Sample Implementation of Intel MIC User Space Daemon.

This patch introduces a sample user space daemon which
implements the virtio device backends on the host. The daemon
creates/removes/configures virtio device backends by communicating with
the Intel MIC Host Driver. The virtio devices currently supported are
virtio net, virtio console and virtio block. Virtio net supports TSO/GSO.
The daemon also monitors card shutdown status and takes appropriate actions
like killing the virtio backends and resetting the card upon card shutdown
and crashes.

Co-author: Ashutosh Dixit <ashutosh.dixit@intel.com>
Co-author: Sudeep Dutt <sudeep.dutt@intel.com>
Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Signed-off-by: Caz Yokoyama <Caz.Yokoyama@intel.com>
Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com>
Signed-off-by: Nikhil Rao <nikhil.rao@intel.com>
Signed-off-by: Harshavardhan R Kharche <harshavardhan.r.kharche@intel.com>
Signed-off-by: Sudeep Dutt <sudeep.dutt@intel.com>
Acked-by: Yaozu (Eddie) Dong <eddie.dong@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

8 files changed, 2347 insertions, 0 deletions

diff --git a/Documentation/mic/mic_overview.txt b/Documentation/mic/mic_overview.txt
new file mode 100644
index 000000000000..c4424ed1b746
--- /dev/null
+++ b/Documentation/mic/mic_overview.txt
@@ -0,0 +1,49 @@
+An Intel MIC X100 device is a PCIe form factor add-in coprocessor
+card based on the Intel Many Integrated Core (MIC) architecture
+that runs a Linux OS. It is a PCIe endpoint in a platform and therefore
+implements the three required standard address spaces i.e. configuration,
+memory and I/O. The host OS loads a device driver as is typical for
+PCIe devices. The card itself runs a bootstrap after reset that
+transfers control to the card OS downloaded from the host driver.
+The card OS as shipped by Intel is a Linux kernel with modifications
+for the X100 devices.
+
+Since it is a PCIe card, it does not have the ability to host hardware
+devices for networking, storage and console. We provide these devices
+on X100 coprocessors thus enabling a self-bootable equivalent environment
+for applications. A key benefit of our solution is that it leverages
+the standard virtio framework for network, disk and console devices,
+though in our case the virtio framework is used across a PCIe bus.
+
+Here is a block diagram of the various components described above. The
+virtio backends are situated on the host rather than the card given better
+single threaded performance for the host compared to MIC, the ability of
+the host to initiate DMA's to/from the card using the MIC DMA engine and
+the fact that the virtio block storage backend can only be on the host.
+
+                              |
+       +----------+           |             +----------+
+       | Card OS  |           |             | Host OS  |
+       +----------+           |             +----------+
+                              |
++-------+ +--------+ +------+ | +---------+  +--------+ +--------+
+| Virtio| |Virtio  | |Virtio| | |Virtio   |  |Virtio  | |Virtio  |
+| Net   | |Console | |Block | | |Net      |  |Console | |Block   |
+| Driver| |Driver  | |Driver| | |backend  |  |backend | |backend |
++-------+ +--------+ +------+ | +---------+  +--------+ +--------+
+    |         |         |     |      |            |         |
+    |         |         |     |User  |            |         |
+    |         |         |     |------|------------|---------|-------
+    +-------------------+     |Kernel +--------------------------+
+              |               |       | Virtio over PCIe IOCTLs  |
+              |               |       +--------------------------+
+      +--------------+        |                   |
+      |Intel MIC     |        |            +---------------+
+      |Card Driver   |        |            |Intel MIC      |
+      +--------------+        |            |Host Driver    |
+              |               |            +---------------+
+              |               |                   |
+     +-------------------------------------------------------------+
+     |                                                             |
+     |                    PCIe Bus                                 |
+     +-------------------------------------------------------------+
diff --git a/Documentation/mic/mpssd/.gitignore b/Documentation/mic/mpssd/.gitignore
new file mode 100644
index 000000000000..8b7c72f07c92
--- /dev/null
+++ b/Documentation/mic/mpssd/.gitignore
@@ -0,0 +1 @@
+mpssd
diff --git a/Documentation/mic/mpssd/Makefile b/Documentation/mic/mpssd/Makefile
new file mode 100644
index 000000000000..eb860a7d152e
--- /dev/null
+++ b/Documentation/mic/mpssd/Makefile
@@ -0,0 +1,19 @@
+#
+# Makefile - Intel MIC User Space Tools.
+# Copyright(c) 2013, Intel Corporation.
+#
+ifdef DEBUG
+CFLAGS += $(USERWARNFLAGS) -I. -g -Wall -DDEBUG=$(DEBUG)
+else
+CFLAGS += $(USERWARNFLAGS) -I. -g -Wall
+endif
+
+mpssd: mpssd.o sysfs.o
+	$(CC) $(CFLAGS) -o $@ $^ -lpthread
+
+install:
+	install mpssd /usr/sbin/mpssd
+	install micctrl /usr/sbin/micctrl
+
+clean:
+	rm -f mpssd *.o
diff --git a/Documentation/mic/mpssd/micctrl b/Documentation/mic/mpssd/micctrl
new file mode 100755
index 000000000000..8f2629b41c5f
--- /dev/null
+++ b/Documentation/mic/mpssd/micctrl
@@ -0,0 +1,173 @@
+#!/bin/bash
+# Intel MIC Platform Software Stack (MPSS)
+#
+# Copyright(c) 2013 Intel Corporation.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License, version 2, as
+# published by the Free Software Foundation.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+# General Public License for more details.
+#
+# The full GNU General Public License is included in this distribution in
+# the file called "COPYING".
+#
+# Intel MIC User Space Tools.
+#
+# micctrl - Controls MIC boot/start/stop.
+#
+# chkconfig: 2345 95 05
+# description: start MPSS stack processing.
+#
+### BEGIN INIT INFO
+# Provides: micctrl
+### END INIT INFO
+
+# Source function library.
+. /etc/init.d/functions
+
+sysfs="/sys/class/mic"
+
+_status()
+{
+	f=$sysfs/$1
+	echo -e $1 state: "`cat $f/state`" shutdown_status: "`cat $f/shutdown_status`"
+}
+
+status()
+{
+	if [ "`echo $1 | head -c3`" == "mic" ]; then
+		_status $1
+		return $?
+	fi
+	for f in $sysfs/*
+	do
+		_status `basename $f`
+		RETVAL=$?
+		[ $RETVAL -ne 0 ] && return $RETVAL
+	done
+	return 0
+}
+
+_reset()
+{
+	f=$sysfs/$1
+	echo reset > $f/state
+}
+
+reset()
+{
+	if [ "`echo $1 | head -c3`" == "mic" ]; then
+		_reset $1
+		return $?
+	fi
+	for f in $sysfs/*
+	do
+		_reset `basename $f`
+		RETVAL=$?
+		[ $RETVAL -ne 0 ] && return $RETVAL
+	done
+	return 0
+}
+
+_boot()
+{
+	f=$sysfs/$1
+	echo "linux" > $f/bootmode
+	echo "mic/uos.img" > $f/firmware
+	echo "mic/$1.image" > $f/ramdisk
+	echo "boot" > $f/state
+}
+
+boot()
+{
+	if [ "`echo $1 | head -c3`" == "mic" ]; then
+		_boot $1
+		return $?
+	fi
+	for f in $sysfs/*
+	do
+		_boot `basename $f`
+		RETVAL=$?
+		[ $RETVAL -ne 0 ] && return $RETVAL
+	done
+	return 0
+}
+
+_shutdown()
+{
+	f=$sysfs/$1
+	echo shutdown > $f/state
+}
+
+shutdown()
+{
+	if [ "`echo $1 | head -c3`" == "mic" ]; then
+		_shutdown $1
+		return $?
+	fi
+	for f in $sysfs/*
+	do
+		_shutdown `basename $f`
+		RETVAL=$?
+		[ $RETVAL -ne 0 ] && return $RETVAL
+	done
+	return 0
+}
+
+_wait()
+{
+	f=$sysfs/$1
+	while [ "`cat $f/state`" != "offline" -a "`cat $f/state`" != "online" ]
+	do
+		sleep 1
+		echo -e "Waiting for $1 to go offline"
+	done
+}
+
+wait()
+{
+	if [ "`echo $1 | head -c3`" == "mic" ]; then
+		_wait $1
+		return $?
+	fi
+	# Wait for the cards to go offline
+	for f in $sysfs/*
+	do
+		_wait `basename $f`
+		RETVAL=$?
+		[ $RETVAL -ne 0 ] && return $RETVAL
+	done
+	return 0
+}
+
+if [ ! -d "$sysfs" ]; then
+	echo -e $"Module unloaded "
+	exit 3
+fi
+
+case $1 in
+	-s)
+		status $2
+		;;
+	-r)
+		reset $2
+		;;
+	-b)
+		boot $2
+		;;
+	-S)
+		shutdown $2
+		;;
+	-w)
+		wait $2
+		;;
+	*)
+		echo $"Usage: $0 {-s (status) |-r (reset) |-b (boot) |-S (shutdown) |-w (wait)}"
+		exit 2
+esac
+
+exit $?
diff --git a/Documentation/mic/mpssd/mpss b/Documentation/mic/mpssd/mpss
new file mode 100755
index 000000000000..3136c68dad0b
--- /dev/null
+++ b/Documentation/mic/mpssd/mpss
@@ -0,0 +1,202 @@
+#!/bin/bash
+# Intel MIC Platform Software Stack (MPSS)
+#
+# Copyright(c) 2013 Intel Corporation.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License, version 2, 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 for more details.
+#
+# The full GNU General Public License is included in this distribution in
+# the file called "COPYING".
+#
+# Intel MIC User Space Tools.
+#
+# mpss	Start mpssd.
+#
+# chkconfig: 2345 95 05
+# description: start MPSS stack processing.
+#
+### BEGIN INIT INFO
+# Provides: mpss
+# Required-Start:
+# Required-Stop:
+# Short-Description: MPSS stack control
+# Description: MPSS stack control
+### END INIT INFO
+
+# Source function library.
+. /etc/init.d/functions
+
+exec=/usr/sbin/mpssd
+sysfs="/sys/class/mic"
+
+start()
+{
+	[ -x $exec ] || exit 5
+
+	if [ "`ps -e | awk '{print $4}' | grep mpssd | head -1`" = "mpssd" ]; then
+		echo -e $"MPSSD already running! "
+		success
+		echo
+		return 0
+	fi
+
+	echo -e $"Starting MPSS Stack"
+	echo -e $"Loading MIC_HOST Module"
+
+	# Ensure the driver is loaded
+	if [ ! -d "$sysfs" ]; then
+		modprobe mic_host
+		RETVAL=$?
+		if [ $RETVAL -ne 0 ]; then
+			failure
+			echo
+			return $RETVAL
+		fi
+	fi
+
+	# Start the daemon
+	echo -n $"Starting MPSSD "
+	$exec
+	RETVAL=$?
+	if [ $RETVAL -ne 0 ]; then
+		failure
+		echo
+		return $RETVAL
+	fi
+	success
+	echo
+
+	sleep 5
+
+	# Boot the cards
+	micctrl -b
+
+	# Wait till ping works
+	for f in $sysfs/*
+	do
+		count=100
+		ipaddr=`cat $f/cmdline`
+		ipaddr=${ipaddr#*address,}
+		ipaddr=`echo $ipaddr | cut -d, -f1 | cut -d\; -f1`
+		while [ $count -ge 0 ]
+		do
+			echo -e "Pinging "`basename $f`" "
+			ping -c 1 $ipaddr &> /dev/null
+			RETVAL=$?
+			if [ $RETVAL -eq 0 ]; then
+				success
+				break
+			fi
+			sleep 1
+			count=`expr $count - 1`
+		done
+		[ $RETVAL -ne 0 ] && failure || success
+		echo
+	done
+	return $RETVAL
+}
+
+stop()
+{
+	echo -e $"Shutting down MPSS Stack: "
+
+	# Bail out if module is unloaded
+	if [ ! -d "$sysfs" ]; then
+		echo -n $"Module unloaded "
+		success
+		echo
+		return 0
+	fi
+
+	# Shut down the cards.
+	micctrl -S
+
+	# Wait for the cards to go offline
+	for f in $sysfs/*
+	do
+		while [ "`cat $f/state`" != "offline" ]
+		do
+			sleep 1
+			echo -e "Waiting for "`basename $f`" to go offline"
+		done
+	done
+
+	# Display the status of the cards
+	micctrl -s
+
+	# Kill MPSSD now
+	echo -n $"Killing MPSSD"
+	killall -9 mpssd 2>/dev/null
+	RETVAL=$?
+	[ $RETVAL -ne 0 ] && failure || success
+	echo
+	return $RETVAL
+}
+
+restart()
+{
+	stop
+	sleep 5
+	start
+}
+
+status()
+{
+	micctrl -s
+	if [ "`ps -e | awk '{print $4}' | grep mpssd | head -n 1`" = "mpssd" ]; then
+		echo "mpssd is running"
+	else
+		echo "mpssd is stopped"
+	fi
+	return 0
+}
+
+unload()
+{
+	if [ ! -d "$sysfs" ]; then
+		echo -n $"No MIC_HOST Module: "
+		success
+		echo
+		return
+	fi
+
+	stop
+
+	sleep 5
+	echo -n $"Removing MIC_HOST Module: "
+	modprobe -r mic_host
+	RETVAL=$?
+	[ $RETVAL -ne 0 ] && failure || success
+	echo
+	return $RETVAL
+}
+
+case $1 in
+	start)
+		start
+		;;
+	stop)
+		stop
+		;;
+	restart)
+		restart
+		;;
+	status)
+		status
+		;;
+	unload)
+		unload
+		;;
+	*)
+		echo $"Usage: $0 {start|stop|restart|status|unload}"
+		exit 2
+esac
+
+exit $?
diff --git a/Documentation/mic/mpssd/mpssd.c b/Documentation/mic/mpssd/mpssd.c
new file mode 100644
index 000000000000..8064804cdac3
--- /dev/null
+++ b/Documentation/mic/mpssd/mpssd.c
@@ -0,0 +1,1701 @@
+/*
+ * Intel MIC Platform Software Stack (MPSS)
+ *
+ * Copyright(c) 2013 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, 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 for more details.
+ *
+ * The full GNU General Public License is included in this distribution in
+ * the file called "COPYING".
+ *
+ * Intel MIC User Space Tools.
+ */
+
+#define _GNU_SOURCE
+
+#include <stdlib.h>
+#include <fcntl.h>
+#include <getopt.h>
+#include <assert.h>
+#include <unistd.h>
+#include <stdbool.h>
+#include <signal.h>
+#include <poll.h>
+#include <features.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/mman.h>
+#include <sys/socket.h>
+#include <linux/virtio_ring.h>
+#include <linux/virtio_net.h>
+#include <linux/virtio_console.h>
+#include <linux/virtio_blk.h>
+#include <linux/version.h>
+#include "mpssd.h"
+#include <linux/mic_ioctl.h>
+#include <linux/mic_common.h>
+
+static void init_mic(struct mic_info *mic);
+
+static FILE *logfp;
+static struct mic_info mic_list;
+
+#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
+
+#define min_t(type, x, y) ({				\
+		type __min1 = (x);                      \
+		type __min2 = (y);                      \
+		__min1 < __min2 ? __min1 : __min2; })
+
+/* align addr on a size boundary - adjust address up/down if needed */
+#define _ALIGN_DOWN(addr, size)  ((addr)&(~((size)-1)))
+#define _ALIGN_UP(addr, size)    _ALIGN_DOWN(addr + size - 1, size)
+
+/* align addr on a size boundary - adjust address up if needed */
+#define _ALIGN(addr, size)     _ALIGN_UP(addr, size)
+
+/* to align the pointer to the (next) page boundary */
+#define PAGE_ALIGN(addr)        _ALIGN(addr, PAGE_SIZE)
+
+#define ACCESS_ONCE(x) (*(volatile typeof(x) *)&(x))
+
+#define GSO_ENABLED		1
+#define MAX_GSO_SIZE		(64 * 1024)
+#define ETH_H_LEN		14
+#define MAX_NET_PKT_SIZE	(_ALIGN_UP(MAX_GSO_SIZE + ETH_H_LEN, 64))
+#define MIC_DEVICE_PAGE_END	0x1000
+
+#ifndef VIRTIO_NET_HDR_F_DATA_VALID
+#define VIRTIO_NET_HDR_F_DATA_VALID	2	/* Csum is valid */
+#endif
+
+static struct {
+	struct mic_device_desc dd;
+	struct mic_vqconfig vqconfig[2];
+	__u32 host_features, guest_acknowledgements;
+	struct virtio_console_config cons_config;
+} virtcons_dev_page = {
+	.dd = {
+		.type = VIRTIO_ID_CONSOLE,
+		.num_vq = ARRAY_SIZE(virtcons_dev_page.vqconfig),
+		.feature_len = sizeof(virtcons_dev_page.host_features),
+		.config_len = sizeof(virtcons_dev_page.cons_config),
+	},
+	.vqconfig[0] = {
+		.num = htole16(MIC_VRING_ENTRIES),
+	},
+	.vqconfig[1] = {
+		.num = htole16(MIC_VRING_ENTRIES),
+	},
+};
+
+static struct {
+	struct mic_device_desc dd;
+	struct mic_vqconfig vqconfig[2];
+	__u32 host_features, guest_acknowledgements;
+	struct virtio_net_config net_config;
+} virtnet_dev_page = {
+	.dd = {
+		.type = VIRTIO_ID_NET,
+		.num_vq = ARRAY_SIZE(virtnet_dev_page.vqconfig),
+		.feature_len = sizeof(virtnet_dev_page.host_features),
+		.config_len = sizeof(virtnet_dev_page.net_config),
+	},
+	.vqconfig[0] = {
+		.num = htole16(MIC_VRING_ENTRIES),
+	},
+	.vqconfig[1] = {
+		.num = htole16(MIC_VRING_ENTRIES),
+	},
+#if GSO_ENABLED
+		.host_features = htole32(
+		1 << VIRTIO_NET_F_CSUM |
+		1 << VIRTIO_NET_F_GSO |
+		1 << VIRTIO_NET_F_GUEST_TSO4 |
+		1 << VIRTIO_NET_F_GUEST_TSO6 |
+		1 << VIRTIO_NET_F_GUEST_ECN |
+		1 << VIRTIO_NET_F_GUEST_UFO),
+#else
+		.host_features = 0,
+#endif
+};
+
+static const char *mic_config_dir = "/etc/sysconfig/mic";
+static const char *virtblk_backend = "VIRTBLK_BACKEND";
+static struct {
+	struct mic_device_desc dd;
+	struct mic_vqconfig vqconfig[1];
+	__u32 host_features, guest_acknowledgements;
+	struct virtio_blk_config blk_config;
+} virtblk_dev_page = {
+	.dd = {
+		.type = VIRTIO_ID_BLOCK,
+		.num_vq = ARRAY_SIZE(virtblk_dev_page.vqconfig),
+		.feature_len = sizeof(virtblk_dev_page.host_features),
+		.config_len = sizeof(virtblk_dev_page.blk_config),
+	},
+	.vqconfig[0] = {
+		.num = htole16(MIC_VRING_ENTRIES),
+	},
+	.host_features =
+		htole32(1<<VIRTIO_BLK_F_SEG_MAX),
+	.blk_config = {
+		.seg_max = htole32(MIC_VRING_ENTRIES - 2),
+		.capacity = htole64(0),
+	 }
+};
+
+static char *myname;
+
+static int
+tap_configure(struct mic_info *mic, char *dev)
+{
+	pid_t pid;
+	char *ifargv[7];
+	char ipaddr[IFNAMSIZ];
+	int ret = 0;
+
+	pid = fork();
+	if (pid == 0) {
+		ifargv[0] = "ip";
+		ifargv[1] = "link";
+		ifargv[2] = "set";
+		ifargv[3] = dev;
+		ifargv[4] = "up";
+		ifargv[5] = NULL;
+		mpsslog("Configuring %s\n", dev);
+		ret = execvp("ip", ifargv);
+		if (ret < 0) {
+			mpsslog("%s execvp failed errno %s\n",
+				mic->name, strerror(errno));
+			return ret;
+		}
+	}
+	if (pid < 0) {
+		mpsslog("%s fork failed errno %s\n",
+			mic->name, strerror(errno));
+		return ret;
+	}
+
+	ret = waitpid(pid, NULL, 0);
+	if (ret < 0) {
+		mpsslog("%s waitpid failed errno %s\n",
+			mic->name, strerror(errno));
+		return ret;
+	}
+
+	snprintf(ipaddr, IFNAMSIZ, "172.31.%d.254/24", mic->id);
+
+	pid = fork();
+	if (pid == 0) {
+		ifargv[0] = "ip";
+		ifargv[1] = "addr";
+		ifargv[2] = "add";
+		ifargv[3] = ipaddr;
+		ifargv[4] = "dev";
+		ifargv[5] = dev;
+		ifargv[6] = NULL;
+		mpsslog("Configuring %s ipaddr %s\n", dev, ipaddr);
+		ret = execvp("ip", ifargv);
+		if (ret < 0) {
+			mpsslog("%s execvp failed errno %s\n",
+				mic->name, strerror(errno));
+			return ret;
+		}
+	}
+	if (pid < 0) {
+		mpsslog("%s fork failed errno %s\n",
+			mic->name, strerror(errno));
+		return ret;
+	}
+
+	ret = waitpid(pid, NULL, 0);
+	if (ret < 0) {
+		mpsslog("%s waitpid failed errno %s\n",
+			mic->name, strerror(errno));
+		return ret;
+	}
+	mpsslog("MIC name %s %s %d DONE!\n",
+		mic->name, __func__, __LINE__);
+	return 0;
+}
+
+static int tun_alloc(struct mic_info *mic, char *dev)
+{
+	struct ifreq ifr;
+	int fd, err;
+#if GSO_ENABLED
+	unsigned offload;
+#endif
+	fd = open("/dev/net/tun", O_RDWR);
+	if (fd < 0) {
+		mpsslog("Could not open /dev/net/tun %s\n", strerror(errno));
+		goto done;
+	}
+
+	memset(&ifr, 0, sizeof(ifr));
+
+	ifr.ifr_flags = IFF_TAP | IFF_NO_PI | IFF_VNET_HDR;
+	if (*dev)
+		strncpy(ifr.ifr_name, dev, IFNAMSIZ);
+
+	err = ioctl(fd, TUNSETIFF, (void *) &ifr);
+	if (err < 0) {
+		mpsslog("%s %s %d TUNSETIFF failed %s\n",
+			mic->name, __func__, __LINE__, strerror(errno));
+		close(fd);
+		return err;
+	}
+#if GSO_ENABLED
+	offload = TUN_F_CSUM | TUN_F_TSO4 | TUN_F_TSO6 |
+		TUN_F_TSO_ECN | TUN_F_UFO;
+
+	err = ioctl(fd, TUNSETOFFLOAD, offload);
+	if (err < 0) {
+		mpsslog("%s %s %d TUNSETOFFLOAD failed %s\n",
+			mic->name, __func__, __LINE__, strerror(errno));
+		close(fd);
+		return err;
+	}
+#endif
+	strcpy(dev, ifr.ifr_name);
+	mpsslog("Created TAP %s\n", dev);
+done:
+	return fd;
+}
+
+#define NET_FD_VIRTIO_NET 0
+#define NET_FD_TUN 1
+#define MAX_NET_FD 2
+
+static void set_dp(struct mic_info *mic, int type, void *dp)
+{
+	switch (type) {
+	case VIRTIO_ID_CONSOLE:
+		mic->mic_console.console_dp = dp;
+		return;
+	case VIRTIO_ID_NET:
+		mic->mic_net.net_dp = dp;
+		return;
+	case VIRTIO_ID_BLOCK:
+		mic->mic_virtblk.block_dp = dp;
+		return;
+	}
+	mpsslog("%s %s %d not found\n", mic->name, __func__, type);
+	assert(0);
+}
+
+static void *get_dp(struct mic_info *mic, int type)
+{
+	switch (type) {
+	case VIRTIO_ID_CONSOLE:
+		return mic->mic_console.console_dp;
+	case VIRTIO_ID_NET:
+		return mic->mic_net.net_dp;
+	case VIRTIO_ID_BLOCK:
+		return mic->mic_virtblk.block_dp;
+	}
+	mpsslog("%s %s %d not found\n", mic->name, __func__, type);
+	assert(0);
+	return NULL;
+}
+
+static struct mic_device_desc *get_device_desc(struct mic_info *mic, int type)
+{
+	struct mic_device_desc *d;
+	int i;
+	void *dp = get_dp(mic, type);
+
+	for (i = mic_aligned_size(struct mic_bootparam); i < PAGE_SIZE;
+		i += mic_total_desc_size(d)) {
+		d = dp + i;
+
+		/* End of list */
+		if (d->type == 0)
+			break;
+
+		if (d->type == -1)
+			continue;
+
+		mpsslog("%s %s d-> type %d d %p\n",
+			mic->name, __func__, d->type, d);
+
+		if (d->type == (__u8)type)
+			return d;
+	}
+	mpsslog("%s %s %d not found\n", mic->name, __func__, type);
+	assert(0);
+	return NULL;
+}
+
+/* See comments in vhost.c for explanation of next_desc() */
+static unsigned next_desc(struct vring_desc *desc)
+{
+	unsigned int next;
+
+	if (!(le16toh(desc->flags) & VRING_DESC_F_NEXT))
+		return -1U;
+	next = le16toh(desc->next);
+	return next;
+}
+
+/* Sum up all the IOVEC length */
+static ssize_t
+sum_iovec_len(struct mic_copy_desc *copy)
+{
+	ssize_t sum = 0;
+	int i;
+
+	for (i = 0; i < copy->iovcnt; i++)
+		sum += copy->iov[i].iov_len;
+	return sum;
+}
+
+static inline void verify_out_len(struct mic_info *mic,
+	struct mic_copy_desc *copy)
+{
+	if (copy->out_len != sum_iovec_len(copy)) {
+		mpsslog("%s %s %d BUG copy->out_len 0x%x len 0x%x\n",
+				mic->name, __func__, __LINE__,
+				copy->out_len, sum_iovec_len(copy));
+		assert(copy->out_len == sum_iovec_len(copy));
+	}
+}
+
+/* Display an iovec */
+static void
+disp_iovec(struct mic_info *mic, struct mic_copy_desc *copy,
+	const char *s, int line)
+{
+	int i;
+
+	for (i = 0; i < copy->iovcnt; i++)
+		mpsslog("%s %s %d copy->iov[%d] addr %p len 0x%lx\n",
+			mic->name, s, line, i,
+			copy->iov[i].iov_base, copy->iov[i].iov_len);
+}
+
+static inline __u16 read_avail_idx(struct mic_vring *vr)
+{
+	return ACCESS_ONCE(vr->info->avail_idx);
+}
+
+static inline void txrx_prepare(int type, bool tx, struct mic_vring *vr,
+				struct mic_copy_desc *copy, ssize_t len)
+{
+	copy->vr_idx = tx ? 0 : 1;
+	copy->update_used = true;
+	if (type == VIRTIO_ID_NET)
+		copy->iov[1].iov_len = len - sizeof(struct virtio_net_hdr);
+	else
+		copy->iov[0].iov_len = len;
+}
+
+/* Central API which triggers the copies */
+static int
+mic_virtio_copy(struct mic_info *mic, int fd,
+	struct mic_vring *vr, struct mic_copy_desc *copy)
+{
+	int ret;
+
+	ret = ioctl(fd, MIC_VIRTIO_COPY_DESC, copy);
+	if (ret) {
+		mpsslog("%s %s %d errno %s ret %d\n",
+			mic->name, __func__, __LINE__,
+			strerror(errno), ret);
+	}
+	return ret;
+}
+
+/*
+ * This initialization routine requires at least one
+ * vring i.e. vr0. vr1 is optional.
+ */
+static void *
+init_vr(struct mic_info *mic, int fd, int type,
+	struct mic_vring *vr0, struct mic_vring *vr1, int num_vq)
+{
+	int vr_size;
+	char *va;
+
+	vr_size = PAGE_ALIGN(vring_size(MIC_VRING_ENTRIES,
+		MIC_VIRTIO_RING_ALIGN) + sizeof(struct _mic_vring_info));
+	va = mmap(NULL, MIC_DEVICE_PAGE_END + vr_size * num_vq,
+		PROT_READ, MAP_SHARED, fd, 0);
+	if (MAP_FAILED == va) {
+		mpsslog("%s %s %d mmap failed errno %s\n",
+			mic->name, __func__, __LINE__,
+			strerror(errno));
+		goto done;
+	}
+	set_dp(mic, type, va);
+	vr0->va = (struct mic_vring *)&va[MIC_DEVICE_PAGE_END];
+	vr0->info = vr0->va +
+		vring_size(MIC_VRING_ENTRIES, MIC_VIRTIO_RING_ALIGN);
+	vring_init(&vr0->vr,
+		MIC_VRING_ENTRIES, vr0->va, MIC_VIRTIO_RING_ALIGN);
+	mpsslog("%s %s vr0 %p vr0->info %p vr_size 0x%x vring 0x%x ",
+		__func__, mic->name, vr0->va, vr0->info, vr_size,
+		vring_size(MIC_VRING_ENTRIES, MIC_VIRTIO_RING_ALIGN));
+	mpsslog("magic 0x%x expected 0x%x\n",
+		vr0->info->magic, MIC_MAGIC + type);
+	assert(vr0->info->magic == MIC_MAGIC + type);
+	if (vr1) {
+		vr1->va = (struct mic_vring *)
+			&va[MIC_DEVICE_PAGE_END + vr_size];
+		vr1->info = vr1->va + vring_size(MIC_VRING_ENTRIES,
+			MIC_VIRTIO_RING_ALIGN);
+		vring_init(&vr1->vr,
+			MIC_VRING_ENTRIES, vr1->va, MIC_VIRTIO_RING_ALIGN);
+		mpsslog("%s %s vr1 %p vr1->info %p vr_size 0x%x vring 0x%x ",
+			__func__, mic->name, vr1->va, vr1->info, vr_size,
+			vring_size(MIC_VRING_ENTRIES, MIC_VIRTIO_RING_ALIGN));
+		mpsslog("magic 0x%x expected 0x%x\n",
+			vr1->info->magic, MIC_MAGIC + type + 1);
+		assert(vr1->info->magic == MIC_MAGIC + type + 1);
+	}
+done:
+	return va;
+}
+
+static void
+wait_for_card_driver(struct mic_info *mic, int fd, int type)
+{
+	struct pollfd pollfd;
+	int err;
+	struct mic_device_desc *desc = get_device_desc(mic, type);
+
+	pollfd.fd = fd;
+	mpsslog("%s %s Waiting .... desc-> type %d status 0x%x\n",
+		mic->name, __func__, type, desc->status);
+	while (1) {
+		pollfd.events = POLLIN;
+		pollfd.revents = 0;
+		err = poll(&pollfd, 1, -1);
+		if (err < 0) {
+			mpsslog("%s %s poll failed %s\n",
+				mic->name, __func__, strerror(errno));
+			continue;
+		}
+
+		if (pollfd.revents) {
+			mpsslog("%s %s Waiting... desc-> type %d status 0x%x\n",
+				mic->name, __func__, type, desc->status);
+			if (desc->status & VIRTIO_CONFIG_S_DRIVER_OK) {
+				mpsslog("%s %s poll.revents %d\n",
+					mic->name, __func__, pollfd.revents);
+				mpsslog("%s %s desc-> type %d status 0x%x\n",
+					mic->name, __func__, type,
+					desc->status);
+				break;
+			}
+		}
+	}
+}
+
+/* Spin till we have some descriptors */
+static void
+spin_for_descriptors(struct mic_info *mic, struct mic_vring *vr)
+{
+	__u16 avail_idx = read_avail_idx(vr);
+
+	while (avail_idx == le16toh(ACCESS_ONCE(vr->vr.avail->idx))) {
+#ifdef DEBUG
+		mpsslog("%s %s waiting for desc avail %d info_avail %d\n",
+			mic->name, __func__,
+			le16toh(vr->vr.avail->idx), vr->info->avail_idx);
+#endif
+		sched_yield();
+	}
+}
+
+static void *
+virtio_net(void *arg)
+{
+	static __u8 vnet_hdr[2][sizeof(struct virtio_net_hdr)];
+	static __u8 vnet_buf[2][MAX_NET_PKT_SIZE] __aligned(64);
+	struct iovec vnet_iov[2][2] = {
+		{ { .iov_base = vnet_hdr[0], .iov_len = sizeof(vnet_hdr[0]) },
+		  { .iov_base = vnet_buf[0], .iov_len = sizeof(vnet_buf[0]) } },
+		{ { .iov_base = vnet_hdr[1], .iov_len = sizeof(vnet_hdr[1]) },
+		  { .iov_base = vnet_buf[1], .iov_len = sizeof(vnet_buf[1]) } },
+	};
+	struct iovec *iov0 = vnet_iov[0], *iov1 = vnet_iov[1];
+	struct mic_info *mic = (struct mic_info *)arg;
+	char if_name[IFNAMSIZ];
+	struct pollfd net_poll[MAX_NET_FD];
+	struct mic_vring tx_vr, rx_vr;
+	struct mic_copy_desc copy;
+	struct mic_device_desc *desc;
+	int err;
+
+	snprintf(if_name, IFNAMSIZ, "mic%d", mic->id);
+	mic->mic_net.tap_fd = tun_alloc(mic, if_name);
+	if (mic->mic_net.tap_fd < 0)
+		goto done;
+
+	if (tap_configure(mic, if_name))
+		goto done;
+	mpsslog("MIC name %s id %d\n", mic->name, mic->id);
+
+	net_poll[NET_FD_VIRTIO_NET].fd = mic->mic_net.virtio_net_fd;
+	net_poll[NET_FD_VIRTIO_NET].events = POLLIN;
+	net_poll[NET_FD_TUN].fd = mic->mic_net.tap_fd;
+	net_poll[NET_FD_TUN].events = POLLIN;
+
+	if (MAP_FAILED == init_vr(mic, mic->mic_net.virtio_net_fd,
+		VIRTIO_ID_NET, &tx_vr, &rx_vr,
+		virtnet_dev_page.dd.num_vq)) {
+		mpsslog("%s init_vr failed %s\n",
+			mic->name, strerror(errno));
+		goto done;
+	}
+
+	copy.iovcnt = 2;
+	desc = get_device_desc(mic, VIRTIO_ID_NET);
+
+	while (1) {
+		ssize_t len;
+
+		net_poll[NET_FD_VIRTIO_NET].revents = 0;
+		net_poll[NET_FD_TUN].revents = 0;
+
+		/* Start polling for data from tap and virtio net */
+		err = poll(net_poll, 2, -1);
+		if (err < 0) {
+			mpsslog("%s poll failed %s\n",
+				__func__, strerror(errno));
+			continue;
+		}
+		if (!(desc->status & VIRTIO_CONFIG_S_DRIVER_OK))
+			wait_for_card_driver(mic, mic->mic_net.virtio_net_fd,
+					VIRTIO_ID_NET);
+		/*
+		 * Check if there is data to be read from TUN and write to
+		 * virtio net fd if there is.
+		 */
+		if (net_poll[NET_FD_TUN].revents & POLLIN) {
+			copy.iov = iov0;
+			len = readv(net_poll[NET_FD_TUN].fd,
+				copy.iov, copy.iovcnt);
+			if (len > 0) {
+				struct virtio_net_hdr *hdr
+					= (struct virtio_net_hdr *) vnet_hdr[0];
+
+				/* Disable checksums on the card since we are on
+				   a reliable PCIe link */
+				hdr->flags |= VIRTIO_NET_HDR_F_DATA_VALID;
+#ifdef DEBUG
+				mpsslog("%s %s %d hdr->flags 0x%x ", mic->name,
+					__func__, __LINE__, hdr->flags);
+				mpsslog("copy.out_len %d hdr->gso_type 0x%x\n",
+					copy.out_len, hdr->gso_type);
+#endif
+#ifdef DEBUG
+				disp_iovec(mic, copy, __func__, __LINE__);
+				mpsslog("%s %s %d read from tap 0x%lx\n",
+					mic->name, __func__, __LINE__,
+					len);
+#endif
+				spin_for_descriptors(mic, &tx_vr);
+				txrx_prepare(VIRTIO_ID_NET, 1, &tx_vr, &copy,
+					len);
+
+				err = mic_virtio_copy(mic,
+					mic->mic_net.virtio_net_fd, &tx_vr,
+					&copy);
+				if (err < 0) {
+					mpsslog("%s %s %d mic_virtio_copy %s\n",
+						mic->name, __func__, __LINE__,
+						strerror(errno));
+				}
+				if (!err)
+					verify_out_len(mic, &copy);
+#ifdef DEBUG
+				disp_iovec(mic, copy, __func__, __LINE__);
+				mpsslog("%s %s %d wrote to net 0x%lx\n",
+					mic->name, __func__, __LINE__,
+					sum_iovec_len(&copy));
+#endif
+				/* Reinitialize IOV for next run */
+				iov0[1].iov_len = MAX_NET_PKT_SIZE;
+			} else if (len < 0) {
+				disp_iovec(mic, &copy, __func__, __LINE__);
+				mpsslog("%s %s %d read failed %s ", mic->name,
+					__func__, __LINE__, strerror(errno));
+				mpsslog("cnt %d sum %d\n",
+					copy.iovcnt, sum_iovec_len(&copy));
+			}
+		}
+
+		/*
+		 * Check if there is data to be read from virtio net and
+		 * write to TUN if there is.
+		 */
+		if (net_poll[NET_FD_VIRTIO_NET].revents & POLLIN) {
+			while (rx_vr.info->avail_idx !=
+				le16toh(rx_vr.vr.avail->idx)) {
+				copy.iov = iov1;
+				txrx_prepare(VIRTIO_ID_NET, 0, &rx_vr, &copy,
+					MAX_NET_PKT_SIZE
+					+ sizeof(struct virtio_net_hdr));
+
+				err = mic_virtio_copy(mic,
+					mic->mic_net.virtio_net_fd, &rx_vr,
+					&copy);
+				if (!err) {
+#ifdef DEBUG
+					struct virtio_net_hdr *hdr
+						= (struct virtio_net_hdr *)
+							vnet_hdr[1];
+
+					mpsslog("%s %s %d hdr->flags 0x%x, ",
+						mic->name, __func__, __LINE__,
+						hdr->flags);
+					mpsslog("out_len %d gso_type 0x%x\n",
+						copy.out_len,
+						hdr->gso_type);
+#endif
+					/* Set the correct output iov_len */
+					iov1[1].iov_len = copy.out_len -
+						sizeof(struct virtio_net_hdr);
+					verify_out_len(mic, &copy);
+#ifdef DEBUG
+					disp_iovec(mic, copy, __func__,
+						__LINE__);
+					mpsslog("%s %s %d ",
+						mic->name, __func__, __LINE__);
+					mpsslog("read from net 0x%lx\n",