159. Tx Preparation Forwarding Tests¶

The support of TX preparation forwarding feature consists in:

Do necessary preparations of packet burst to be safely transmitted on device for desired HW offloads: Set/reset checksum field according to the hardware requirements. Check HW constraints (number of segments per packet, etc).
Provide information about max segments of TSO and non-TSO packets accepted by device.

APPLICATION (CASE OF USE):

Application should initialize burst of packets to send, set required tx offload flags and required fields, like l2_len, l3_len, l4_len and tso_segsz.
Application passes burst to check required conditions to send packets through the NIC.
The result can be used to send valid packets and restore invalid packets if function fails.

159.1. Prerequisites¶

Support igb_uio, test txprep forwarding features on e1000, i40e, ice, igc and ixgbe drivers.Send packets from tester platform through the interface eth1 to the tested port 0, then testpmd sends back packet using same port and uses tcpdump to capture packet information:

Tester      DUT
eth1  <---> port 0

Turn off all hardware offloads on tester machine:

ethtool -K eth1 rx off tx off tso off gso off gro off lro off

Change mtu for large packet:

ifconfig eth1 mtu 9000

Launch the testpmd with the following arguments, set --tx-offloads=0x8000 to let TX checksum offloads, TSO mode in the “Full Featured” TX path, add –max-pkt-len for large packet:

./app/dpdk-testpmd -c 0x6 -n 4  -- -i --tx-offloads=0x8000 --port-topology=chained
--max-pkt-len=9000

Set the csum forwarding mode:

testpmd> set fwd csum

Set the verbose level to 1 to display information for each received packet:

testpmd> set verbose 1

Enable hardware checksum for IP/TCP/UDP packets:

testpmd> csum set ip hw 0
testpmd> csum set tcp hw 0
testpmd> csum set udp hw 0

Due to DPDK 236bc417e2da(app/testpmd: fix MAC header in checksum forward engine) changes the checksum functions adds switches to control whether to exchange MAC address. Currently, our test scripts are based on not exchanging MAC addresses, mac-swap needs to be disabled:

testpmd> csum mac-swap off 0

159.2. Test Case: TX preparation forwarding of non-TSO packets¶

Set TSO turned off:

testpmd> tso set 0 0

Start the packet forwarding:

testpmd> start

Send few IP/TCP/UDP packets from tester machine to DUT. Check IP/TCP/UDP checksum correctness in captured packet, such as correct as below:

Transmitted packet:

03:06:36.569730 3c:fd:fe:9d:64:30 > 90:e2:ba:63:22:e8, ethertype IPv4
(0x0800), length 104: (tos 0x0, ttl 64, id 1, offset 0, flags [none],
proto TCP (6), length 90)
  127.0.0.1.ftp-data > 127.0.0.1.http: Flags [.], cksum 0x1998 (correct),
seq 0:50, ack 0, win 8192, length 50: HTTP

Captured packet:

03:06:36.569816 90:e2:ba:63:22:e8 > 02:00:00:00:00:00, ethertype IPv4
(0x0800), length 104: (tos 0x0, ttl 64, id 1, offset 0, flags [none],
proto TCP (6), length 90)
  127.0.0.1.ftp-data > 127.0.0.1.http: Flags [.], cksum 0x1998 (correct),
seq 0:50, ack 1, win 8192, length 50: HTTP

159.3. Test Case: TX preparation forwarding of TSO packets¶

Set TSO turned on:

testpmd> tso set 1460 0
TSO segment size for non-tunneled packets is 1460

Start the packet forwarding:

testpmd> start

Send few IP/TCP packets from tester machine to DUT. Check IP/TCP checksum correctness in captured packet and verify correctness of HW TSO offload for large packets. One large TCP packet (5214 bytes + headers) segmented to four fragments (1460 bytes+header,1460 bytes+header,1460 bytes+header and 834 bytes + headers), checksums are also ok:

Transmitted packet:

21:48:24.214136 00:00:00:00:00:00 > 3c:fd:fe:9d:69:68, ethertype IPv6
(0x86dd), length 5288: (hlim 64, next-header TCP (6) payload length: 5234)
 ::1.ftp-data > ::1.http: Flags [.], cksum 0xac95 (correct), seq 0:5214,
ack 1, win 8192, length 5214: HTTP

Captured packet:

21:48:24.214207 3c:fd:fe:9d:69:68 > 02:00:00:00:00:00, ethertype IPv6
(0x86dd), length 1534: (hlim 64, next-header TCP (6) payload length: 1480)
::1.ftp-data > ::1.http: Flags [.], cksum 0xa641 (correct), seq 0:1460,
ack 1, win 8192, length 1460: HTTP
21:48:24.214212 3c:fd:fe:9d:69:68 > 02:00:00:00:00:00, ethertype IPv6
(0x86dd), length 1534: (hlim 64, next-header TCP (6) payload length: 1480)
::1.ftp-data > ::1.http: Flags [.], cksum 0xae89 (correct), seq 1460:2920,
ack 1, win 8192, length 1460: HTTP
21:48:24.214213 3c:fd:fe:9d:69:68 > 02:00:00:00:00:00, ethertype IPv6
(0x86dd), length 1534: (hlim 64, next-header TCP (6) payload length: 1480)
::1.ftp-data > ::1.http: Flags [.], cksum 0xfdb6 (correct), seq 2920:4380,
ack 1, win 8192, length 1460: HTTP
21:48:24.214215 3c:fd:fe:9d:69:68 > 02:00:00:00:00:00, ethertype IPv6
(0x86dd), length 908: (hlim 64, next-header TCP (6) payload length: 854)
::1.ftp-data > ::1.http: Flags [.], cksum 0xe629 (correct), seq 4380:5214,
ack 1, win 8192, length 834: HTTP

Note: Generally TSO only supports TCP packets but doesn’t support UDP packets due to hardware segmentation limitation, for example packets are sent on 82599 NIC, but not segmented.

Packet:

########
# IPv4 #
########

# checksum TCP
p=Ether()/IP()/TCP(flags=0x10)/Raw(RandString(50))

# bad IP checksum
p=Ether()/IP(chksum=0x1234)/TCP(flags=0x10)/Raw(RandString(50))

# bad TCP checksum
p=Ether()/IP()/TCP(flags=0x10, chksum=0x1234)/Raw(RandString(50))

# large packet
p=Ether()/IP()/TCP(flags=0x10)/Raw(RandString(length))

# bad checksum and large packet
p=Ether()/IP(chksum=0x1234)/TCP(flags=0x10,chksum=0x1234)/
Raw(RandString(length))


########
# IPv6 #
########

# checksum TCP
p=Ether()/IPv6()/TCP(flags=0x10)/Raw(RandString(50))

# checksum UDP
p=Ether()/IPv6()/UDP()/Raw(RandString(50))

# bad TCP checksum
p=Ether()/IPv6()/TCP(flags=0x10, chksum=0x1234)/Raw(RandString(50))

# large packet
p=Ether()/IPv6()/TCP(flags=0x10)/Raw(RandString(length))