blob: 67b7a701ce9e15a0bde78072569f1c31d7959295 [file] [log] [blame]
.. SPDX-License-Identifier: GPL-2.0+
Linux Base Driver for the Intel(R) Ethernet Controller 800 Series
Intel ice Linux driver.
Copyright(c) 2018-2021 Intel Corporation.
- Overview
- Identifying Your Adapter
- Important Notes
- Additional Features & Configurations
- Performance Optimization
The associated Virtual Function (VF) driver for this driver is iavf.
Driver information can be obtained using ethtool and lspci.
For questions related to hardware requirements, refer to the documentation
supplied with your Intel adapter. All hardware requirements listed apply to use
with Linux.
This driver supports XDP (Express Data Path) and AF_XDP zero-copy. Note that
XDP is blocked for frame sizes larger than 3KB.
Identifying Your Adapter
For information on how to identify your adapter, and for the latest Intel
network drivers, refer to the Intel Support website:
Important Notes
Packet drops may occur under receive stress
Devices based on the Intel(R) Ethernet Controller 800 Series are designed to
tolerate a limited amount of system latency during PCIe and DMA transactions.
If these transactions take longer than the tolerated latency, it can impact the
length of time the packets are buffered in the device and associated memory,
which may result in dropped packets. These packets drops typically do not have
a noticeable impact on throughput and performance under standard workloads.
If these packet drops appear to affect your workload, the following may improve
the situation:
1) Make sure that your system's physical memory is in a high-performance
configuration, as recommended by the platform vendor. A common
recommendation is for all channels to be populated with a single DIMM
2) In your system's BIOS/UEFI settings, select the "Performance" profile.
3) Your distribution may provide tools like "tuned," which can help tweak
kernel settings to achieve better standard settings for different workloads.
Configuring SR-IOV for improved network security
In a virtualized environment, on Intel(R) Ethernet Network Adapters that
support SR-IOV, the virtual function (VF) may be subject to malicious behavior.
Software-generated layer two frames, like IEEE 802.3x (link flow control), IEEE
802.1Qbb (priority based flow-control), and others of this type, are not
expected and can throttle traffic between the host and the virtual switch,
reducing performance. To resolve this issue, and to ensure isolation from
unintended traffic streams, configure all SR-IOV enabled ports for VLAN tagging
from the administrative interface on the PF. This configuration allows
unexpected, and potentially malicious, frames to be dropped.
See "Configuring VLAN Tagging on SR-IOV Enabled Adapter Ports" later in this
README for configuration instructions.
Do not unload port driver if VF with active VM is bound to it
Do not unload a port's driver if a Virtual Function (VF) with an active Virtual
Machine (VM) is bound to it. Doing so will cause the port to appear to hang.
Once the VM shuts down, or otherwise releases the VF, the command will
Important notes for SR-IOV and Link Aggregation
Link Aggregation is mutually exclusive with SR-IOV.
- If Link Aggregation is active, SR-IOV VFs cannot be created on the PF.
- If SR-IOV is active, you cannot set up Link Aggregation on the interface.
Bridging and MACVLAN are also affected by this. If you wish to use bridging or
MACVLAN with SR-IOV, you must set up bridging or MACVLAN before enabling
SR-IOV. If you are using bridging or MACVLAN in conjunction with SR-IOV, and
you want to remove the interface from the bridge or MACVLAN, you must follow
these steps:
1. Destroy SR-IOV VFs if they exist
2. Remove the interface from the bridge or MACVLAN
3. Recreate SRIOV VFs as needed
Additional Features and Configurations
The driver utilizes the ethtool interface for driver configuration and
diagnostics, as well as displaying statistical information. The latest ethtool
version is required for this functionality. Download it at:
NOTE: The rx_bytes value of ethtool does not match the rx_bytes value of
Netdev, due to the 4-byte CRC being stripped by the device. The difference
between the two rx_bytes values will be 4 x the number of Rx packets. For
example, if Rx packets are 10 and Netdev (software statistics) displays
rx_bytes as "X", then ethtool (hardware statistics) will display rx_bytes as
"X+40" (4 bytes CRC x 10 packets).
Viewing Link Messages
Link messages will not be displayed to the console if the distribution is
restricting system messages. In order to see network driver link messages on
your console, set dmesg to eight by entering the following::
# dmesg -n 8
NOTE: This setting is not saved across reboots.
Dynamic Device Personalization
Dynamic Device Personalization (DDP) allows you to change the packet processing
pipeline of a device by applying a profile package to the device at runtime.
Profiles can be used to, for example, add support for new protocols, change
existing protocols, or change default settings. DDP profiles can also be rolled
back without rebooting the system.
The DDP package loads during device initialization. The driver looks for
``intel/ice/ddp/ice.pkg`` in your firmware root (typically ``/lib/firmware/``
or ``/lib/firmware/updates/``) and checks that it contains a valid DDP package
NOTE: Your distribution should likely have provided the latest DDP file, but if
ice.pkg is missing, you can find it in the linux-firmware repository or from
If the driver is unable to load the DDP package, the device will enter Safe
Mode. Safe Mode disables advanced and performance features and supports only
basic traffic and minimal functionality, such as updating the NVM or
downloading a new driver or DDP package. Safe Mode only applies to the affected
physical function and does not impact any other PFs. See the "Intel(R) Ethernet
Adapters and Devices User Guide" for more details on DDP and Safe Mode.
- If you encounter issues with the DDP package file, you may need to download
an updated driver or DDP package file. See the log messages for more
- The ice.pkg file is a symbolic link to the default DDP package file.
- You cannot update the DDP package if any PF drivers are already loaded. To
overwrite a package, unload all PFs and then reload the driver with the new
- Only the first loaded PF per device can download a package for that device.
You can install specific DDP package files for different physical devices in
the same system. To install a specific DDP package file:
1. Download the DDP package file you want for your device.
2. Rename the file ice-xxxxxxxxxxxxxxxx.pkg, where 'xxxxxxxxxxxxxxxx' is the
unique 64-bit PCI Express device serial number (in hex) of the device you
want the package downloaded on. The filename must include the complete
serial number (including leading zeros) and be all lowercase. For example,
if the 64-bit serial number is b887a3ffffca0568, then the file name would be
To find the serial number from the PCI bus address, you can use the
following command::
# lspci -vv -s af:00.0 | grep -i Serial
Capabilities: [150 v1] Device Serial Number b8-87-a3-ff-ff-ca-05-68
You can use the following command to format the serial number without the
# lspci -vv -s af:00.0 | grep -i Serial | awk '{print $7}' | sed s/-//g
3. Copy the renamed DDP package file to
``/lib/firmware/updates/intel/ice/ddp/``. If the directory does not yet
exist, create it before copying the file.
4. Unload all of the PFs on the device.
5. Reload the driver with the new package.
NOTE: The presence of a device-specific DDP package file overrides the loading
of the default DDP package file (ice.pkg).
Intel(R) Ethernet Flow Director
The Intel Ethernet Flow Director performs the following tasks:
- Directs receive packets according to their flows to different queues
- Enables tight control on routing a flow in the platform
- Matches flows and CPU cores for flow affinity
NOTE: This driver supports the following flow types:
- IPv4
- TCPv4
- UDPv4
- SCTPv4
- IPv6
- TCPv6
- UDPv6
- SCTPv6
Each flow type supports valid combinations of IP addresses (source or
destination) and UDP/TCP/SCTP ports (source and destination). You can supply
only a source IP address, a source IP address and a destination port, or any
combination of one or more of these four parameters.
NOTE: This driver allows you to filter traffic based on a user-defined flexible
two-byte pattern and offset by using the ethtool user-def and mask fields. Only
L3 and L4 flow types are supported for user-defined flexible filters. For a
given flow type, you must clear all Intel Ethernet Flow Director filters before
changing the input set (for that flow type).
Flow Director Filters
Flow Director filters are used to direct traffic that matches specified
characteristics. They are enabled through ethtool's ntuple interface. To enable
or disable the Intel Ethernet Flow Director and these filters::
# ethtool -K <ethX> ntuple <off|on>
NOTE: When you disable ntuple filters, all the user programmed filters are
flushed from the driver cache and hardware. All needed filters must be re-added
when ntuple is re-enabled.
To display all of the active filters::
# ethtool -u <ethX>
To add a new filter::
# ethtool -U <ethX> flow-type <type> src-ip <ip> [m <ip_mask>] dst-ip <ip>
[m <ip_mask>] src-port <port> [m <port_mask>] dst-port <port> [m <port_mask>]
action <queue>
<ethX> - the Ethernet device to program
<type> - can be ip4, tcp4, udp4, sctp4, ip6, tcp6, udp6, sctp6
<ip> - the IP address to match on
<ip_mask> - the IPv4 address to mask on
NOTE: These filters use inverted masks.
<port> - the port number to match on
<port_mask> - the 16-bit integer for masking
NOTE: These filters use inverted masks.
<queue> - the queue to direct traffic toward (-1 discards the
matched traffic)
To delete a filter::
# ethtool -U <ethX> delete <N>
Where <N> is the filter ID displayed when printing all the active filters,
and may also have been specified using "loc <N>" when adding the filter.
To add a filter that directs packet to queue 2::
# ethtool -U <ethX> flow-type tcp4 src-ip dst-ip \ src-port 2000 dst-port 2001 action 2 [loc 1]
To set a filter using only the source and destination IP address::
# ethtool -U <ethX> flow-type tcp4 src-ip dst-ip \ action 2 [loc 1]
To set a filter based on a user-defined pattern and offset::
# ethtool -U <ethX> flow-type tcp4 src-ip dst-ip \ user-def 0x4FFFF action 2 [loc 1]
where the value of the user-def field contains the offset (4 bytes) and
the pattern (0xffff).
To match TCP traffic sent from, port 5300, directed to,
port 80, and then send it to queue 7::
# ethtool -U enp130s0 flow-type tcp4 src-ip dst-ip
src-port 5300 dst-port 80 action 7
To add a TCPv4 filter with a partial mask for a source IP subnet::
# ethtool -U <ethX> flow-type tcp4 src-ip m dst-ip src-port 12600 dst-port 31 action 12
For each flow-type, the programmed filters must all have the same matching
input set. For example, issuing the following two commands is acceptable::
# ethtool -U enp130s0 flow-type ip4 src-ip src-port 5300 action 7
# ethtool -U enp130s0 flow-type ip4 src-ip src-port 55 action 10
Issuing the next two commands, however, is not acceptable, since the first
specifies src-ip and the second specifies dst-ip::
# ethtool -U enp130s0 flow-type ip4 src-ip src-port 5300 action 7
# ethtool -U enp130s0 flow-type ip4 dst-ip src-port 55 action 10
The second command will fail with an error. You may program multiple filters
with the same fields, using different values, but, on one device, you may not
program two tcp4 filters with different matching fields.
The ice driver does not support matching on a subportion of a field, thus
partial mask fields are not supported.
Flex Byte Flow Director Filters
The driver also supports matching user-defined data within the packet payload.
This flexible data is specified using the "user-def" field of the ethtool
command in the following way:
.. table::
============================== ============================
``31 28 24 20 16`` ``15 12 8 4 0``
``offset into packet payload`` ``2 bytes of flexible data``
============================== ============================
For example,
... user-def 0x4FFFF ...
tells the filter to look 4 bytes into the payload and match that value against
0xFFFF. The offset is based on the beginning of the payload, and not the
beginning of the packet. Thus
flow-type tcp4 ... user-def 0x8BEAF ...
would match TCP/IPv4 packets which have the value 0xBEAF 8 bytes into the
TCP/IPv4 payload.
Note that ICMP headers are parsed as 4 bytes of header and 4 bytes of payload.
Thus to match the first byte of the payload, you must actually add 4 bytes to
the offset. Also note that ip4 filters match both ICMP frames as well as raw
(unknown) ip4 frames, where the payload will be the L3 payload of the IP4
The maximum offset is 64. The hardware will only read up to 64 bytes of data
from the payload. The offset must be even because the flexible data is 2 bytes
long and must be aligned to byte 0 of the packet payload.
The user-defined flexible offset is also considered part of the input set and
cannot be programmed separately for multiple filters of the same type. However,
the flexible data is not part of the input set and multiple filters may use the
same offset but match against different data.
RSS Hash Flow
Allows you to set the hash bytes per flow type and any combination of one or
more options for Receive Side Scaling (RSS) hash byte configuration.
# ethtool -N <ethX> rx-flow-hash <type> <option>
Where <type> is:
tcp4 signifying TCP over IPv4
udp4 signifying UDP over IPv4
tcp6 signifying TCP over IPv6
udp6 signifying UDP over IPv6
And <option> is one or more of:
s Hash on the IP source address of the Rx packet.
d Hash on the IP destination address of the Rx packet.
f Hash on bytes 0 and 1 of the Layer 4 header of the Rx packet.
n Hash on bytes 2 and 3 of the Layer 4 header of the Rx packet.
Accelerated Receive Flow Steering (aRFS)
Devices based on the Intel(R) Ethernet Controller 800 Series support
Accelerated Receive Flow Steering (aRFS) on the PF. aRFS is a load-balancing
mechanism that allows you to direct packets to the same CPU where an
application is running or consuming the packets in that flow.
- aRFS requires that ntuple filtering is enabled via ethtool.
- aRFS support is limited to the following packet types:
- TCP over IPv4 and IPv6
- UDP over IPv4 and IPv6
- Nonfragmented packets
- aRFS only supports Flow Director filters, which consist of the
source/destination IP addresses and source/destination ports.
- aRFS and ethtool's ntuple interface both use the device's Flow Director. aRFS
and ntuple features can coexist, but you may encounter unexpected results if
there's a conflict between aRFS and ntuple requests. See "Intel(R) Ethernet
Flow Director" for additional information.
To set up aRFS:
1. Enable the Intel Ethernet Flow Director and ntuple filters using ethtool.
# ethtool -K <ethX> ntuple on
2. Set up the number of entries in the global flow table. For example:
# echo $NUM_RPS_ENTRIES > /proc/sys/net/core/rps_sock_flow_entries
3. Set up the number of entries in the per-queue flow table. For example:
# for file in /sys/class/net/$IFACE/queues/rx-*/rps_flow_cnt; do
# echo $(($NUM_RPS_ENTRIES/$NUM_RX_QUEUES)) > $file;
# done
4. Disable the IRQ balance daemon (this is only a temporary stop of the service
until the next reboot).
# systemctl stop irqbalance
5. Configure the interrupt affinity.
See ``/Documentation/core-api/irq/irq-affinity.rst``
To disable aRFS using ethtool::
# ethtool -K <ethX> ntuple off
NOTE: This command will disable ntuple filters and clear any aRFS filters in
software and hardware.
Example Use Case:
1. Set the server application on the desired CPU (e.g., CPU 4).
# taskset -c 4 netserver
2. Use netperf to route traffic from the client to CPU 4 on the server with
aRFS configured. This example uses TCP over IPv4.
# netperf -H <Host IPv4 Address> -t TCP_STREAM
Enabling Virtual Functions (VFs)
Use sysfs to enable virtual functions (VF).
For example, you can create 4 VFs as follows::
# echo 4 > /sys/class/net/<ethX>/device/sriov_numvfs
To disable VFs, write 0 to the same file::
# echo 0 > /sys/class/net/<ethX>/device/sriov_numvfs
The maximum number of VFs for the ice driver is 256 total (all ports). To check
how many VFs each PF supports, use the following command::
# cat /sys/class/net/<ethX>/device/sriov_totalvfs
Note: You cannot use SR-IOV when link aggregation (LAG)/bonding is active, and
vice versa. To enforce this, the driver checks for this mutual exclusion.
Displaying VF Statistics on the PF
Use the following command to display the statistics for the PF and its VFs::
# ip -s link show dev <ethX>
NOTE: The output of this command can be very large due to the maximum number of
possible VFs.
The PF driver will display a subset of the statistics for the PF and for all
VFs that are configured. The PF will always print a statistics block for each
of the possible VFs, and it will show zero for all unconfigured VFs.
Configuring VLAN Tagging on SR-IOV Enabled Adapter Ports
To configure VLAN tagging for the ports on an SR-IOV enabled adapter, use the
following command. The VLAN configuration should be done before the VF driver
is loaded or the VM is booted. The VF is not aware of the VLAN tag being
inserted on transmit and removed on received frames (sometimes called "port
VLAN" mode).
# ip link set dev <ethX> vf <id> vlan <vlan id>
For example, the following will configure PF eth0 and the first VF on VLAN 10::
# ip link set dev eth0 vf 0 vlan 10
Enabling a VF link if the port is disconnected
If the physical function (PF) link is down, you can force link up (from the
host PF) on any virtual functions (VF) bound to the PF.
For example, to force link up on VF 0 bound to PF eth0::
# ip link set eth0 vf 0 state enable
Note: If the command does not work, it may not be supported by your system.
Setting the MAC Address for a VF
To change the MAC address for the specified VF::
# ip link set <ethX> vf 0 mac <address>
For example::
# ip link set <ethX> vf 0 mac 00:01:02:03:04:05
This setting lasts until the PF is reloaded.
NOTE: Assigning a MAC address for a VF from the host will disable any
subsequent requests to change the MAC address from within the VM. This is a
security feature. The VM is not aware of this restriction, so if this is
attempted in the VM, it will trigger MDD events.
Trusted VFs and VF Promiscuous Mode
This feature allows you to designate a particular VF as trusted and allows that
trusted VF to request selective promiscuous mode on the Physical Function (PF).
To set a VF as trusted or untrusted, enter the following command in the
# ip link set dev <ethX> vf 1 trust [on|off]
NOTE: It's important to set the VF to trusted before setting promiscuous mode.
If the VM is not trusted, the PF will ignore promiscuous mode requests from the
VF. If the VM becomes trusted after the VF driver is loaded, you must make a
new request to set the VF to promiscuous.
Once the VF is designated as trusted, use the following commands in the VM to
set the VF to promiscuous mode.
For promiscuous all::
# ip link set <ethX> promisc on
Where <ethX> is a VF interface in the VM
For promiscuous Multicast::
# ip link set <ethX> allmulticast on
Where <ethX> is a VF interface in the VM
NOTE: By default, the ethtool private flag vf-true-promisc-support is set to
"off," meaning that promiscuous mode for the VF will be limited. To set the
promiscuous mode for the VF to true promiscuous and allow the VF to see all
ingress traffic, use the following command::
# ethtool --set-priv-flags <ethX> vf-true-promisc-support on
The vf-true-promisc-support private flag does not enable promiscuous mode;
rather, it designates which type of promiscuous mode (limited or true) you will
get when you enable promiscuous mode using the ip link commands above. Note
that this is a global setting that affects the entire device. However, the
vf-true-promisc-support private flag is only exposed to the first PF of the
device. The PF remains in limited promiscuous mode regardless of the
vf-true-promisc-support setting.
Next, add a VLAN interface on the VF interface. For example::
# ip link add link eth2 name eth2.100 type vlan id 100
Note that the order in which you set the VF to promiscuous mode and add the
VLAN interface does not matter (you can do either first). The result in this
example is that the VF will get all traffic that is tagged with VLAN 100.
Malicious Driver Detection (MDD) for VFs
Some Intel Ethernet devices use Malicious Driver Detection (MDD) to detect
malicious traffic from the VF and disable Tx/Rx queues or drop the offending
packet until a VF driver reset occurs. You can view MDD messages in the PF's
system log using the dmesg command.
- If the PF driver logs MDD events from the VF, confirm that the correct VF
driver is installed.
- To restore functionality, you can manually reload the VF or VM or enable
automatic VF resets.
- When automatic VF resets are enabled, the PF driver will immediately reset
the VF and reenable queues when it detects MDD events on the receive path.
- If automatic VF resets are disabled, the PF will not automatically reset the
VF when it detects MDD events.
To enable or disable automatic VF resets, use the following command::
# ethtool --set-priv-flags <ethX> mdd-auto-reset-vf on|off
MAC and VLAN Anti-Spoofing Feature for VFs
When a malicious driver on a Virtual Function (VF) interface attempts to send a
spoofed packet, it is dropped by the hardware and not transmitted.
NOTE: This feature can be disabled for a specific VF::
# ip link set <ethX> vf <vf id> spoofchk {off|on}
Jumbo Frames
Jumbo Frames support is enabled by changing the Maximum Transmission Unit (MTU)
to a value larger than the default value of 1500.
Use the ifconfig command to increase the MTU size. For example, enter the
following where <ethX> is the interface number::
# ifconfig <ethX> mtu 9000 up
Alternatively, you can use the ip command as follows::
# ip link set mtu 9000 dev <ethX>
# ip link set up dev <ethX>
This setting is not saved across reboots.
NOTE: The maximum MTU setting for jumbo frames is 9702. This corresponds to the
maximum jumbo frame size of 9728 bytes.
NOTE: This driver will attempt to use multiple page sized buffers to receive
each jumbo packet. This should help to avoid buffer starvation issues when
allocating receive packets.
NOTE: Packet loss may have a greater impact on throughput when you use jumbo
frames. If you observe a drop in performance after enabling jumbo frames,
enabling flow control may mitigate the issue.
Speed and Duplex Configuration
In addressing speed and duplex configuration issues, you need to distinguish
between copper-based adapters and fiber-based adapters.
In the default mode, an Intel(R) Ethernet Network Adapter using copper
connections will attempt to auto-negotiate with its link partner to determine
the best setting. If the adapter cannot establish link with the link partner
using auto-negotiation, you may need to manually configure the adapter and link
partner to identical settings to establish link and pass packets. This should
only be needed when attempting to link with an older switch that does not
support auto-negotiation or one that has been forced to a specific speed or
duplex mode. Your link partner must match the setting you choose. 1 Gbps speeds
and higher cannot be forced. Use the autonegotiation advertising setting to
manually set devices for 1 Gbps and higher.
Speed, duplex, and autonegotiation advertising are configured through the
ethtool utility. For the latest version, download and install ethtool from the
following website:
To see the speed configurations your device supports, run the following::
# ethtool <ethX>
Caution: Only experienced network administrators should force speed and duplex
or change autonegotiation advertising manually. The settings at the switch must
always match the adapter settings. Adapter performance may suffer or your
adapter may not operate if you configure the adapter differently from your
Data Center Bridging (DCB)
NOTE: The kernel assumes that TC0 is available, and will disable Priority Flow
Control (PFC) on the device if TC0 is not available. To fix this, ensure TC0 is
enabled when setting up DCB on your switch.
DCB is a configuration Quality of Service implementation in hardware. It uses
the VLAN priority tag (802.1p) to filter traffic. That means that there are 8
different priorities that traffic can be filtered into. It also enables
priority flow control (802.1Qbb) which can limit or eliminate the number of
dropped packets during network stress. Bandwidth can be allocated to each of
these priorities, which is enforced at the hardware level (802.1Qaz).
DCB is normally configured on the network using the DCBX protocol (802.1Qaz), a
specialization of LLDP (802.1AB). The ice driver supports the following
mutually exclusive variants of DCBX support:
1) Firmware-based LLDP Agent
2) Software-based LLDP Agent
In firmware-based mode, firmware intercepts all LLDP traffic and handles DCBX
negotiation transparently for the user. In this mode, the adapter operates in
"willing" DCBX mode, receiving DCB settings from the link partner (typically a
switch). The local user can only query the negotiated DCB configuration. For
information on configuring DCBX parameters on a switch, please consult the
switch manufacturer's documentation.
In software-based mode, LLDP traffic is forwarded to the network stack and user
space, where a software agent can handle it. In this mode, the adapter can
operate in either "willing" or "nonwilling" DCBX mode and DCB configuration can
be both queried and set locally. This mode requires the FW-based LLDP Agent to
be disabled.
- You can enable and disable the firmware-based LLDP Agent using an ethtool
private flag. Refer to the "FW-LLDP (Firmware Link Layer Discovery Protocol)"
section in this README for more information.
- In software-based DCBX mode, you can configure DCB parameters using software
LLDP/DCBX agents that interface with the Linux kernel's DCB Netlink API. We
recommend using OpenLLDP as the DCBX agent when running in software mode. For
more information, see the OpenLLDP man pages and
- The driver implements the DCB netlink interface layer to allow the user space
to communicate with the driver and query DCB configuration for the port.
- iSCSI with DCB is not supported.
FW-LLDP (Firmware Link Layer Discovery Protocol)
Use ethtool to change FW-LLDP settings. The FW-LLDP setting is per port and
persists across boots.
To enable LLDP::
# ethtool --set-priv-flags <ethX> fw-lldp-agent on
To disable LLDP::
# ethtool --set-priv-flags <ethX> fw-lldp-agent off
To check the current LLDP setting::
# ethtool --show-priv-flags <ethX>
NOTE: You must enable the UEFI HII "LLDP Agent" attribute for this setting to
take effect. If "LLDP AGENT" is set to disabled, you cannot enable it from the
Flow Control
Ethernet Flow Control (IEEE 802.3x) can be configured with ethtool to enable
receiving and transmitting pause frames for ice. When transmit is enabled,
pause frames are generated when the receive packet buffer crosses a predefined
threshold. When receive is enabled, the transmit unit will halt for the time
delay specified when a pause frame is received.
NOTE: You must have a flow control capable link partner.
Flow Control is disabled by default.
Use ethtool to change the flow control settings.
To enable or disable Rx or Tx Flow Control::
# ethtool -A <ethX> rx <on|off> tx <on|off>
Note: This command only enables or disables Flow Control if auto-negotiation is
disabled. If auto-negotiation is enabled, this command changes the parameters
used for auto-negotiation with the link partner.
Note: Flow Control auto-negotiation is part of link auto-negotiation. Depending
on your device, you may not be able to change the auto-negotiation setting.
- The ice driver requires flow control on both the port and link partner. If
flow control is disabled on one of the sides, the port may appear to hang on
heavy traffic.
- You may encounter issues with link-level flow control (LFC) after disabling
DCB. The LFC status may show as enabled but traffic is not paused. To resolve
this issue, disable and reenable LFC using ethtool::
# ethtool -A <ethX> rx off tx off
# ethtool -A <ethX> rx on tx on
This driver supports NAPI (Rx polling mode).
For more information on NAPI, see
This driver supports MACVLAN. Kernel support for MACVLAN can be tested by
checking if the MACVLAN driver is loaded. You can run 'lsmod | grep macvlan' to
see if the MACVLAN driver is loaded or run 'modprobe macvlan' to try to load
the MACVLAN driver.
- In passthru mode, you can only set up one MACVLAN device. It will inherit the
MAC address of the underlying PF (Physical Function) device.
IEEE 802.1ad (QinQ) Support
The IEEE 802.1ad standard, informally known as QinQ, allows for multiple VLAN
IDs within a single Ethernet frame. VLAN IDs are sometimes referred to as
"tags," and multiple VLAN IDs are thus referred to as a "tag stack." Tag stacks
allow L2 tunneling and the ability to segregate traffic within a particular
VLAN ID, among other uses.
- Receive checksum offloads and VLAN acceleration are not supported for 802.1ad
(QinQ) packets.
- 0x88A8 traffic will not be received unless VLAN stripping is disabled with
the following command::
# ethtool -K <ethX> rxvlan off
- 0x88A8/0x8100 double VLANs cannot be used with 0x8100 or 0x8100/0x8100 VLANS
configured on the same port. 0x88a8/0x8100 traffic will not be received if
0x8100 VLANs are configured.
- The VF can only transmit 0x88A8/0x8100 (i.e., 802.1ad/802.1Q) traffic if:
1) The VF is not assigned a port VLAN.
2) spoofchk is disabled from the PF. If you enable spoofchk, the VF will
not transmit 0x88A8/0x8100 traffic.
- The VF may not receive all network traffic based on the Inner VLAN header
when VF true promiscuous mode (vf-true-promisc-support) and double VLANs are
enabled in SR-IOV mode.
The following are examples of how to configure 802.1ad (QinQ)::
# ip link add link eth0 eth0.24 type vlan proto 802.1ad id 24
# ip link add link eth0.24 eth0.24.371 type vlan proto 802.1Q id 371
Where "24" and "371" are example VLAN IDs.
Tunnel/Overlay Stateless Offloads
Supported tunnels and overlays include VXLAN, GENEVE, and others depending on
hardware and software configuration. Stateless offloads are enabled by default.
To view the current state of all offloads::
# ethtool -k <ethX>
UDP Segmentation Offload
Allows the adapter to offload transmit segmentation of UDP packets with
payloads up to 64K into valid Ethernet frames. Because the adapter hardware is
able to complete data segmentation much faster than operating system software,
this feature may improve transmission performance.
In addition, the adapter may use fewer CPU resources.
- The application sending UDP packets must support UDP segmentation offload.
To enable/disable UDP Segmentation Offload, issue the following command::
# ethtool -K <ethX> tx-udp-segmentation [off|on]
Performance Optimization
Driver defaults are meant to fit a wide variety of workloads, but if further
optimization is required, we recommend experimenting with the following
Rx Descriptor Ring Size
To reduce the number of Rx packet discards, increase the number of Rx
descriptors for each Rx ring using ethtool.
Check if the interface is dropping Rx packets due to buffers being full
(rx_dropped.nic can mean that there is no PCIe bandwidth)::
# ethtool -S <ethX> | grep "rx_dropped"
If the previous command shows drops on queues, it may help to increase
the number of descriptors using 'ethtool -G'::
# ethtool -G <ethX> rx <N>
Where <N> is the desired number of ring entries/descriptors
This can provide temporary buffering for issues that create latency while
the CPUs process descriptors.
Interrupt Rate Limiting
This driver supports an adaptive interrupt throttle rate (ITR) mechanism that
is tuned for general workloads. The user can customize the interrupt rate
control for specific workloads, via ethtool, adjusting the number of
microseconds between interrupts.
To set the interrupt rate manually, you must disable adaptive mode::
# ethtool -C <ethX> adaptive-rx off adaptive-tx off
For lower CPU utilization:
Disable adaptive ITR and lower Rx and Tx interrupts. The examples below
affect every queue of the specified interface.
Setting rx-usecs and tx-usecs to 80 will limit interrupts to about
12,500 interrupts per second per queue::
# ethtool -C <ethX> adaptive-rx off adaptive-tx off rx-usecs 80 tx-usecs 80
For reduced latency:
Disable adaptive ITR and ITR by setting rx-usecs and tx-usecs to 0
using ethtool::
# ethtool -C <ethX> adaptive-rx off adaptive-tx off rx-usecs 0 tx-usecs 0
Per-queue interrupt rate settings:
The following examples are for queues 1 and 3, but you can adjust other
To disable Rx adaptive ITR and set static Rx ITR to 10 microseconds or
about 100,000 interrupts/second, for queues 1 and 3::
# ethtool --per-queue <ethX> queue_mask 0xa --coalesce adaptive-rx off
rx-usecs 10
To show the current coalesce settings for queues 1 and 3::
# ethtool --per-queue <ethX> queue_mask 0xa --show-coalesce
Bounding interrupt rates using rx-usecs-high:
:Valid Range: 0-236 (0=no limit)
The range of 0-236 microseconds provides an effective range of 4,237 to
250,000 interrupts per second. The value of rx-usecs-high can be set
independently of rx-usecs and tx-usecs in the same ethtool command, and is
also independent of the adaptive interrupt moderation algorithm. The
underlying hardware supports granularity in 4-microsecond intervals, so
adjacent values may result in the same interrupt rate.
The following command would disable adaptive interrupt moderation, and allow
a maximum of 5 microseconds before indicating a receive or transmit was
complete. However, instead of resulting in as many as 200,000 interrupts per
second, it limits total interrupts per second to 50,000 via the rx-usecs-high
# ethtool -C <ethX> adaptive-rx off adaptive-tx off rx-usecs-high 20
rx-usecs 5 tx-usecs 5
Virtualized Environments
In addition to the other suggestions in this section, the following may be
helpful to optimize performance in VMs.
Using the appropriate mechanism (vcpupin) in the VM, pin the CPUs to
individual LCPUs, making sure to use a set of CPUs included in the
device's local_cpulist: ``/sys/class/net/<ethX>/device/local_cpulist``.
Configure as many Rx/Tx queues in the VM as available. (See the iavf driver
documentation for the number of queues supported.) For example::
# ethtool -L <virt_interface> rx <max> tx <max>
For general information, go to the Intel support website at:
or the Intel Wired Networking project hosted by Sourceforge at:
If an issue is identified with the released source code on a supported kernel
with a supported adapter, email the specific information related to the issue
Intel is a trademark or registered trademark of Intel Corporation or its
subsidiaries in the United States and/or other countries.
* Other names and brands may be claimed as the property of others.