Merge pull request #1628 from kmroz/wip-btl-tcp-ethtool-speed

btl/tcp: autodetect bandwidth and latency if unset by the user

configure.ac

@@ -577,7 +577,8 @@ AC_CHECK_HEADERS([alloca.h aio.h arpa/inet.h dirent.h \
     dlfcn.h execinfo.h err.h fcntl.h grp.h libgen.h \
     libutil.h memory.h netdb.h netinet/in.h netinet/tcp.h \
     poll.h pthread.h pty.h pwd.h sched.h \
-    strings.h stropts.h sys/fcntl.h sys/ipc.h sys/shm.h \
+    strings.h stropts.h linux/ethtool.h linux/sockios.h \
+    net/if.h sys/fcntl.h sys/ipc.h sys/shm.h \
     sys/ioctl.h sys/mman.h sys/param.h sys/queue.h \
     sys/resource.h sys/select.h sys/socket.h sys/sockio.h \
     sys/stat.h sys/statfs.h sys/statvfs.h sys/time.h sys/tree.h \
@@ -628,13 +629,28 @@ AC_CACHE_SAVE
 opal_show_title "Type tests"
 
 AC_CHECK_TYPES([socklen_t, struct sockaddr_in, struct sockaddr_in6,
-                struct sockaddr_storage],
+                struct sockaddr_storage, struct ifreq, struct ethtool_cmd],
                [], [], [AC_INCLUDES_DEFAULT
 #if HAVE_SYS_SOCKET_H
 #include <sys/socket.h>
 #endif
 #ifdef HAVE_NETINET_IN_H
 #include <netinet/in.h>
+#endif
+#ifdef HAVE_LINUX_ETHTOOL_H
+#include <linux/ethtool.h>
+#endif
+#ifdef HAVE_NET_IF_H
+#include <net/if.h>
+#endif])
+
+AC_CHECK_DECLS([ethtool_cmd_speed, SIOCETHTOOL],
+               [], [], [AC_INCLUDES_DEFAULT
+#ifdef HAVE_LINUX_ETHTOOL_H
+#include <linux/ethtool.h>
+#endif
+#ifdef HAVE_LINUX_SOCKIOS_H
+#include <linux/sockios.h>
 #endif])
 
 AC_CHECK_DECLS([AF_UNSPEC, PF_UNSPEC, AF_INET6, PF_INET6],

opal/mca/btl/tcp/btl_tcp_component.c

@@ -56,6 +56,7 @@
 #include <limits.h>
 
 #include "opal/mca/event/event.h"
+#include "opal/util/ethtool.h"
 #include "opal/util/if.h"
 #include "opal/util/output.h"
 #include "opal/util/argv.h"
@@ -85,6 +86,9 @@
 #include "opal/mca/common/cuda/common_cuda.h"
 #endif /* OPAL_CUDA_SUPPORT */
 
+#define MCA_BTL_TCP_BTL_BANDWIDTH 100
+#define MCA_BTL_TCP_BTL_LATENCY 100
+
 /*
  * Local functions
  */
@@ -320,8 +324,11 @@ static int mca_btl_tcp_component_register(void)
                                        MCA_BTL_FLAGS_HETEROGENEOUS_RDMA |
                                        MCA_BTL_FLAGS_SEND;
 
-    mca_btl_tcp_module.super.btl_bandwidth = 100;
+    /* Bandwidth and latency initially set to 0. May be overridden during
-    mca_btl_tcp_module.super.btl_latency = 100;
+     * mca_btl_tcp_create().
+     */
+    mca_btl_tcp_module.super.btl_bandwidth = 0;
+    mca_btl_tcp_module.super.btl_latency = 0;
 
     mca_btl_base_param_register(&mca_btl_tcp_component.super.btl_version,
                                 &mca_btl_tcp_module.super);
@@ -513,6 +520,27 @@ static int mca_btl_tcp_create(int if_kindex, const char* if_name)
         /* allow user to override/specify latency ranking */
         sprintf(param, "latency_%s:%d", if_name, i);
         mca_btl_tcp_param_register_uint(param, NULL, btl->super.btl_latency, OPAL_INFO_LVL_5, &btl->super.btl_latency);
+
+        /* Only attempt to auto-detect bandwidth and/or latency if it is 0.
+         *
+         * If detection fails to return anything other than 0, set a default
+         * bandwidth and latency.
+         */
+        if (0 == btl->super.btl_bandwidth) {
+            unsigned int speed = opal_ethtool_get_speed(if_name);
+            btl->super.btl_bandwidth = (speed == 0) ? MCA_BTL_TCP_BTL_BANDWIDTH : speed;
+            if (i > 0) {
+                btl->super.btl_bandwidth >>= 1;
+            }
+        }
+        /* We have no runtime btl latency detection mechanism. Just set a default. */
+        if (0 == btl->super.btl_latency) {
+            btl->super.btl_latency = MCA_BTL_TCP_BTL_LATENCY;
+            if (i > 0) {
+                btl->super.btl_latency <<= 1;
+            }
+        }
+
 #if 0 && OPAL_ENABLE_DEBUG
         BTL_OUTPUT(("interface %s instance %i: bandwidth %d latency %d\n", if_name, i,
                     btl->super.btl_bandwidth, btl->super.btl_latency));

opal/util/Makefile.am

@@ -43,6 +43,7 @@ headers = \
         cmd_line.h \
         crc.h \
         daemon_init.h \
+	ethtool.h \
         error.h \
         fd.h \
         few.h \
@@ -78,6 +79,7 @@ libopalutil_la_SOURCES = \
         cmd_line.c \
         crc.c \
         daemon_init.c \
+	ethtool.c \
         error.c \
         fd.c \
         few.c \

opal/util/ethtool.c

@@ -0,0 +1,83 @@
+/*
+ * Copyright (c) 2016 Karol Mroz.  All rights reserved.
+ * $COPYRIGHT$
+ *
+ * Additional copyrights may follow
+ *
+ * $HEADER$
+ */
+
+#include "opal_config.h"
+
+#include <string.h>
+#include <limits.h>
+
+#ifdef HAVE_UNISTD_H
+#include <unistd.h>
+#endif
+#ifdef HAVE_NET_IF_H
+#include <net/if.h>
+#endif
+#ifdef HAVE_LINUX_ETHTOOL_H
+#include <linux/ethtool.h>
+#endif
+#ifdef HAVE_SYS_IOCTL_H
+#include <sys/ioctl.h>
+#endif
+#ifdef HAVE_LINUX_SOCKIOS_H
+#include <linux/sockios.h>
+#endif
+
+#include "opal/util/ethtool.h"
+#include "opal/util/if.h"
+
+#if !defined(HAVE_DECL_ETHTOOL_CMD_SPEED)
+static inline unsigned int
+ethtool_cmd_speed(const struct ethtool_cmd *ep)
+{
+    return (ep->speed_hi << 16) | ep->speed;
+}
+#endif
+
+/*
+ * Obtain an appropriate bandwidth for the interface if_name. On Linux, we
+ * get this via an ioctl(). Elsewhere or in the error case, we return the
+ * speed as 0.
+ */
+unsigned int
+opal_ethtool_get_speed (const char *if_name)
+{
+    unsigned int speed = 0;
+
+#if defined(HAVE_DECL_SIOCETHTOOL) && defined(HAVE_STRUCT_IFREQ) && defined(HAVE_STRUCT_ETHTOOL_CMD)
+    int sockfd;
+    struct ifreq ifr;
+    struct ethtool_cmd edata = {
+        .cmd = ETHTOOL_GSET,
+    };
+
+    sockfd = socket(PF_INET, SOCK_DGRAM, 0);
+    if (sockfd < 0) {
+        goto out;
+    }
+
+    memset(&ifr, 0, sizeof(struct ifreq));
+    strncpy(ifr.ifr_name, if_name, IF_NAMESIZE);
+    ifr.ifr_data = (char *)&edata;
+
+    if (ioctl(sockfd, SIOCETHTOOL, &ifr) < 0) {
+        goto out;
+    }
+
+    speed = ethtool_cmd_speed(&edata);
+    if (UINT_MAX == speed) {
+        speed = 0;
+    }
+
+out:
+    close(sockfd);
+    return speed;
+#else
+    return speed;
+#endif
+}

opal/util/ethtool.h

@@ -0,0 +1,20 @@
+/*
+ * Copyright (c) 2016 Karol Mroz.  All rights reserved.
+ * $COPYRIGHT$
+ *
+ * Additional copyrights may follow
+ *
+ * $HEADER$
+ */
+
+#ifndef OPAL_ETHTOOL_H
+#define OPAL_ETHTOOL_H
+
+/*
+ * Obtain an appropriate bandwidth for the interface if_name. On Linux, we
+ * get this via an ioctl(). Elsewhere or in the error case, we return the
+ * speed as 0.
+ */
+unsigned int opal_ethtool_get_speed(const char *if_name);
+
+#endif