Add vfio support to scripts.

cleanup.sh and unbind.sh have been combined into a single
setup.sh that takes one optional parameter (reset). If no
parameter is given, the script will automatically bind
all NVMe and IOAT devices to either uio_pci_generic
or vfio-pci, as appropriate based on IOMMU settings. If
the reset parameter is given, the devices will be bound back
to the appropriate kernel drivers.

Change-Id: I25db3234f1ecfb352a281e5093f4c1aa455152ae
Signed-off-by: Ben Walker <benjamin.walker@intel.com>

README.md

@@ -88,4 +88,4 @@ any NVMe and I/OAT devices must be unbound from the native kernel drivers.
 SPDK includes scripts to automate this process on both Linux and FreeBSD.
 
     1) scripts/configure_hugepages.sh
-    2) scripts/unbind.sh
+    2) scripts/setup.sh

autotest.sh

@@ -10,7 +10,7 @@ if [ $EUID -ne 0 ]; then
 	exit 1
 fi
 
-trap "process_core; $rootdir/scripts/cleanup.sh; exit 1" SIGINT SIGTERM EXIT
+trap "process_core; $rootdir/scripts/setup.sh reset; exit 1" SIGINT SIGTERM EXIT
 
 timing_enter autotest
 
@@ -38,7 +38,7 @@ timing_enter afterboot
 ./scripts/configure_hugepages.sh 1024
 timing_exit afterboot
 
-./scripts/unbind.sh
+./scripts/setup.sh
 
 #####################
 # Unit Tests
@@ -52,7 +52,7 @@ time test/lib/ioat/ioat.sh
 
 timing_exit lib
 
-./scripts/cleanup.sh
+./scripts/setup.sh reset
 ./scripts/build_kmod.sh clean
 
 timing_exit autotest

lib/util/pci.c

@@ -448,7 +448,8 @@ spdk_pci_device_has_non_uio_driver(struct spdk_pci_device *dev)
 		driver_begin = driver;
 	}
 
-	return strcmp(driver_begin, "uio_pci_generic") != 0;
+	return (strcmp(driver_begin, "uio_pci_generic") != 0 &&
+		strcmp(driver_begin, "vfio-pci") != 0);
 }
 #endif
 

scripts/cleanup.sh

@@ -1,28 +0,0 @@
-#!/usr/bin/env bash
-
-set -e
-
-function cleanup_linux() {
-	# detach pci devices from uio driver
-	grep -q "^uio_pci_generic" /proc/modules && rmmod uio_pci_generic
-
-	# bind NVMe devices to NVMe driver if no kernel device
-	if [ -d "/sys/bus/pci/drivers/nvme" ]; then
-		device=`find /sys/bus/pci/drivers/nvme -name "0000*" -print`
-		if [ -z "$device" ]; then
-			rmmod nvme
-			modprobe nvme
-		fi
-	fi
-}
-
-function cleanup_freebsd {
-	kldunload contigmem.ko || true
-	kldunload nic_uio.ko || true
-}
-
-if [ `uname` = Linux ]; then
-	cleanup_linux
-else
-	cleanup_freebsd
-fi

scripts/setup.sh

@@ -0,0 +1,138 @@
+#!/usr/bin/env bash
+
+set -e
+
+rootdir=$(readlink -f $(dirname $0))/..
+
+function linux_iter_pci {
+	# Argument is the class code
+	# TODO: More specifically match against only class codes in the grep
+	# step.
+	lspci -mm -n | grep $1 | tr -d '"' | awk -F " " '{print "0000:"$1}'
+}
+
+function configure_linux {
+	driver_name=vfio-pci
+	if [ -z "$(ls /sys/kernel/iommu_groups)" ]; then
+		# No IOMMU. Use uio.
+		driver_name=uio_pci_generic
+	fi
+
+	# NVMe
+	modprobe $driver_name || true
+	for bdf in $(linux_iter_pci 0108); do
+		if [ -e "/sys/bus/pci/devices/$bdf/driver" ]; then
+			# Unbind the device from whatever driver it is currently bound to
+			echo $bdf > "/sys/bus/pci/devices/$bdf/driver/unbind"
+		fi
+		# Bind this device to the new driver
+		ven_dev_id=$(lspci -n -s $bdf | cut -d' ' -f3 | sed 's/:/ /')
+		echo "Binding $bdf ($ven_dev_id) to $driver_name"
+		echo $ven_dev_id > "/sys/bus/pci/drivers/$driver_name/new_id"
+	done
+
+
+	# IOAT
+	TMP=`mktemp`
+	#collect all the device_id info of ioat devices.
+	grep "PCI_DEVICE_ID_INTEL_IOAT" $rootdir/lib/ioat/ioat_pci.h \
+	| awk -F"x" '{print $2}' > $TMP
+
+	for dev_id in `cat $TMP`; do
+		# Abuse linux_iter_pci by giving it a device ID instead of a class code
+		for bdf in $(linux_iter_pci $dev_id); do
+			if [ -e "/sys/bus/pci/devices/$bdf/driver" ]; then
+				# Unbind the device from whatever driver it is currently bound to
+				echo $bdf > "/sys/bus/pci/devices/$bdf/driver/unbind"
+			fi
+			# Bind this device to the new driver
+			ven_dev_id=$(lspci -n -s $bdf | cut -d' ' -f3 | sed 's/:/ /')
+			echo "Binding $bdf ($ven_dev_id) to $driver_name"
+			echo $ven_dev_id > "/sys/bus/pci/drivers/$driver_name/new_id"
+		done
+	done
+	rm $TMP
+
+	echo "1" > "/sys/bus/pci/rescan"
+}
+
+function reset_linux {
+	# NVMe
+	modprobe nvme || true
+	for bdf in $(linux_iter_pci 0108); do
+		ven_dev_id=$(lspci -n -s $bdf | cut -d' ' -f3 | sed 's/:/ /')
+		if [ -e "/sys/bus/pci/devices/$bdf/driver" ]; then
+			# Unregister this device from the driver it was bound to.
+			echo $ven_dev_id > "/sys/bus/pci/devices/$bdf/driver/remove_id"
+			# Unbind the device from whatever driver it is currently bound to
+			echo $bdf > "/sys/bus/pci/devices/$bdf/driver/unbind"
+		fi
+		# Bind this device back to the nvme driver
+		echo "Binding $bdf ($ven_dev_id) to nvme"
+		echo $bdf > "/sys/bus/pci/drivers/nvme/bind"
+	done
+
+
+	# IOAT
+	TMP=`mktemp`
+	#collect all the device_id info of ioat devices.
+	grep "PCI_DEVICE_ID_INTEL_IOAT" $rootdir/lib/ioat/ioat_pci.h \
+	| awk -F"x" '{print $2}' > $TMP
+
+	modprobe ioatdma || true
+	for dev_id in `cat $TMP`; do
+		# Abuse linux_iter_pci by giving it a device ID instead of a class code
+		for bdf in $(linux_iter_pci $dev_id); do
+			ven_dev_id=$(lspci -n -s $bdf | cut -d' ' -f3 | sed 's/:/ /')
+			if [ -e "/sys/bus/pci/devices/$bdf/driver" ]; then
+				# Unregister this device from the driver it was bound to.
+				echo $ven_dev_id > "/sys/bus/pci/devices/$bdf/driver/remove_id"
+				# Unbind the device from whatever driver it is currently bound to
+				echo $bdf > "/sys/bus/pci/devices/$bdf/driver/unbind"
+			fi
+			# Bind this device back to the nvme driver
+			echo "Binding $bdf ($ven_dev_id) to ioatdma"
+			echo $bdf > "/sys/bus/pci/drivers/ioatdma/bind"
+		done
+	done
+	rm $TMP
+
+	echo "1" > "/sys/bus/pci/rescan"
+}
+
+function configure_freebsd {
+	TMP=`mktemp`
+	AWK_PROG="{if (count > 0) printf \",\"; printf \"%s:%s:%s\",\$2,\$3,\$4; count++}"
+	echo $AWK_PROG > $TMP
+	PCICONF=`pciconf -l | grep 'class=0x010802\|^ioat'`
+	BDFS=`echo $PCICONF | awk -F: -f $TMP`
+	kldunload nic_uio.ko || true
+	kenv hw.nic_uio.bdfs=$BDFS
+	kldload nic_uio.ko
+	rm $TMP
+}
+
+function reset_freebsd {
+	kldunload contigmem.ko || true
+	kldunload nic_uio.ko || true
+}
+
+mode=$1
+if [ "$mode" == "" ]; then
+	mode="config"
+fi
+
+if [ `uname` = Linux ]; then
+	if [ "$mode" == "config" ]; then
+		configure_linux
+	elif [ "$mode" == "reset" ]; then
+		reset_linux
+	fi
+else
+	if [ "$mode" == "config" ]; then
+		configure_freebsd
+	elif [ "$mode" == "reset" ]; then
+		reset_freebsd
+	fi
+fi
+

scripts/unbind.sh

@@ -1,63 +0,0 @@
-#!/usr/bin/env bash
-
-set -e
-
-rootdir=$(readlink -f $(dirname $0))/..
-
-function prep_nvme {
-	TMP=`mktemp`
-	# Get vendor_id:device_id by nvme's class_id and subcalss_id
-	lspci -n | awk -F " " '{if ($2 == "0108:") {print $3}}' \
-	| awk -F ":" '{print $1" "$2}' > $TMP
-
-	cat $TMP | while read device
-	do
-		echo $device > /sys/bus/pci/drivers/uio_pci_generic/new_id
-	done
-	rm $TMP
-}
-
-function prep_ioat {
-	TMP=`mktemp`
-	#collect all the device_id info of ioat devices.
-	grep "PCI_DEVICE_ID_INTEL_IOAT" $rootdir/lib/ioat/ioat_pci.h \
-	| awk -F"x" '{print $2}' > $TMP
-	vendor=8086
-
-	for device in `cat $TMP`
-	do
-		if lspci -n | grep "$vendor:$device"
-		then
-			echo $vendor $device > /sys/bus/pci/drivers/uio_pci_generic/new_id
-		fi
-	done
-	rm $TMP
-}
-
-function configure_linux {
-	rmmod nvme || true
-	rmmod uio_pci_generic || true
-	rmmod ioatdma || true
-	modprobe uio_pci_generic || true
-	prep_nvme
-	prep_ioat
-}
-
-function configure_freebsd {
-	TMP=`mktemp`
-	AWK_PROG="{if (count > 0) printf \",\"; printf \"%s:%s:%s\",\$2,\$3,\$4; count++}"
-	echo $AWK_PROG > $TMP
-	PCICONF=`pciconf -l | grep 'class=0x010802\|^ioat'`
-	BDFS=`echo $PCICONF | awk -F: -f $TMP`
-	kldunload nic_uio.ko || true
-	kenv hw.nic_uio.bdfs=$BDFS
-	kldload nic_uio.ko
-	rm $TMP
-}
-
-if [ `uname` = Linux ]; then
-	configure_linux
-else
-	configure_freebsd
-fi
-