merge master

experiments/cc/two_flow_sweep.sh

+#!/bin/bash
+
+DUR=30
+# K is DCTCP threshold in the unit of cells (208 bytes for BCMXXXXX)
+K_START=0
+K_END=32
+K_INTERVAL=32
+
+MTU=1500
+
+set -x
+
+for ((K=$K_START; K<=$K_END; K+=$K_INTERVAL))
+do
+    ssh honey1.kaist.ac.kr -p 2222 ~/change_k.sh $K > /dev/null
+    EXP=$MTU-$K
+    sleep 1
+    ssh honey2.kaist.ac.kr -p 2222 sudo taskset 0x02 iperf -c 10.9.9.11 -i 1 -Z dctcp -w 400K -t $DUR > $EXP-1.txt &
+    ssh honey3.kaist.ac.kr -p 2222 sudo taskset 0x02 iperf -c 10.9.9.11 -i 1 -Z dctcp -w 400K -t $DUR > $EXP-2.txt 
+    wait
+    cat $EXP-*.txt
+    sleep 3
+done
+

experiments/common-functions.sh

@@ -5,6 +5,8 @@ QEMU_CMD="$EHSIM_BASE/qemu/x86_64-softmmu/qemu-system-x86_64"
 QEMU_IMG="$EHSIM_BASE/qemu/qemu-img"
 GEM5_BASE="$EHSIM_BASE/gem5"
 NS3_BASE="$EHSIM_BASE/ns-3"
+OUTDIR_BASE="$EHSIM_BASE/experiments/out"
+WORKDIR_BASE="$OUTDIR_BASE"
 
 if [ -f local-config.sh ] ; then
     source local-config.sh
@@ -40,9 +42,10 @@ GEM5_KERNEL=$EHSIM_BASE/images/vmlinux
 # Args:
 #   - experiment name
 init_out() {
-  export OUTDIR=./out/$1/$2
-  rm -rf $OUTDIR
-  mkdir -p $OUTDIR
+  export OUTDIR=$OUTDIR_BASE/$1/$2
+  export WORKDIR=$WORKDIR_BASE/$1/$2
+  rm -rf $OUTDIR $WORKDIR
+  mkdir -p $OUTDIR $WORKDIR
   date > $OUTDIR/starttime
 }
 
@@ -51,10 +54,11 @@ init_out() {
 #   - Cosim instance
 #   - secondary hard drive
 #   - [optional primary image name: default ubuntu1804-base]
+#   - [optional: additinoal qemu arguments]
 run_qemu() {
-    img_a="$OUTDIR/qemu.hd.a.$1"
-    img_b="$OUTDIR/qemu.hd.b.$1"
-    pcisock="$OUTDIR/pci.$2"
+    img_a="$WORKDIR/qemu.hd.a.$1"
+    img_b="$WORKDIR/qemu.hd.b.$1"
+    pcisock="$WORKDIR/pci.$2"
     rm -f $img_a $img_b
     echo Creating disk for qemu $1
     if [ -z "$4" ]; then
@@ -74,7 +78,8 @@ run_qemu() {
         -serial mon:stdio -m $((16 * 1024)) -smp 1 -display none -enable-kvm \
         -nic none \
         -chardev socket,path=$pcisock,id=cosimcd \
-        -device cosim-pci,chardev=cosimcd &>$OUTDIR/qemu.$1.log &
+        -device cosim-pci,chardev=cosimcd \
+        $5 &>$OUTDIR/qemu.$1.log &
     pid=$!
     ALL_PIDS="$ALL_PIDS $pid"
     return $pid
@@ -90,9 +95,9 @@ run_qemu() {
 #   - [optional primary image name: default ubuntu1804-base]
 run_gem5() {
     echo Starting gem5 $1
-    pcisock="$OUTDIR/pci.$2"
-    shm="$OUTDIR/shm.$2"
-    cpdir="$OUTDIR/../checkpoint/checkpoints.$5"
+    pcisock="$WORKDIR/pci.$2"
+    shm="$WORKDIR/shm.$2"
+    cpdir="$WORKDIR/../checkpoint/checkpoints.$5"
     mkdir -p $cpdir
 
     if [ -z "$7" ]; then
@@ -126,7 +131,7 @@ run_gem5() {
 run_corundum_verilator() {
     echo Starting corundum_verilator $1
     $EHSIM_BASE/corundum/corundum_verilator \
-        $OUTDIR/pci.$1 $OUTDIR/eth.$1 $OUTDIR/shm.$1 \
+        $WORKDIR/pci.$1 $WORKDIR/eth.$1 $WORKDIR/shm.$1 \
             &>$OUTDIR/corundum_verilator.$1.log &
     pid=$!
     ALL_PIDS="$ALL_PIDS $pid"
@@ -138,7 +143,7 @@ run_corundum_verilator() {
 run_corundum_bm() {
     echo Starting corundum_bm $1
     $EHSIM_BASE/corundum_bm/corundum_bm \
-        $OUTDIR/pci.$1 $OUTDIR/eth.$1 $OUTDIR/shm.$1 \
+        $WORKDIR/pci.$1 $WORKDIR/eth.$1 $WORKDIR/shm.$1 \
         &>$OUTDIR/corundum_bm.$1.log &
     pid=$!
     ALL_PIDS="$ALL_PIDS $pid"
@@ -150,7 +155,7 @@ run_corundum_bm() {
 run_i40e_bm() {
     echo Starting i40e $1
     $EHSIM_BASE/i40e_bm/i40e_bm \
-        $OUTDIR/pci.$1 $OUTDIR/eth.$1 $OUTDIR/shm.$1 \
+        $WORKDIR/pci.$1 $WORKDIR/eth.$1 $WORKDIR/shm.$1 \
         &>$OUTDIR/i40e_bm.$1.log &
     pid=$!
     ALL_PIDS="$ALL_PIDS $pid"
@@ -171,7 +176,7 @@ run_wire() {
     fi
 
     $EHSIM_BASE/net_wire/net_wire \
-        $OUTDIR/eth.$2 $OUTDIR/eth.$3 $pcap &>$OUTDIR/wire.$1.log &
+        $WORKDIR/eth.$2 $WORKDIR/eth.$3 $pcap &>$OUTDIR/wire.$1.log &
     pid=$!
     ALL_PIDS="$ALL_PIDS $pid"
     return $pid
@@ -183,7 +188,7 @@ run_wire() {
 run_ns3_bridge() {
     ports=""
     for p in $2; do
-        epath="`readlink -f $OUTDIR/eth.$p`"
+        epath="`readlink -f $WORKDIR/eth.$p`"
         ports="$ports --CosimPort=$epath"
     done
     $NS3_BASE/cosim-run.sh cosim cosim-bridge-example \
@@ -201,11 +206,11 @@ run_ns3_bridge() {
 run_ns3_dumbbell() {
     ports=""
     for p in $2; do
-        epath="`readlink -f $OUTDIR/eth.$p`"
+        epath="`readlink -f $WORKDIR/eth.$p`"
         ports="$ports --CosimPortLeft=$epath"
     done
     for p in $3; do
-        epath="`readlink -f $OUTDIR/eth.$p`"
+        epath="`readlink -f $WORKDIR/eth.$p`"
         ports="$ports --CosimPortRight=$epath"
     done
 
@@ -224,11 +229,11 @@ run_ns3_dumbbell() {
 run_ns3_sequencer() {
     ports=""
     for p in $2; do
-        epath="`readlink -f $OUTDIR/eth.$p`"
+        epath="`readlink -f $WORKDIR/eth.$p`"
         ports="$ports --ClientPort=$epath"
     done
     for p in $3; do
-        epath="`readlink -f $OUTDIR/eth.$p`"
+        epath="`readlink -f $WORKDIR/eth.$p`"
         ports="$ports --ServerPort=$epath"
     done
     for p in $4; do
@@ -253,7 +258,11 @@ cleanup() {
         kill -KILL $p &>/dev/null
     done
 
-    rm -f $OUTDIR/{qemu.hd.*,shm.*,pci.*,eth.*}
+    if [ "$OUTDIR" != "$WORKDIR" ]; then
+        rm -rf $WORKDIR
+    else
+        rm -f $WORKDIR/{qemu.hd.*,shm.*,pci.*,eth.*}
+    fi
     date >>$OUTDIR/endtime
 }
 

experiments/experiments/gem5-i40e-cc-mck.sh

+#!/bin/bash
+
+source common-functions.sh
+
+init_out gem5-i40e-cc checkpoint
+
+# first run to checkpoint with fast CPU
+run_i40e_bm a
+run_i40e_bm b
+run_i40e_bm c
+sleep 2
+run_ns3_bridge bridge "a b c"
+run_gem5 a a build/gem5-pair-i40e-server.tar X86KvmCPU server "--cosim-type=i40e"
+run_gem5 b b build/gem5-pair-i40e-client.tar X86KvmCPU client "--cosim-type=i40e"
+run_gem5 c c build/gem5-pair-i40e-client-2.tar X86KvmCPU client2 "--cosim-type=i40e"
+client_pid=$!
+wait $client_pid
+cleanup
+
+echo "Took checkpoint successfully"

experiments/experiments/gem5-i40e-cc.sh

+#!/bin/bash
+
+source common-functions.sh
+
+init_out gem5-i40e-cc $1
+
+echo "Restoring from checkpoint"
+
+# then run with timing CPU
+run_i40e_bm a
+run_i40e_bm b
+run_i40e_bm c
+sleep 0.5
+run_ns3_bridge bridge "a b c"
+run_gem5 a a build/gem5-pair-i40e-server.tar TimingSimpleCPU server "-r 0 --cosim-sync --cosim-type=i40e"
+run_gem5 b b build/gem5-pair-i40e-client.tar TimingSimpleCPU client "-r 0 --cosim-sync --cosim-type=i40e"
+run_gem5 c c build/gem5-pair-i40e-client-2.tar TimingSimpleCPU client2 "-r 0 --cosim-sync --cosim-type=i40e"
+client_pid=$!
+wait $client_pid
+cleanup

experiments/experiments/qemu-i40e-bm-cc.sh

+#!/bin/bash
+
+source common-functions.sh
+
+init_out qemu-i40e-bm-cc $1
+run_i40e_bm a
+run_i40e_bm b
+run_i40e_bm c
+sleep 0.5
+run_ns3_bridge bridge "a b c"
+run_qemu a a build/qemu-pair-i40e-server.tar
+sleep 2
+run_qemu b b build/qemu-pair-i40e-client.tar
+run_qemu c c build/qemu-pair-i40e-client-2.tar
+client_pid=$!
+wait $client_pid
+cleanup

experiments/experiments/qemu-i40e-bm-pair.sh

@@ -7,9 +7,8 @@ run_i40e_bm a
 run_i40e_bm b
 sleep 0.5
 run_wire ab a b
-run_qemu a a build/qemu-pair-server.tar
-run_qemu b b build/qemu-pair-client.tar
-#client_pid=$!
-#wait $client_pid
-sleep 30
+run_qemu a a build/qemu-pair-i40e-server.tar
+run_qemu b b build/qemu-pair-i40e-client.tar
+client_pid=$!
+wait $client_pid
 cleanup

experiments/experiments/qemu-i40e-bm-rpc.sh

+#!/bin/bash
+
+source common-functions.sh
+
+init_out qemu-i40e-bm-rpc $1
+run_i40e_bm a
+run_i40e_bm b
+sleep 0.5
+run_wire ab a b
+run_qemu a a build/qemu-i40e-rpc-server.tar tas
+run_qemu b b build/qemu-i40e-rpc-client.tar tas
+client_pid=$!
+wait $client_pid
+cleanup

experiments/experiments/qemu-i40e-bm-tas.sh

+#!/bin/bash
+
+source common-functions.sh
+
+init_out qemu-i40e-bm-tas $1
+run_i40e_bm a
+run_i40e_bm b
+sleep 0.5
+run_wire ab a b
+run_qemu a a build/qemu-tas-server.tar tas "-smp 4"
+run_qemu b b build/qemu-tas-client.tar tas "-smp 4"
+client_pid=$!
+wait $client_pid
+cleanup

experiments/guests/gem5-pair-i40e-client-2/run.sh

+#!/bin/bash
+mount -t proc proc /proc
+mount -t sysfs sysfs /sys
+sysctl -w net.core.rmem_default=31457280
+sysctl -w net.core.rmem_max=31457280
+sysctl -w net.core.wmem_default=31457280
+sysctl -w net.core.wmem_max=31457280
+sysctl -w net.core.optmem_max=25165824
+sysctl -w net.ipv4.tcp_mem="786432 1048576 26777216"
+sysctl -w net.ipv4.tcp_rmem="8192 87380 33554432"
+sysctl -w net.ipv4.tcp_wmem="8192 87380 33554432"
+
+
+m5 checkpoint
+modprobe i40e
+ethtool -G eth0 rx 4096 tx 4096
+ethtool -K eth0 tso off
+ip link set eth0 txqueuelen 13888
+ip link set dev eth0 mtu 9000 up
+ip addr add 192.168.64.3/24 dev eth0
+sleep 0.2
+iperf -l 32M -w 32M -c 192.168.64.1 -i 1 -P 4
+m5 exit

experiments/guests/gem5-pair-i40e-client/run.sh

 #!/bin/bash
+mount -t proc proc /proc
+mount -t sysfs sysfs /sys
+sysctl -w net.core.rmem_default=31457280
+sysctl -w net.core.rmem_max=31457280
+sysctl -w net.core.wmem_default=31457280
+sysctl -w net.core.wmem_max=31457280
+sysctl -w net.core.optmem_max=25165824
+sysctl -w net.ipv4.tcp_mem="786432 1048576 26777216"
+sysctl -w net.ipv4.tcp_rmem="8192 87380 33554432"
+sysctl -w net.ipv4.tcp_wmem="8192 87380 33554432"
+
+
 m5 checkpoint
 modprobe i40e
-ip link set dev eth0 up
-ip addr add 192.168.64.2/24 dev eth0
+ethtool -G eth0 rx 4096 tx 4096
 ethtool -K eth0 tso off
-sleep 2
-iperf -l 1M -w 1M -c 192.168.64.1 -i 1 -P 4
+ip link set eth0 txqueuelen 13888
+ip link set dev eth0 mtu 9000 up
+ip addr add 192.168.64.2/24 dev eth0
+sleep 0.2
+iperf -l 32M -w 32M -c 192.168.64.1 -i 1 -P 4
 m5 exit

experiments/guests/gem5-pair-i40e-server/run.sh

 #!/bin/bash
+mount -t proc proc /proc
+mount -t sysfs sysfs /sys
+sysctl -w net.core.rmem_default=31457280
+sysctl -w net.core.rmem_max=31457280
+sysctl -w net.core.wmem_default=31457280
+sysctl -w net.core.wmem_max=31457280
+sysctl -w net.core.optmem_max=25165824
+sysctl -w net.ipv4.tcp_mem="786432 1048576 26777216"
+sysctl -w net.ipv4.tcp_rmem="8192 87380 33554432"
+sysctl -w net.ipv4.tcp_wmem="8192 87380 33554432"
+
+
 m5 checkpoint
 modprobe i40e
-ip link set dev eth0 up
-ip addr add 192.168.64.1/24 dev eth0
+ethtool -G eth0 rx 4096 tx 4096
 ethtool -K eth0 tso off
-iperf -s -l 1M -w 1M -P 4
+ip link set eth0 txqueuelen 13888
+ip link set dev eth0 mtu 9000 up
+ip addr add 192.168.64.1/24 dev eth0
+iperf -s -l 32M -w 32M -P 8
 m5 exit

experiments/guests/qemu-i40e-rpc-client/run.sh

+#!/bin/bash
+modprobe i40e
+ip link set dev eth0 up
+ip addr add 192.168.64.2/24 dev eth0
+
+sleep 2
+cd /root/tasbench/micro_rpc
+./testclient_linux 192.168.64.1 1234 1 /tmp/guest/mtcp.conf 1024 1 128 2 0 8 &
+sleep 25
+
+poweroff -f

experiments/guests/qemu-i40e-rpc-server/run.sh

+#!/bin/bash
+modprobe i40e
+ip link set dev eth0 up
+ip addr add 192.168.64.1/24 dev eth0
+
+cd /root/tasbench/micro_rpc
+./echoserver_linux 1234 1 /tmp/guest/mtcp.conf 1024 1024
+poweroff -f

experiments/guests/qemu-pair-i40e-client-2/run.sh

+#!/bin/bash
+mount -t proc proc /proc
+mount -t sysfs sysfs /sys
+#sysctl -w net.core.busy_poll=50
+#sysctl -w net.core.busy_read=50
+sysctl -w net.core.rmem_default=31457280
+sysctl -w net.core.rmem_max=31457280
+sysctl -w net.core.wmem_default=31457280
+sysctl -w net.core.wmem_max=31457280
+sysctl -w net.core.optmem_max=25165824
+sysctl -w net.ipv4.tcp_mem="786432 1048576 26777216"
+sysctl -w net.ipv4.tcp_rmem="8192 87380 33554432"
+sysctl -w net.ipv4.tcp_wmem="8192 87380 33554432"
+
+modprobe i40e
+ethtool -G eth0 rx 4096 tx 4096
+ethtool -K eth0 tso off
+echo 13888 > /proc/sys/net/core/netdev_max_backlog
+ip link set eth0 txqueuelen 13888
+ip link set dev eth0 mtu 9000 up
+ip addr add 192.168.64.3/24 dev eth0
+sleep 2
+iperf -l 32M -w 32M  -c 192.168.64.1 -i 1 -P 4
+poweroff -f

experiments/guests/qemu-pair-i40e-client/run.sh

+#!/bin/bash
+mount -t proc proc /proc
+mount -t sysfs sysfs /sys
+#sysctl -w net.core.busy_poll=50
+#sysctl -w net.core.busy_read=50
+sysctl -w net.core.rmem_default=31457280
+sysctl -w net.core.rmem_max=31457280
+sysctl -w net.core.wmem_default=31457280
+sysctl -w net.core.wmem_max=31457280
+sysctl -w net.core.optmem_max=25165824
+sysctl -w net.ipv4.tcp_mem="786432 1048576 26777216"
+sysctl -w net.ipv4.tcp_rmem="8192 87380 33554432"
+sysctl -w net.ipv4.tcp_wmem="8192 87380 33554432"
+
+modprobe i40e
+ethtool -G eth0 rx 4096 tx 4096
+ethtool -K eth0 tso off
+echo 13888 > /proc/sys/net/core/netdev_max_backlog
+ip link set eth0 txqueuelen 13888
+ip link set dev eth0 mtu 9000 up
+ip addr add 192.168.64.2/24 dev eth0
+sleep 2
+iperf -l 32M -w 32M  -c 192.168.64.1 -i 1 -P 4
+poweroff -f

experiments/guests/qemu-pair-i40e-server/run.sh

+#!/bin/bash
+mount -t proc proc /proc
+mount -t sysfs sysfs /sys
+#sysctl -w net.core.busy_poll=50
+#sysctl -w net.core.busy_read=50
+sysctl -w net.core.rmem_default=31457280
+sysctl -w net.core.rmem_max=31457280
+sysctl -w net.core.wmem_default=31457280
+sysctl -w net.core.wmem_max=31457280
+sysctl -w net.core.optmem_max=25165824
+sysctl -w net.ipv4.tcp_mem="786432 1048576 26777216"
+sysctl -w net.ipv4.tcp_rmem="8192 87380 33554432"
+sysctl -w net.ipv4.tcp_wmem="8192 87380 33554432"
+
+modprobe i40e
+ethtool -G eth0 rx 4096 tx 4096
+ethtool -K eth0 tso off
+echo 13888 > /proc/sys/net/core/netdev_max_backlog
+ip link set eth0 txqueuelen 13888
+ip link set dev eth0 mtu 9000 up
+ip addr add 192.168.64.1/24 dev eth0
+iperf -s -l 32M -w 32M -P 8
+poweroff -f

experiments/guests/qemu-tas-client/run.sh

+#!/bin/bash
+export HOME=/root
+export LANG=en_US
+export PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games"
+
+mount -t proc proc /proc
+mount -t sysfs sysfs /sys
+mkdir -p /dev/hugepages
+mount -t hugetlbfs nodev /dev/hugepages
+mkdir -p /dev/shm
+mount -t tmpfs tmpfs /dev/shm
+insmod /root/dpdk/lib/modules/5.4.46/extra/dpdk/igb_uio.ko
+/root/dpdk/sbin/dpdk-devbind -b igb_uio 0000:00:02.0
+echo 4096 > /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages
+
+cd /root/tas
+tas/tas --ip-addr=192.168.64.2/24 --fp-cores-max=1 --fp-no-ints &
+sleep 3
+
+cd /root/tasbench/micro_rpc
+LD_PRELOAD=/root/tas/lib/libtas_interpose.so ./testclient_linux 192.168.64.1 1234 1 /tmp/guest/mtcp.conf 1024 1 128 2 0 8 &
+sleep 25
+
+poweroff -f

experiments/guests/qemu-tas-server/run.sh

+#!/bin/bash
+export HOME=/root
+export LANG=en_US
+export PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games"
+
+mount -t proc proc /proc
+mount -t sysfs sysfs /sys
+mkdir -p /dev/hugepages
+mount -t hugetlbfs nodev /dev/hugepages
+mkdir -p /dev/shm
+mount -t tmpfs tmpfs /dev/shm
+insmod /root/dpdk/lib/modules/5.4.46/extra/dpdk/igb_uio.ko
+/root/dpdk/sbin/dpdk-devbind -b igb_uio 0000:00:02.0
+echo 4096 > /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages
+
+cd /root/tas
+tas/tas --ip-addr=192.168.64.1/24 --fp-cores-max=1 --fp-no-ints &
+sleep 3
+
+cd /root/tasbench/micro_rpc
+LD_PRELOAD=/root/tas/lib/libtas_interpose.so ./echoserver_linux 1234 1 /tmp/guest/mtcp.conf 1024 1024
+poweroff -f

i40e_bm/i40e_bm.cc

@@ -219,6 +219,9 @@ uint32_t i40e_bm::reg_mem_read32(uint64_t addr)
             case I40E_GLGEN_RSTCTL:
                 val = regs.glgen_rstctl;
                 break;
+            case I40E_GLGEN_STAT:
+                val = regs.glgen_stat;
+                break;
 
             case I40E_GLVFGEN_TIMER:
                 val = runner->time_ps() / 1000000;
@@ -238,6 +241,20 @@ uint32_t i40e_bm::reg_mem_read32(uint64_t addr)
                 regs.pfint_icr0 = 0;
                 break;
 
+            case I40E_PFINT_DYN_CTL0:
+                val = regs.pfint_dyn_ctl0;
+                break;
+
+            case I40E_PFINT_ITR0(0):
+                val = regs.pfint_itr0[0];
+                break;
+            case I40E_PFINT_ITR0(1):
+                val = regs.pfint_itr0[1];
+                break;
+            case I40E_PFINT_ITR0(2):
+                val = regs.pfint_itr0[2];
+                break;
+
             case I40E_GLPCI_CNF2:
                 val = ((NUM_PFINTS - 2) << I40E_GLPCI_CNF2_MSI_X_PF_N_SHIFT) |
                     (2 << I40E_GLPCI_CNF2_MSI_X_VF_N_SHIFT); /* that is ugly,
@@ -478,6 +495,18 @@ void i40e_bm::reg_mem_write32(uint64_t addr, uint32_t val)
             case I40E_PFINT_ICR0:
                 regs.pfint_icr0 = val;
                 break;
+            case I40E_PFINT_DYN_CTL0:
+                regs.pfint_dyn_ctl0 = val;
+                break;
+            case I40E_PFINT_ITR0(0):
+                regs.pfint_itr0[0] = val;
+                break;
+            case I40E_PFINT_ITR0(1):
+                regs.pfint_itr0[1] = val;
+                break;
+            case I40E_PFINT_ITR0(2):
+                regs.pfint_itr0[2] = val;
+                break;
 
             case I40E_PFHMC_SDCMD:
                 regs.pfhmc_sdcmd = val;
@@ -545,6 +574,64 @@ void i40e_bm::reg_mem_write32(uint64_t addr, uint32_t val)
     }
 }
 
+void i40e_bm::timed_event(nicbm::TimedEvent &ev)
+{
+    int_ev &iev = *((int_ev *) &ev);
+#ifdef DEBUG_DEV
+    log << "timed_event: triggering interrupt (" << iev.vector << ")" <<
+        logger::endl;
+#endif
+    iev.armed = false;
+    runner->msi_issue(0);
+}
+
+void i40e_bm::signal_interrupt(uint16_t vector, uint8_t itr)
+{
+    if (vector != 0) {
+        log << "signal_interrupt() only supports vector 0" << logger::endl;
+        abort();
+    }
+
+    int_ev &iev = intevs[vector];
+
+    uint64_t mindelay;
+    if (itr <= 2) {
+        // itr 0-2
+        mindelay = regs.pfint_itr0[itr];
+        mindelay *= 2000000ULL;
+    } else if (itr == 3) {
+        // noitr
+        mindelay = 0;
+    } else {
+        log << "signal_interrupt() invalid itr (" << itr << ")" << logger::endl;
+        abort();
+    }
+
+    uint64_t curtime = runner->time_ps();
+    uint64_t newtime = curtime + mindelay;
+    if (iev.armed && iev.time <= newtime) {
+        // already armed and this is not scheduled sooner
+#ifdef DEBUG_DEV
+        log << "signal_interrupt: vector " << vector << " already scheduled" <<
+            logger::endl;
+#endif
+        return;
+    } else if (iev.armed) {
+        // need to reschedule
+        runner->event_cancel(iev);
+    }
+
+    iev.armed = true;
+    iev.time = newtime;
+
+#ifdef DEBUG_DEV
+    log << "signal_interrupt: scheduled vector " << vector << " for time=" <<
+        newtime << " (itr " << itr << ")" << logger::endl;
+#endif
+
+    runner->event_schedule(iev);
+}
+
 void i40e_bm::reset(bool indicate_done)
 {
 #ifdef DEBUG_DEV
@@ -558,6 +645,15 @@ void i40e_bm::reset(bool indicate_done)
     memset(&regs, 0, sizeof(regs));
     if (indicate_done)
         regs.glnvm_srctl = I40E_GLNVM_SRCTL_DONE_MASK;
+
+    for (uint16_t i = 0; i < NUM_PFINTS; i++) {
+        intevs[i].vector = i;
+        if (intevs[i].armed) {
+            runner->event_cancel(intevs[i]);
+            intevs[i].armed = false;
+        }
+        intevs[i].time = 0;
+    }
 }
 
 shadow_ram::shadow_ram(i40e_bm &dev_)
@@ -633,6 +729,12 @@ void shadow_ram::write(uint16_t addr, uint16_t val)
 #endif
 }
 
+int_ev::int_ev()
+{
+    armed = false;
+    time = 0;
+}
+
 } //namespace i40e
 
 using namespace i40e;

i40e_bm/i40e_bm.h

@@ -28,6 +28,15 @@ class dma_base : public nicbm::DMAOp {
         virtual void done() = 0;
 };
 
+class int_ev : public nicbm::TimedEvent {
+    public:
+        uint16_t vector;
+        bool armed;
+
+        int_ev();
+};
+
+
 class logger : public std::ostream {
     public:
         static const char endl = '\n';
@@ -141,6 +150,8 @@ class queue_base {
                 desc_ctx &ctx;
 
             public:
+                size_t total_len;
+                size_t part_offset;
                 dma_data_fetch(desc_ctx &ctx_, size_t len, void *buffer);
                 virtual ~dma_data_fetch();
                 virtual void done();
@@ -150,6 +161,8 @@ class queue_base {
             protected:
                 desc_ctx &ctx;
             public:
+                size_t total_len;
+                size_t part_offset;
                 dma_data_wb(desc_ctx &ctx_, size_t len);
                 virtual ~dma_data_wb();
                 virtual void done();
@@ -311,7 +324,7 @@ class lan_queue_base : public queue_base {
 
 class lan_queue_tx : public lan_queue_base {
     protected:
-        static const uint16_t MTU = 2048;
+        static const uint16_t MTU = 9024;
 
         class tx_desc_ctx : public desc_ctx {
             protected:
@@ -342,6 +355,8 @@ class lan_queue_tx : public lan_queue_base {
         };
 
         uint8_t pktbuf[MTU];
+        uint32_t tso_off;
+        uint32_t tso_len;
         std::deque<tx_desc_ctx *> ready_segments;
 
         bool hwb;
@@ -372,7 +387,7 @@ class lan_queue_rx : public lan_queue_base {
             public:
                 rx_desc_ctx(lan_queue_rx &queue_);
                 virtual void process();
-                void packet_received(const void *data, size_t len);
+                void packet_received(const void *data, size_t len, bool last);
         };
 
         uint16_t dbuff_size;
@@ -443,14 +458,18 @@ protected:
     static const uint32_t NUM_PFINTS = 512;
     static const uint32_t NUM_VSIS = 384;
     static const uint16_t MAX_MTU = 2048;
+    static const uint8_t NUM_ITR = 3;
 
     struct i40e_regs {
         uint32_t glgen_rstctl;
+        uint32_t glgen_stat;
         uint32_t gllan_rctl_0;
         uint32_t pfint_lnklst0;
         uint32_t pfint_icr0_ena;
         uint32_t pfint_icr0;
+        uint32_t pfint_itr0[NUM_ITR];
 
+        uint32_t pfint_dyn_ctl0;
         uint32_t pfint_dyn_ctln[NUM_PFINTS - 1];
         uint32_t pfint_lnklstn[NUM_PFINTS - 1];
         uint32_t pfint_raten[NUM_PFINTS - 1];
@@ -492,8 +511,6 @@ protected:
     };
 
 public:
-    nicbm::Runner *runner;
-
     i40e_bm();
     ~i40e_bm();
 
@@ -505,6 +522,9 @@ public:
     virtual void reg_write32(uint8_t bar, uint64_t addr, uint32_t val);
     virtual void dma_complete(nicbm::DMAOp &op);
     virtual void eth_rx(uint8_t port, const void *data, size_t len);
+    virtual void timed_event(nicbm::TimedEvent &ev);
+
+    void signal_interrupt(uint16_t vector, uint8_t itr);
 
 protected:
     logger log;
@@ -514,6 +534,8 @@ protected:
     shadow_ram shram;
     lan lanmgr;
 
+    int_ev intevs[NUM_PFINTS];
+
     /** Read from the I/O bar */
     virtual uint32_t reg_io_read(uint64_t addr);
     /** Write to the I/O bar */
@@ -530,4 +552,12 @@ protected:
 // places the tcp checksum in the packet (assuming ipv4)
 void xsum_tcp(void *tcphdr, size_t l4len);
 
+// calculates the full ipv4 & tcp checksum without assuming any pseudo header
+// xsums
+void xsum_tcpip_tso(void *iphdr, uint8_t iplen, uint8_t l4len,
+        uint16_t paylen);
+
+void tso_postupdate_header(void *iphdr, uint8_t iplen, uint8_t l4len,
+        uint16_t paylen);
+
 } // namespace corundum