prov/udp: detect and use MTU to set max_msg_size and inject_size

For each interface detected by the udp provider, determine the MTU of
the interface, and use that value to set the max_msg_size field of the
fi_ep_attr and fi_tx_attr values of the fi_info element. When the MTU
cannot be determined, the MTU value assumed by previous code
versions (1500) is used.

include/freebsd/osd.h

@@ -76,6 +76,11 @@ static inline size_t ofi_ifaddr_get_speed(struct ifaddrs *ifa)
 	return 0;
 }
 
+static inline int ofi_ifaddr_get_mtu(struct ifaddrs *ifa)
+{
+	return -1;
+}
+
 static inline ssize_t ofi_process_vm_readv(pid_t pid,
 			const struct iovec *local_iov,
 			unsigned long liovcnt,
@@ -185,5 +190,3 @@ ofi_recvv_socket(SOCKET fd, const struct iovec *iov, size_t cnt, int flags)
 }
 
 #endif /* _FREEBSD_OSD_H_ */
-
-

include/linux/osd.h

@@ -92,6 +92,8 @@ static inline int ofi_hugepage_enabled(void)
 
 size_t ofi_ifaddr_get_speed(struct ifaddrs *ifa);
 
+int ofi_ifaddr_get_mtu(struct ifaddrs *ifa);
+
 #ifndef __NR_process_vm_readv
 # define __NR_process_vm_readv 310
 #endif

include/osx/osd.h

@@ -99,6 +99,11 @@ static inline size_t ofi_ifaddr_get_speed(struct ifaddrs *ifa)
 	return 0;
 }
 
+static inline int ofi_ifaddr_get_mtu(struct ifaddrs *ifa)
+{
+	return -1;
+}
+
 static inline int ofi_hugepage_enabled(void)
 {
 	return 0;

include/windows/ifaddrs.h

@@ -34,8 +34,8 @@ struct ifaddrs {
 
 	char		   ad_name[16];
 	size_t		   speed;
+	int		   mtu;
 };
 
 int getifaddrs(struct ifaddrs **ifap);
 void freeifaddrs(struct ifaddrs *ifa);
-

include/windows/osd.h

@@ -1006,6 +1006,8 @@ static inline int ofi_is_loopback_addr(struct sockaddr *addr) {
 
 size_t ofi_ifaddr_get_speed(struct ifaddrs *ifa);
 
+int ofi_ifaddr_get_mtu(struct ifaddrs *ifa);
+
 #define file2unix_time	10000000i64
 #define win2unix_epoch	116444736000000000i64
 #define CLOCK_REALTIME 0

man/fi_udp.7.md

@@ -41,9 +41,8 @@ receiving datagram messages over an unreliable endpoint.
 
 # LIMITATIONS
 
-The UDP provider has hard-coded maximums for supported queue sizes and data
-transfers.  These values are reflected in the related fabric attribute
-structures
+The UDP provider has a hard-coded maximum for supported queue sizes.
+This value is reflected in the related fabric attribute structures.
 
 EPs must be bound to both RX and TX CQs.
 
@@ -53,7 +52,10 @@ No support for counters.
 
 # RUNTIME PARAMETERS
 
-No runtime parameters are currently defined.
+The UDP provider checks for the following environment variables -
+
+*FI_UDP_IFACE*
+: An string value that specifies the name of the interface.
 
 # SEE ALSO
 

prov/udp/src/udpx.h

@@ -63,22 +63,22 @@
 #ifndef _UDPX_H_
 #define _UDPX_H_
 
-
 extern struct fi_provider udpx_prov;
 extern struct util_prov udpx_util_prov;
 extern struct fi_info udpx_info;
 
-
 int udpx_fabric(struct fi_fabric_attr *attr, struct fid_fabric **fabric,
 		void *context);
 int udpx_domain_open(struct fid_fabric *fabric, struct fi_info *info,
 		struct fid_domain **dom, void *context);
 int udpx_eq_open(struct fid_fabric *fabric, struct fi_eq_attr *attr,
 		struct fid_eq **eq, void *context);
-
+int udpx_util_prov_init(uint32_t version, const char *node, const char *service,
+			uint64_t flags);
 
 #define UDPX_FLAG_MULTI_RECV	1
 #define UDPX_IOV_LIMIT		4
+#define UDPX_MTU		1500
 
 struct udpx_ep_entry {
 	void			*context;
@@ -88,6 +88,8 @@ struct udpx_ep_entry {
 	uint8_t			resv[sizeof(size_t) - 2];
 };
 
+#define UDPX_MAX_MSG_SIZE(mtu) ((mtu) - 28)
+
 OFI_DECLARE_CIRQUE(struct udpx_ep_entry, udpx_rx_cirq);
 
 struct udpx_ep;

prov/udp/src/udpx_attr.c

@@ -31,14 +31,15 @@
  */
 
 #include "udpx.h"
+#include "ofi_osd.h"
 
 #define UDPX_TX_CAPS (OFI_TX_MSG_CAPS | FI_MULTICAST)
 #define UDPX_RX_CAPS (FI_SOURCE | OFI_RX_MSG_CAPS)
 #define UDPX_DOMAIN_CAPS (FI_LOCAL_COMM | FI_REMOTE_COMM)
 
 struct fi_tx_attr udpx_tx_attr = {
 	.caps = UDPX_TX_CAPS,
-	.inject_size = 1472,
+	.inject_size = UDPX_MAX_MSG_SIZE(UDPX_MTU),
 	.size = 1024,
 	.iov_limit = UDPX_IOV_LIMIT
 };
@@ -53,7 +54,7 @@ struct fi_ep_attr udpx_ep_attr = {
 	.type = FI_EP_DGRAM,
 	.protocol = FI_PROTO_UDP,
 	.protocol_version = 0,
-	.max_msg_size = 1472,
+	.max_msg_size = UDPX_MAX_MSG_SIZE(UDPX_MTU),
 	.tx_ctx_cnt = 1,
 	.rx_ctx_cnt = 1
 };
@@ -93,6 +94,39 @@ struct fi_info udpx_info = {
 
 struct util_prov udpx_util_prov = {
 	.prov = &udpx_prov,
-	.info = &udpx_info,
-	.flags = 0,
+	.info = NULL,
+        .flags = 0,
 };
+
+static int detect_mtu(const struct fi_info* info) {
+
+	struct ifaddrs ifaddrs;
+	ifaddrs.ifa_next = NULL;
+	ifaddrs.ifa_flags = 0;
+	ifaddrs.ifa_netmask = NULL;
+	ifaddrs.ifa_name = info->domain_attr->name;
+	ifaddrs.ifa_addr = info->src_addr;
+	return ofi_ifaddr_get_mtu(&ifaddrs);
+}
+
+int udpx_util_prov_init(uint32_t version, const char *node, const char *service,
+			uint64_t flags) {
+
+	struct fi_info* cur;
+	struct fi_info* info;
+	int max_msg_size;
+        if (udpx_util_prov.info == NULL) {
+		udpx_util_prov.info = &udpx_info;
+		info = fi_allocinfo();
+		ofi_ip_getinfo(&udpx_util_prov, version, node, service, flags,
+			       NULL, &info);
+		for (cur = info; cur; cur = cur->next) {
+			max_msg_size = UDPX_MAX_MSG_SIZE(detect_mtu(cur));
+			if (max_msg_size > 0) {
+				cur->tx_attr->inject_size = max_msg_size;
+				cur->ep_attr->max_msg_size = max_msg_size;
+			}
+		}
+		udpx_util_prov.info = info;
+	}
+}

prov/udp/src/udpx_init.c

@@ -37,20 +37,26 @@
 
 #include <sys/types.h>
 
+static ofi_mutex_t init_lock;
 
 static int udpx_getinfo(uint32_t version, const char *node, const char *service,
 			uint64_t flags, const struct fi_info *hints,
 			struct fi_info **info)
 {
-	return ofi_ip_getinfo(&udpx_util_prov, version, node, service, flags,
-			      hints, info);
+	ofi_mutex_lock(&init_lock);
+	udpx_util_prov_init(version, node, service, flags);
+	ofi_mutex_unlock(&init_lock);
+	return util_getinfo(&udpx_util_prov, version, node, service, flags,
+			    hints, info);
 }
 
 static void udpx_fini(void)
 {
-	/* yawn */
+	if (udpx_util_prov.info != NULL)
+		fi_freeinfo(udpx_util_prov.info);
 }
 
+
 struct fi_provider udpx_prov = {
 	.name = "udp",
 	.version = OFI_VERSION_DEF_PROV,
@@ -65,5 +71,6 @@ UDP_INI
 	fi_param_define(&udpx_prov, "iface", FI_PARAM_STRING,
 			"Specify interface name");
 
+	ofi_mutex_init(&init_lock);
 	return &udpx_prov;
 }

src/linux/osd.c

@@ -257,3 +257,49 @@ size_t ofi_ifaddr_get_speed(struct ifaddrs *ifa)
 }
 
 #endif /* HAVE_ETHTOOL */
+
+int ofi_ifaddr_get_mtu(struct ifaddrs *ifa)
+{
+	FILE *fd;
+	char *line = NULL;
+	size_t len = 0;
+	char *mtu_filename_prefix = "/sys/class/net/";
+	char *mtu_filename_suffix = "/mtu";
+	char *mtu_filename;
+	size_t mtu;
+	/* IF_NAMESIZE includes NULL-terminated symbol */
+	size_t filename_len = strlen(mtu_filename_prefix) +
+				     strlen(mtu_filename_prefix) +
+				     IF_NAMESIZE;
+
+	mtu_filename = calloc(1, filename_len);
+	if (!mtu_filename)
+		return 0;
+
+	snprintf(mtu_filename, filename_len, "%s%s%s",
+		 mtu_filename_prefix, ifa->ifa_name, mtu_filename_suffix);
+
+	fd = fopen(mtu_filename, "r");
+	if (!fd)
+		goto err1;
+
+	if (getline(&line, &len, fd) == -1) {
+		goto err2;
+	}
+
+	if (sscanf(line, "%d", &mtu) != 1)
+		goto err3;
+
+	free(line);
+	fclose(fd);
+	free(mtu_filename);
+
+	return mtu;
+ err3:
+	free(line);
+ err2:
+	fclose(fd);
+ err1:
+	free(mtu_filename);
+	return 0;
+}

src/windows/osd.c

@@ -477,6 +477,7 @@ int getifaddrs(struct ifaddrs **ifap)
 				(*addr6) = *(struct sockaddr_in6 *) pSockAddr;
 			}
 			fa->speed = aa->TransmitLinkSpeed;
+			fa->mtu = aa->Mtu;
 			/* Generate fake Unix-like device names */
 			sprintf_s(fa->ad_name, sizeof(fa->ad_name), "eth%d", i++);
 		}