28. ICE Poll Mode Driver

The ice PMD (librte_net_ice) provides poll mode driver support for 10/25/50/100 Gbps Intel® Ethernet 800 Series Network Adapters based on the Intel Ethernet Controller E810 and Intel Ethernet Connection E822/E823.

28.1. Linux Prerequisites

28.2. Windows Prerequisites

  • Follow the guide for Windows to setup the basic DPDK environment.
  • Identify the Intel® Ethernet adapter and get the latest NVM/FW version.
  • To access any Intel® Ethernet hardware, load the NetUIO driver in place of existing built-in (inbox) driver.
  • To load NetUIO driver, follow the steps mentioned in dpdk-kmods repository.
  • Loading of private Dynamic Device Personalization (DDP) package is not supported on Windows.

28.4. Configuration

28.4.1. Runtime Configuration

  • Safe Mode Support (default 0)

    If driver failed to load OS package, by default driver’s initialization failed. But if user intend to use the device without OS package, user can take devargs parameter safe-mode-support, for example:

    -a 80:00.0,safe-mode-support=1
    

    Then the driver will be initialized successfully and the device will enter Safe Mode. NOTE: In Safe mode, only very limited features are available, features like RSS, checksum, fdir, tunneling … are all disabled.

  • Generic Flow Pipeline Mode Support (default 0)

    In pipeline mode, a flow can be set at one specific stage by setting parameter priority. Currently, we support two stages: priority = 0 or !0. Flows with priority 0 located at the first pipeline stage which typically be used as a firewall to drop the packet on a blocklist(we called it permission stage). At this stage, flow rules are created for the device’s exact match engine: switch. Flows with priority !0 located at the second stage, typically packets are classified here and be steered to specific queue or queue group (we called it distribution stage), At this stage, flow rules are created for device’s flow director engine. For none-pipeline mode, priority is ignored, a flow rule can be created as a flow director rule or a switch rule depends on its pattern/action and the resource allocation situation, all flows are virtually at the same pipeline stage. By default, generic flow API is enabled in none-pipeline mode, user can choose to use pipeline mode by setting devargs parameter pipeline-mode-support, for example:

    -a 80:00.0,pipeline-mode-support=1
    
  • Default MAC Disable (default 0)

    Disable the default MAC make the device drop all packets by default, only packets hit on filter rules will pass.

    Default MAC can be disabled by setting the devargs parameter default-mac-disable, for example:

    -a 80:00.0,default-mac-disable=1
    
  • Protocol extraction for per queue

    Configure the RX queues to do protocol extraction into mbuf for protocol handling acceleration, like checking the TCP SYN packets quickly.

    The argument format is:

    18:00.0,proto_xtr=<queues:protocol>[<queues:protocol>...],field_offs=<offset>, \
    field_name=<name>
    18:00.0,proto_xtr=<protocol>,field_offs=<offset>,field_name=<name>
    

    Queues are grouped by ( and ) within the group. The - character is used as a range separator and , is used as a single number separator. The grouping () can be omitted for single element group. If no queues are specified, PMD will use this protocol extraction type for all queues. field_offs is the offset of mbuf dynamic field for protocol extraction data. field_name is the name of mbuf dynamic field for protocol extraction data. field_offs and field_name will be checked whether it is valid. If invalid, an error print will be returned: Invalid field offset or name, no match dynfield, and the proto_ext function will not be enabled.

    Protocol is : vlan, ipv4, ipv6, ipv6_flow, tcp, ip_offset.

    dpdk-testpmd -c 0xff -- -i
    port stop 0
    port detach 0
    port attach 18:00.0,proto_xtr='[(1,2-3,8-9):tcp,10-13:vlan]',field_offs=92,field_name=pmd_dyn
    

    This setting means queues 1, 2-3, 8-9 are TCP extraction, queues 10-13 are VLAN extraction, other queues run with no protocol extraction. The offset of mbuf dynamic field is 92 for all queues with protocol extraction.

    dpdk-testpmd -c 0xff -- -i
    port stop 0
    port detach 0
    port attach 18:00.0,proto_xtr=vlan,proto_xtr='[(1,2-3,8-9):tcp,10-23:ipv6]', \
    field_offs=92,field_name=pmd_dyn
    

    This setting means queues 1, 2-3, 8-9 are TCP extraction, queues 10-23 are IPv6 extraction, other queues use the default VLAN extraction. The offset of mbuf dynamic field is 92 for all queues with protocol extraction.

    The extraction metadata is copied into the registered dynamic mbuf field, and the related dynamic mbuf flags is set.

    Table 28.4 Protocol extraction : vlan
    VLAN2 VLAN1
    PCP D VID PCP D VID

    VLAN1 - single or EVLAN (first for QinQ).

    VLAN2 - C-VLAN (second for QinQ).

    Table 28.5 Protocol extraction : ipv4
    IPHDR2 IPHDR1
    Ver Hdr Len ToS TTL Protocol

    IPHDR1 - IPv4 header word 4, “TTL” and “Protocol” fields.

    IPHDR2 - IPv4 header word 0, “Ver”, “Hdr Len” and “Type of Service” fields.

    Table 28.6 Protocol extraction : ipv6
    IPHDR2 IPHDR1
    Ver Traffic class Flow Next Header Hop Limit

    IPHDR1 - IPv6 header word 3, “Next Header” and “Hop Limit” fields.

    IPHDR2 - IPv6 header word 0, “Ver”, “Traffic class” and high 4 bits of “Flow Label” fields.

    Table 28.7 Protocol extraction : ipv6_flow
    IPHDR2 IPHDR1
    Ver Traffic class Flow Label

    IPHDR1 - IPv6 header word 1, 16 low bits of the “Flow Label” field.

    IPHDR2 - IPv6 header word 0, “Ver”, “Traffic class” and high 4 bits of “Flow Label” fields.

    Table 28.8 Protocol extraction : tcp
    TCPHDR2 TCPHDR1
    Reserved Offset RSV Flags

    TCPHDR1 - TCP header word 6, “Data Offset” and “Flags” fields.

    TCPHDR2 - Reserved

    Table 28.9 Protocol extraction : ip_offset
    IPHDR2 IPHDR1
    IPv6 HDR Offset IPv4 HDR Offset

    IPHDR1 - Outer/Single IPv4 Header offset.

    IPHDR2 - Outer/Single IPv6 Header offset.

  • Hardware debug mask log support (default 0)

    User can enable the related hardware debug mask such as ICE_DBG_NVM:

    -a 0000:88:00.0,hw_debug_mask=0x80 --log-level=pmd.net.ice.driver:8
    

    These ICE_DBG_XXX are defined in drivers/net/ice/base/ice_type.h.

  • 1PPS out support

    The E810 supports four single-ended GPIO signals (SDP[20:23]). The 1PPS signal outputs via SDP[20:23]. User can select GPIO pin index flexibly. Pin index 0 means SDP20, 1 means SDP21 and so on. For example:

    -a af:00.0,pps_out='[pin:0]'
    
  • Low Rx latency (default 0)

    vRAN workloads require low latency DPDK interface for the front haul interface connection to Radio. By specifying 1 for parameter rx_low_latency, each completed Rx descriptor can be written immediately to host memory and the Rx interrupt latency can be reduced to 2us:

    -a 0000:88:00.0,rx_low_latency=1
    

    As a trade-off, this configuration may cause the packet processing performance degradation due to the PCI bandwidth limitation.

28.5. Driver compilation and testing

Refer to the document compiling and testing a PMD for a NIC for details.

28.6. Features

28.6.1. Vector PMD

Vector PMD for RX and TX path are selected automatically. The paths are chosen based on 2 conditions.

  • CPU On the X86 platform, the driver checks if the CPU supports AVX2. If it’s supported, AVX2 paths will be chosen. If not, SSE is chosen. If the CPU supports AVX512 and EAL argument --force-max-simd-bitwidth is set to 512, AVX512 paths will be chosen.
  • Offload features The supported HW offload features are described in the document ice.ini, A value “P” means the offload feature is not supported by vector path. If any not supported features are used, ICE vector PMD is disabled and the normal paths are chosen.

28.6.2. Malicious driver detection (MDD)

It’s not appropriate to send a packet, if this packet’s destination MAC address is just this port’s MAC address. If SW tries to send such packets, HW will report a MDD event and drop the packets.

The APPs based on DPDK should avoid providing such packets.

28.6.3. Device Config Function (DCF)

This section demonstrates ICE DCF PMD, which shares the core module with ICE PMD and iAVF PMD.

A DCF (Device Config Function) PMD bounds to the device’s trusted VF with ID 0, it can act as a sole controlling entity to exercise advance functionality (such as switch, ACL) for the rest VFs.

The DCF PMD needs to advertise and acquire DCF capability which allows DCF to send AdminQ commands that it would like to execute over to the PF and receive responses for the same from PF.

28.6.3.1. Additional Options

  • Disable ACL Engine (default enabled)

    By default, all flow engines are enabled. But if user does not need the ACL engine related functions, user can set devargs parameter acl=off to disable the ACL engine and shorten the startup time.

    -a 18:01.0,cap=dcf,acl=off

../_images/ice_dcf.svg

Fig. 28.2 DCF Communication flow.

  1. Create the VFs:

    echo 4 > /sys/bus/pci/devices/0000\:18\:00.0/sriov_numvfs
    
  2. Enable the VF0 trust on:

    ip link set dev enp24s0f0 vf 0 trust on
    
  3. Bind the VF0, and run testpmd with ‘cap=dcf’ with port representor for VF 1 and 2:

    dpdk-testpmd -l 22-25 -n 4 -a 18:01.0,cap=dcf,representor=vf[1-2] -- -i
    
  4. Monitor the VF2 interface network traffic:

    tcpdump -e -nn -i enp24s1f2
    
  5. Create one flow to redirect the traffic to VF2 by DCF (assume the representor port ID is 5):

    flow create 0 priority 0 ingress pattern eth / ipv4 src is 192.168.0.2 \
    dst is 192.168.0.3 / end actions represented_port ethdev_port_id 5 / end
    
  6. Send the packet, and it should be displayed on tcpdump:

    sendp(Ether(src='3c:fd:fe:aa:bb:78', dst='00:00:00:01:02:03')/IP(src=' \
    192.168.0.2', dst="192.168.0.3")/TCP(flags='S')/Raw(load='XXXXXXXXXX'), \
    iface="enp24s0f0", count=10)
    

28.7. Sample Application Notes

28.7.1. Vlan filter

Vlan filter only works when Promiscuous mode is off.

To start testpmd, and add vlan 10 to port 0:

./app/dpdk-testpmd -l 0-15 -n 4 -- -i
...

testpmd> rx_vlan add 10 0

28.8. Limitations or Known issues

The Intel E810 requires a programmable pipeline package be downloaded by the driver to support normal operations. The E810 has a limited functionality built in to allow PXE boot and other use cases, but the driver must download a package file during the driver initialization stage.

The default DDP package file name is ice.pkg. For a specific NIC, the DDP package supposed to be loaded can have a filename: ice-xxxxxx.pkg, where ‘xxxxxx’ is the 64-bit PCIe Device Serial Number of the NIC. For example, if the NIC’s device serial number is 00-CC-BB-FF-FF-AA-05-68, the device-specific DDP package filename is ice-00ccbbffffaa0568.pkg (in hex and all low case). During initialization, the driver searches in the following paths in order: /lib/firmware/updates/intel/ice/ddp and /lib/firmware/intel/ice/ddp. The corresponding device-specific DDP package will be downloaded first if the file exists. If not, then the driver tries to load the default package. The type of loaded package is stored in ice_adapter->active_pkg_type.

A symbolic link to the DDP package file is also ok. The same package file is used by both the kernel driver and the DPDK PMD.

Note

Windows support: The DDP package is not supported on Windows so, loading of the package is disabled on Windows.