-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
+/*
+ * GPL HEADER START
*
- * Copyright (C) 2006 Cluster File Systems, Inc.
- * Author: Eric Barton <eric@bartonsoftware.com>
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
- * This file is part of Lustre, http://www.lustre.org.
+ * 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.
*
- * Lustre 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).
*
- * Lustre 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.sun.com/software/products/lustre/docs/GPLv2.pdf
*
- * You should have received a copy of the GNU General Public License
- * along with Lustre; if not, write to the Free Software
- * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
*
+ * GPL HEADER END
+ */
+/*
+ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Use is subject to license terms.
+ *
+ * Copyright (c) 2012, Intel Corporation.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
+ *
+ * lnet/klnds/o2iblnd/o2iblnd_modparams.c
+ *
+ * Author: Eric Barton <eric@bartonsoftware.com>
*/
#include "o2iblnd.h"
static int cksum = 0;
CFS_MODULE_PARM(cksum, "i", int, 0644,
- "set non-zero to enable message (not RDMA) checksums");
+ "set non-zero to enable message (not RDMA) checksums");
static int timeout = 50;
CFS_MODULE_PARM(timeout, "i", int, 0644,
- "timeout (seconds)");
+ "timeout (seconds)");
-static int ntx = 256;
+/* Number of threads in each scheduler pool which is percpt,
+ * we will estimate reasonable value based on CPUs if it's set to zero. */
+static int nscheds;
+CFS_MODULE_PARM(nscheds, "i", int, 0444,
+ "number of threads in each scheduler pool");
+
+/* NB: this value is shared by all CPTs, it can grow at runtime */
+static int ntx = 512;
CFS_MODULE_PARM(ntx, "i", int, 0444,
- "# of message descriptors");
+ "# of message descriptors allocated for each pool");
-static int credits = 64;
+/* NB: this value is shared by all CPTs */
+static int credits = 256;
CFS_MODULE_PARM(credits, "i", int, 0444,
- "# concurrent sends");
+ "# concurrent sends");
static int peer_credits = 8;
CFS_MODULE_PARM(peer_credits, "i", int, 0444,
- "# concurrent sends to 1 peer");
+ "# concurrent sends to 1 peer");
+
+static int peer_credits_hiw = 0;
+CFS_MODULE_PARM(peer_credits_hiw, "i", int, 0444,
+ "when eagerly to return credits");
+
+static int peer_buffer_credits = 0;
+CFS_MODULE_PARM(peer_buffer_credits, "i", int, 0444,
+ "# per-peer router buffer credits");
+
+static int peer_timeout = 180;
+CFS_MODULE_PARM(peer_timeout, "i", int, 0444,
+ "Seconds without aliveness news to declare peer dead (<=0 to disable)");
static char *ipif_name = "ib0";
CFS_MODULE_PARM(ipif_name, "s", charp, 0444,
CFS_MODULE_PARM(ib_mtu, "i", int, 0444,
"IB MTU 256/512/1024/2048/4096");
-#if IBLND_MAP_ON_DEMAND
-static int concurrent_sends = IBLND_RX_MSGS;
-#else
-static int concurrent_sends = IBLND_MSG_QUEUE_SIZE;
-#endif
+static int concurrent_sends = 0;
CFS_MODULE_PARM(concurrent_sends, "i", int, 0444,
"send work-queue sizing");
-#if IBLND_MAP_ON_DEMAND
+static int map_on_demand = 0;
+CFS_MODULE_PARM(map_on_demand, "i", int, 0444,
+ "map on demand");
+
+/* NB: this value is shared by all CPTs, it can grow at runtime */
static int fmr_pool_size = 512;
CFS_MODULE_PARM(fmr_pool_size, "i", int, 0444,
- "size of the fmr pool (>= ntx)");
+ "size of fmr pool on each CPT (>= ntx / 4)");
+/* NB: this value is shared by all CPTs, it can grow at runtime */
static int fmr_flush_trigger = 384;
CFS_MODULE_PARM(fmr_flush_trigger, "i", int, 0444,
- "# dirty FMRs that triggers pool flush");
+ "# dirty FMRs that triggers pool flush");
static int fmr_cache = 1;
CFS_MODULE_PARM(fmr_cache, "i", int, 0444,
- "non-zero to enable FMR caching");
-#endif
+ "non-zero to enable FMR caching");
+
+/*
+ * 0: disable failover
+ * 1: enable failover if necessary
+ * 2: force to failover (for debug)
+ */
+static int dev_failover = 0;
+CFS_MODULE_PARM(dev_failover, "i", int, 0444,
+ "HCA failover for bonding (0 off, 1 on, other values reserved)");
+
+
+static int require_privileged_port = 0;
+CFS_MODULE_PARM(require_privileged_port, "i", int, 0644,
+ "require privileged port when accepting connection");
+
+static int use_privileged_port = 1;
+CFS_MODULE_PARM(use_privileged_port, "i", int, 0644,
+ "use privileged port when initiating connection");
kib_tunables_t kiblnd_tunables = {
+ .kib_dev_failover = &dev_failover,
.kib_service = &service,
.kib_cksum = &cksum,
.kib_timeout = &timeout,
.kib_keepalive = &keepalive,
.kib_ntx = &ntx,
.kib_credits = &credits,
- .kib_peercredits = &peer_credits,
+ .kib_peertxcredits = &peer_credits,
+ .kib_peercredits_hiw = &peer_credits_hiw,
+ .kib_peerrtrcredits = &peer_buffer_credits,
+ .kib_peertimeout = &peer_timeout,
.kib_default_ipif = &ipif_name,
.kib_retry_count = &retry_count,
.kib_rnr_retry_count = &rnr_retry_count,
.kib_concurrent_sends = &concurrent_sends,
.kib_ib_mtu = &ib_mtu,
-#if IBLND_MAP_ON_DEMAND
+ .kib_map_on_demand = &map_on_demand,
.kib_fmr_pool_size = &fmr_pool_size,
.kib_fmr_flush_trigger = &fmr_flush_trigger,
.kib_fmr_cache = &fmr_cache,
-#endif
+ .kib_require_priv_port = &require_privileged_port,
+ .kib_use_priv_port = &use_privileged_port,
+ .kib_nscheds = &nscheds
};
-#if CONFIG_SYSCTL && !CFS_SYSFS_MODULE_PARM
+#if defined(CONFIG_SYSCTL) && !CFS_SYSFS_MODULE_PARM
static char ipif_basename_space[32];
-static ctl_table kiblnd_ctl_table[] = {
- {1, "service", &service,
- sizeof(int), 0444, NULL, &proc_dointvec},
- {2, "cksum", &cksum,
- sizeof(int), 0644, NULL, &proc_dointvec},
- {3, "timeout", &timeout,
- sizeof(int), 0644, NULL, &proc_dointvec},
- {4, "ntx", &ntx,
- sizeof(int), 0444, NULL, &proc_dointvec},
- {5, "credits", &credits,
- sizeof(int), 0444, NULL, &proc_dointvec},
- {6, "peer_credits", &peer_credits,
- sizeof(int), 0444, NULL, &proc_dointvec},
- {7, "ipif_name", ipif_basename_space,
- sizeof(ipif_basename_space), 0444, NULL, &proc_dostring},
- {8, "retry_count", &retry_count,
- sizeof(int), 0644, NULL, &proc_dointvec},
- {9, "rnr_retry_count", &rnr_retry_count,
- sizeof(int), 0644, NULL, &proc_dointvec},
- {10, "keepalive", &keepalive,
- sizeof(int), 0644, NULL, &proc_dointvec},
- {11, "concurrent_sends", &concurrent_sends,
- sizeof(int), 0644, NULL, &proc_dointvec},
- {12, "ib_mtu", &ib_mtu,
- sizeof(int), 0444, NULL, &proc_dointvec},
-#if IBLND_MAP_ON_DEMAND
- {12, "fmr_pool_size", &fmr_pool_size,
- sizeof(int), 0444, NULL, &proc_dointvec},
- {13, "fmr_flush_trigger", &fmr_flush_trigger,
- sizeof(int), 0444, NULL, &proc_dointvec},
- {14, "fmr_cache", &fmr_cache,
- sizeof(int), 0444, NULL, &proc_dointvec},
-#endif
- {0}
+static struct ctl_table kiblnd_ctl_table[] = {
+ {
+ INIT_CTL_NAME
+ .procname = "service",
+ .data = &service,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "cksum",
+ .data = &cksum,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "timeout",
+ .data = &timeout,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "ntx",
+ .data = &ntx,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "credits",
+ .data = &credits,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "peer_credits",
+ .data = &peer_credits,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "peer_credits_hiw",
+ .data = &peer_credits_hiw,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "peer_buffer_credits",
+ .data = &peer_buffer_credits,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "peer_timeout",
+ .data = &peer_timeout,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "ipif_name",
+ .data = ipif_basename_space,
+ .maxlen = sizeof(ipif_basename_space),
+ .mode = 0444,
+ .proc_handler = &proc_dostring
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "retry_count",
+ .data = &retry_count,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "rnr_retry_count",
+ .data = &rnr_retry_count,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "keepalive",
+ .data = &keepalive,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "concurrent_sends",
+ .data = &concurrent_sends,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "ib_mtu",
+ .data = &ib_mtu,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "map_on_demand",
+ .data = &map_on_demand,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "fmr_pool_size",
+ .data = &fmr_pool_size,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "fmr_flush_trigger",
+ .data = &fmr_flush_trigger,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "fmr_cache",
+ .data = &fmr_cache,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ {
+ INIT_CTL_NAME
+ .procname = "dev_failover",
+ .data = &dev_failover,
+ .maxlen = sizeof(int),
+ .mode = 0444,
+ .proc_handler = &proc_dointvec
+ },
+ { 0 }
};
-static ctl_table kiblnd_top_ctl_table[] = {
- {203, "o2iblnd", NULL, 0, 0555, kiblnd_ctl_table},
- {0}
+static struct ctl_table kiblnd_top_ctl_table[] = {
+ {
+ INIT_CTL_NAME
+ .procname = "o2iblnd",
+ .data = NULL,
+ .maxlen = 0,
+ .mode = 0555,
+ .child = kiblnd_ctl_table
+ },
+ { 0 }
};
void
space[size-1] = 0;
}
-void
+static void
kiblnd_sysctl_init (void)
{
- kiblnd_initstrtunable(ipif_basename_space, ipif_name,
- sizeof(ipif_basename_space));
+ kiblnd_initstrtunable(ipif_basename_space, ipif_name,
+ sizeof(ipif_basename_space));
kiblnd_tunables.kib_sysctl =
- cfs_register_sysctl_table(kiblnd_top_ctl_table, 0);
+ register_sysctl_table(kiblnd_top_ctl_table);
if (kiblnd_tunables.kib_sysctl == NULL)
CWARN("Can't setup /proc tunables\n");
}
-void
+static void
kiblnd_sysctl_fini (void)
{
if (kiblnd_tunables.kib_sysctl != NULL)
- cfs_unregister_sysctl_table(kiblnd_tunables.kib_sysctl);
+ unregister_sysctl_table(kiblnd_tunables.kib_sysctl);
}
#else
-void
+static void
kiblnd_sysctl_init (void)
{
}
-void
+static void
kiblnd_sysctl_fini (void)
{
}
int
kiblnd_tunables_init (void)
{
- kiblnd_sysctl_init();
+ if (kiblnd_translate_mtu(*kiblnd_tunables.kib_ib_mtu) < 0) {
+ CERROR("Invalid ib_mtu %d, expected 256/512/1024/2048/4096\n",
+ *kiblnd_tunables.kib_ib_mtu);
+ return -EINVAL;
+ }
+
+ if (*kiblnd_tunables.kib_peertxcredits < IBLND_CREDITS_DEFAULT)
+ *kiblnd_tunables.kib_peertxcredits = IBLND_CREDITS_DEFAULT;
+
+ if (*kiblnd_tunables.kib_peertxcredits > IBLND_CREDITS_MAX)
+ *kiblnd_tunables.kib_peertxcredits = IBLND_CREDITS_MAX;
+
+ if (*kiblnd_tunables.kib_peertxcredits > *kiblnd_tunables.kib_credits)
+ *kiblnd_tunables.kib_peertxcredits = *kiblnd_tunables.kib_credits;
+
+ if (*kiblnd_tunables.kib_peercredits_hiw < *kiblnd_tunables.kib_peertxcredits / 2)
+ *kiblnd_tunables.kib_peercredits_hiw = *kiblnd_tunables.kib_peertxcredits / 2;
+
+ if (*kiblnd_tunables.kib_peercredits_hiw >= *kiblnd_tunables.kib_peertxcredits)
+ *kiblnd_tunables.kib_peercredits_hiw = *kiblnd_tunables.kib_peertxcredits - 1;
- if (*kiblnd_tunables.kib_concurrent_sends > IBLND_RX_MSGS)
- *kiblnd_tunables.kib_concurrent_sends = IBLND_RX_MSGS;
- if (*kiblnd_tunables.kib_concurrent_sends < IBLND_MSG_QUEUE_SIZE)
- *kiblnd_tunables.kib_concurrent_sends = IBLND_MSG_QUEUE_SIZE;
+ if (*kiblnd_tunables.kib_map_on_demand < 0 ||
+ *kiblnd_tunables.kib_map_on_demand > IBLND_MAX_RDMA_FRAGS)
+ *kiblnd_tunables.kib_map_on_demand = 0; /* disable map-on-demand */
- return 0;
+ if (*kiblnd_tunables.kib_map_on_demand == 1)
+ *kiblnd_tunables.kib_map_on_demand = 2; /* don't make sense to create map if only one fragment */
+
+ if (*kiblnd_tunables.kib_concurrent_sends == 0) {
+ if (*kiblnd_tunables.kib_map_on_demand > 0 &&
+ *kiblnd_tunables.kib_map_on_demand <= IBLND_MAX_RDMA_FRAGS / 8)
+ *kiblnd_tunables.kib_concurrent_sends = (*kiblnd_tunables.kib_peertxcredits) * 2;
+ else
+ *kiblnd_tunables.kib_concurrent_sends = (*kiblnd_tunables.kib_peertxcredits);
+ }
+
+ if (*kiblnd_tunables.kib_concurrent_sends > *kiblnd_tunables.kib_peertxcredits * 2)
+ *kiblnd_tunables.kib_concurrent_sends = *kiblnd_tunables.kib_peertxcredits * 2;
+
+ if (*kiblnd_tunables.kib_concurrent_sends < *kiblnd_tunables.kib_peertxcredits / 2)
+ *kiblnd_tunables.kib_concurrent_sends = *kiblnd_tunables.kib_peertxcredits / 2;
+
+ if (*kiblnd_tunables.kib_concurrent_sends < *kiblnd_tunables.kib_peertxcredits) {
+ CWARN("Concurrent sends %d is lower than message queue size: %d, "
+ "performance may drop slightly.\n",
+ *kiblnd_tunables.kib_concurrent_sends, *kiblnd_tunables.kib_peertxcredits);
+ }
+
+ kiblnd_sysctl_init();
+ return 0;
}
void
{
kiblnd_sysctl_fini();
}
-
-
-