selftests/bpf: Xsk selftests framework
authorWeqaar Janjua <weqaar.janjua@gmail.com>
Mon, 7 Dec 2020 21:53:29 +0000 (21:53 +0000)
committerDaniel Borkmann <daniel@iogearbox.net>
Wed, 9 Dec 2020 15:44:44 +0000 (16:44 +0100)
This patch adds AF_XDP selftests framework under selftests/bpf.

Topology:
---------
     -----------           -----------
     |  xskX   | --------- |  xskY   |
     -----------     |     -----------
          |          |          |
     -----------     |     ----------
     |  vethX  | --------- |  vethY |
     -----------   peer    ----------
          |          |          |
     namespaceX      |     namespaceY

Prerequisites setup by script test_xsk.sh:

   Set up veth interfaces as per the topology shown ^^:
   * setup two veth interfaces and one namespace
   ** veth<xxxx> in root namespace
   ** veth<yyyy> in af_xdp<xxxx> namespace
   ** namespace af_xdp<xxxx>
   * create a spec file veth.spec that includes this run-time configuration
   *** xxxx and yyyy are randomly generated 4 digit numbers used to avoid
       conflict with any existing interface
   * tests the veth and xsk layers of the topology

Signed-off-by: Weqaar Janjua <weqaar.a.janjua@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Yonghong Song <yhs@fb.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/20201207215333.11586-2-weqaar.a.janjua@intel.com
tools/testing/selftests/bpf/Makefile
tools/testing/selftests/bpf/test_xsk.sh [new file with mode: 0755]
tools/testing/selftests/bpf/xsk_prereqs.sh [new file with mode: 0755]

index ac25ba5..6a1ddfe 100644 (file)
@@ -46,7 +46,8 @@ endif
 
 TEST_GEN_FILES =
 TEST_FILES = test_lwt_ip_encap.o \
-       test_tc_edt.o
+       test_tc_edt.o \
+       xsk_prereqs.sh
 
 # Order correspond to 'make run_tests' order
 TEST_PROGS := test_kmod.sh \
@@ -70,6 +71,7 @@ TEST_PROGS := test_kmod.sh \
        test_bpftool_build.sh \
        test_bpftool.sh \
        test_bpftool_metadata.sh \
+       test_xsk.sh
 
 TEST_PROGS_EXTENDED := with_addr.sh \
        with_tunnels.sh \
diff --git a/tools/testing/selftests/bpf/test_xsk.sh b/tools/testing/selftests/bpf/test_xsk.sh
new file mode 100755 (executable)
index 0000000..cae4c55
--- /dev/null
@@ -0,0 +1,152 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright(c) 2020 Intel Corporation, Weqaar Janjua <weqaar.a.janjua@intel.com>
+
+# AF_XDP selftests based on veth
+#
+# End-to-end AF_XDP over Veth test
+#
+# Topology:
+# ---------
+#      -----------           -----------
+#      |  xskX   | --------- |  xskY   |
+#      -----------     |     -----------
+#           |          |          |
+#      -----------     |     ----------
+#      |  vethX  | --------- |  vethY |
+#      -----------   peer    ----------
+#           |          |          |
+#      namespaceX      |     namespaceY
+#
+# AF_XDP is an address family optimized for high performance packet processing,
+# it is XDP’s user-space interface.
+#
+# An AF_XDP socket is linked to a single UMEM which is a region of virtual
+# contiguous memory, divided into equal-sized frames.
+#
+# Refer to AF_XDP Kernel Documentation for detailed information:
+# https://www.kernel.org/doc/html/latest/networking/af_xdp.html
+#
+# Prerequisites setup by script:
+#
+#   Set up veth interfaces as per the topology shown ^^:
+#   * setup two veth interfaces and one namespace
+#   ** veth<xxxx> in root namespace
+#   ** veth<yyyy> in af_xdp<xxxx> namespace
+#   ** namespace af_xdp<xxxx>
+#   * create a spec file veth.spec that includes this run-time configuration
+#   *** xxxx and yyyy are randomly generated 4 digit numbers used to avoid
+#       conflict with any existing interface
+#   * tests the veth and xsk layers of the topology
+#
+# Kernel configuration:
+# ---------------------
+# See "config" file for recommended kernel config options.
+#
+# Turn on XDP sockets and veth support when compiling i.e.
+#      Networking support -->
+#              Networking options -->
+#                      [ * ] XDP sockets
+#
+# Executing Tests:
+# ----------------
+# Must run with CAP_NET_ADMIN capability.
+#
+# Run (full color-coded output):
+#   sudo ./test_xsk.sh -c
+#
+# If running from kselftests:
+#   sudo make colorconsole=1 run_tests
+#
+# Run (full output without color-coding):
+#   sudo ./test_xsk.sh
+
+. xsk_prereqs.sh
+
+while getopts c flag
+do
+       case "${flag}" in
+               c) colorconsole=1;;
+       esac
+done
+
+TEST_NAME="PREREQUISITES"
+
+URANDOM=/dev/urandom
+[ ! -e "${URANDOM}" ] && { echo "${URANDOM} not found. Skipping tests."; test_exit 1 1; }
+
+VETH0_POSTFIX=$(cat ${URANDOM} | tr -dc '0-9' | fold -w 256 | head -n 1 | head --bytes 4)
+VETH0=ve${VETH0_POSTFIX}
+VETH1_POSTFIX=$(cat ${URANDOM} | tr -dc '0-9' | fold -w 256 | head -n 1 | head --bytes 4)
+VETH1=ve${VETH1_POSTFIX}
+NS0=root
+NS1=af_xdp${VETH1_POSTFIX}
+MTU=1500
+
+setup_vethPairs() {
+       echo "setting up ${VETH0}: namespace: ${NS0}"
+       ip netns add ${NS1}
+       ip link add ${VETH0} type veth peer name ${VETH1}
+       if [ -f /proc/net/if_inet6 ]; then
+               echo 1 > /proc/sys/net/ipv6/conf/${VETH0}/disable_ipv6
+       fi
+       echo "setting up ${VETH1}: namespace: ${NS1}"
+       ip link set ${VETH1} netns ${NS1}
+       ip netns exec ${NS1} ip link set ${VETH1} mtu ${MTU}
+       ip link set ${VETH0} mtu ${MTU}
+       ip netns exec ${NS1} ip link set ${VETH1} up
+       ip link set ${VETH0} up
+}
+
+validate_root_exec
+validate_veth_support ${VETH0}
+validate_ip_utility
+setup_vethPairs
+
+retval=$?
+if [ $retval -ne 0 ]; then
+       test_status $retval "${TEST_NAME}"
+       cleanup_exit ${VETH0} ${VETH1} ${NS1}
+       exit $retval
+fi
+
+echo "${VETH0}:${VETH1},${NS1}" > ${SPECFILE}
+
+validate_veth_spec_file
+
+echo "Spec file created: ${SPECFILE}"
+
+test_status $retval "${TEST_NAME}"
+
+## START TESTS
+
+statusList=()
+
+### TEST 1
+TEST_NAME="XSK KSELFTEST FRAMEWORK"
+
+echo "Switching interfaces [${VETH0}, ${VETH1}] to XDP Generic mode"
+vethXDPgeneric ${VETH0} ${VETH1} ${NS1}
+
+retval=$?
+if [ $retval -eq 0 ]; then
+       echo "Switching interfaces [${VETH0}, ${VETH1}] to XDP Native mode"
+       vethXDPnative ${VETH0} ${VETH1} ${NS1}
+fi
+
+retval=$?
+test_status $retval "${TEST_NAME}"
+statusList+=($retval)
+
+## END TESTS
+
+cleanup_exit ${VETH0} ${VETH1} ${NS1}
+
+for _status in "${statusList[@]}"
+do
+       if [ $_status -ne 0 ]; then
+               test_exit $ksft_fail 0
+       fi
+done
+
+test_exit $ksft_pass 0
diff --git a/tools/testing/selftests/bpf/xsk_prereqs.sh b/tools/testing/selftests/bpf/xsk_prereqs.sh
new file mode 100755 (executable)
index 0000000..2976273
--- /dev/null
@@ -0,0 +1,119 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright(c) 2020 Intel Corporation.
+
+ksft_pass=0
+ksft_fail=1
+ksft_xfail=2
+ksft_xpass=3
+ksft_skip=4
+
+GREEN='\033[0;92m'
+YELLOW='\033[0;93m'
+RED='\033[0;31m'
+NC='\033[0m'
+STACK_LIM=131072
+SPECFILE=veth.spec
+
+validate_root_exec()
+{
+       msg="skip all tests:"
+       if [ $UID != 0 ]; then
+               echo $msg must be run as root >&2
+               test_exit $ksft_fail 2
+       else
+               return $ksft_pass
+       fi
+}
+
+validate_veth_support()
+{
+       msg="skip all tests:"
+       if [ $(ip link add $1 type veth 2>/dev/null; echo $?;) != 0 ]; then
+               echo $msg veth kernel support not available >&2
+               test_exit $ksft_skip 1
+       else
+               ip link del $1
+               return $ksft_pass
+       fi
+}
+
+validate_veth_spec_file()
+{
+       if [ ! -f ${SPECFILE} ]; then
+               test_exit $ksft_skip 1
+       fi
+}
+
+test_status()
+{
+       statusval=$1
+       if [ -n "${colorconsole+set}" ]; then
+               if [ $statusval -eq 2 ]; then
+                       echo -e "${YELLOW}$2${NC}: [ ${RED}FAIL${NC} ]"
+               elif [ $statusval -eq 1 ]; then
+                       echo -e "${YELLOW}$2${NC}: [ ${RED}SKIPPED${NC} ]"
+               elif [ $statusval -eq 0 ]; then
+                       echo -e "${YELLOW}$2${NC}: [ ${GREEN}PASS${NC} ]"
+               fi
+       else
+               if [ $statusval -eq 2 ]; then
+                       echo -e "$2: [ FAIL ]"
+               elif [ $statusval -eq 1 ]; then
+                       echo -e "$2: [ SKIPPED ]"
+               elif [ $statusval -eq 0 ]; then
+                       echo -e "$2: [ PASS ]"
+               fi
+       fi
+}
+
+test_exit()
+{
+       retval=$1
+       if [ $2 -ne 0 ]; then
+               test_status $2 $(basename $0)
+       fi
+       exit $retval
+}
+
+clear_configs()
+{
+       if [ $(ip netns show | grep $3 &>/dev/null; echo $?;) == 0 ]; then
+               [ $(ip netns exec $3 ip link show $2 &>/dev/null; echo $?;) == 0 ] &&
+                       { echo "removing link $1:$2"; ip netns exec $3 ip link del $2; }
+               echo "removing ns $3"
+               ip netns del $3
+       fi
+       #Once we delete a veth pair node, the entire veth pair is removed,
+       #this is just to be cautious just incase the NS does not exist then
+       #veth node inside NS won't get removed so we explicitly remove it
+       [ $(ip link show $1 &>/dev/null; echo $?;) == 0 ] &&
+               { echo "removing link $1"; ip link del $1; }
+       if [ -f ${SPECFILE} ]; then
+               echo "removing spec file:" ${SPECFILE}
+               rm -f ${SPECFILE}
+       fi
+}
+
+cleanup_exit()
+{
+       echo "cleaning up..."
+       clear_configs $1 $2 $3
+}
+
+validate_ip_utility()
+{
+       [ ! $(type -P ip) ] && { echo "'ip' not found. Skipping tests."; test_exit $ksft_skip 1; }
+}
+
+vethXDPgeneric()
+{
+       ip link set dev $1 xdpdrv off
+       ip netns exec $3 ip link set dev $2 xdpdrv off
+}
+
+vethXDPnative()
+{
+       ip link set dev $1 xdpgeneric off
+       ip netns exec $3 ip link set dev $2 xdpgeneric off
+}