct ksock_route,
+ ksnr_list);
+
+ *id = peer_ni->ksnp_id;
+ *myip = route->ksnr_myipaddr;
+ *peer_ip = route->ksnr_ipaddr;
+ *port = route->ksnr_port;
+ *conn_count = route->ksnr_conn_count;
+ *share_count = route->ksnr_share_count;
+ rc = 0;
+ goto out;
+ }
+ }
+ }
+out:
+ read_unlock(&ksocknal_data.ksnd_global_lock);
+ return rc;
}
-void
-ksocknal_associate_route_conn_locked(ksock_route_t *route, ksock_conn_t *conn)
+static void
+ksocknal_associate_route_conn_locked(struct ksock_route *route, struct ksock_conn *conn)
{
- ksock_peer_t *peer = route->ksnr_peer;
- int type = conn->ksnc_type;
- ksock_interface_t *iface;
-
- conn->ksnc_route = route;
- atomic_inc (&route->ksnr_refcount);
-
- if (route->ksnr_myipaddr != conn->ksnc_myipaddr) {
- if (route->ksnr_myipaddr == 0) {
- /* route wasn't bound locally yet (the initial route) */
- CWARN("Binding "LPX64" %u.%u.%u.%u to %u.%u.%u.%u\n",
- peer->ksnp_nid,
- HIPQUAD(route->ksnr_ipaddr),
- HIPQUAD(conn->ksnc_myipaddr));
- } else {
- CWARN("Rebinding "LPX64" %u.%u.%u.%u from "
- "%u.%u.%u.%u to %u.%u.%u.%u\n",
- peer->ksnp_nid,
- HIPQUAD(route->ksnr_ipaddr),
- HIPQUAD(route->ksnr_myipaddr),
- HIPQUAD(conn->ksnc_myipaddr));
-
- iface = ksocknal_ip2iface(route->ksnr_myipaddr);
+ struct ksock_peer_ni *peer_ni = route->ksnr_peer;
+ int type = conn->ksnc_type;
+ struct ksock_interface *iface;
+
+ conn->ksnc_route = route;
+ ksocknal_route_addref(route);
+
+ if (route->ksnr_myipaddr != conn->ksnc_myipaddr) {
+ if (route->ksnr_myipaddr == 0) {
+ /* route wasn't bound locally yet (the initial route) */
+ CDEBUG(D_NET, "Binding %s %pI4h to %pI4h\n",
+ libcfs_id2str(peer_ni->ksnp_id),
+ &route->ksnr_ipaddr,
+ &conn->ksnc_myipaddr);
+ } else {
+ CDEBUG(D_NET, "Rebinding %s %pI4h from %pI4h "
+ "to %pI4h\n", libcfs_id2str(peer_ni->ksnp_id),
+ &route->ksnr_ipaddr,
+ &route->ksnr_myipaddr,
+ &conn->ksnc_myipaddr);
+
+ iface = ksocknal_ip2iface(route->ksnr_peer->ksnp_ni,
+ route->ksnr_myipaddr);
if (iface != NULL)
iface->ksni_nroutes--;
}
route->ksnr_myipaddr = conn->ksnc_myipaddr;
- iface = ksocknal_ip2iface(route->ksnr_myipaddr);
+ iface = ksocknal_ip2iface(route->ksnr_peer->ksnp_ni,
+ route->ksnr_myipaddr);
if (iface != NULL)
iface->ksni_nroutes++;
}
/* Successful connection => further attempts can
* proceed immediately */
- route->ksnr_timeout = cfs_time_current();
- route->ksnr_retry_interval = SOCKNAL_MIN_RECONNECT_INTERVAL;
+ route->ksnr_retry_interval = 0;
}
-void
-ksocknal_add_route_locked (ksock_peer_t *peer, ksock_route_t *route)
+static void
+ksocknal_add_route_locked(struct ksock_peer_ni *peer_ni, struct ksock_route *route)
{
- struct list_head *tmp;
- ksock_conn_t *conn;
- int type;
- ksock_route_t *route2;
-
- LASSERT (route->ksnr_peer == NULL);
- LASSERT (!route->ksnr_connecting);
- LASSERT (route->ksnr_connected == 0);
-
- /* LASSERT(unique) */
- list_for_each(tmp, &peer->ksnp_routes) {
- route2 = list_entry(tmp, ksock_route_t, ksnr_list);
-
- if (route2->ksnr_ipaddr == route->ksnr_ipaddr) {
- CERROR ("Duplicate route "LPX64" %u.%u.%u.%u\n",
- peer->ksnp_nid, HIPQUAD(route->ksnr_ipaddr));
- LBUG();
- }
- }
-
- route->ksnr_peer = peer;
- atomic_inc (&peer->ksnp_refcount);
- /* peer's routelist takes over my ref on 'route' */
- list_add_tail(&route->ksnr_list, &peer->ksnp_routes);
-
- list_for_each(tmp, &peer->ksnp_conns) {
- conn = list_entry(tmp, ksock_conn_t, ksnc_list);
- type = conn->ksnc_type;
-
- if (conn->ksnc_ipaddr != route->ksnr_ipaddr)
- continue;
-
- ksocknal_associate_route_conn_locked(route, conn);
- /* keep going (typed routes) */
- }
+ struct list_head *tmp;
+ struct ksock_conn *conn;
+ struct ksock_route *route2;
+
+ LASSERT(!peer_ni->ksnp_closing);
+ LASSERT(route->ksnr_peer == NULL);
+ LASSERT(!route->ksnr_scheduled);
+ LASSERT(!route->ksnr_connecting);
+ LASSERT(route->ksnr_connected == 0);
+
+ /* LASSERT(unique) */
+ list_for_each(tmp, &peer_ni->ksnp_routes) {
+ route2 = list_entry(tmp, struct ksock_route, ksnr_list);
+
+ if (route2->ksnr_ipaddr == route->ksnr_ipaddr) {
+ CERROR("Duplicate route %s %pI4h\n",
+ libcfs_id2str(peer_ni->ksnp_id),
+ &route->ksnr_ipaddr);
+ LBUG();
+ }
+ }
+
+ route->ksnr_peer = peer_ni;
+ ksocknal_peer_addref(peer_ni);
+ /* peer_ni's routelist takes over my ref on 'route' */
+ list_add_tail(&route->ksnr_list, &peer_ni->ksnp_routes);
+
+ list_for_each(tmp, &peer_ni->ksnp_conns) {
+ conn = list_entry(tmp, struct ksock_conn, ksnc_list);
+
+ if (conn->ksnc_ipaddr != route->ksnr_ipaddr)
+ continue;
+
+ ksocknal_associate_route_conn_locked(route, conn);
+ /* keep going (typed routes) */
+ }
}
-void
-ksocknal_del_route_locked (ksock_route_t *route)
+static void
+ksocknal_del_route_locked(struct ksock_route *route)
{
- ksock_peer_t *peer = route->ksnr_peer;
- ksock_interface_t *iface;
- ksock_conn_t *conn;
- struct list_head *ctmp;
- struct list_head *cnxt;
-
- LASSERT (!route->ksnr_deleted);
-
- /* Close associated conns */
- list_for_each_safe (ctmp, cnxt, &peer->ksnp_conns) {
- conn = list_entry(ctmp, ksock_conn_t, ksnc_list);
-
- if (conn->ksnc_route != route)
- continue;
-
- ksocknal_close_conn_locked (conn, 0);
- }
-
- if (route->ksnr_myipaddr != 0) {
- iface = ksocknal_ip2iface(route->ksnr_myipaddr);
- if (iface != NULL)
- iface->ksni_nroutes--;
- }
-
- route->ksnr_deleted = 1;
- list_del (&route->ksnr_list);
- ksocknal_put_route (route); /* drop peer's ref */
-
- if (list_empty (&peer->ksnp_routes) &&
- list_empty (&peer->ksnp_conns)) {
- /* I've just removed the last autoconnect route of a peer
- * with no active connections */
- ksocknal_unlink_peer_locked (peer);
- }
+ struct ksock_peer_ni *peer_ni = route->ksnr_peer;
+ struct ksock_interface *iface;
+ struct ksock_conn *conn;
+ struct list_head *ctmp;
+ struct list_head *cnxt;
+
+ LASSERT(!route->ksnr_deleted);
+
+ /* Close associated conns */
+ list_for_each_safe(ctmp, cnxt, &peer_ni->ksnp_conns) {
+ conn = list_entry(ctmp, struct ksock_conn, ksnc_list);
+
+ if (conn->ksnc_route != route)
+ continue;
+
+ ksocknal_close_conn_locked(conn, 0);
+ }
+
+ if (route->ksnr_myipaddr != 0) {
+ iface = ksocknal_ip2iface(route->ksnr_peer->ksnp_ni,
+ route->ksnr_myipaddr);
+ if (iface != NULL)
+ iface->ksni_nroutes--;
+ }
+
+ route->ksnr_deleted = 1;
+ list_del(&route->ksnr_list);
+ ksocknal_route_decref(route); /* drop peer_ni's ref */
+
+ if (list_empty(&peer_ni->ksnp_routes) &&
+ list_empty(&peer_ni->ksnp_conns)) {
+ /* I've just removed the last route to a peer_ni with no active
+ * connections */
+ ksocknal_unlink_peer_locked(peer_ni);
+ }
}
int
-ksocknal_add_peer (ptl_nid_t nid, __u32 ipaddr, int port)
+ksocknal_add_peer(struct lnet_ni *ni, struct lnet_process_id id, __u32 ipaddr,
+ int port)
{
- unsigned long flags;
- struct list_head *tmp;
- ksock_peer_t *peer;
- ksock_peer_t *peer2;
- ksock_route_t *route;
- ksock_route_t *route2;
-
- if (nid == PTL_NID_ANY)
+ struct list_head *tmp;
+ struct ksock_peer_ni *peer_ni;
+ struct ksock_peer_ni *peer2;
+ struct ksock_route *route;
+ struct ksock_route *route2;
+ int rc;
+
+ if (id.nid == LNET_NID_ANY ||
+ id.pid == LNET_PID_ANY)
return (-EINVAL);
- /* Have a brand new peer ready... */
- peer = ksocknal_create_peer (nid);
- if (peer == NULL)
- return (-ENOMEM);
+ /* Have a brand new peer_ni ready... */
+ rc = ksocknal_create_peer(&peer_ni, ni, id);
+ if (rc != 0)
+ return rc;
route = ksocknal_create_route (ipaddr, port);
if (route == NULL) {
- ksocknal_put_peer (peer);
+ ksocknal_peer_decref(peer_ni);
return (-ENOMEM);
}
- write_lock_irqsave (&ksocknal_data.ksnd_global_lock, flags);
+ write_lock_bh(&ksocknal_data.ksnd_global_lock);
- peer2 = ksocknal_find_peer_locked (nid);
- if (peer2 != NULL) {
- ksocknal_put_peer (peer);
- peer = peer2;
- } else {
- /* peer table takes my ref on peer */
- list_add_tail (&peer->ksnp_list,
- ksocknal_nid2peerlist (nid));
- }
+ /* always called with a ref on ni, so shutdown can't have started */
+ LASSERT(((struct ksock_net *) ni->ni_data)->ksnn_shutdown == 0);
- route2 = NULL;
- list_for_each (tmp, &peer->ksnp_routes) {
- route2 = list_entry(tmp, ksock_route_t, ksnr_list);
+ peer2 = ksocknal_find_peer_locked(ni, id);
+ if (peer2 != NULL) {
+ ksocknal_peer_decref(peer_ni);
+ peer_ni = peer2;
+ } else {
+ /* peer_ni table takes my ref on peer_ni */
+ list_add_tail(&peer_ni->ksnp_list,
+ ksocknal_nid2peerlist(id.nid));
+ }
- if (route2->ksnr_ipaddr == ipaddr)
- break;
+ route2 = NULL;
+ list_for_each(tmp, &peer_ni->ksnp_routes) {
+ route2 = list_entry(tmp, struct ksock_route, ksnr_list);
- route2 = NULL;
- }
- if (route2 == NULL) {
- ksocknal_add_route_locked(peer, route);
- route->ksnr_share_count++;
- } else {
- ksocknal_put_route(route);
- route2->ksnr_share_count++;
- }
+ if (route2->ksnr_ipaddr == ipaddr)
+ break;
+
+ route2 = NULL;
+ }
+ if (route2 == NULL) {
+ ksocknal_add_route_locked(peer_ni, route);
+ route->ksnr_share_count++;
+ } else {
+ ksocknal_route_decref(route);
+ route2->ksnr_share_count++;
+ }
- write_unlock_irqrestore (&ksocknal_data.ksnd_global_lock, flags);
+ write_unlock_bh(&ksocknal_data.ksnd_global_lock);
- return (0);
+ return 0;
}
-void
-ksocknal_del_peer_locked (ksock_peer_t *peer, __u32 ip, int single_share)
+static void
+ksocknal_del_peer_locked(struct ksock_peer_ni *peer_ni, __u32 ip)
{
- ksock_conn_t *conn;
- ksock_route_t *route;
- struct list_head *tmp;
- struct list_head *nxt;
- int nshared;
+ struct ksock_conn *conn;
+ struct ksock_route *route;
+ struct list_head *tmp;
+ struct list_head *nxt;
+ int nshared;
- LASSERT (!peer->ksnp_closing);
+ LASSERT(!peer_ni->ksnp_closing);
- /* Extra ref prevents peer disappearing until I'm done with it */
- atomic_inc(&peer->ksnp_refcount);
+ /* Extra ref prevents peer_ni disappearing until I'm done with it */
+ ksocknal_peer_addref(peer_ni);
- list_for_each_safe (tmp, nxt, &peer->ksnp_routes) {
- route = list_entry(tmp, ksock_route_t, ksnr_list);
+ list_for_each_safe(tmp, nxt, &peer_ni->ksnp_routes) {
+ route = list_entry(tmp, struct ksock_route, ksnr_list);
- if (single_share && route->ksnr_share_count == 0)
- continue;
+ /* no match */
+ if (!(ip == 0 || route->ksnr_ipaddr == ip))
+ continue;
- /* no match */
- if (!(ip == 0 || route->ksnr_ipaddr == ip))
- continue;
+ route->ksnr_share_count = 0;
+ /* This deletes associated conns too */
+ ksocknal_del_route_locked(route);
+ }
- if (!single_share)
- route->ksnr_share_count = 0;
- else if (route->ksnr_share_count > 0)
- route->ksnr_share_count--;
+ nshared = 0;
+ list_for_each_safe(tmp, nxt, &peer_ni->ksnp_routes) {
+ route = list_entry(tmp, struct ksock_route, ksnr_list);
+ nshared += route->ksnr_share_count;
+ }
- if (route->ksnr_share_count == 0) {
- /* This deletes associated conns too */
- ksocknal_del_route_locked (route);
- }
+ if (nshared == 0) {
+ /* remove everything else if there are no explicit entries
+ * left */
- if (single_share)
- break;
- }
+ list_for_each_safe(tmp, nxt, &peer_ni->ksnp_routes) {
+ route = list_entry(tmp, struct ksock_route, ksnr_list);
- nshared = 0;
- list_for_each_safe (tmp, nxt, &peer->ksnp_routes) {
- route = list_entry(tmp, ksock_route_t, ksnr_list);
- nshared += route->ksnr_share_count;
- }
-
- if (nshared == 0) {
- /* remove everything else if there are no explicit entries
- * left */
+ /* we should only be removing auto-entries */
+ LASSERT(route->ksnr_share_count == 0);
+ ksocknal_del_route_locked(route);
+ }
- list_for_each_safe (tmp, nxt, &peer->ksnp_routes) {
- route = list_entry(tmp, ksock_route_t, ksnr_list);
-
- /* we should only be removing auto-entries */
- LASSERT(route->ksnr_share_count == 0);
- ksocknal_del_route_locked (route);
- }
+ list_for_each_safe(tmp, nxt, &peer_ni->ksnp_conns) {
+ conn = list_entry(tmp, struct ksock_conn, ksnc_list);
- list_for_each_safe (tmp, nxt, &peer->ksnp_conns) {
- conn = list_entry(tmp, ksock_conn_t, ksnc_list);
-
- ksocknal_close_conn_locked(conn, 0);
- }
- }
+ ksocknal_close_conn_locked(conn, 0);
+ }
+ }
- ksocknal_put_peer(peer);
- /* NB peer unlinks itself when last conn/route is removed */
+ ksocknal_peer_decref(peer_ni);
+ /* NB peer_ni unlinks itself when last conn/route is removed */
}
-int
-ksocknal_del_peer (ptl_nid_t nid, __u32 ip, int single_share)
+static int
+ksocknal_del_peer(struct lnet_ni *ni, struct lnet_process_id id, __u32 ip)
{
- unsigned long flags;
- struct list_head *ptmp;
- struct list_head *pnxt;
- ksock_peer_t *peer;
- int lo;
- int hi;
- int i;
- int rc = -ENOENT;
-
- write_lock_irqsave (&ksocknal_data.ksnd_global_lock, flags);
-
- if (nid != PTL_NID_ANY)
- lo = hi = ksocknal_nid2peerlist(nid) - ksocknal_data.ksnd_peers;
- else {
- lo = 0;
- hi = ksocknal_data.ksnd_peer_hash_size - 1;
- }
+ struct list_head zombies = LIST_HEAD_INIT(zombies);
+ struct list_head *ptmp;
+ struct list_head *pnxt;
+ struct ksock_peer_ni *peer_ni;
+ int lo;
+ int hi;
+ int i;
+ int rc = -ENOENT;
- for (i = lo; i <= hi; i++) {
- list_for_each_safe (ptmp, pnxt, &ksocknal_data.ksnd_peers[i]) {
- peer = list_entry (ptmp, ksock_peer_t, ksnp_list);
+ write_lock_bh(&ksocknal_data.ksnd_global_lock);
- if (!(nid == PTL_NID_ANY || peer->ksnp_nid == nid))
- continue;
+ if (id.nid != LNET_NID_ANY) {
+ hi = (int)(ksocknal_nid2peerlist(id.nid) -
+ ksocknal_data.ksnd_peers);
+ lo = hi;
+ } else {
+ lo = 0;
+ hi = ksocknal_data.ksnd_peer_hash_size - 1;
+ }
- ksocknal_del_peer_locked (peer, ip, single_share);
- rc = 0; /* matched! */
+ for (i = lo; i <= hi; i++) {
+ list_for_each_safe(ptmp, pnxt,
+ &ksocknal_data.ksnd_peers[i]) {
+ peer_ni = list_entry(ptmp, struct ksock_peer_ni, ksnp_list);
- if (single_share)
- break;
- }
- }
+ if (peer_ni->ksnp_ni != ni)
+ continue;
- write_unlock_irqrestore (&ksocknal_data.ksnd_global_lock, flags);
+ if (!((id.nid == LNET_NID_ANY ||
+ peer_ni->ksnp_id.nid == id.nid) &&
+ (id.pid == LNET_PID_ANY ||
+ peer_ni->ksnp_id.pid == id.pid)))
+ continue;
- return (rc);
-}
+ ksocknal_peer_addref(peer_ni); /* a ref for me... */
-ksock_conn_t *
-ksocknal_get_conn_by_idx (int index)
-{
- ksock_peer_t *peer;
- struct list_head *ptmp;
- ksock_conn_t *conn;
- struct list_head *ctmp;
- int i;
+ ksocknal_del_peer_locked(peer_ni, ip);
- read_lock (&ksocknal_data.ksnd_global_lock);
+ if (peer_ni->ksnp_closing &&
+ !list_empty(&peer_ni->ksnp_tx_queue)) {
+ LASSERT(list_empty(&peer_ni->ksnp_conns));
+ LASSERT(list_empty(&peer_ni->ksnp_routes));
- for (i = 0; i < ksocknal_data.ksnd_peer_hash_size; i++) {
- list_for_each (ptmp, &ksocknal_data.ksnd_peers[i]) {
- peer = list_entry (ptmp, ksock_peer_t, ksnp_list);
+ list_splice_init(&peer_ni->ksnp_tx_queue,
+ &zombies);
+ }
- LASSERT (!peer->ksnp_closing);
+ ksocknal_peer_decref(peer_ni); /* ...till here */
- list_for_each (ctmp, &peer->ksnp_conns) {
- if (index-- > 0)
- continue;
+ rc = 0; /* matched! */
+ }
+ }
- conn = list_entry (ctmp, ksock_conn_t, ksnc_list);
- atomic_inc (&conn->ksnc_refcount);
- read_unlock (&ksocknal_data.ksnd_global_lock);
- return (conn);
- }
- }
- }
+ write_unlock_bh(&ksocknal_data.ksnd_global_lock);
+
+ ksocknal_txlist_done(ni, &zombies, -ENETDOWN);
- read_unlock (&ksocknal_data.ksnd_global_lock);
- return (NULL);
+ return rc;
}
-ksock_sched_t *
-ksocknal_choose_scheduler_locked (unsigned int irq)
+static struct ksock_conn *
+ksocknal_get_conn_by_idx(struct lnet_ni *ni, int index)
{
- ksock_sched_t *sched;
- ksock_irqinfo_t *info;
- int i;
-
- LASSERT (irq < NR_IRQS);
- info = &ksocknal_data.ksnd_irqinfo[irq];
-
- if (irq != 0 && /* hardware NIC */
- info->ksni_valid) { /* already set up */
- return (&ksocknal_data.ksnd_schedulers[info->ksni_sched]);
- }
-
- /* software NIC (irq == 0) || not associated with a scheduler yet.
- * Choose the CPU with the fewest connections... */
- sched = &ksocknal_data.ksnd_schedulers[0];
- for (i = 1; i < ksocknal_data.ksnd_nschedulers; i++)
- if (sched->kss_nconns >
- ksocknal_data.ksnd_schedulers[i].kss_nconns)
- sched = &ksocknal_data.ksnd_schedulers[i];
-
- if (irq != 0) { /* Hardware NIC */
- info->ksni_valid = 1;
- info->ksni_sched = sched - ksocknal_data.ksnd_schedulers;
-
- /* no overflow... */
- LASSERT (info->ksni_sched == sched - ksocknal_data.ksnd_schedulers);
- }
-
- return (sched);
+ struct ksock_peer_ni *peer_ni;
+ struct list_head *ptmp;
+ struct ksock_conn *conn;
+ struct list_head *ctmp;
+ int i;
+
+ read_lock(&ksocknal_data.ksnd_global_lock);
+
+ for (i = 0; i < ksocknal_data.ksnd_peer_hash_size; i++) {
+ list_for_each(ptmp, &ksocknal_data.ksnd_peers[i]) {
+ peer_ni = list_entry(ptmp, struct ksock_peer_ni, ksnp_list);
+
+ LASSERT(!peer_ni->ksnp_closing);
+
+ if (peer_ni->ksnp_ni != ni)
+ continue;
+
+ list_for_each(ctmp, &peer_ni->ksnp_conns) {
+ if (index-- > 0)
+ continue;
+
+ conn = list_entry(ctmp, struct ksock_conn,
+ ksnc_list);
+ ksocknal_conn_addref(conn);
+ read_unlock(&ksocknal_data. \
+ ksnd_global_lock);
+ return conn;
+ }
+ }
+ }
+
+ read_unlock(&ksocknal_data.ksnd_global_lock);
+ return NULL;
}
-int
-ksocknal_local_ipvec (__u32 *ipaddrs)
+static struct ksock_sched *
+ksocknal_choose_scheduler_locked(unsigned int cpt)
{
- int i;
- int nip;
-
- read_lock (&ksocknal_data.ksnd_global_lock);
-
- nip = ksocknal_data.ksnd_ninterfaces;
- for (i = 0; i < nip; i++) {
- LASSERT (i < SOCKNAL_MAX_INTERFACES);
-
- ipaddrs[i] = ksocknal_data.ksnd_interfaces[i].ksni_ipaddr;
- LASSERT (ipaddrs[i] != 0);
- }
+ struct ksock_sched *sched = ksocknal_data.ksnd_schedulers[cpt];
+ int i;
+
+ if (sched->kss_nthreads == 0) {
+ cfs_percpt_for_each(sched, i, ksocknal_data.ksnd_schedulers) {
+ if (sched->kss_nthreads > 0) {
+ CDEBUG(D_NET, "scheduler[%d] has no threads. selected scheduler[%d]\n",
+ cpt, sched->kss_cpt);
+ return sched;
+ }
+ }
+ return NULL;
+ }
+
+ return sched;
+}
- read_unlock (&ksocknal_data.ksnd_global_lock);
- return (nip);
+static int
+ksocknal_local_ipvec(struct lnet_ni *ni, __u32 *ipaddrs)
+{
+ struct ksock_net *net = ni->ni_data;
+ int i;
+ int nip;
+
+ read_lock(&ksocknal_data.ksnd_global_lock);
+
+ nip = net->ksnn_ninterfaces;
+ LASSERT(nip <= LNET_INTERFACES_NUM);
+
+ /*
+ * Only offer interfaces for additional connections if I have
+ * more than one.
+ */
+ if (nip < 2) {
+ read_unlock(&ksocknal_data.ksnd_global_lock);
+ return 0;
+ }
+
+ for (i = 0; i < nip; i++) {
+ ipaddrs[i] = net->ksnn_interfaces[i].ksni_ipaddr;
+ LASSERT(ipaddrs[i] != 0);
+ }
+
+ read_unlock(&ksocknal_data.ksnd_global_lock);
+ return nip;
}
-int
-ksocknal_match_peerip (ksock_interface_t *iface, __u32 *ips, int nips)
+static int
+ksocknal_match_peerip(struct ksock_interface *iface, __u32 *ips, int nips)
{
- int best_netmatch = 0;
- int best_xor = 0;
- int best = -1;
- int this_xor;
- int this_netmatch;
- int i;
+ int best_netmatch = 0;
+ int best_xor = 0;
+ int best = -1;
+ int this_xor;
+ int this_netmatch;
+ int i;
for (i = 0; i < nips; i++) {
if (ips[i] == 0)
return (best);
}
-int
-ksocknal_select_ips(ksock_peer_t *peer, __u32 *peerips, int n_peerips)
+static int
+ksocknal_select_ips(struct ksock_peer_ni *peer_ni, __u32 *peerips, int n_peerips)
{
- rwlock_t *global_lock = &ksocknal_data.ksnd_global_lock;
- unsigned long flags;
- ksock_interface_t *iface;
- ksock_interface_t *best_iface;
- int n_ips;
- int i;
- int j;
- int k;
- __u32 ip;
- __u32 xor;
- int this_netmatch;
- int best_netmatch;
- int best_npeers;
+ rwlock_t *global_lock = &ksocknal_data.ksnd_global_lock;
+ struct ksock_net *net = peer_ni->ksnp_ni->ni_data;
+ struct ksock_interface *iface;
+ struct ksock_interface *best_iface;
+ int n_ips;
+ int i;
+ int j;
+ int k;
+ u32 ip;
+ u32 xor;
+ int this_netmatch;
+ int best_netmatch;
+ int best_npeers;
/* CAVEAT EMPTOR: We do all our interface matching with an
* exclusive hold of global lock at IRQ priority. We're only
/* Also note that I'm not going to return more than n_peerips
* interfaces, even if I have more myself */
- write_lock_irqsave(global_lock, flags);
+ write_lock_bh(global_lock);
- LASSERT (n_peerips <= SOCKNAL_MAX_INTERFACES);
- LASSERT (ksocknal_data.ksnd_ninterfaces <= SOCKNAL_MAX_INTERFACES);
+ LASSERT(n_peerips <= LNET_INTERFACES_NUM);
+ LASSERT(net->ksnn_ninterfaces <= LNET_INTERFACES_NUM);
- n_ips = MIN(n_peerips, ksocknal_data.ksnd_ninterfaces);
+ /* Only match interfaces for additional connections
+ * if I have > 1 interface */
+ n_ips = (net->ksnn_ninterfaces < 2) ? 0 :
+ MIN(n_peerips, net->ksnn_ninterfaces);
- for (i = 0; peer->ksnp_n_passive_ips < n_ips; i++) {
+ for (i = 0; peer_ni->ksnp_n_passive_ips < n_ips; i++) {
/* ^ yes really... */
/* If we have any new interfaces, first tick off all the
- * peer IPs that match old interfaces, then choose new
- * interfaces to match the remaining peer IPS.
+ * peer_ni IPs that match old interfaces, then choose new
+ * interfaces to match the remaining peer_ni IPS.
* We don't forget interfaces we've stopped using; we might
* start using them again... */
- if (i < peer->ksnp_n_passive_ips) {
+ if (i < peer_ni->ksnp_n_passive_ips) {
/* Old interface. */
- ip = peer->ksnp_passive_ips[i];
- best_iface = ksocknal_ip2iface(ip);
+ ip = peer_ni->ksnp_passive_ips[i];
+ best_iface = ksocknal_ip2iface(peer_ni->ksnp_ni, ip);
- /* peer passive ips are kept up to date */
+ /* peer_ni passive ips are kept up to date */
LASSERT(best_iface != NULL);
} else {
/* choose a new interface */
- LASSERT (i == peer->ksnp_n_passive_ips);
+ LASSERT (i == peer_ni->ksnp_n_passive_ips);
best_iface = NULL;
best_netmatch = 0;
best_npeers = 0;
- for (j = 0; j < ksocknal_data.ksnd_ninterfaces; j++) {
- iface = &ksocknal_data.ksnd_interfaces[j];
+ for (j = 0; j < net->ksnn_ninterfaces; j++) {
+ iface = &net->ksnn_interfaces[j];
ip = iface->ksni_ipaddr;
- for (k = 0; k < peer->ksnp_n_passive_ips; k++)
- if (peer->ksnp_passive_ips[k] == ip)
+ for (k = 0; k < peer_ni->ksnp_n_passive_ips; k++)
+ if (peer_ni->ksnp_passive_ips[k] == ip)
break;
- if (k < peer->ksnp_n_passive_ips) /* using it already */
+ if (k < peer_ni->ksnp_n_passive_ips) /* using it already */
continue;
k = ksocknal_match_peerip(iface, peerips, n_peerips);
best_npeers = iface->ksni_npeers;
}
+ LASSERT(best_iface != NULL);
+
best_iface->ksni_npeers++;
ip = best_iface->ksni_ipaddr;
- peer->ksnp_passive_ips[i] = ip;
- peer->ksnp_n_passive_ips = i+1;
+ peer_ni->ksnp_passive_ips[i] = ip;
+ peer_ni->ksnp_n_passive_ips = i+1;
}
- LASSERT (best_iface != NULL);
-
- /* mark the best matching peer IP used */
+ /* mark the best matching peer_ni IP used */
j = ksocknal_match_peerip(best_iface, peerips, n_peerips);
peerips[j] = 0;
}
- /* Overwrite input peer IP addresses */
- memcpy(peerips, peer->ksnp_passive_ips, n_ips * sizeof(*peerips));
+ /* Overwrite input peer_ni IP addresses */
+ memcpy(peerips, peer_ni->ksnp_passive_ips, n_ips * sizeof(*peerips));
- write_unlock_irqrestore(global_lock, flags);
+ write_unlock_bh(global_lock);
return (n_ips);
}
-void
-ksocknal_create_routes(ksock_peer_t *peer, int port,
+static void
+ksocknal_create_routes(struct ksock_peer_ni *peer_ni, int port,
__u32 *peer_ipaddrs, int npeer_ipaddrs)
{
- ksock_route_t *newroute = NULL;
- rwlock_t *global_lock = &ksocknal_data.ksnd_global_lock;
- unsigned long flags;
- struct list_head *rtmp;
- ksock_route_t *route;
- ksock_interface_t *iface;
- ksock_interface_t *best_iface;
- int best_netmatch;
- int this_netmatch;
- int best_nroutes;
- int i;
- int j;
+ struct ksock_route *newroute = NULL;
+ rwlock_t *global_lock = &ksocknal_data.ksnd_global_lock;
+ struct lnet_ni *ni = peer_ni->ksnp_ni;
+ struct ksock_net *net = ni->ni_data;
+ struct list_head *rtmp;
+ struct ksock_route *route;
+ struct ksock_interface *iface;
+ struct ksock_interface *best_iface;
+ int best_netmatch;
+ int this_netmatch;
+ int best_nroutes;
+ int i;
+ int j;
/* CAVEAT EMPTOR: We do all our interface matching with an
* exclusive hold of global lock at IRQ priority. We're only
* expecting to be dealing with small numbers of interfaces, so the
* O(n**3)-ness here shouldn't matter */
- write_lock_irqsave(global_lock, flags);
+ write_lock_bh(global_lock);
+
+ if (net->ksnn_ninterfaces < 2) {
+ /* Only create additional connections
+ * if I have > 1 interface */
+ write_unlock_bh(global_lock);
+ return;
+ }
- LASSERT (npeer_ipaddrs <= SOCKNAL_MAX_INTERFACES);
+ LASSERT(npeer_ipaddrs <= LNET_INTERFACES_NUM);
for (i = 0; i < npeer_ipaddrs; i++) {
if (newroute != NULL) {
newroute->ksnr_ipaddr = peer_ipaddrs[i];
} else {
- write_unlock_irqrestore(global_lock, flags);
+ write_unlock_bh(global_lock);
newroute = ksocknal_create_route(peer_ipaddrs[i], port);
if (newroute == NULL)
return;
- write_lock_irqsave(global_lock, flags);
+ write_lock_bh(global_lock);
}
- /* Already got a route? */
- route = NULL;
- list_for_each(rtmp, &peer->ksnp_routes) {
- route = list_entry(rtmp, ksock_route_t, ksnr_list);
+ if (peer_ni->ksnp_closing) {
+ /* peer_ni got closed under me */
+ break;
+ }
- if (route->ksnr_ipaddr == newroute->ksnr_ipaddr)
- break;
+ /* Already got a route? */
+ route = NULL;
+ list_for_each(rtmp, &peer_ni->ksnp_routes) {
+ route = list_entry(rtmp, struct ksock_route, ksnr_list);
- route = NULL;
- }
- if (route != NULL)
- continue;
+ if (route->ksnr_ipaddr == newroute->ksnr_ipaddr)
+ break;
- best_iface = NULL;
- best_nroutes = 0;
- best_netmatch = 0;
+ route = NULL;
+ }
+ if (route != NULL)
+ continue;
- LASSERT (ksocknal_data.ksnd_ninterfaces <= SOCKNAL_MAX_INTERFACES);
+ best_iface = NULL;
+ best_nroutes = 0;
+ best_netmatch = 0;
- /* Select interface to connect from */
- for (j = 0; j < ksocknal_data.ksnd_ninterfaces; j++) {
- iface = &ksocknal_data.ksnd_interfaces[j];
+ LASSERT(net->ksnn_ninterfaces <= LNET_INTERFACES_NUM);
- /* Using this interface already? */
- list_for_each(rtmp, &peer->ksnp_routes) {
- route = list_entry(rtmp, ksock_route_t, ksnr_list);
+ /* Select interface to connect from */
+ for (j = 0; j < net->ksnn_ninterfaces; j++) {
+ iface = &net->ksnn_interfaces[j];
- if (route->ksnr_myipaddr == iface->ksni_ipaddr)
- break;
+ /* Using this interface already? */
+ list_for_each(rtmp, &peer_ni->ksnp_routes) {
+ route = list_entry(rtmp, struct ksock_route,
+ ksnr_list);
- route = NULL;
- }
- if (route != NULL)
- continue;
+ if (route->ksnr_myipaddr == iface->ksni_ipaddr)
+ break;
+
+ route = NULL;
+ }
+ if (route != NULL)
+ continue;
this_netmatch = (((iface->ksni_ipaddr ^
newroute->ksnr_ipaddr) &
newroute->ksnr_myipaddr = best_iface->ksni_ipaddr;
best_iface->ksni_nroutes++;
- ksocknal_add_route_locked(peer, newroute);
+ ksocknal_add_route_locked(peer_ni, newroute);
newroute = NULL;
}
- write_unlock_irqrestore(global_lock, flags);
+ write_unlock_bh(global_lock);
if (newroute != NULL)
- ksocknal_put_route(newroute);
+ ksocknal_route_decref(newroute);
}
int
-ksocknal_create_conn (ksock_route_t *route, struct socket *sock, int type)
+ksocknal_accept(struct lnet_ni *ni, struct socket *sock)
{
- int passive = (type == SOCKNAL_CONN_NONE);
- rwlock_t *global_lock = &ksocknal_data.ksnd_global_lock;
- __u32 ipaddrs[SOCKNAL_MAX_INTERFACES];
- int nipaddrs;
- ptl_nid_t nid;
- struct list_head *tmp;
- __u64 incarnation;
- unsigned long flags;
- ksock_conn_t *conn;
- ksock_conn_t *conn2;
- ksock_peer_t *peer = NULL;
- ksock_peer_t *peer2;
- ksock_sched_t *sched;
- unsigned int irq;
- ksock_tx_t *tx;
- int rc;
-
- /* NB, sock has an associated file since (a) this connection might
- * have been created in userland and (b) we need to refcount the
- * socket so that we don't close it while I/O is being done on
- * it, and sock->file has that pre-cooked... */
- LASSERT (KSN_SOCK2FILE(sock) != NULL);
- LASSERT (cfs_file_count(KSN_SOCK2FILE(sock)) > 0);
- LASSERT (route == NULL || !passive);
-
- rc = ksocknal_lib_setup_sock (sock);
- if (rc != 0)
- return (rc);
+ struct ksock_connreq *cr;
+ int rc;
+ u32 peer_ip;
+ int peer_port;
- irq = ksocknal_lib_sock_irq (sock);
+ rc = lnet_sock_getaddr(sock, true, &peer_ip, &peer_port);
+ LASSERT(rc == 0); /* we succeeded before */
- PORTAL_ALLOC(conn, sizeof(*conn));
- if (conn == NULL)
- return (-ENOMEM);
+ LIBCFS_ALLOC(cr, sizeof(*cr));
+ if (cr == NULL) {
+ LCONSOLE_ERROR_MSG(0x12f, "Dropping connection request from "
+ "%pI4h: memory exhausted\n", &peer_ip);
+ return -ENOMEM;
+ }
+
+ lnet_ni_addref(ni);
+ cr->ksncr_ni = ni;
+ cr->ksncr_sock = sock;
+
+ spin_lock_bh(&ksocknal_data.ksnd_connd_lock);
+
+ list_add_tail(&cr->ksncr_list, &ksocknal_data.ksnd_connd_connreqs);
+ wake_up(&ksocknal_data.ksnd_connd_waitq);
+
+ spin_unlock_bh(&ksocknal_data.ksnd_connd_lock);
+ return 0;
+}
+
+static int
+ksocknal_connecting(struct ksock_peer_ni *peer_ni, __u32 ipaddr)
+{
+ struct ksock_route *route;
+
+ list_for_each_entry(route, &peer_ni->ksnp_routes, ksnr_list) {
+ if (route->ksnr_ipaddr == ipaddr)
+ return route->ksnr_connecting;
+ }
+ return 0;
+}
+
+int
+ksocknal_create_conn(struct lnet_ni *ni, struct ksock_route *route,
+ struct socket *sock, int type)
+{
+ rwlock_t *global_lock = &ksocknal_data.ksnd_global_lock;
+ struct list_head zombies = LIST_HEAD_INIT(zombies);
+ struct lnet_process_id peerid;
+ struct list_head *tmp;
+ u64 incarnation;
+ struct ksock_conn *conn;
+ struct ksock_conn *conn2;
+ struct ksock_peer_ni *peer_ni = NULL;
+ struct ksock_peer_ni *peer2;
+ struct ksock_sched *sched;
+ struct ksock_hello_msg *hello;
+ int cpt;
+ struct ksock_tx *tx;
+ struct ksock_tx *txtmp;
+ int rc;
+ int rc2;
+ int active;
+ char *warn = NULL;
+
+ active = (route != NULL);
+
+ LASSERT (active == (type != SOCKLND_CONN_NONE));
+
+ LIBCFS_ALLOC(conn, sizeof(*conn));
+ if (conn == NULL) {
+ rc = -ENOMEM;
+ goto failed_0;
+ }
- memset (conn, 0, sizeof (*conn));
conn->ksnc_peer = NULL;
conn->ksnc_route = NULL;
conn->ksnc_sock = sock;
- conn->ksnc_type = type;
- ksocknal_lib_save_callback(sock, conn);
- atomic_set (&conn->ksnc_refcount, 1); /* 1 ref for me */
-
- conn->ksnc_rx_ready = 0;
- conn->ksnc_rx_scheduled = 0;
- ksocknal_new_packet (conn, 0);
-
- CFS_INIT_LIST_HEAD (&conn->ksnc_tx_queue);
- conn->ksnc_tx_ready = 0;
- conn->ksnc_tx_scheduled = 0;
- atomic_set (&conn->ksnc_tx_nob, 0);
+ /* 2 ref, 1 for conn, another extra ref prevents socket
+ * being closed before establishment of connection */
+ atomic_set (&conn->ksnc_sock_refcount, 2);
+ conn->ksnc_type = type;
+ ksocknal_lib_save_callback(sock, conn);
+ atomic_set (&conn->ksnc_conn_refcount, 1); /* 1 ref for me */
+
+ conn->ksnc_rx_ready = 0;
+ conn->ksnc_rx_scheduled = 0;
+
+ INIT_LIST_HEAD(&conn->ksnc_tx_queue);
+ conn->ksnc_tx_ready = 0;
+ conn->ksnc_tx_scheduled = 0;
+ conn->ksnc_tx_carrier = NULL;
+ atomic_set (&conn->ksnc_tx_nob, 0);
+
+ LIBCFS_ALLOC(hello, offsetof(struct ksock_hello_msg,
+ kshm_ips[LNET_INTERFACES_NUM]));
+ if (hello == NULL) {
+ rc = -ENOMEM;
+ goto failed_1;
+ }
/* stash conn's local and remote addrs */
rc = ksocknal_lib_get_conn_addrs (conn);
if (rc != 0)
- goto failed_0;
+ goto failed_1;
+
+ /* Find out/confirm peer_ni's NID and connection type and get the
+ * vector of interfaces she's willing to let me connect to.
+ * Passive connections use the listener timeout since the peer_ni sends
+ * eagerly */
+
+ if (active) {
+ peer_ni = route->ksnr_peer;
+ LASSERT(ni == peer_ni->ksnp_ni);
- if (!passive) {
/* Active connection sends HELLO eagerly */
- rc = ksocknal_local_ipvec(ipaddrs);
- if (rc < 0)
- goto failed_0;
- nipaddrs = rc;
+ hello->kshm_nips = ksocknal_local_ipvec(ni, hello->kshm_ips);
+ peerid = peer_ni->ksnp_id;
+
+ write_lock_bh(global_lock);
+ conn->ksnc_proto = peer_ni->ksnp_proto;
+ write_unlock_bh(global_lock);
+
+ if (conn->ksnc_proto == NULL) {
+ conn->ksnc_proto = &ksocknal_protocol_v3x;
+#if SOCKNAL_VERSION_DEBUG
+ if (*ksocknal_tunables.ksnd_protocol == 2)
+ conn->ksnc_proto = &ksocknal_protocol_v2x;
+ else if (*ksocknal_tunables.ksnd_protocol == 1)
+ conn->ksnc_proto = &ksocknal_protocol_v1x;
+#endif
+ }
- rc = ksocknal_send_hello (conn, ipaddrs, nipaddrs);
+ rc = ksocknal_send_hello (ni, conn, peerid.nid, hello);
if (rc != 0)
- goto failed_0;
+ goto failed_1;
+ } else {
+ peerid.nid = LNET_NID_ANY;
+ peerid.pid = LNET_PID_ANY;
+
+ /* Passive, get protocol from peer_ni */
+ conn->ksnc_proto = NULL;
}
- /* Find out/confirm peer's NID and connection type and get the
- * vector of interfaces she's willing to let me connect to */
- nid = (route == NULL) ? PTL_NID_ANY : route->ksnr_peer->ksnp_nid;
- rc = ksocknal_recv_hello (conn, &nid, &incarnation, ipaddrs);
+ rc = ksocknal_recv_hello (ni, conn, hello, &peerid, &incarnation);
if (rc < 0)
- goto failed_0;
- nipaddrs = rc;
- LASSERT (nid != PTL_NID_ANY);
+ goto failed_1;
+
+ LASSERT (rc == 0 || active);
+ LASSERT (conn->ksnc_proto != NULL);
+ LASSERT (peerid.nid != LNET_NID_ANY);
- if (route != NULL) {
- peer = route->ksnr_peer;
- atomic_inc(&peer->ksnp_refcount);
+ cpt = lnet_cpt_of_nid(peerid.nid, ni);
+
+ if (active) {
+ ksocknal_peer_addref(peer_ni);
+ write_lock_bh(global_lock);
} else {
- peer = ksocknal_create_peer(nid);
- if (peer == NULL) {
- rc = -ENOMEM;
- goto failed_0;
- }
+ rc = ksocknal_create_peer(&peer_ni, ni, peerid);
+ if (rc != 0)
+ goto failed_1;
- write_lock_irqsave(global_lock, flags);
-
- peer2 = ksocknal_find_peer_locked(nid);
- if (peer2 == NULL) {
- /* NB this puts an "empty" peer in the peer
- * table (which takes my ref) */
- list_add_tail(&peer->ksnp_list,
- ksocknal_nid2peerlist(nid));
- } else {
- ksocknal_put_peer(peer);
- peer = peer2;
- }
- /* +1 ref for me */
- atomic_inc(&peer->ksnp_refcount);
+ write_lock_bh(global_lock);
+
+ /* called with a ref on ni, so shutdown can't have started */
+ LASSERT(((struct ksock_net *) ni->ni_data)->ksnn_shutdown == 0);
- write_unlock_irqrestore(global_lock, flags);
+ peer2 = ksocknal_find_peer_locked(ni, peerid);
+ if (peer2 == NULL) {
+ /* NB this puts an "empty" peer_ni in the peer_ni
+ * table (which takes my ref) */
+ list_add_tail(&peer_ni->ksnp_list,
+ ksocknal_nid2peerlist(peerid.nid));
+ } else {
+ ksocknal_peer_decref(peer_ni);
+ peer_ni = peer2;
+ }
+
+ /* +1 ref for me */
+ ksocknal_peer_addref(peer_ni);
+ peer_ni->ksnp_accepting++;
+
+ /* Am I already connecting to this guy? Resolve in
+ * favour of higher NID... */
+ if (peerid.nid < ni->ni_nid &&
+ ksocknal_connecting(peer_ni, conn->ksnc_ipaddr)) {
+ rc = EALREADY;
+ warn = "connection race resolution";
+ goto failed_2;
+ }
}
- if (!passive) {
- ksocknal_create_routes(peer, conn->ksnc_port,
- ipaddrs, nipaddrs);
- rc = 0;
- } else {
- rc = ksocknal_select_ips(peer, ipaddrs, nipaddrs);
- LASSERT (rc >= 0);
- rc = ksocknal_send_hello (conn, ipaddrs, rc);
+ if (peer_ni->ksnp_closing ||
+ (active && route->ksnr_deleted)) {
+ /* peer_ni/route got closed under me */
+ rc = -ESTALE;
+ warn = "peer_ni/route removed";
+ goto failed_2;
}
- if (rc < 0)
- goto failed_1;
- write_lock_irqsave (global_lock, flags);
+ if (peer_ni->ksnp_proto == NULL) {
+ /* Never connected before.
+ * NB recv_hello may have returned EPROTO to signal my peer_ni
+ * wants a different protocol than the one I asked for.
+ */
+ LASSERT(list_empty(&peer_ni->ksnp_conns));
- if (peer->ksnp_closing ||
- (route != NULL && route->ksnr_deleted)) {
- /* route/peer got closed under me */
- rc = -ESTALE;
+ peer_ni->ksnp_proto = conn->ksnc_proto;
+ peer_ni->ksnp_incarnation = incarnation;
+ }
+
+ if (peer_ni->ksnp_proto != conn->ksnc_proto ||
+ peer_ni->ksnp_incarnation != incarnation) {
+ /* peer_ni rebooted or I've got the wrong protocol version */
+ ksocknal_close_peer_conns_locked(peer_ni, 0, 0);
+
+ peer_ni->ksnp_proto = NULL;
+ rc = ESTALE;
+ warn = peer_ni->ksnp_incarnation != incarnation ?
+ "peer_ni rebooted" :
+ "wrong proto version";
+ goto failed_2;
+ }
+
+ switch (rc) {
+ default:
+ LBUG();
+ case 0:
+ break;
+ case EALREADY:
+ warn = "lost conn race";
+ goto failed_2;
+ case EPROTO:
+ warn = "retry with different protocol version";
goto failed_2;
}
- /* Refuse to duplicate an existing connection (both sides might
- * autoconnect at once), unless this is a loopback connection */
- if (conn->ksnc_ipaddr != conn->ksnc_myipaddr) {
- list_for_each(tmp, &peer->ksnp_conns) {
- conn2 = list_entry(tmp, ksock_conn_t, ksnc_list);
+ /* Refuse to duplicate an existing connection, unless this is a
+ * loopback connection */
+ if (conn->ksnc_ipaddr != conn->ksnc_myipaddr) {
+ list_for_each(tmp, &peer_ni->ksnp_conns) {
+ conn2 = list_entry(tmp, struct ksock_conn, ksnc_list);
if (conn2->ksnc_ipaddr != conn->ksnc_ipaddr ||
conn2->ksnc_myipaddr != conn->ksnc_myipaddr ||
- conn2->ksnc_type != conn->ksnc_type ||
- conn2->ksnc_incarnation != incarnation)
+ conn2->ksnc_type != conn->ksnc_type)
continue;
- CWARN("Not creating duplicate connection to "
- "%u.%u.%u.%u type %d\n",
- HIPQUAD(conn->ksnc_ipaddr), conn->ksnc_type);
- rc = -EALREADY;
+ /* Reply on a passive connection attempt so the peer_ni
+ * realises we're connected. */
+ LASSERT (rc == 0);
+ if (!active)
+ rc = EALREADY;
+
+ warn = "duplicate";
goto failed_2;
}
}
/* If the connection created by this route didn't bind to the IP
* address the route connected to, the connection/route matching
* code below probably isn't going to work. */
- if (route != NULL &&
+ if (active &&
route->ksnr_ipaddr != conn->ksnc_ipaddr) {
- CERROR("Route "LPX64" %u.%u.%u.%u connected to %u.%u.%u.%u\n",
- peer->ksnp_nid,
- HIPQUAD(route->ksnr_ipaddr),
- HIPQUAD(conn->ksnc_ipaddr));
- }
-
- /* Search for a route corresponding to the new connection and
- * create an association. This allows incoming connections created
- * by routes in my peer to match my own route entries so I don't
- * continually create duplicate routes. */
- list_for_each (tmp, &peer->ksnp_routes) {
- route = list_entry(tmp, ksock_route_t, ksnr_list);
-
- if (route->ksnr_ipaddr != conn->ksnc_ipaddr)
- continue;
+ CERROR("Route %s %pI4h connected to %pI4h\n",
+ libcfs_id2str(peer_ni->ksnp_id),
+ &route->ksnr_ipaddr,
+ &conn->ksnc_ipaddr);
+ }
+
+ /* Search for a route corresponding to the new connection and
+ * create an association. This allows incoming connections created
+ * by routes in my peer_ni to match my own route entries so I don't
+ * continually create duplicate routes. */
+ list_for_each(tmp, &peer_ni->ksnp_routes) {
+ route = list_entry(tmp, struct ksock_route, ksnr_list);
+
+ if (route->ksnr_ipaddr != conn->ksnc_ipaddr)
+ continue;
+
+ ksocknal_associate_route_conn_locked(route, conn);
+ break;
+ }
+
+ conn->ksnc_peer = peer_ni; /* conn takes my ref on peer_ni */
+ peer_ni->ksnp_last_alive = ktime_get_seconds();
+ peer_ni->ksnp_send_keepalive = 0;
+ peer_ni->ksnp_error = 0;
+
+ sched = ksocknal_choose_scheduler_locked(cpt);
+ if (!sched) {
+ CERROR("no schedulers available. node is unhealthy\n");
+ goto failed_2;
+ }
+ /*
+ * The cpt might have changed if we ended up selecting a non cpt
+ * native scheduler. So use the scheduler's cpt instead.
+ */
+ cpt = sched->kss_cpt;
+ sched->kss_nconns++;
+ conn->ksnc_scheduler = sched;
- ksocknal_associate_route_conn_locked(route, conn);
- break;
+ conn->ksnc_tx_last_post = ktime_get_seconds();
+ /* Set the deadline for the outgoing HELLO to drain */
+ conn->ksnc_tx_bufnob = sock->sk->sk_wmem_queued;
+ conn->ksnc_tx_deadline = ktime_get_seconds() +
+ lnet_get_lnd_timeout();
+ smp_mb(); /* order with adding to peer_ni's conn list */
+
+ list_add(&conn->ksnc_list, &peer_ni->ksnp_conns);
+ ksocknal_conn_addref(conn);
+
+ ksocknal_new_packet(conn, 0);
+
+ conn->ksnc_zc_capable = ksocknal_lib_zc_capable(conn);
+
+ /* Take packets blocking for this connection. */
+ list_for_each_entry_safe(tx, txtmp, &peer_ni->ksnp_tx_queue, tx_list) {
+ if (conn->ksnc_proto->pro_match_tx(conn, tx, tx->tx_nonblk) ==
+ SOCKNAL_MATCH_NO)
+ continue;
+
+ list_del(&tx->tx_list);
+ ksocknal_queue_tx_locked(tx, conn);
+ }
+
+ write_unlock_bh(global_lock);
+
+ /* We've now got a new connection. Any errors from here on are just
+ * like "normal" comms errors and we close the connection normally.
+ * NB (a) we still have to send the reply HELLO for passive
+ * connections,
+ * (b) normal I/O on the conn is blocked until I setup and call the
+ * socket callbacks.
+ */
+
+ CDEBUG(D_NET, "New conn %s p %d.x %pI4h -> %pI4h/%d"
+ " incarnation:%lld sched[%d]\n",
+ libcfs_id2str(peerid), conn->ksnc_proto->pro_version,
+ &conn->ksnc_myipaddr, &conn->ksnc_ipaddr,
+ conn->ksnc_port, incarnation, cpt);
+
+ if (active) {
+ /* additional routes after interface exchange? */
+ ksocknal_create_routes(peer_ni, conn->ksnc_port,
+ hello->kshm_ips, hello->kshm_nips);
+ } else {
+ hello->kshm_nips = ksocknal_select_ips(peer_ni, hello->kshm_ips,
+ hello->kshm_nips);
+ rc = ksocknal_send_hello(ni, conn, peerid.nid, hello);
}
- /* Give conn a ref on sock->file since we're going to return success */
- cfs_get_file(KSN_SOCK2FILE(sock));
-
- conn->ksnc_peer = peer; /* conn takes my ref on peer */
- conn->ksnc_incarnation = incarnation;
- peer->ksnp_last_alive = cfs_time_current();
- peer->ksnp_error = 0;
+ LIBCFS_FREE(hello, offsetof(struct ksock_hello_msg,
+ kshm_ips[LNET_INTERFACES_NUM]));
- sched = ksocknal_choose_scheduler_locked (irq);
- sched->kss_nconns++;
- conn->ksnc_scheduler = sched;
+ /* setup the socket AFTER I've received hello (it disables
+ * SO_LINGER). I might call back to the acceptor who may want
+ * to send a protocol version response and then close the
+ * socket; this ensures the socket only tears down after the
+ * response has been sent. */
+ if (rc == 0)
+ rc = ksocknal_lib_setup_sock(sock);
- /* Set the deadline for the outgoing HELLO to drain */
- conn->ksnc_tx_bufnob = SOCK_WMEM_QUEUED(sock);
- conn->ksnc_tx_deadline = cfs_time_shift(ksocknal_tunables.ksnd_io_timeout);
- mb(); /* order with adding to peer's conn list */
-
- list_add (&conn->ksnc_list, &peer->ksnp_conns);
- atomic_inc (&conn->ksnc_refcount);
+ write_lock_bh(global_lock);
/* NB my callbacks block while I hold ksnd_global_lock */
ksocknal_lib_set_callback(sock, conn);
- /* Take all the packets blocking for a connection.
- * NB, it might be nicer to share these blocked packets among any
- * other connections that are becoming established. */
- while (!list_empty (&peer->ksnp_tx_queue)) {
- tx = list_entry (peer->ksnp_tx_queue.next,
- ksock_tx_t, tx_list);
+ if (!active)
+ peer_ni->ksnp_accepting--;
- list_del (&tx->tx_list);
- ksocknal_queue_tx_locked (tx, conn);
- }
+ write_unlock_bh(global_lock);
- rc = ksocknal_close_stale_conns_locked(peer, incarnation);
- if (rc != 0)
- CDEBUG(D_HA,
- "Closed %d stale conns to nid "LPX64" ip %d.%d.%d.%d\n",
- rc, conn->ksnc_peer->ksnp_nid,
- HIPQUAD(conn->ksnc_ipaddr));
+ if (rc != 0) {
+ write_lock_bh(global_lock);
+ if (!conn->ksnc_closing) {
+ /* could be closed by another thread */
+ ksocknal_close_conn_locked(conn, rc);
+ }
+ write_unlock_bh(global_lock);
+ } else if (ksocknal_connsock_addref(conn) == 0) {
+ /* Allow I/O to proceed. */
+ ksocknal_read_callback(conn);
+ ksocknal_write_callback(conn);
+ ksocknal_connsock_decref(conn);
+ }
- write_unlock_irqrestore (global_lock, flags);
+ ksocknal_connsock_decref(conn);
+ ksocknal_conn_decref(conn);
+ return rc;
- ksocknal_lib_bind_irq (irq);
+failed_2:
+ if (!peer_ni->ksnp_closing &&
+ list_empty(&peer_ni->ksnp_conns) &&
+ list_empty(&peer_ni->ksnp_routes)) {
+ list_add(&zombies, &peer_ni->ksnp_tx_queue);
+ list_del_init(&peer_ni->ksnp_tx_queue);
+ ksocknal_unlink_peer_locked(peer_ni);
+ }
- /* Call the callbacks right now to get things going. */
- if (ksocknal_getconnsock(conn) == 0) {
- ksocknal_lib_act_callback(sock, conn);
- ksocknal_putconnsock(conn);
- }
+ write_unlock_bh(global_lock);
- CDEBUG(D_HA, "New conn nid:"LPX64" %u.%u.%u.%u -> %u.%u.%u.%u/%d "
- "incarnation:"LPX64" sched[%d]/%d\n",
- nid, HIPQUAD(conn->ksnc_myipaddr),
- HIPQUAD(conn->ksnc_ipaddr), conn->ksnc_port, incarnation,
- (int)(conn->ksnc_scheduler - ksocknal_data.ksnd_schedulers),irq);
+ if (warn != NULL) {
+ if (rc < 0)
+ CERROR("Not creating conn %s type %d: %s\n",
+ libcfs_id2str(peerid), conn->ksnc_type, warn);
+ else
+ CDEBUG(D_NET, "Not creating conn %s type %d: %s\n",
+ libcfs_id2str(peerid), conn->ksnc_type, warn);
+ }
+
+ if (!active) {
+ if (rc > 0) {
+ /* Request retry by replying with CONN_NONE
+ * ksnc_proto has been set already */
+ conn->ksnc_type = SOCKLND_CONN_NONE;
+ hello->kshm_nips = 0;
+ ksocknal_send_hello(ni, conn, peerid.nid, hello);
+ }
- ksocknal_put_conn (conn);
- return (0);
+ write_lock_bh(global_lock);
+ peer_ni->ksnp_accepting--;
+ write_unlock_bh(global_lock);
+ }
- failed_2:
- if (!peer->ksnp_closing &&
- list_empty (&peer->ksnp_conns) &&
- list_empty (&peer->ksnp_routes))
- ksocknal_unlink_peer_locked(peer);
- write_unlock_irqrestore(global_lock, flags);
+ /*
+ * If we get here without an error code, just use -EALREADY.
+ * Depending on how we got here, the error may be positive
+ * or negative. Normalize the value for ksocknal_txlist_done().
+ */
+ rc2 = (rc == 0 ? -EALREADY : (rc > 0 ? -rc : rc));
+ ksocknal_txlist_done(ni, &zombies, rc2);
+ ksocknal_peer_decref(peer_ni);
- failed_1:
- ksocknal_put_peer (peer);
+failed_1:
+ if (hello != NULL)
+ LIBCFS_FREE(hello, offsetof(struct ksock_hello_msg,
+ kshm_ips[LNET_INTERFACES_NUM]));
- failed_0:
- PORTAL_FREE (conn, sizeof(*conn));
+ LIBCFS_FREE(conn, sizeof(*conn));
- LASSERT (rc != 0);
- return (rc);
+failed_0:
+ sock_release(sock);
+ return rc;
}
void
-ksocknal_close_conn_locked (ksock_conn_t *conn, int error)
+ksocknal_close_conn_locked(struct ksock_conn *conn, int error)
{
/* This just does the immmediate housekeeping, and queues the
* connection for the reaper to terminate.
* Caller holds ksnd_global_lock exclusively in irq context */
- ksock_peer_t *peer = conn->ksnc_peer;
- ksock_route_t *route;
- ksock_conn_t *conn2;
- struct list_head *tmp;
-
- LASSERT (peer->ksnp_error == 0);
- LASSERT (!conn->ksnc_closing);
- conn->ksnc_closing = 1;
- atomic_inc (&ksocknal_data.ksnd_nclosing_conns);
-
- /* ksnd_deathrow_conns takes over peer's ref */
- list_del (&conn->ksnc_list);
-
- route = conn->ksnc_route;
- if (route != NULL) {
- /* dissociate conn from route... */
- LASSERT (!route->ksnr_deleted);
- LASSERT ((route->ksnr_connected & (1 << conn->ksnc_type)) != 0);
-
- conn2 = NULL;
- list_for_each(tmp, &peer->ksnp_conns) {
- conn2 = list_entry(tmp, ksock_conn_t, ksnc_list);
-
- if (conn2->ksnc_route == route &&
- conn2->ksnc_type == conn->ksnc_type)
- break;
+ struct ksock_peer_ni *peer_ni = conn->ksnc_peer;
+ struct ksock_route *route;
+ struct ksock_conn *conn2;
+ struct list_head *tmp;
- conn2 = NULL;
- }
- if (conn2 == NULL)
- route->ksnr_connected &= ~(1 << conn->ksnc_type);
+ LASSERT(peer_ni->ksnp_error == 0);
+ LASSERT(!conn->ksnc_closing);
+ conn->ksnc_closing = 1;
- conn->ksnc_route = NULL;
+ /* ksnd_deathrow_conns takes over peer_ni's ref */
+ list_del(&conn->ksnc_list);
-#if 0 /* irrelevent with only eager routes */
- list_del (&route->ksnr_list); /* make route least favourite */
- list_add_tail (&route->ksnr_list, &peer->ksnp_routes);
-#endif
- ksocknal_put_route (route); /* drop conn's ref on route */
- }
+ route = conn->ksnc_route;
+ if (route != NULL) {
+ /* dissociate conn from route... */
+ LASSERT(!route->ksnr_deleted);
+ LASSERT((route->ksnr_connected & (1 << conn->ksnc_type)) != 0);
- if (list_empty (&peer->ksnp_conns)) {
- /* No more connections to this peer */
+ conn2 = NULL;
+ list_for_each(tmp, &peer_ni->ksnp_conns) {
+ conn2 = list_entry(tmp, struct ksock_conn, ksnc_list);
- peer->ksnp_error = error; /* stash last conn close reason */
+ if (conn2->ksnc_route == route &&
+ conn2->ksnc_type == conn->ksnc_type)
+ break;
- if (list_empty (&peer->ksnp_routes)) {
- /* I've just closed last conn belonging to a
- * non-autoconnecting peer */
- ksocknal_unlink_peer_locked (peer);
- }
- }
+ conn2 = NULL;
+ }
+ if (conn2 == NULL)
+ route->ksnr_connected &= ~(1 << conn->ksnc_type);
- spin_lock (&ksocknal_data.ksnd_reaper_lock);
+ conn->ksnc_route = NULL;
- list_add_tail (&conn->ksnc_list, &ksocknal_data.ksnd_deathrow_conns);
- cfs_waitq_signal (&ksocknal_data.ksnd_reaper_waitq);
+ ksocknal_route_decref(route); /* drop conn's ref on route */
+ }
- spin_unlock (&ksocknal_data.ksnd_reaper_lock);
+ if (list_empty(&peer_ni->ksnp_conns)) {
+ /* No more connections to this peer_ni */
+
+ if (!list_empty(&peer_ni->ksnp_tx_queue)) {
+ struct ksock_tx *tx;
+
+ LASSERT(conn->ksnc_proto == &ksocknal_protocol_v3x);
+
+ /* throw them to the last connection...,
+ * these TXs will be send to /dev/null by scheduler */
+ list_for_each_entry(tx, &peer_ni->ksnp_tx_queue,
+ tx_list)
+ ksocknal_tx_prep(conn, tx);
+
+ spin_lock_bh(&conn->ksnc_scheduler->kss_lock);
+ list_splice_init(&peer_ni->ksnp_tx_queue,
+ &conn->ksnc_tx_queue);
+ spin_unlock_bh(&conn->ksnc_scheduler->kss_lock);
+ }
+
+ /* renegotiate protocol version */
+ peer_ni->ksnp_proto = NULL;
+ /* stash last conn close reason */
+ peer_ni->ksnp_error = error;
+
+ if (list_empty(&peer_ni->ksnp_routes)) {
+ /* I've just closed last conn belonging to a
+ * peer_ni with no routes to it */
+ ksocknal_unlink_peer_locked(peer_ni);
+ }
+ }
+
+ spin_lock_bh(&ksocknal_data.ksnd_reaper_lock);
+
+ list_add_tail(&conn->ksnc_list,
+ &ksocknal_data.ksnd_deathrow_conns);
+ wake_up(&ksocknal_data.ksnd_reaper_waitq);
+
+ spin_unlock_bh(&ksocknal_data.ksnd_reaper_lock);
}
void
-ksocknal_terminate_conn (ksock_conn_t *conn)
+ksocknal_peer_failed(struct ksock_peer_ni *peer_ni)
+{
+ int notify = 0;
+ time64_t last_alive = 0;
+
+ /* There has been a connection failure or comms error; but I'll only
+ * tell LNET I think the peer_ni is dead if it's to another kernel and
+ * there are no connections or connection attempts in existence. */
+
+ read_lock(&ksocknal_data.ksnd_global_lock);
+
+ if ((peer_ni->ksnp_id.pid & LNET_PID_USERFLAG) == 0 &&
+ list_empty(&peer_ni->ksnp_conns) &&
+ peer_ni->ksnp_accepting == 0 &&
+ ksocknal_find_connecting_route_locked(peer_ni) == NULL) {
+ notify = 1;
+ last_alive = peer_ni->ksnp_last_alive;
+ }
+
+ read_unlock(&ksocknal_data.ksnd_global_lock);
+
+ if (notify)
+ lnet_notify(peer_ni->ksnp_ni, peer_ni->ksnp_id.nid, 0,
+ last_alive);
+}
+
+void
+ksocknal_finalize_zcreq(struct ksock_conn *conn)
+{
+ struct ksock_peer_ni *peer_ni = conn->ksnc_peer;
+ struct ksock_tx *tx;
+ struct ksock_tx *tmp;
+ struct list_head zlist = LIST_HEAD_INIT(zlist);
+
+ /* NB safe to finalize TXs because closing of socket will
+ * abort all buffered data */
+ LASSERT(conn->ksnc_sock == NULL);
+
+ spin_lock(&peer_ni->ksnp_lock);
+
+ list_for_each_entry_safe(tx, tmp, &peer_ni->ksnp_zc_req_list, tx_zc_list) {
+ if (tx->tx_conn != conn)
+ continue;
+
+ LASSERT(tx->tx_msg.ksm_zc_cookies[0] != 0);
+
+ tx->tx_msg.ksm_zc_cookies[0] = 0;
+ tx->tx_zc_aborted = 1; /* mark it as not-acked */
+ list_del(&tx->tx_zc_list);
+ list_add(&tx->tx_zc_list, &zlist);
+ }
+
+ spin_unlock(&peer_ni->ksnp_lock);
+
+ while (!list_empty(&zlist)) {
+ tx = list_entry(zlist.next, struct ksock_tx, tx_zc_list);
+
+ list_del(&tx->tx_zc_list);
+ ksocknal_tx_decref(tx);
+ }
+}
+
+void
+ksocknal_terminate_conn(struct ksock_conn *conn)
{
/* This gets called by the reaper (guaranteed thread context) to
* disengage the socket from its callbacks and close it.
* ksnc_refcount will eventually hit zero, and then the reaper will
* destroy it. */
- unsigned long flags;
- ksock_peer_t *peer = conn->ksnc_peer;
- ksock_sched_t *sched = conn->ksnc_scheduler;
- struct timeval now;
- time_t then = 0;
- int notify = 0;
+ struct ksock_peer_ni *peer_ni = conn->ksnc_peer;
+ struct ksock_sched *sched = conn->ksnc_scheduler;
+ int failed = 0;
LASSERT(conn->ksnc_closing);
/* wake up the scheduler to "send" all remaining packets to /dev/null */
- spin_lock_irqsave(&sched->kss_lock, flags);
+ spin_lock_bh(&sched->kss_lock);
+
+ /* a closing conn is always ready to tx */
+ conn->ksnc_tx_ready = 1;
if (!conn->ksnc_tx_scheduled &&
- !list_empty(&conn->ksnc_tx_queue)){
- list_add_tail (&conn->ksnc_tx_list,
+ !list_empty(&conn->ksnc_tx_queue)) {
+ list_add_tail(&conn->ksnc_tx_list,
&sched->kss_tx_conns);
- /* a closing conn is always ready to tx */
- conn->ksnc_tx_ready = 1;
conn->ksnc_tx_scheduled = 1;
/* extra ref for scheduler */
- atomic_inc (&conn->ksnc_refcount);
+ ksocknal_conn_addref(conn);
- cfs_waitq_signal (&sched->kss_waitq);
- }
+ wake_up (&sched->kss_waitq);
+ }
- spin_unlock_irqrestore (&sched->kss_lock, flags);
+ spin_unlock_bh(&sched->kss_lock);
- /* serialise with callbacks */
- write_lock_irqsave (&ksocknal_data.ksnd_global_lock, flags);
+ /* serialise with callbacks */
+ write_lock_bh(&ksocknal_data.ksnd_global_lock);
ksocknal_lib_reset_callback(conn->ksnc_sock, conn);
* scheduler yet, but it _has_ committed to terminate... */
conn->ksnc_scheduler->kss_nconns--;
- if (peer->ksnp_error != 0) {
- /* peer's last conn closed in error */
- LASSERT (list_empty (&peer->ksnp_conns));
-
- /* convert peer's last-known-alive timestamp from jiffies */
- do_gettimeofday (&now);
- then = now.tv_sec - cfs_duration_sec(cfs_time_sub(cfs_time_current(),
- peer->ksnp_last_alive));
- notify = 1;
+ if (peer_ni->ksnp_error != 0) {
+ /* peer_ni's last conn closed in error */
+ LASSERT(list_empty(&peer_ni->ksnp_conns));
+ failed = 1;
+ peer_ni->ksnp_error = 0; /* avoid multiple notifications */
}
- write_unlock_irqrestore (&ksocknal_data.ksnd_global_lock, flags);
+ write_unlock_bh(&ksocknal_data.ksnd_global_lock);
+
+ if (failed)
+ ksocknal_peer_failed(peer_ni);
/* The socket is closed on the final put; either here, or in
* ksocknal_{send,recv}msg(). Since we set up the linger2 option
* when the connection was established, this will close the socket
* immediately, aborting anything buffered in it. Any hung
* zero-copy transmits will therefore complete in finite time. */
- ksocknal_putconnsock (conn);
+ ksocknal_connsock_decref(conn);
+}
+
+void
+ksocknal_queue_zombie_conn(struct ksock_conn *conn)
+{
+ /* Queue the conn for the reaper to destroy */
+ LASSERT(atomic_read(&conn->ksnc_conn_refcount) == 0);
+ spin_lock_bh(&ksocknal_data.ksnd_reaper_lock);
+
+ list_add_tail(&conn->ksnc_list, &ksocknal_data.ksnd_zombie_conns);
+ wake_up(&ksocknal_data.ksnd_reaper_waitq);
- if (notify)
- kpr_notify (&ksocknal_data.ksnd_router, peer->ksnp_nid,
- 0, then);
+ spin_unlock_bh(&ksocknal_data.ksnd_reaper_lock);
}
void
-ksocknal_destroy_conn (ksock_conn_t *conn)
+ksocknal_destroy_conn(struct ksock_conn *conn)
{
- /* Final coup-de-grace of the reaper */
- CDEBUG (D_NET, "connection %p\n", conn);
+ time64_t last_rcv;
- LASSERT (atomic_read (&conn->ksnc_refcount) == 0);
- LASSERT (conn->ksnc_route == NULL);
- LASSERT (!conn->ksnc_tx_scheduled);
- LASSERT (!conn->ksnc_rx_scheduled);
- LASSERT (list_empty(&conn->ksnc_tx_queue));
+ /* Final coup-de-grace of the reaper */
+ CDEBUG (D_NET, "connection %p\n", conn);
+
+ LASSERT (atomic_read (&conn->ksnc_conn_refcount) == 0);
+ LASSERT (atomic_read (&conn->ksnc_sock_refcount) == 0);
+ LASSERT (conn->ksnc_sock == NULL);
+ LASSERT (conn->ksnc_route == NULL);
+ LASSERT (!conn->ksnc_tx_scheduled);
+ LASSERT (!conn->ksnc_rx_scheduled);
+ LASSERT(list_empty(&conn->ksnc_tx_queue));
/* complete current receive if any */
switch (conn->ksnc_rx_state) {
- case SOCKNAL_RX_BODY:
- CERROR("Completing partial receive from "LPX64
- ", ip %d.%d.%d.%d:%d, with error\n",
- conn->ksnc_peer->ksnp_nid,
- HIPQUAD(conn->ksnc_ipaddr), conn->ksnc_port);
- lib_finalize (&ksocknal_lib, NULL, conn->ksnc_cookie, PTL_FAIL);
+ case SOCKNAL_RX_LNET_PAYLOAD:
+ last_rcv = conn->ksnc_rx_deadline -
+ lnet_get_lnd_timeout();
+ CERROR("Completing partial receive from %s[%d], "
+ "ip %pI4h:%d, with error, wanted: %d, left: %d, "
+ "last alive is %lld secs ago\n",
+ libcfs_id2str(conn->ksnc_peer->ksnp_id), conn->ksnc_type,
+ &conn->ksnc_ipaddr, conn->ksnc_port,
+ conn->ksnc_rx_nob_wanted, conn->ksnc_rx_nob_left,
+ ktime_get_seconds() - last_rcv);
+ if (conn->ksnc_lnet_msg)
+ conn->ksnc_lnet_msg->msg_health_status =
+ LNET_MSG_STATUS_REMOTE_ERROR;
+ lnet_finalize(conn->ksnc_lnet_msg, -EIO);
+ break;
+ case SOCKNAL_RX_LNET_HEADER:
+ if (conn->ksnc_rx_started)
+ CERROR("Incomplete receive of lnet header from %s, "
+ "ip %pI4h:%d, with error, protocol: %d.x.\n",
+ libcfs_id2str(conn->ksnc_peer->ksnp_id),
+ &conn->ksnc_ipaddr, conn->ksnc_port,
+ conn->ksnc_proto->pro_version);
break;
- case SOCKNAL_RX_BODY_FWD:
- ksocknal_fmb_callback (conn->ksnc_cookie, -ECONNABORTED);
+ case SOCKNAL_RX_KSM_HEADER:
+ if (conn->ksnc_rx_started)
+ CERROR("Incomplete receive of ksock message from %s, "
+ "ip %pI4h:%d, with error, protocol: %d.x.\n",
+ libcfs_id2str(conn->ksnc_peer->ksnp_id),
+ &conn->ksnc_ipaddr, conn->ksnc_port,
+ conn->ksnc_proto->pro_version);
break;
- case SOCKNAL_RX_HEADER:
case SOCKNAL_RX_SLOP:
- break;
+ if (conn->ksnc_rx_started)
+ CERROR("Incomplete receive of slops from %s, "
+ "ip %pI4h:%d, with error\n",
+ libcfs_id2str(conn->ksnc_peer->ksnp_id),
+ &conn->ksnc_ipaddr, conn->ksnc_port);
+ break;
default:
LBUG ();
break;
}
- ksocknal_put_peer (conn->ksnc_peer);
+ ksocknal_peer_decref(conn->ksnc_peer);
- PORTAL_FREE (conn, sizeof (*conn));
- atomic_dec (&ksocknal_data.ksnd_nclosing_conns);
-}
-
-void
-ksocknal_put_conn (ksock_conn_t *conn)
-{
- unsigned long flags;
-
- CDEBUG (D_OTHER, "putting conn[%p] -> "LPX64" (%d)\n",
- conn, conn->ksnc_peer->ksnp_nid,
- atomic_read (&conn->ksnc_refcount));
-
- LASSERT (atomic_read (&conn->ksnc_refcount) > 0);
- if (!atomic_dec_and_test (&conn->ksnc_refcount))
- return;
-
- spin_lock_irqsave (&ksocknal_data.ksnd_reaper_lock, flags);
-
- list_add (&conn->ksnc_list, &ksocknal_data.ksnd_zombie_conns);
- cfs_waitq_signal (&ksocknal_data.ksnd_reaper_waitq);
-
- spin_unlock_irqrestore (&ksocknal_data.ksnd_reaper_lock, flags);
+ LIBCFS_FREE (conn, sizeof (*conn));
}
int
-ksocknal_close_peer_conns_locked (ksock_peer_t *peer, __u32 ipaddr, int why)
+ksocknal_close_peer_conns_locked(struct ksock_peer_ni *peer_ni, __u32 ipaddr, int why)
{
- ksock_conn_t *conn;
- struct list_head *ctmp;
- struct list_head *cnxt;
- int count = 0;
+ struct ksock_conn *conn;
+ struct list_head *ctmp;
+ struct list_head *cnxt;
+ int count = 0;
- list_for_each_safe (ctmp, cnxt, &peer->ksnp_conns) {
- conn = list_entry (ctmp, ksock_conn_t, ksnc_list);
+ list_for_each_safe(ctmp, cnxt, &peer_ni->ksnp_conns) {
+ conn = list_entry(ctmp, struct ksock_conn, ksnc_list);
if (ipaddr == 0 ||
conn->ksnc_ipaddr == ipaddr) {
}
int
-ksocknal_close_stale_conns_locked (ksock_peer_t *peer, __u64 incarnation)
-{
- ksock_conn_t *conn;
- struct list_head *ctmp;
- struct list_head *cnxt;
- int count = 0;
-
- list_for_each_safe (ctmp, cnxt, &peer->ksnp_conns) {
- conn = list_entry (ctmp, ksock_conn_t, ksnc_list);
-
- if (conn->ksnc_incarnation == incarnation)
- continue;
-
- CWARN("Closing stale conn nid:"LPX64" ip:%08x/%d "
- "incarnation:"LPX64"("LPX64")\n",
- peer->ksnp_nid, conn->ksnc_ipaddr, conn->ksnc_port,
- conn->ksnc_incarnation, incarnation);
-
- count++;
- ksocknal_close_conn_locked (conn, -ESTALE);
- }
-
- return (count);
-}
-
-int
-ksocknal_close_conn_and_siblings (ksock_conn_t *conn, int why)
+ksocknal_close_conn_and_siblings(struct ksock_conn *conn, int why)
{
- ksock_peer_t *peer = conn->ksnc_peer;
- __u32 ipaddr = conn->ksnc_ipaddr;
- unsigned long flags;
- int count;
+ struct ksock_peer_ni *peer_ni = conn->ksnc_peer;
+ u32 ipaddr = conn->ksnc_ipaddr;
+ int count;
- write_lock_irqsave (&ksocknal_data.ksnd_global_lock, flags);
+ write_lock_bh(&ksocknal_data.ksnd_global_lock);
- count = ksocknal_close_peer_conns_locked (peer, ipaddr, why);
+ count = ksocknal_close_peer_conns_locked (peer_ni, ipaddr, why);
- write_unlock_irqrestore (&ksocknal_data.ksnd_global_lock, flags);
+ write_unlock_bh(&ksocknal_data.ksnd_global_lock);
return (count);
}
int
-ksocknal_close_matching_conns (ptl_nid_t nid, __u32 ipaddr)
+ksocknal_close_matching_conns(struct lnet_process_id id, __u32 ipaddr)
{
- unsigned long flags;
- ksock_peer_t *peer;
- struct list_head *ptmp;
- struct list_head *pnxt;
- int lo;
- int hi;
- int i;
- int count = 0;
-
- write_lock_irqsave (&ksocknal_data.ksnd_global_lock, flags);
-
- if (nid != PTL_NID_ANY)
- lo = hi = ksocknal_nid2peerlist(nid) - ksocknal_data.ksnd_peers;
+ struct ksock_peer_ni *peer_ni;
+ struct list_head *ptmp;
+ struct list_head *pnxt;
+ int lo;
+ int hi;
+ int i;
+ int count = 0;
+
+ write_lock_bh(&ksocknal_data.ksnd_global_lock);
+
+ if (id.nid != LNET_NID_ANY)
+ lo = hi = (int)(ksocknal_nid2peerlist(id.nid) - ksocknal_data.ksnd_peers);
else {
lo = 0;
hi = ksocknal_data.ksnd_peer_hash_size - 1;
}
for (i = lo; i <= hi; i++) {
- list_for_each_safe (ptmp, pnxt, &ksocknal_data.ksnd_peers[i]) {
+ list_for_each_safe(ptmp, pnxt, &ksocknal_data.ksnd_peers[i]) {
- peer = list_entry (ptmp, ksock_peer_t, ksnp_list);
+ peer_ni = list_entry(ptmp, struct ksock_peer_ni, ksnp_list);
- if (!(nid == PTL_NID_ANY || nid == peer->ksnp_nid))
+ if (!((id.nid == LNET_NID_ANY || id.nid == peer_ni->ksnp_id.nid) &&
+ (id.pid == LNET_PID_ANY || id.pid == peer_ni->ksnp_id.pid)))
continue;
- count += ksocknal_close_peer_conns_locked (peer, ipaddr, 0);
+ count += ksocknal_close_peer_conns_locked (peer_ni, ipaddr, 0);
}
}
- write_unlock_irqrestore (&ksocknal_data.ksnd_global_lock, flags);
+ write_unlock_bh(&ksocknal_data.ksnd_global_lock);
/* wildcards always succeed */
- if (nid == PTL_NID_ANY || ipaddr == 0)
+ if (id.nid == LNET_NID_ANY || id.pid == LNET_PID_ANY || ipaddr == 0)
return (0);
return (count == 0 ? -ENOENT : 0);
}
void
-ksocknal_notify (void *arg, ptl_nid_t gw_nid, int alive)
+ksocknal_notify(struct lnet_ni *ni, lnet_nid_t gw_nid, int alive)
{
- /* The router is telling me she's been notified of a change in
- * gateway state.... */
+ /* The router is telling me she's been notified of a change in
+ * gateway state....
+ */
+ struct lnet_process_id id = {
+ .nid = gw_nid,
+ .pid = LNET_PID_ANY,
+ };
- CDEBUG (D_NET, "gw "LPX64" %s\n", gw_nid, alive ? "up" : "down");
+ CDEBUG (D_NET, "gw %s %s\n", libcfs_nid2str(gw_nid),
+ alive ? "up" : "down");
if (!alive) {
/* If the gateway crashed, close all open connections... */
- ksocknal_close_matching_conns (gw_nid, 0);
+ ksocknal_close_matching_conns (id, 0);
return;
}
}
void
-ksocknal_push_peer (ksock_peer_t *peer)
+ksocknal_query(struct lnet_ni *ni, lnet_nid_t nid, time64_t *when)
{
- int index;
- int i;
- struct list_head *tmp;
- ksock_conn_t *conn;
-
- for (index = 0; ; index++) {
- read_lock (&ksocknal_data.ksnd_global_lock);
-
- i = 0;
- conn = NULL;
-
- list_for_each (tmp, &peer->ksnp_conns) {
- if (i++ == index) {
- conn = list_entry (tmp, ksock_conn_t, ksnc_list);
- atomic_inc (&conn->ksnc_refcount);
- break;
+ int connect = 1;
+ time64_t last_alive = 0;
+ time64_t now = ktime_get_seconds();
+ struct ksock_peer_ni *peer_ni = NULL;
+ rwlock_t *glock = &ksocknal_data.ksnd_global_lock;
+ struct lnet_process_id id = {
+ .nid = nid,
+ .pid = LNET_PID_LUSTRE,
+ };
+
+ read_lock(glock);
+
+ peer_ni = ksocknal_find_peer_locked(ni, id);
+ if (peer_ni != NULL) {
+ struct list_head *tmp;
+ struct ksock_conn *conn;
+ int bufnob;
+
+ list_for_each(tmp, &peer_ni->ksnp_conns) {
+ conn = list_entry(tmp, struct ksock_conn, ksnc_list);
+ bufnob = conn->ksnc_sock->sk->sk_wmem_queued;
+
+ if (bufnob < conn->ksnc_tx_bufnob) {
+ /* something got ACKed */
+ conn->ksnc_tx_deadline = ktime_get_seconds() +
+ lnet_get_lnd_timeout();
+ peer_ni->ksnp_last_alive = now;
+ conn->ksnc_tx_bufnob = bufnob;
}
}
- read_unlock (&ksocknal_data.ksnd_global_lock);
-
- if (conn == NULL)
- break;
-
- ksocknal_lib_push_conn (conn);
- ksocknal_put_conn (conn);
+ last_alive = peer_ni->ksnp_last_alive;
+ if (ksocknal_find_connectable_route_locked(peer_ni) == NULL)
+ connect = 0;
}
-}
-int
-ksocknal_push (ptl_nid_t nid)
-{
- ksock_peer_t *peer;
- struct list_head *tmp;
- int index;
- int i;
- int j;
- int rc = -ENOENT;
+ read_unlock(glock);
- if (nid != PTL_NID_ANY) {
- peer = ksocknal_get_peer (nid);
+ if (last_alive != 0)
+ *when = last_alive;
- if (peer != NULL) {
- rc = 0;
- ksocknal_push_peer (peer);
- ksocknal_put_peer (peer);
- }
- return (rc);
- }
+ CDEBUG(D_NET, "peer_ni %s %p, alive %lld secs ago, connect %d\n",
+ libcfs_nid2str(nid), peer_ni,
+ last_alive ? now - last_alive : -1,
+ connect);
- for (i = 0; i < ksocknal_data.ksnd_peer_hash_size; i++) {
- for (j = 0; ; j++) {
- read_lock (&ksocknal_data.ksnd_global_lock);
-
- index = 0;
- peer = NULL;
-
- list_for_each (tmp, &ksocknal_data.ksnd_peers[i]) {
- if (index++ == j) {
- peer = list_entry(tmp, ksock_peer_t,
- ksnp_list);
- atomic_inc (&peer->ksnp_refcount);
- break;
- }
- }
+ if (!connect)
+ return;
- read_unlock (&ksocknal_data.ksnd_global_lock);
+ ksocknal_add_peer(ni, id, LNET_NIDADDR(nid), lnet_acceptor_port());
- if (peer != NULL) {
- rc = 0;
- ksocknal_push_peer (peer);
- ksocknal_put_peer (peer);
- }
- }
+ write_lock_bh(glock);
- }
+ peer_ni = ksocknal_find_peer_locked(ni, id);
+ if (peer_ni != NULL)
+ ksocknal_launch_all_connections_locked(peer_ni);
- return (rc);
+ write_unlock_bh(glock);
+ return;
}
-int
-ksocknal_add_interface(__u32 ipaddress, __u32 netmask)
+static void
+ksocknal_push_peer(struct ksock_peer_ni *peer_ni)
{
- unsigned long flags;
- ksock_interface_t *iface;
- int rc;
- int i;
- int j;
- struct list_head *ptmp;
- ksock_peer_t *peer;
- struct list_head *rtmp;
- ksock_route_t *route;
-
- if (ipaddress == 0 ||
- netmask == 0)
- return (-EINVAL);
+ int index;
+ int i;
+ struct list_head *tmp;
+ struct ksock_conn *conn;
- write_lock_irqsave(&ksocknal_data.ksnd_global_lock, flags);
-
- iface = ksocknal_ip2iface(ipaddress);
- if (iface != NULL) {
- /* silently ignore dups */
- rc = 0;
- } else if (ksocknal_data.ksnd_ninterfaces == SOCKNAL_MAX_INTERFACES) {
- rc = -ENOSPC;
- } else {
- iface = &ksocknal_data.ksnd_interfaces[ksocknal_data.ksnd_ninterfaces++];
-
- iface->ksni_ipaddr = ipaddress;
- iface->ksni_netmask = netmask;
- iface->ksni_nroutes = 0;
- iface->ksni_npeers = 0;
-
- for (i = 0; i < ksocknal_data.ksnd_peer_hash_size; i++) {
- list_for_each(ptmp, &ksocknal_data.ksnd_peers[i]) {
- peer = list_entry(ptmp, ksock_peer_t, ksnp_list);
-
- for (j = 0; i < peer->ksnp_n_passive_ips; j++)
- if (peer->ksnp_passive_ips[j] == ipaddress)
- iface->ksni_npeers++;
+ for (index = 0; ; index++) {
+ read_lock(&ksocknal_data.ksnd_global_lock);
- list_for_each(rtmp, &peer->ksnp_routes) {
- route = list_entry(rtmp, ksock_route_t, ksnr_list);
+ i = 0;
+ conn = NULL;
- if (route->ksnr_myipaddr == ipaddress)
- iface->ksni_nroutes++;
- }
+ list_for_each(tmp, &peer_ni->ksnp_conns) {
+ if (i++ == index) {
+ conn = list_entry(tmp, struct ksock_conn,
+ ksnc_list);
+ ksocknal_conn_addref(conn);
+ break;
}
}
- rc = 0;
- /* NB only new connections will pay attention to the new interface! */
+ read_unlock(&ksocknal_data.ksnd_global_lock);
+
+ if (conn == NULL)
+ break;
+
+ ksocknal_lib_push_conn (conn);
+ ksocknal_conn_decref(conn);
}
+}
- write_unlock_irqrestore(&ksocknal_data.ksnd_global_lock, flags);
+static int
+ksocknal_push(struct lnet_ni *ni, struct lnet_process_id id)
+{
+ struct list_head *start;
+ struct list_head *end;
+ struct list_head *tmp;
+ int rc = -ENOENT;
+ unsigned int hsize = ksocknal_data.ksnd_peer_hash_size;
+
+ if (id.nid == LNET_NID_ANY) {
+ start = &ksocknal_data.ksnd_peers[0];
+ end = &ksocknal_data.ksnd_peers[hsize - 1];
+ } else {
+ start = end = ksocknal_nid2peerlist(id.nid);
+ }
+
+ for (tmp = start; tmp <= end; tmp++) {
+ int peer_off; /* searching offset in peer_ni hash table */
+
+ for (peer_off = 0; ; peer_off++) {
+ struct ksock_peer_ni *peer_ni;
+ int i = 0;
+
+ read_lock(&ksocknal_data.ksnd_global_lock);
+ list_for_each_entry(peer_ni, tmp, ksnp_list) {
+ if (!((id.nid == LNET_NID_ANY ||
+ id.nid == peer_ni->ksnp_id.nid) &&
+ (id.pid == LNET_PID_ANY ||
+ id.pid == peer_ni->ksnp_id.pid)))
+ continue;
+
+ if (i++ == peer_off) {
+ ksocknal_peer_addref(peer_ni);
+ break;
+ }
+ }
+ read_unlock(&ksocknal_data.ksnd_global_lock);
+
+ if (i == 0) /* no match */
+ break;
+
+ rc = 0;
+ ksocknal_push_peer(peer_ni);
+ ksocknal_peer_decref(peer_ni);
+ }
+ }
+ return rc;
+}
- return (rc);
+static int
+ksocknal_add_interface(struct lnet_ni *ni, __u32 ipaddress, __u32 netmask)
+{
+ struct ksock_net *net = ni->ni_data;
+ struct ksock_interface *iface;
+ int rc;
+ int i;
+ int j;
+ struct list_head *ptmp;
+ struct ksock_peer_ni *peer_ni;
+ struct list_head *rtmp;
+ struct ksock_route *route;
+
+ if (ipaddress == 0 ||
+ netmask == 0)
+ return -EINVAL;
+
+ write_lock_bh(&ksocknal_data.ksnd_global_lock);
+
+ iface = ksocknal_ip2iface(ni, ipaddress);
+ if (iface != NULL) {
+ /* silently ignore dups */
+ rc = 0;
+ } else if (net->ksnn_ninterfaces == LNET_INTERFACES_NUM) {
+ rc = -ENOSPC;
+ } else {
+ iface = &net->ksnn_interfaces[net->ksnn_ninterfaces++];
+
+ iface->ksni_ipaddr = ipaddress;
+ iface->ksni_netmask = netmask;
+ iface->ksni_nroutes = 0;
+ iface->ksni_npeers = 0;
+
+ for (i = 0; i < ksocknal_data.ksnd_peer_hash_size; i++) {
+ list_for_each(ptmp, &ksocknal_data.ksnd_peers[i]) {
+ peer_ni = list_entry(ptmp, struct ksock_peer_ni,
+ ksnp_list);
+
+ for (j = 0; j < peer_ni->ksnp_n_passive_ips; j++)
+ if (peer_ni->ksnp_passive_ips[j] == ipaddress)
+ iface->ksni_npeers++;
+
+ list_for_each(rtmp, &peer_ni->ksnp_routes) {
+ route = list_entry(rtmp,
+ struct ksock_route,
+ ksnr_list);
+
+ if (route->ksnr_myipaddr == ipaddress)
+ iface->ksni_nroutes++;
+ }
+ }
+ }
+
+ rc = 0;
+ /* NB only new connections will pay attention to the new interface! */
+ }
+
+ write_unlock_bh(&ksocknal_data.ksnd_global_lock);
+
+ return rc;
}
-void
-ksocknal_peer_del_interface_locked(ksock_peer_t *peer, __u32 ipaddr)
+static void
+ksocknal_peer_del_interface_locked(struct ksock_peer_ni *peer_ni, __u32 ipaddr)
{
- struct list_head *tmp;
- struct list_head *nxt;
- ksock_route_t *route;
- ksock_conn_t *conn;
- int i;
- int j;
-
- for (i = 0; i < peer->ksnp_n_passive_ips; i++)
- if (peer->ksnp_passive_ips[i] == ipaddr) {
- for (j = i+1; j < peer->ksnp_n_passive_ips; j++)
- peer->ksnp_passive_ips[j-1] =
- peer->ksnp_passive_ips[j];
- peer->ksnp_n_passive_ips--;
+ struct list_head *tmp;
+ struct list_head *nxt;
+ struct ksock_route *route;
+ struct ksock_conn *conn;
+ int i;
+ int j;
+
+ for (i = 0; i < peer_ni->ksnp_n_passive_ips; i++)
+ if (peer_ni->ksnp_passive_ips[i] == ipaddr) {
+ for (j = i+1; j < peer_ni->ksnp_n_passive_ips; j++)
+ peer_ni->ksnp_passive_ips[j-1] =
+ peer_ni->ksnp_passive_ips[j];
+ peer_ni->ksnp_n_passive_ips--;
break;
}
- list_for_each_safe(tmp, nxt, &peer->ksnp_routes) {
- route = list_entry (tmp, ksock_route_t, ksnr_list);
+ list_for_each_safe(tmp, nxt, &peer_ni->ksnp_routes) {
+ route = list_entry(tmp, struct ksock_route, ksnr_list);
if (route->ksnr_myipaddr != ipaddr)
continue;
}
}
- list_for_each_safe(tmp, nxt, &peer->ksnp_conns) {
- conn = list_entry(tmp, ksock_conn_t, ksnc_list);
+ list_for_each_safe(tmp, nxt, &peer_ni->ksnp_conns) {
+ conn = list_entry(tmp, struct ksock_conn, ksnc_list);
if (conn->ksnc_myipaddr == ipaddr)
ksocknal_close_conn_locked (conn, 0);
}
}
-int
-ksocknal_del_interface(__u32 ipaddress)
+static int
+ksocknal_del_interface(struct lnet_ni *ni, __u32 ipaddress)
{
- int rc = -ENOENT;
- unsigned long flags;
- struct list_head *tmp;
- struct list_head *nxt;
- ksock_peer_t *peer;
- __u32 this_ip;
- int i;
- int j;
+ struct ksock_net *net = ni->ni_data;
+ int rc = -ENOENT;
+ struct list_head *tmp;
+ struct list_head *nxt;
+ struct ksock_peer_ni *peer_ni;
+ u32 this_ip;
+ int i;
+ int j;
- write_lock_irqsave(&ksocknal_data.ksnd_global_lock, flags);
+ write_lock_bh(&ksocknal_data.ksnd_global_lock);
- for (i = 0; i < ksocknal_data.ksnd_ninterfaces; i++) {
- this_ip = ksocknal_data.ksnd_interfaces[i].ksni_ipaddr;
+ for (i = 0; i < net->ksnn_ninterfaces; i++) {
+ this_ip = net->ksnn_interfaces[i].ksni_ipaddr;
if (!(ipaddress == 0 ||
ipaddress == this_ip))
rc = 0;
- for (j = i+1; j < ksocknal_data.ksnd_ninterfaces; j++)
- ksocknal_data.ksnd_interfaces[j-1] =
- ksocknal_data.ksnd_interfaces[j];
+ for (j = i+1; j < net->ksnn_ninterfaces; j++)
+ net->ksnn_interfaces[j-1] =
+ net->ksnn_interfaces[j];
- ksocknal_data.ksnd_ninterfaces--;
+ net->ksnn_ninterfaces--;
for (j = 0; j < ksocknal_data.ksnd_peer_hash_size; j++) {
- list_for_each_safe(tmp, nxt, &ksocknal_data.ksnd_peers[j]) {
- peer = list_entry(tmp, ksock_peer_t, ksnp_list);
+ list_for_each_safe(tmp, nxt,
+ &ksocknal_data.ksnd_peers[j]) {
+ peer_ni = list_entry(tmp, struct ksock_peer_ni,
+ ksnp_list);
+
+ if (peer_ni->ksnp_ni != ni)
+ continue;
- ksocknal_peer_del_interface_locked(peer, this_ip);
+ ksocknal_peer_del_interface_locked(peer_ni, this_ip);
}
}
}
- write_unlock_irqrestore(&ksocknal_data.ksnd_global_lock, flags);
+ write_unlock_bh(&ksocknal_data.ksnd_global_lock);
return (rc);
}
int
-ksocknal_cmd(struct portals_cfg *pcfg, void * private)
+ksocknal_ctl(struct lnet_ni *ni, unsigned int cmd, void *arg)
{
+ struct lnet_process_id id = {0};
+ struct libcfs_ioctl_data *data = arg;
int rc;
- switch(pcfg->pcfg_command) {
- case NAL_CMD_GET_INTERFACE: {
- ksock_interface_t *iface;
+ switch(cmd) {
+ case IOC_LIBCFS_GET_INTERFACE: {
+ struct ksock_net *net = ni->ni_data;
+ struct ksock_interface *iface;
- read_lock (&ksocknal_data.ksnd_global_lock);
+ read_lock(&ksocknal_data.ksnd_global_lock);
- if (pcfg->pcfg_count < 0 ||
- pcfg->pcfg_count >= ksocknal_data.ksnd_ninterfaces) {
+ if (data->ioc_count >= (__u32)net->ksnn_ninterfaces) {
rc = -ENOENT;
} else {
rc = 0;
- iface = &ksocknal_data.ksnd_interfaces[pcfg->pcfg_count];
+ iface = &net->ksnn_interfaces[data->ioc_count];
- pcfg->pcfg_id = iface->ksni_ipaddr;
- pcfg->pcfg_misc = iface->ksni_netmask;
- pcfg->pcfg_fd = iface->ksni_npeers;
- pcfg->pcfg_count = iface->ksni_nroutes;
+ data->ioc_u32[0] = iface->ksni_ipaddr;
+ data->ioc_u32[1] = iface->ksni_netmask;
+ data->ioc_u32[2] = iface->ksni_npeers;
+ data->ioc_u32[3] = iface->ksni_nroutes;
}
- read_unlock (&ksocknal_data.ksnd_global_lock);
- break;
- }
- case NAL_CMD_ADD_INTERFACE: {
- rc = ksocknal_add_interface(pcfg->pcfg_id, /* IP address */
- pcfg->pcfg_misc); /* net mask */
- break;
+ read_unlock(&ksocknal_data.ksnd_global_lock);
+ return rc;
}
- case NAL_CMD_DEL_INTERFACE: {
- rc = ksocknal_del_interface(pcfg->pcfg_id); /* IP address */
- break;
- }
- case NAL_CMD_GET_PEER: {
- ptl_nid_t nid = 0;
- __u32 myip = 0;
- __u32 ip = 0;
- int port = 0;
- int conn_count = 0;
- int share_count = 0;
-
- rc = ksocknal_get_peer_info(pcfg->pcfg_count, &nid,
- &myip, &ip, &port,
- &conn_count, &share_count);
- pcfg->pcfg_nid = nid;
- pcfg->pcfg_size = myip;
- pcfg->pcfg_id = ip;
- pcfg->pcfg_misc = port;
- pcfg->pcfg_count = conn_count;
- pcfg->pcfg_wait = share_count;
- break;
- }
- case NAL_CMD_ADD_PEER: {
- rc = ksocknal_add_peer (pcfg->pcfg_nid,
- pcfg->pcfg_id, /* IP */
- pcfg->pcfg_misc); /* port */
- break;
- }
- case NAL_CMD_DEL_PEER: {
- rc = ksocknal_del_peer (pcfg->pcfg_nid,
- pcfg->pcfg_id, /* IP */
- pcfg->pcfg_flags); /* single_share? */
- break;
- }
- case NAL_CMD_GET_CONN: {
- ksock_conn_t *conn = ksocknal_get_conn_by_idx (pcfg->pcfg_count);
- if (conn == NULL)
- rc = -ENOENT;
- else {
- int txmem;
- int rxmem;
- int nagle;
+ case IOC_LIBCFS_ADD_INTERFACE:
+ return ksocknal_add_interface(ni,
+ data->ioc_u32[0], /* IP address */
+ data->ioc_u32[1]); /* net mask */
- ksocknal_lib_get_conn_tunables(conn, &txmem, &rxmem, &nagle);
+ case IOC_LIBCFS_DEL_INTERFACE:
+ return ksocknal_del_interface(ni,
+ data->ioc_u32[0]); /* IP address */
- rc = 0;
- pcfg->pcfg_nid = conn->ksnc_peer->ksnp_nid;
- pcfg->pcfg_id = conn->ksnc_ipaddr;
- pcfg->pcfg_misc = conn->ksnc_port;
- pcfg->pcfg_fd = conn->ksnc_myipaddr;
- pcfg->pcfg_flags = conn->ksnc_type;
- pcfg->pcfg_gw_nal = conn->ksnc_scheduler -
- ksocknal_data.ksnd_schedulers;
- pcfg->pcfg_count = txmem;
- pcfg->pcfg_size = rxmem;
- pcfg->pcfg_wait = nagle;
- ksocknal_put_conn (conn);
- }
- break;
- }
- case NAL_CMD_REGISTER_PEER_FD: {
- struct socket *sock = sockfd_lookup (pcfg->pcfg_fd, &rc);
- int type = pcfg->pcfg_misc;
+ case IOC_LIBCFS_GET_PEER: {
+ __u32 myip = 0;
+ __u32 ip = 0;
+ int port = 0;
+ int conn_count = 0;
+ int share_count = 0;
- if (sock == NULL)
- break;
+ rc = ksocknal_get_peer_info(ni, data->ioc_count,
+ &id, &myip, &ip, &port,
+ &conn_count, &share_count);
+ if (rc != 0)
+ return rc;
+
+ data->ioc_nid = id.nid;
+ data->ioc_count = share_count;
+ data->ioc_u32[0] = ip;
+ data->ioc_u32[1] = port;
+ data->ioc_u32[2] = myip;
+ data->ioc_u32[3] = conn_count;
+ data->ioc_u32[4] = id.pid;
+ return 0;
+ }
+
+ case IOC_LIBCFS_ADD_PEER:
+ id.nid = data->ioc_nid;
+ id.pid = LNET_PID_LUSTRE;
+ return ksocknal_add_peer (ni, id,
+ data->ioc_u32[0], /* IP */
+ data->ioc_u32[1]); /* port */
+
+ case IOC_LIBCFS_DEL_PEER:
+ id.nid = data->ioc_nid;
+ id.pid = LNET_PID_ANY;
+ return ksocknal_del_peer (ni, id,
+ data->ioc_u32[0]); /* IP */
+
+ case IOC_LIBCFS_GET_CONN: {
+ int txmem;
+ int rxmem;
+ int nagle;
+ struct ksock_conn *conn = ksocknal_get_conn_by_idx(ni, data->ioc_count);
+
+ if (conn == NULL)
+ return -ENOENT;
+
+ ksocknal_lib_get_conn_tunables(conn, &txmem, &rxmem, &nagle);
+
+ data->ioc_count = txmem;
+ data->ioc_nid = conn->ksnc_peer->ksnp_id.nid;
+ data->ioc_flags = nagle;
+ data->ioc_u32[0] = conn->ksnc_ipaddr;
+ data->ioc_u32[1] = conn->ksnc_port;
+ data->ioc_u32[2] = conn->ksnc_myipaddr;
+ data->ioc_u32[3] = conn->ksnc_type;
+ data->ioc_u32[4] = conn->ksnc_scheduler->kss_cpt;
+ data->ioc_u32[5] = rxmem;
+ data->ioc_u32[6] = conn->ksnc_peer->ksnp_id.pid;
+ ksocknal_conn_decref(conn);
+ return 0;
+ }
+
+ case IOC_LIBCFS_CLOSE_CONNECTION:
+ id.nid = data->ioc_nid;
+ id.pid = LNET_PID_ANY;
+ return ksocknal_close_matching_conns (id,
+ data->ioc_u32[0]);
+
+ case IOC_LIBCFS_REGISTER_MYNID:
+ /* Ignore if this is a noop */
+ if (data->ioc_nid == ni->ni_nid)
+ return 0;
+
+ CERROR("obsolete IOC_LIBCFS_REGISTER_MYNID: %s(%s)\n",
+ libcfs_nid2str(data->ioc_nid),
+ libcfs_nid2str(ni->ni_nid));
+ return -EINVAL;
+
+ case IOC_LIBCFS_PUSH_CONNECTION:
+ id.nid = data->ioc_nid;
+ id.pid = LNET_PID_ANY;
+ return ksocknal_push(ni, id);
- switch (type) {
- case SOCKNAL_CONN_NONE:
- case SOCKNAL_CONN_ANY:
- case SOCKNAL_CONN_CONTROL:
- case SOCKNAL_CONN_BULK_IN:
- case SOCKNAL_CONN_BULK_OUT:
- rc = ksocknal_create_conn(NULL, sock, type);
- break;
- default:
- rc = -EINVAL;
- break;
- }
- cfs_put_file (KSN_SOCK2FILE(sock));
- break;
- }
- case NAL_CMD_CLOSE_CONNECTION: {
- rc = ksocknal_close_matching_conns (pcfg->pcfg_nid,
- pcfg->pcfg_id);
- break;
- }
- case NAL_CMD_REGISTER_MYNID: {
- rc = ksocknal_set_mynid (pcfg->pcfg_nid);
- break;
- }
- case NAL_CMD_PUSH_CONNECTION: {
- rc = ksocknal_push (pcfg->pcfg_nid);
- break;
- }
default:
- rc = -EINVAL;
- break;
+ return -EINVAL;
}
-
- return rc;
+ /* not reached */
}
-void
-ksocknal_free_fmbs (ksock_fmb_pool_t *p)
+static void
+ksocknal_free_buffers (void)
{
- int npages = p->fmp_buff_pages;
- ksock_fmb_t *fmb;
- int i;
-
- LASSERT (list_empty(&p->fmp_blocked_conns));
- LASSERT (p->fmp_nactive_fmbs == 0);
-
- while (!list_empty(&p->fmp_idle_fmbs)) {
-
- fmb = list_entry(p->fmp_idle_fmbs.next,
- ksock_fmb_t, fmb_list);
+ LASSERT (atomic_read(&ksocknal_data.ksnd_nactive_txs) == 0);
- for (i = 0; i < npages; i++)
- if (fmb->fmb_kiov[i].kiov_page != NULL)
- cfs_free_page(fmb->fmb_kiov[i].kiov_page);
+ if (ksocknal_data.ksnd_schedulers != NULL)
+ cfs_percpt_free(ksocknal_data.ksnd_schedulers);
- list_del(&fmb->fmb_list);
- PORTAL_FREE(fmb, offsetof(ksock_fmb_t, fmb_kiov[npages]));
- }
-}
+ LIBCFS_FREE (ksocknal_data.ksnd_peers,
+ sizeof(struct list_head) *
+ ksocknal_data.ksnd_peer_hash_size);
-void
-ksocknal_free_buffers (void)
-{
- ksocknal_free_fmbs(&ksocknal_data.ksnd_small_fmp);
- ksocknal_free_fmbs(&ksocknal_data.ksnd_large_fmp);
+ spin_lock(&ksocknal_data.ksnd_tx_lock);
- LASSERT (atomic_read(&ksocknal_data.ksnd_nactive_ltxs) == 0);
+ if (!list_empty(&ksocknal_data.ksnd_idle_noop_txs)) {
+ struct list_head zlist;
+ struct ksock_tx *tx;
- if (ksocknal_data.ksnd_schedulers != NULL)
- PORTAL_FREE (ksocknal_data.ksnd_schedulers,
- sizeof (ksock_sched_t) * ksocknal_data.ksnd_nschedulers);
+ list_add(&zlist, &ksocknal_data.ksnd_idle_noop_txs);
+ list_del_init(&ksocknal_data.ksnd_idle_noop_txs);
+ spin_unlock(&ksocknal_data.ksnd_tx_lock);
- PORTAL_FREE (ksocknal_data.ksnd_peers,
- sizeof (struct list_head) *
- ksocknal_data.ksnd_peer_hash_size);
+ while (!list_empty(&zlist)) {
+ tx = list_entry(zlist.next, struct ksock_tx, tx_list);
+ list_del(&tx->tx_list);
+ LIBCFS_FREE(tx, tx->tx_desc_size);
+ }
+ } else {
+ spin_unlock(&ksocknal_data.ksnd_tx_lock);
+ }
}
-void
-ksocknal_api_shutdown (nal_t *nal)
+static void
+ksocknal_base_shutdown(void)
{
- ksock_sched_t *sched;
- int i;
-
- if (nal->nal_refct != 0) {
- /* This module got the first ref */
- PORTAL_MODULE_UNUSE;
- return;
- }
+ struct ksock_sched *sched;
+ int i;
- CDEBUG(D_MALLOC, "before NAL cleanup: kmem %d\n",
- atomic_read (&portal_kmemory));
-
- LASSERT(nal == &ksocknal_api);
+ CDEBUG(D_MALLOC, "before NAL cleanup: kmem %d\n",
+ atomic_read (&libcfs_kmemory));
+ LASSERT (ksocknal_data.ksnd_nnets == 0);
switch (ksocknal_data.ksnd_init) {
default:
LASSERT (0);
case SOCKNAL_INIT_ALL:
- libcfs_nal_cmd_unregister(SOCKNAL);
-
- ksocknal_data.ksnd_init = SOCKNAL_INIT_LIB;
- /* fall through */
-
- case SOCKNAL_INIT_LIB:
- /* No more calls to ksocknal_cmd() to create new
- * autoroutes/connections since we're being unloaded. */
-
- /* Delete all peers */
- ksocknal_del_peer(PTL_NID_ANY, 0, 0);
-
- /* Wait for all peer state to clean up */
- i = 2;
- while (atomic_read (&ksocknal_data.ksnd_npeers) != 0) {
- i++;
- CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET, /* power of 2? */
- "waiting for %d peers to disconnect\n",
- atomic_read (&ksocknal_data.ksnd_npeers));
- set_current_state (TASK_UNINTERRUPTIBLE);
- schedule_timeout (cfs_time_seconds(1));
- }
-
- /* Tell lib we've stopped calling into her. */
- lib_fini(&ksocknal_lib);
-
- ksocknal_data.ksnd_init = SOCKNAL_INIT_DATA;
- /* fall through */
-
case SOCKNAL_INIT_DATA:
- LASSERT (atomic_read (&ksocknal_data.ksnd_npeers) == 0);
LASSERT (ksocknal_data.ksnd_peers != NULL);
for (i = 0; i < ksocknal_data.ksnd_peer_hash_size; i++) {
- LASSERT (list_empty (&ksocknal_data.ksnd_peers[i]));
- }
- LASSERT (list_empty (&ksocknal_data.ksnd_enomem_conns));
- LASSERT (list_empty (&ksocknal_data.ksnd_zombie_conns));
- LASSERT (list_empty (&ksocknal_data.ksnd_autoconnectd_routes));
- LASSERT (list_empty (&ksocknal_data.ksnd_small_fmp.fmp_blocked_conns));
- LASSERT (list_empty (&ksocknal_data.ksnd_large_fmp.fmp_blocked_conns));
-
- if (ksocknal_data.ksnd_schedulers != NULL)
- for (i = 0; i < ksocknal_data.ksnd_nschedulers; i++) {
- ksock_sched_t *kss =
- &ksocknal_data.ksnd_schedulers[i];
-
- LASSERT (list_empty (&kss->kss_tx_conns));
- LASSERT (list_empty (&kss->kss_rx_conns));
- LASSERT (kss->kss_nconns == 0);
- }
-
- /* stop router calling me */
- kpr_shutdown (&ksocknal_data.ksnd_router);
-
- /* flag threads to terminate; wake and wait for them to die */
- ksocknal_data.ksnd_shuttingdown = 1;
- cfs_waitq_broadcast (&ksocknal_data.ksnd_autoconnectd_waitq);
- cfs_waitq_broadcast (&ksocknal_data.ksnd_reaper_waitq);
-
- for (i = 0; i < ksocknal_data.ksnd_nschedulers; i++) {
- sched = &ksocknal_data.ksnd_schedulers[i];
- cfs_waitq_broadcast(&sched->kss_waitq);
- }
-
- i = 4;
- read_lock(&ksocknal_data.ksnd_global_lock);
- while (ksocknal_data.ksnd_nthreads != 0) {
- i++;
- CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET, /* power of 2? */
- "waiting for %d threads to terminate\n",
- ksocknal_data.ksnd_nthreads);
- read_unlock(&ksocknal_data.ksnd_global_lock);
- set_current_state (TASK_UNINTERRUPTIBLE);
- schedule_timeout (cfs_time_seconds(1));
- read_lock(&ksocknal_data.ksnd_global_lock);
+ LASSERT(list_empty(&ksocknal_data.ksnd_peers[i]));
}
- read_unlock(&ksocknal_data.ksnd_global_lock);
- kpr_deregister (&ksocknal_data.ksnd_router);
+ LASSERT(list_empty(&ksocknal_data.ksnd_nets));
+ LASSERT(list_empty(&ksocknal_data.ksnd_enomem_conns));
+ LASSERT(list_empty(&ksocknal_data.ksnd_zombie_conns));
+ LASSERT(list_empty(&ksocknal_data.ksnd_connd_connreqs));
+ LASSERT(list_empty(&ksocknal_data.ksnd_connd_routes));
+
+ if (ksocknal_data.ksnd_schedulers != NULL) {
+ cfs_percpt_for_each(sched, i,
+ ksocknal_data.ksnd_schedulers) {
+
+ LASSERT(list_empty(&sched->kss_tx_conns));
+ LASSERT(list_empty(&sched->kss_rx_conns));
+ LASSERT(list_empty(&sched->kss_zombie_noop_txs));
+ LASSERT(sched->kss_nconns == 0);
+ }
+ }
+
+ /* flag threads to terminate; wake and wait for them to die */
+ ksocknal_data.ksnd_shuttingdown = 1;
+ wake_up_all(&ksocknal_data.ksnd_connd_waitq);
+ wake_up_all(&ksocknal_data.ksnd_reaper_waitq);
+
+ if (ksocknal_data.ksnd_schedulers != NULL) {
+ cfs_percpt_for_each(sched, i,
+ ksocknal_data.ksnd_schedulers)
+ wake_up_all(&sched->kss_waitq);
+ }
+
+ i = 4;
+ read_lock(&ksocknal_data.ksnd_global_lock);
+ while (ksocknal_data.ksnd_nthreads != 0) {
+ i++;
+ /* power of 2? */
+ CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET,
+ "waiting for %d threads to terminate\n",
+ ksocknal_data.ksnd_nthreads);
+ read_unlock(&ksocknal_data.ksnd_global_lock);
+ set_current_state(TASK_UNINTERRUPTIBLE);
+ schedule_timeout(cfs_time_seconds(1));
+ read_lock(&ksocknal_data.ksnd_global_lock);
+ }
+ read_unlock(&ksocknal_data.ksnd_global_lock);
ksocknal_free_buffers();
ksocknal_data.ksnd_init = SOCKNAL_INIT_NOTHING;
- /* fall through */
-
- case SOCKNAL_INIT_NOTHING:
break;
}
- CDEBUG(D_MALLOC, "after NAL cleanup: kmem %d\n",
- atomic_read (&portal_kmemory));
+ CDEBUG(D_MALLOC, "after NAL cleanup: kmem %d\n",
+ atomic_read (&libcfs_kmemory));
- printk(KERN_INFO "Lustre: Routing socket NAL unloaded (final mem %d)\n",
- atomic_read(&portal_kmemory));
+ module_put(THIS_MODULE);
}
-
-void
-ksocknal_init_incarnation (void)
+static int
+ksocknal_base_startup(void)
{
- struct timeval tv;
-
- /* The incarnation number is the time this module loaded and it
- * identifies this particular instance of the socknal. Hopefully
- * we won't be able to reboot more frequently than 1MHz for the
- * forseeable future :) */
-
- do_gettimeofday(&tv);
-
- ksocknal_data.ksnd_incarnation =
- (((__u64)tv.tv_sec) * 1000000) + tv.tv_usec;
-}
-
-int
-ksocknal_api_startup (nal_t *nal, ptl_pid_t requested_pid,
- ptl_ni_limits_t *requested_limits,
- ptl_ni_limits_t *actual_limits)
-{
- ptl_process_id_t process_id;
- int pkmem = atomic_read(&portal_kmemory);
- int rc;
- int i;
- int j;
-
- LASSERT (nal == &ksocknal_api);
-
- if (nal->nal_refct != 0) {
- if (actual_limits != NULL)
- *actual_limits = ksocknal_lib.libnal_ni.ni_actual_limits;
- /* This module got the first ref */
- PORTAL_MODULE_USE;
- return (PTL_OK);
- }
+ struct ksock_sched *sched;
+ int rc;
+ int i;
LASSERT (ksocknal_data.ksnd_init == SOCKNAL_INIT_NOTHING);
+ LASSERT (ksocknal_data.ksnd_nnets == 0);
memset (&ksocknal_data, 0, sizeof (ksocknal_data)); /* zero pointers */
- ksocknal_init_incarnation();
-
ksocknal_data.ksnd_peer_hash_size = SOCKNAL_PEER_HASH_SIZE;
- PORTAL_ALLOC (ksocknal_data.ksnd_peers,
- sizeof (struct list_head) * ksocknal_data.ksnd_peer_hash_size);
+ LIBCFS_ALLOC(ksocknal_data.ksnd_peers,
+ sizeof(struct list_head) *
+ ksocknal_data.ksnd_peer_hash_size);
if (ksocknal_data.ksnd_peers == NULL)
- return (-ENOMEM);
+ return -ENOMEM;
for (i = 0; i < ksocknal_data.ksnd_peer_hash_size; i++)
- CFS_INIT_LIST_HEAD(&ksocknal_data.ksnd_peers[i]);
-
- rwlock_init(&ksocknal_data.ksnd_global_lock);
+ INIT_LIST_HEAD(&ksocknal_data.ksnd_peers[i]);
+
+ rwlock_init(&ksocknal_data.ksnd_global_lock);
+ INIT_LIST_HEAD(&ksocknal_data.ksnd_nets);
+
+ spin_lock_init(&ksocknal_data.ksnd_reaper_lock);
+ INIT_LIST_HEAD(&ksocknal_data.ksnd_enomem_conns);
+ INIT_LIST_HEAD(&ksocknal_data.ksnd_zombie_conns);
+ INIT_LIST_HEAD(&ksocknal_data.ksnd_deathrow_conns);
+ init_waitqueue_head(&ksocknal_data.ksnd_reaper_waitq);
+
+ spin_lock_init(&ksocknal_data.ksnd_connd_lock);
+ INIT_LIST_HEAD(&ksocknal_data.ksnd_connd_connreqs);
+ INIT_LIST_HEAD(&ksocknal_data.ksnd_connd_routes);
+ init_waitqueue_head(&ksocknal_data.ksnd_connd_waitq);
+
+ spin_lock_init(&ksocknal_data.ksnd_tx_lock);
+ INIT_LIST_HEAD(&ksocknal_data.ksnd_idle_noop_txs);
+
+ /* NB memset above zeros whole of ksocknal_data */
+
+ /* flag lists/ptrs/locks initialised */
+ ksocknal_data.ksnd_init = SOCKNAL_INIT_DATA;
+ try_module_get(THIS_MODULE);
+
+ /* Create a scheduler block per available CPT */
+ ksocknal_data.ksnd_schedulers = cfs_percpt_alloc(lnet_cpt_table(),
+ sizeof(*sched));
+ if (ksocknal_data.ksnd_schedulers == NULL)
+ goto failed;
+
+ cfs_percpt_for_each(sched, i, ksocknal_data.ksnd_schedulers) {
+ int nthrs;
+
+ /*
+ * make sure not to allocate more threads than there are
+ * cores/CPUs in teh CPT
+ */
+ nthrs = cfs_cpt_weight(lnet_cpt_table(), i);
+ if (*ksocknal_tunables.ksnd_nscheds > 0) {
+ nthrs = min(nthrs, *ksocknal_tunables.ksnd_nscheds);
+ } else {
+ /*
+ * max to half of CPUs, assume another half should be
+ * reserved for upper layer modules
+ */
+ nthrs = min(max(SOCKNAL_NSCHEDS, nthrs >> 1), nthrs);
+ }
+
+ sched->kss_nthreads_max = nthrs;
+ sched->kss_cpt = i;
+
+ spin_lock_init(&sched->kss_lock);
+ INIT_LIST_HEAD(&sched->kss_rx_conns);
+ INIT_LIST_HEAD(&sched->kss_tx_conns);
+ INIT_LIST_HEAD(&sched->kss_zombie_noop_txs);
+ init_waitqueue_head(&sched->kss_waitq);
+ }
+
+ ksocknal_data.ksnd_connd_starting = 0;
+ ksocknal_data.ksnd_connd_failed_stamp = 0;
+ ksocknal_data.ksnd_connd_starting_stamp = ktime_get_real_seconds();
+ /* must have at least 2 connds to remain responsive to accepts while
+ * connecting */
+ if (*ksocknal_tunables.ksnd_nconnds < SOCKNAL_CONND_RESV + 1)
+ *ksocknal_tunables.ksnd_nconnds = SOCKNAL_CONND_RESV + 1;
+
+ if (*ksocknal_tunables.ksnd_nconnds_max <
+ *ksocknal_tunables.ksnd_nconnds) {
+ ksocknal_tunables.ksnd_nconnds_max =
+ ksocknal_tunables.ksnd_nconnds;
+ }
+
+ for (i = 0; i < *ksocknal_tunables.ksnd_nconnds; i++) {
+ char name[16];
+ spin_lock_bh(&ksocknal_data.ksnd_connd_lock);
+ ksocknal_data.ksnd_connd_starting++;
+ spin_unlock_bh(&ksocknal_data.ksnd_connd_lock);
+
+
+ snprintf(name, sizeof(name), "socknal_cd%02d", i);
+ rc = ksocknal_thread_start(ksocknal_connd,
+ (void *)((uintptr_t)i), name);
+ if (rc != 0) {
+ spin_lock_bh(&ksocknal_data.ksnd_connd_lock);
+ ksocknal_data.ksnd_connd_starting--;
+ spin_unlock_bh(&ksocknal_data.ksnd_connd_lock);
+ CERROR("Can't spawn socknal connd: %d\n", rc);
+ goto failed;
+ }
+ }
- spin_lock_init(&ksocknal_data.ksnd_small_fmp.fmp_lock);
- CFS_INIT_LIST_HEAD(&ksocknal_data.ksnd_small_fmp.fmp_idle_fmbs);
- CFS_INIT_LIST_HEAD(&ksocknal_data.ksnd_small_fmp.fmp_blocked_conns);
- ksocknal_data.ksnd_small_fmp.fmp_buff_pages = SOCKNAL_SMALL_FWD_PAGES;
+ rc = ksocknal_thread_start(ksocknal_reaper, NULL, "socknal_reaper");
+ if (rc != 0) {
+ CERROR ("Can't spawn socknal reaper: %d\n", rc);
+ goto failed;
+ }
- spin_lock_init(&ksocknal_data.ksnd_large_fmp.fmp_lock);
- CFS_INIT_LIST_HEAD(&ksocknal_data.ksnd_large_fmp.fmp_idle_fmbs);
- CFS_INIT_LIST_HEAD(&ksocknal_data.ksnd_large_fmp.fmp_blocked_conns);
- ksocknal_data.ksnd_large_fmp.fmp_buff_pages = SOCKNAL_LARGE_FWD_PAGES;
+ /* flag everything initialised */
+ ksocknal_data.ksnd_init = SOCKNAL_INIT_ALL;
- spin_lock_init (&ksocknal_data.ksnd_reaper_lock);
- CFS_INIT_LIST_HEAD (&ksocknal_data.ksnd_enomem_conns);
- CFS_INIT_LIST_HEAD (&ksocknal_data.ksnd_zombie_conns);
- CFS_INIT_LIST_HEAD (&ksocknal_data.ksnd_deathrow_conns);
- cfs_waitq_init(&ksocknal_data.ksnd_reaper_waitq);
+ return 0;
- spin_lock_init (&ksocknal_data.ksnd_autoconnectd_lock);
- CFS_INIT_LIST_HEAD (&ksocknal_data.ksnd_autoconnectd_routes);
- cfs_waitq_init(&ksocknal_data.ksnd_autoconnectd_waitq);
+ failed:
+ ksocknal_base_shutdown();
+ return -ENETDOWN;
+}
- /* NB memset above zeros whole of ksocknal_data, including
- * ksocknal_data.ksnd_irqinfo[all].ksni_valid */
+static void
+ksocknal_debug_peerhash(struct lnet_ni *ni)
+{
+ struct ksock_peer_ni *peer_ni = NULL;
+ struct list_head *tmp;
+ int i;
- /* flag lists/ptrs/locks initialised */
- ksocknal_data.ksnd_init = SOCKNAL_INIT_DATA;
+ read_lock(&ksocknal_data.ksnd_global_lock);
- ksocknal_data.ksnd_nschedulers = ksocknal_nsched();
- PORTAL_ALLOC(ksocknal_data.ksnd_schedulers,
- sizeof(ksock_sched_t) * ksocknal_data.ksnd_nschedulers);
- if (ksocknal_data.ksnd_schedulers == NULL) {
- ksocknal_api_shutdown (nal);
- return (-ENOMEM);
- }
+ for (i = 0; i < ksocknal_data.ksnd_peer_hash_size; i++) {
+ list_for_each(tmp, &ksocknal_data.ksnd_peers[i]) {
+ peer_ni = list_entry(tmp, struct ksock_peer_ni, ksnp_list);
- for (i = 0; i < ksocknal_data.ksnd_nschedulers; i++) {
- ksock_sched_t *kss = &ksocknal_data.ksnd_schedulers[i];
+ if (peer_ni->ksnp_ni == ni) break;
- spin_lock_init (&kss->kss_lock);
- CFS_INIT_LIST_HEAD (&kss->kss_rx_conns);
- CFS_INIT_LIST_HEAD (&kss->kss_tx_conns);
-#if SOCKNAL_ZC
- CFS_INIT_LIST_HEAD (&kss->kss_zctxdone_list);
-#endif
- cfs_waitq_init (&kss->kss_waitq);
+ peer_ni = NULL;
+ }
}
- /* NB we have to wait to be told our true NID... */
- process_id.pid = requested_pid;
- process_id.nid = 0;
+ if (peer_ni != NULL) {
+ struct ksock_route *route;
+ struct ksock_conn *conn;
+
+ CWARN ("Active peer_ni on shutdown: %s, ref %d, scnt %d, "
+ "closing %d, accepting %d, err %d, zcookie %llu, "
+ "txq %d, zc_req %d\n", libcfs_id2str(peer_ni->ksnp_id),
+ atomic_read(&peer_ni->ksnp_refcount),
+ peer_ni->ksnp_sharecount, peer_ni->ksnp_closing,
+ peer_ni->ksnp_accepting, peer_ni->ksnp_error,
+ peer_ni->ksnp_zc_next_cookie,
+ !list_empty(&peer_ni->ksnp_tx_queue),
+ !list_empty(&peer_ni->ksnp_zc_req_list));
+
+ list_for_each(tmp, &peer_ni->ksnp_routes) {
+ route = list_entry(tmp, struct ksock_route, ksnr_list);
+ CWARN ("Route: ref %d, schd %d, conn %d, cnted %d, "
+ "del %d\n", atomic_read(&route->ksnr_refcount),
+ route->ksnr_scheduled, route->ksnr_connecting,
+ route->ksnr_connected, route->ksnr_deleted);
+ }
+
+ list_for_each(tmp, &peer_ni->ksnp_conns) {
+ conn = list_entry(tmp, struct ksock_conn, ksnc_list);
+ CWARN ("Conn: ref %d, sref %d, t %d, c %d\n",
+ atomic_read(&conn->ksnc_conn_refcount),
+ atomic_read(&conn->ksnc_sock_refcount),
+ conn->ksnc_type, conn->ksnc_closing);
+ }
+ }
+
+ read_unlock(&ksocknal_data.ksnd_global_lock);
+ return;
+}
+
+void
+ksocknal_shutdown(struct lnet_ni *ni)
+{
+ struct ksock_net *net = ni->ni_data;
+ struct lnet_process_id anyid = {
+ .nid = LNET_NID_ANY,
+ .pid = LNET_PID_ANY,
+ };
+ int i;
- rc = lib_init(&ksocknal_lib, nal, process_id,
- requested_limits, actual_limits);
- if (rc != PTL_OK) {
- CERROR("lib_init failed: error %d\n", rc);
- ksocknal_api_shutdown (nal);
- return (rc);
- }
+ LASSERT(ksocknal_data.ksnd_init == SOCKNAL_INIT_ALL);
+ LASSERT(ksocknal_data.ksnd_nnets > 0);
- ksocknal_data.ksnd_init = SOCKNAL_INIT_LIB; // flag lib_init() called
+ spin_lock_bh(&net->ksnn_lock);
+ net->ksnn_shutdown = 1; /* prevent new peers */
+ spin_unlock_bh(&net->ksnn_lock);
- for (i = 0; i < ksocknal_data.ksnd_nschedulers; i++) {
- rc = ksocknal_thread_start (ksocknal_scheduler,
- &ksocknal_data.ksnd_schedulers[i]);
- if (rc != 0) {
- CERROR("Can't spawn socknal scheduler[%d]: %d\n",
- i, rc);
- ksocknal_api_shutdown (nal);
- return (rc);
- }
- }
+ /* Delete all peers */
+ ksocknal_del_peer(ni, anyid, 0);
- for (i = 0; i < SOCKNAL_N_AUTOCONNECTD; i++) {
- rc = ksocknal_thread_start (ksocknal_autoconnectd, (void *)((long)i));
- if (rc != 0) {
- CERROR("Can't spawn socknal autoconnectd: %d\n", rc);
- ksocknal_api_shutdown (nal);
- return (rc);
- }
- }
+ /* Wait for all peer_ni state to clean up */
+ i = 2;
+ spin_lock_bh(&net->ksnn_lock);
+ while (net->ksnn_npeers != 0) {
+ spin_unlock_bh(&net->ksnn_lock);
- rc = ksocknal_thread_start (ksocknal_reaper, NULL);
- if (rc != 0) {
- CERROR ("Can't spawn socknal reaper: %d\n", rc);
- ksocknal_api_shutdown (nal);
- return (rc);
- }
+ i++;
+ CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET, /* power of 2? */
+ "waiting for %d peers to disconnect\n",
+ net->ksnn_npeers);
+ set_current_state(TASK_UNINTERRUPTIBLE);
+ schedule_timeout(cfs_time_seconds(1));
- rc = kpr_register(&ksocknal_data.ksnd_router,
- &ksocknal_router_interface);
- if (rc != 0) {
- CDEBUG(D_NET, "Can't initialise routing interface "
- "(rc = %d): not routing\n", rc);
- } else {
- /* Only allocate forwarding buffers if there's a router */
+ ksocknal_debug_peerhash(ni);
- for (i = 0; i < (SOCKNAL_SMALL_FWD_NMSGS +
- SOCKNAL_LARGE_FWD_NMSGS); i++) {
- ksock_fmb_t *fmb;
- ksock_fmb_pool_t *pool;
+ spin_lock_bh(&net->ksnn_lock);
+ }
+ spin_unlock_bh(&net->ksnn_lock);
+ for (i = 0; i < net->ksnn_ninterfaces; i++) {
+ LASSERT (net->ksnn_interfaces[i].ksni_npeers == 0);
+ LASSERT (net->ksnn_interfaces[i].ksni_nroutes == 0);
+ }
- if (i < SOCKNAL_SMALL_FWD_NMSGS)
- pool = &ksocknal_data.ksnd_small_fmp;
- else
- pool = &ksocknal_data.ksnd_large_fmp;
+ list_del(&net->ksnn_list);
+ LIBCFS_FREE(net, sizeof(*net));
- PORTAL_ALLOC(fmb, offsetof(ksock_fmb_t,
- fmb_kiov[pool->fmp_buff_pages]));
- if (fmb == NULL) {
- ksocknal_api_shutdown(nal);
- return (-ENOMEM);
- }
+ ksocknal_data.ksnd_nnets--;
+ if (ksocknal_data.ksnd_nnets == 0)
+ ksocknal_base_shutdown();
+}
- fmb->fmb_pool = pool;
+static int
+ksocknal_enumerate_interfaces(struct ksock_net *net, char *iname)
+{
+ struct net_device *dev;
+
+ rtnl_lock();
+ for_each_netdev(&init_net, dev) {
+ /* The iname specified by an user land configuration can
+ * map to an ifa_label so always treat iname as an ifa_label.
+ * If iname is NULL then fall back to the net device name.
+ */
+ const char *name = iname ? iname : dev->name;
+ struct in_device *in_dev;
+
+ if (strcmp(dev->name, "lo") == 0) /* skip the loopback IF */
+ continue;
+
+ if (!(dev_get_flags(dev) & IFF_UP)) {
+ CWARN("Ignoring interface %s (down)\n", dev->name);
+ continue;
+ }
+
+ in_dev = __in_dev_get_rtnl(dev);
+ if (!in_dev) {
+ CWARN("Interface %s has no IPv4 status.\n", dev->name);
+ continue;
+ }
+
+ for_ifa(in_dev)
+ if (strcmp(name, ifa->ifa_label) == 0) {
+ int idx = net->ksnn_ninterfaces;
+ struct ksock_interface *ksi;
+
+ if (idx >= ARRAY_SIZE(net->ksnn_interfaces)) {
+ rtnl_unlock();
+ return -E2BIG;
+ }
+
+ ksi = &net->ksnn_interfaces[idx];
+ ksi->ksni_ipaddr = ntohl(ifa->ifa_local);
+ ksi->ksni_netmask = ifa->ifa_mask;
+ strlcpy(ksi->ksni_name,
+ name, sizeof(ksi->ksni_name));
+ net->ksnn_ninterfaces++;
+ break;
+ }
+ endfor_ifa(in_dev);
+ }
+ rtnl_unlock();
+
+ if (net->ksnn_ninterfaces == 0)
+ CERROR("Can't find any usable interfaces\n");
+
+ return net->ksnn_ninterfaces > 0 ? 0 : -ENOENT;
+}
- for (j = 0; j < pool->fmp_buff_pages; j++) {
- fmb->fmb_kiov[j].kiov_page = cfs_alloc_page(CFS_ALLOC_STD);
+static int
+ksocknal_search_new_ipif(struct ksock_net *net)
+{
+ int new_ipif = 0;
+ int i;
+
+ for (i = 0; i < net->ksnn_ninterfaces; i++) {
+ char *ifnam = &net->ksnn_interfaces[i].ksni_name[0];
+ char *colon = strchr(ifnam, ':');
+ int found = 0;
+ struct ksock_net *tmp;
+ int j;
+
+ if (colon != NULL) /* ignore alias device */
+ *colon = 0;
+
+ list_for_each_entry(tmp, &ksocknal_data.ksnd_nets,
+ ksnn_list) {
+ for (j = 0; !found && j < tmp->ksnn_ninterfaces; j++) {
+ char *ifnam2 = &tmp->ksnn_interfaces[j].\
+ ksni_name[0];
+ char *colon2 = strchr(ifnam2, ':');
+
+ if (colon2 != NULL)
+ *colon2 = 0;
+
+ found = strcmp(ifnam, ifnam2) == 0;
+ if (colon2 != NULL)
+ *colon2 = ':';
+ }
+ if (found)
+ break;
+ }
+
+ new_ipif += !found;
+ if (colon != NULL)
+ *colon = ':';
+ }
+
+ return new_ipif;
+}
- if (fmb->fmb_kiov[j].kiov_page == NULL) {
- ksocknal_api_shutdown (nal);
- return (-ENOMEM);
- }
+static int
+ksocknal_start_schedulers(struct ksock_sched *sched)
+{
+ int nthrs;
+ int rc = 0;
+ int i;
+
+ if (sched->kss_nthreads == 0) {
+ if (*ksocknal_tunables.ksnd_nscheds > 0) {
+ nthrs = sched->kss_nthreads_max;
+ } else {
+ nthrs = cfs_cpt_weight(lnet_cpt_table(),
+ sched->kss_cpt);
+ nthrs = min(max(SOCKNAL_NSCHEDS, nthrs >> 1), nthrs);
+ nthrs = min(SOCKNAL_NSCHEDS_HIGH, nthrs);
+ }
+ nthrs = min(nthrs, sched->kss_nthreads_max);
+ } else {
+ LASSERT(sched->kss_nthreads <= sched->kss_nthreads_max);
+ /* increase two threads if there is new interface */
+ nthrs = min(2, sched->kss_nthreads_max - sched->kss_nthreads);
+ }
+
+ for (i = 0; i < nthrs; i++) {
+ long id;
+ char name[20];
+
+ id = KSOCK_THREAD_ID(sched->kss_cpt, sched->kss_nthreads + i);
+ snprintf(name, sizeof(name), "socknal_sd%02d_%02d",
+ sched->kss_cpt, (int)KSOCK_THREAD_SID(id));
+
+ rc = ksocknal_thread_start(ksocknal_scheduler,
+ (void *)id, name);
+ if (rc == 0)
+ continue;
+
+ CERROR("Can't spawn thread %d for scheduler[%d]: %d\n",
+ sched->kss_cpt, (int) KSOCK_THREAD_SID(id), rc);
+ break;
+ }
+
+ sched->kss_nthreads += i;
+ return rc;
+}
- LASSERT(cfs_page_address(fmb->fmb_kiov[j].kiov_page) != NULL);
- }
+static int
+ksocknal_net_start_threads(struct ksock_net *net, __u32 *cpts, int ncpts)
+{
+ int newif = ksocknal_search_new_ipif(net);
+ int rc;
+ int i;
- list_add(&fmb->fmb_list, &pool->fmp_idle_fmbs);
- }
- }
+ if (ncpts > 0 && ncpts > cfs_cpt_number(lnet_cpt_table()))
+ return -EINVAL;
- rc = libcfs_nal_cmd_register(SOCKNAL, &ksocknal_cmd, NULL);
- if (rc != 0) {
- CERROR ("Can't initialise command interface (rc = %d)\n", rc);
- ksocknal_api_shutdown (nal);
- return (rc);
- }
+ for (i = 0; i < ncpts; i++) {
+ struct ksock_sched *sched;
+ int cpt = (cpts == NULL) ? i : cpts[i];
- /* flag everything initialised */
- ksocknal_data.ksnd_init = SOCKNAL_INIT_ALL;
+ LASSERT(cpt < cfs_cpt_number(lnet_cpt_table()));
+ sched = ksocknal_data.ksnd_schedulers[cpt];
- printk(KERN_INFO "Lustre: Routing socket NAL loaded "
- "(Routing %s, initial mem %d, incarnation "LPX64")\n",
- kpr_routing (&ksocknal_data.ksnd_router) ?
- "enabled" : "disabled", pkmem, ksocknal_data.ksnd_incarnation);
+ if (!newif && sched->kss_nthreads > 0)
+ continue;
- return (0);
+ rc = ksocknal_start_schedulers(sched);
+ if (rc != 0)
+ return rc;
+ }
+ return 0;
}
-void __exit
-ksocknal_module_fini (void)
+int
+ksocknal_startup(struct lnet_ni *ni)
{
-#ifdef CONFIG_SYSCTL
- if (ksocknal_tunables.ksnd_sysctl != NULL)
- unregister_sysctl_table (ksocknal_tunables.ksnd_sysctl);
-#endif
- PtlNIFini(ksocknal_ni);
+ struct ksock_net *net;
+ struct lnet_ioctl_config_lnd_cmn_tunables *net_tunables;
+ int rc;
+ int i;
+ struct net_device *net_dev;
+ int node_id;
- ptl_unregister_nal(SOCKNAL);
+ LASSERT (ni->ni_net->net_lnd == &the_ksocklnd);
+
+ if (ksocknal_data.ksnd_init == SOCKNAL_INIT_NOTHING) {
+ rc = ksocknal_base_startup();
+ if (rc != 0)
+ return rc;
+ }
+
+ LIBCFS_ALLOC(net, sizeof(*net));
+ if (net == NULL)
+ goto fail_0;
+
+ spin_lock_init(&net->ksnn_lock);
+ net->ksnn_incarnation = ktime_get_real_ns();
+ ni->ni_data = net;
+ net_tunables = &ni->ni_net->net_tunables;
+
+ if (net_tunables->lct_peer_timeout == -1)
+ net_tunables->lct_peer_timeout =
+ *ksocknal_tunables.ksnd_peertimeout;
+
+ if (net_tunables->lct_max_tx_credits == -1)
+ net_tunables->lct_max_tx_credits =
+ *ksocknal_tunables.ksnd_credits;
+
+ if (net_tunables->lct_peer_tx_credits == -1)
+ net_tunables->lct_peer_tx_credits =
+ *ksocknal_tunables.ksnd_peertxcredits;
+
+ if (net_tunables->lct_peer_tx_credits >
+ net_tunables->lct_max_tx_credits)
+ net_tunables->lct_peer_tx_credits =
+ net_tunables->lct_max_tx_credits;
+
+ if (net_tunables->lct_peer_rtr_credits == -1)
+ net_tunables->lct_peer_rtr_credits =
+ *ksocknal_tunables.ksnd_peerrtrcredits;
+
+ if (!ni->ni_interfaces[0]) {
+ rc = ksocknal_enumerate_interfaces(net, NULL);
+ if (rc < 0)
+ goto fail_1;
+ } else {
+ /* Before Multi-Rail ksocklnd would manage
+ * multiple interfaces with its own tcp bonding.
+ * If we encounter an old configuration using
+ * this tcp bonding approach then we need to
+ * handle more than one ni_interfaces.
+ *
+ * In Multi-Rail configuration only ONE ni_interface
+ * should exist. Each IP alias should be mapped to
+ * each 'struct net_ni'.
+ */
+ for (i = 0; i < LNET_INTERFACES_NUM; i++) {
+ int j;
+
+ if (!ni->ni_interfaces[i])
+ break;
+
+ for (j = 0; j < net->ksnn_ninterfaces; j++) {
+ struct ksock_interface *ksi;
+
+ ksi = &net->ksnn_interfaces[j];
+
+ if (strcmp(ni->ni_interfaces[i],
+ ksi->ksni_name) == 0) {
+ CERROR("found duplicate %s\n",
+ ksi->ksni_name);
+ rc = -EEXIST;
+ goto fail_1;
+ }
+ }
+
+ rc = ksocknal_enumerate_interfaces(net, ni->ni_interfaces[i]);
+ if (rc < 0)
+ goto fail_1;
+ }
+ }
+
+ net_dev = dev_get_by_name(&init_net,
+ net->ksnn_interfaces[0].ksni_name);
+ if (net_dev != NULL) {
+ node_id = dev_to_node(&net_dev->dev);
+ ni->ni_dev_cpt = cfs_cpt_of_node(lnet_cpt_table(), node_id);
+ dev_put(net_dev);
+ } else {
+ ni->ni_dev_cpt = CFS_CPT_ANY;
+ }
+
+ /* call it before add it to ksocknal_data.ksnd_nets */
+ rc = ksocknal_net_start_threads(net, ni->ni_cpts, ni->ni_ncpts);
+ if (rc != 0)
+ goto fail_1;
+
+ ni->ni_nid = LNET_MKNID(LNET_NIDNET(ni->ni_nid),
+ net->ksnn_interfaces[0].ksni_ipaddr);
+ list_add(&net->ksnn_list, &ksocknal_data.ksnd_nets);
+
+ ksocknal_data.ksnd_nnets++;
+
+ return 0;
+
+ fail_1:
+ LIBCFS_FREE(net, sizeof(*net));
+ fail_0:
+ if (ksocknal_data.ksnd_nnets == 0)
+ ksocknal_base_shutdown();
+
+ return -ENETDOWN;
}
-extern cfs_sysctl_table_t ksocknal_top_ctl_table[];
-int __init
-ksocknal_module_init (void)
+static void __exit ksocklnd_exit(void)
{
- int rc;
-
- /* packet descriptor must fit in a router descriptor's scratchpad */
- LASSERT(sizeof (ksock_tx_t) <= sizeof (kprfd_scratch_t));
- /* the following must be sizeof(int) for proc_dointvec() */
- LASSERT(sizeof (ksocknal_tunables.ksnd_io_timeout) == sizeof (int));
- LASSERT(sizeof (ksocknal_tunables.ksnd_eager_ack) == sizeof (int));
- LASSERT(sizeof (ksocknal_tunables.ksnd_typed_conns) == sizeof (int));
- LASSERT(sizeof (ksocknal_tunables.ksnd_min_bulk) == sizeof (int));
- LASSERT(sizeof (ksocknal_tunables.ksnd_buffer_size) == sizeof (int));
- LASSERT(sizeof (ksocknal_tunables.ksnd_nagle) == sizeof (int));
- LASSERT(sizeof (ksocknal_tunables.ksnd_keepalive_idle) == sizeof (int));
- LASSERT(sizeof (ksocknal_tunables.ksnd_keepalive_count) == sizeof (int));
- LASSERT(sizeof (ksocknal_tunables.ksnd_keepalive_intvl) == sizeof (int));
-#if CPU_AFFINITY
- LASSERT(sizeof (ksocknal_tunables.ksnd_irq_affinity) == sizeof (int));
-#endif
-#if SOCKNAL_ZC
- LASSERT(sizeof (ksocknal_tunables.ksnd_zc_min_frag) == sizeof (int));
-#endif
- /* check ksnr_connected/connecting field large enough */
- LASSERT(SOCKNAL_CONN_NTYPES <= 4);
-
- ksocknal_api.nal_ni_init = ksocknal_api_startup;
- ksocknal_api.nal_ni_fini = ksocknal_api_shutdown;
-
- /* Initialise dynamic tunables to defaults once only */
- ksocknal_tunables.ksnd_io_timeout = SOCKNAL_IO_TIMEOUT;
- ksocknal_tunables.ksnd_eager_ack = SOCKNAL_EAGER_ACK;
- ksocknal_tunables.ksnd_typed_conns = SOCKNAL_TYPED_CONNS;
- ksocknal_tunables.ksnd_min_bulk = SOCKNAL_MIN_BULK;
- ksocknal_tunables.ksnd_buffer_size = SOCKNAL_BUFFER_SIZE;
- ksocknal_tunables.ksnd_nagle = SOCKNAL_NAGLE;
- ksocknal_tunables.ksnd_keepalive_idle = SOCKNAL_KEEPALIVE_IDLE;
- ksocknal_tunables.ksnd_keepalive_count = SOCKNAL_KEEPALIVE_COUNT;
- ksocknal_tunables.ksnd_keepalive_intvl = SOCKNAL_KEEPALIVE_INTVL;
-#if CPU_AFFINITY
- ksocknal_tunables.ksnd_irq_affinity = SOCKNAL_IRQ_AFFINITY;
-#endif
-#if SOCKNAL_ZC
- ksocknal_tunables.ksnd_zc_min_frag = SOCKNAL_ZC_MIN_FRAG;
-#endif
-
- rc = ptl_register_nal(SOCKNAL, &ksocknal_api);
- if (rc != PTL_OK) {
- CERROR("Can't register SOCKNAL: %d\n", rc);
- return (-ENOMEM); /* or something... */
- }
-
- /* Pure gateways want the NAL started up at module load time... */
- rc = PtlNIInit(SOCKNAL, LUSTRE_SRV_PTL_PID, NULL, NULL, &ksocknal_ni);
- if (rc != PTL_OK && rc != PTL_IFACE_DUP) {
- ptl_unregister_nal(SOCKNAL);
- return (-ENODEV);
- }
+ lnet_unregister_lnd(&the_ksocklnd);
+}
-#ifdef CONFIG_SYSCTL
- /* Press on regardless even if registering sysctl doesn't work */
- ksocknal_tunables.ksnd_sysctl =
- register_sysctl_table (ksocknal_top_ctl_table, 0);
-#endif
- return (0);
+static int __init ksocklnd_init(void)
+{
+ int rc;
+
+ /* check ksnr_connected/connecting field large enough */
+ CLASSERT(SOCKLND_CONN_NTYPES <= 4);
+ CLASSERT(SOCKLND_CONN_ACK == SOCKLND_CONN_BULK_IN);
+
+ /* initialize the_ksocklnd */
+ the_ksocklnd.lnd_type = SOCKLND;
+ the_ksocklnd.lnd_startup = ksocknal_startup;
+ the_ksocklnd.lnd_shutdown = ksocknal_shutdown;
+ the_ksocklnd.lnd_ctl = ksocknal_ctl;
+ the_ksocklnd.lnd_send = ksocknal_send;
+ the_ksocklnd.lnd_recv = ksocknal_recv;
+ the_ksocklnd.lnd_notify = ksocknal_notify;
+ the_ksocklnd.lnd_query = ksocknal_query;
+ the_ksocklnd.lnd_accept = ksocknal_accept;
+
+ rc = ksocknal_tunables_init();
+ if (rc != 0)
+ return rc;
+
+ lnet_register_lnd(&the_ksocklnd);
+
+ return 0;
}
-MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");
-MODULE_DESCRIPTION("Kernel TCP Socket NAL v1.0.0");
+MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
+MODULE_DESCRIPTION("TCP Socket LNet Network Driver");
+MODULE_VERSION("2.8.0");
MODULE_LICENSE("GPL");
-cfs_module(ksocknal, "1.0.0", ksocknal_module_init, ksocknal_module_fini);
+module_init(ksocklnd_init);
+module_exit(ksocklnd_exit);
git://git.whamcloud.com / fs / lustre-release.git / blobdiff