NVIDIA Accelerated IO (XLIO) Documentation Rev 3.30.5
NVIDIA Accelerated IO (XLIO) Documentation Rev 3.30.5

XLIO Configuration

You can control the behavior of XLIO by configuring:

  • The libxlio.conf file

  • XLIO configuration parameters, which are Linux OS environment variables

  • XLIO extra API

The installation process creates a default configuration file, /etc/libxlio.conf, in which you can define and change the following settings:

  • The target applications or processes to which the configured control settings apply. By default, XLIO control settings are applied to all applications.

  • The transport to be used for the created sockets.

  • The IP addresses and ports in which you want offload.

By default, the configuration file allows XLIO to offload everything except for the DNS server-side protocol (UDP, port 53) which will be handled by the OS.

In the libxlio.conf file:

  • You can define different XLIO control statements for different processes in a single configuration file. Control statements are always applied to the preceding target process statement in the configuration file.

  • Comments start with # and cause the entire line after it to be ignored.

  • Any beginning whitespace is skipped.

  • Any line that is empty is skipped.

  • It is recommended to add comments when making configuration changes.

The following sections describe configuration options in libxlio.conf. For a sample libxlio.conf file, see Example of XLIO Configuration.

Configuring Target Application or Process

The target process statement specifies the process to which all control statements that appear between this statement and the next target process statement apply.

Each statement specifies a matching rule that all its sub-expressions must evaluate as true (logical and) to apply.

If not provided (default), the statement matches all programs.

The format of the target process statement is:


application-id <program-name|*> <user-defined-id|*>




Define the program name (not including the path) to which the control statements appearing below this statement apply.

Wildcards with the same semantics as "ls" are supported (* and ?).

For example:

  • db2* matches any program with a name starting with db2.

  • t?cp matches ttcp, etc.


Specify the process ID to which the control statements appearing below this statement apply.


You must also set the XLIO_APPLICATION_ID environment variable to the same value as user-defined-id.

Configuring Socket Transport Control

Use socket control statements to specify when libxlio will offload AF_INET/SOCK_STREAM or AF_INET/SOCK_DATAGRAM sockets (currently SOCK_RAW is not supported).

Each control statement specifies a matching rule that all its sub-expressions must evaluate as true (logical and) to apply. Statements are evaluated in order of definition according to "first-match".

Socket control statements use the following format:


use <transport> <role> <address|*>:<port range|*>





Define the mode of transport:

  • xlio – XLIO should be used.

  • os – the socket should be handled by the OS network stack. In this mode, the sockets are not offloaded.

The default is xlio.


Specify one of the following roles:

  • tcp_server – for listen sockets. Accepted sockets follow listen sockets. Defined by local_ip:local_port.

  • tcp_client – for connected sockets. Defined by remote_ip:remote_port:local_ip:local_port

  • udp_sender – for TX flows. Defined by remote_ip:remote_port

  • udp_receiver – for RX flows. Defined by local_ip:local_port

  • udp_connect – for UDP connected sockets. Defined by remote_ip:remote_port:local_ip:local_port


You can specify the local address the server is bind to or the remote server address the client connects to.

The syntax for address matching is:

<IPv4 address>[/<prefix_length>]|*

  • IPv4 address – [0-9]+\.[0-9]+\.[0-9]+\.[0-9]+ each sub number < 255

  • prefix_length – [0-9]+ and with value <= 32. A prefix_length of 24 # matches the subnet mask . A prefix_length of 32 requires matching of the exact IP.

port range

Define the port range as:



Port range: 0-65536

Example of XLIO Configuration

To set the following:

  • Apply the rules to program tcp_lat with ID B1

  • Use XLIO by TCP clients connecting to machines that belong to subnet 192.168.1.*

  • Use OS when TCP server listens to port 5001 of any machine

In libxlio.conf, configure:


application-id tcp-lat B1 use xlio tcp_client*:*:* use os tcp_server *:5001 use os udp_connect *:53


You must also set the XLIO parameter:



XLIO configuration parameters are Linux OS environment variables that are controlled with system environment variables.

It is recommended that you set these parameters prior to loading the application with XLIO. You can set the parameters in a system file, which can be run manually or automatically.

All the parameters have defaults that can be modified.

On default startup, the XLIO library prints the XLIO version information, as well as the configuration parameters being used and their values to stderr.

XLIO always logs the values of the following parameters, even when they are equal to the default value:



For all other parameters, XLIO logs the parameter values only when they are not equal to the default value.


The XLIO version information, parameters, and values are subject to change.

For example:


XLIO INFO : XLIO_VERSION: X.Y.Z-R Release built on MM DD YYYY HH:mm:ss XLIO INFO : Cmd Line: sockperf server -i XLIO INFO : --------------------------------------------------------------------------- XLIO DEBUG : Current Time: DayMonth DD HH:mm:ss YYYY XLIO DEBUG : Pid: 3945469 XLIO INFO : OFED Version: MLNX_OFED_LINUX-X.X-X.X.X.X: XLIO DEBUG : System: 5.4.0-121-generic XLIO DEBUG : Architecture: x86_64 XLIO DEBUG : Node: luna02 XLIO INFO : ---------------------------------------------------------------------------   XLIO INFO : Log Level DETAILS [XLIO_TRACELEVEL] XLIO DETAILS: Log Details 0 [XLIO_LOG_DETAILS] XLIO DETAILS: Log Colors Enabled [XLIO_LOG_COLORS] XLIO DETAILS: Log File [XLIO_LOG_FILE] XLIO DETAILS: Stats File [XLIO_STATS_FILE] XLIO DETAILS: Stats shared memory directory /tmp/xlio [XLIO_STATS_SHMEM_DIR] XLIO DETAILS: SERVICE output directory /tmp/xlio [XLIO_SERVICE_NOTIFY_DIR] XLIO DETAILS: Stats FD Num (max) 100 [XLIO_STATS_FD_NUM] XLIO DETAILS: Conf File /etc/libxlio.conf [XLIO_CONFIG_FILE] XLIO DETAILS: Application ID XLIO_DEFAULT_APPLICATION_ID [XLIO_APPLICATION_ID] XLIO DETAILS: Polling CPU idle usage Disabled [XLIO_CPU_USAGE_STATS] XLIO DETAILS: SigIntr Ctrl-C Handle Enabled [XLIO_HANDLE_SIGINTR] XLIO DETAILS: SegFault Backtrace Disabled [XLIO_HANDLE_SIGSEGV] XLIO INFO : Ring allocation logic TX 20(Ring per thread) [XLIO_RING_ALLOCATION_LOGIC_TX] XLIO INFO : Ring allocation logic RX 20(Ring per thread) [XLIO_RING_ALLOCATION_LOGIC_RX] XLIO DETAILS: Ring migration ratio TX -1 [XLIO_RING_MIGRATION_RATIO_TX] XLIO DETAILS: Ring migration ratio RX 100 [XLIO_RING_MIGRATION_RATIO_RX] XLIO DETAILS: Ring limit per interface 0 (no limit) [XLIO_RING_LIMIT_PER_INTERFACE] XLIO DETAILS: Ring On Device Memory TX 0 [XLIO_RING_DEV_MEM_TX] XLIO DETAILS: TCP max syn rate 0 (no limit) [XLIO_TCP_MAX_SYN_RATE] XLIO DETAILS: Zerocopy Mem Bufs 200000 [XLIO_ZC_BUFS] XLIO DETAILS: Zerocopy Cache Threshold 10240 [XLIO_ZC_CACHE_THRESHOLD] XLIO DETAILS: Tx Mem Segs TCP 1000000 [XLIO_TX_SEGS_TCP] XLIO DETAILS: Tx Mem Bufs 200000 [XLIO_TX_BUFS] XLIO DETAILS: Tx Mem Buf size 0 [XLIO_TX_BUF_SIZE] XLIO DETAILS: ZC TX size 32768 [XLIO_ZC_TX_SIZE] XLIO DETAILS: Tx QP WRE 32768 [XLIO_TX_WRE] XLIO DETAILS: Tx QP WRE Batching 64 [XLIO_TX_WRE_BATCHING] XLIO DETAILS: Tx Max QP INLINE 204 [XLIO_TX_MAX_INLINE] XLIO DETAILS: Tx MC Loopback Enabled [XLIO_TX_MC_LOOPBACK] XLIO DETAILS: Tx non-blocked eagains Disabled [XLIO_TX_NONBLOCKED_EAGAINS] XLIO DETAILS: Tx Prefetch Bytes 256 [XLIO_TX_PREFETCH_BYTES] XLIO INFO : Tx Bufs Batch TCP 1 [XLIO_TX_BUFS_BATCH_TCP] XLIO DETAILS: Tx Segs Batch TCP 64 [XLIO_TX_SEGS_BATCH_TCP] XLIO DETAILS: TCP Send Buffer size 1000000 [XLIO_TCP_SEND_BUFFER_SIZE] XLIO INFO : Rx Mem Bufs 32384 [XLIO_RX_BUFS] XLIO INFO : Rx QP WRE 1024 [XLIO_RX_WRE] XLIO DETAILS: Rx QP WRE Batching 1 [XLIO_RX_WRE_BATCHING] XLIO DETAILS: Rx Byte Min Limit 65536 [XLIO_RX_BYTES_MIN] XLIO DETAILS: Rx Poll Loops 100000 [XLIO_RX_POLL] XLIO DETAILS: Rx Poll Init Loops 0 [XLIO_RX_POLL_INIT] XLIO DETAILS: Rx UDP Poll OS Ratio 100 [XLIO_RX_UDP_POLL_OS_RATIO] XLIO DETAILS: HW TS Conversion 3 [XLIO_HW_TS_CONVERSION] XLIO DETAILS: Rx Poll Yield Disabled [XLIO_RX_POLL_YIELD] XLIO DETAILS: Rx Prefetch Bytes 256 [XLIO_RX_PREFETCH_BYTES] XLIO DETAILS: Rx Prefetch Bytes Before Poll 0 [XLIO_RX_PREFETCH_BYTES_BEFORE_POLL] XLIO DETAILS: Rx CQ Drain Rate Disabled [XLIO_RX_CQ_DRAIN_RATE_NSEC] XLIO INFO : GRO max streams 8192 [XLIO_GRO_STREAMS_MAX] XLIO DETAILS: Disable flow tag 0 [XLIO_DISABLE_FLOW_TAG] XLIO DETAILS: TCP 3T rules Disabled [XLIO_TCP_3T_RULES] XLIO DETAILS: UDP 3T rules Enabled [XLIO_UDP_3T_RULES] XLIO DETAILS: ETH MC L2 only rules Disabled [XLIO_ETH_MC_L2_ONLY_RULES] XLIO DETAILS: Force Flowtag for MC Disabled [XLIO_MC_FORCE_FLOWTAG] XLIO DETAILS: Striding RQ Enabled [XLIO_STRQ] XLIO INFO : STRQ Strides per RWQE 4096 [XLIO_STRQ_NUM_STRIDES] XLIO INFO : STRQ Stride Size (Bytes) 128 [XLIO_STRQ_STRIDE_SIZE_BYTES] XLIO DETAILS: STRQ Initial Strides Per Ring 262144 [XLIO_STRQ_STRIDES_NUM_BUFS] XLIO INFO : STRQ Strides Compensation Level 131072 [XLIO_STRQ_STRIDES_COMPENSATION_LEVEL] XLIO DETAILS: Select Poll (usec) 100000 [XLIO_SELECT_POLL] XLIO DETAILS: Select Poll OS Force Disabled [XLIO_SELECT_POLL_OS_FORCE] XLIO DETAILS: Select Poll OS Ratio 10 [XLIO_SELECT_POLL_OS_RATIO] XLIO DETAILS: Select Skip OS 4 [XLIO_SELECT_SKIP_OS] XLIO INFO : CQ Drain Thread Disabled [XLIO_PROGRESS_ENGINE_INTERVAL] XLIO DETAILS: CQ Interrupts Moderation Enabled [XLIO_CQ_MODERATION_ENABLE] XLIO DETAILS: CQ Moderation Count 48 [XLIO_CQ_MODERATION_COUNT] XLIO DETAILS: CQ Moderation Period (usec) 50 [XLIO_CQ_MODERATION_PERIOD_USEC] XLIO DETAILS: CQ AIM Max Count 560 [XLIO_CQ_AIM_MAX_COUNT] XLIO DETAILS: CQ AIM Max Period (usec) 250 [XLIO_CQ_AIM_MAX_PERIOD_USEC] XLIO DETAILS: CQ AIM Interval (msec) 250 [XLIO_CQ_AIM_INTERVAL_MSEC] XLIO DETAILS: CQ AIM Interrupts Rate (per sec) 5000 [XLIO_CQ_AIM_INTERRUPTS_RATE_PER_SEC] XLIO DETAILS: CQ Poll Batch (max) 16 [XLIO_CQ_POLL_BATCH_MAX] XLIO DETAILS: CQ Keeps QP Full Enabled [XLIO_CQ_KEEP_QP_FULL] XLIO INFO : QP Compensation Level 256 [XLIO_QP_COMPENSATION_LEVEL] XLIO DETAILS: Offloaded Sockets Enabled [XLIO_OFFLOADED_SOCKETS] XLIO DETAILS: Timer Resolution (msec) 10 [XLIO_TIMER_RESOLUTION_MSEC] XLIO DETAILS: TCP Timer Resolution (msec) 100 [XLIO_TCP_TIMER_RESOLUTION_MSEC] XLIO DETAILS: TCP control thread 0 (Disabled) [XLIO_TCP_CTL_THREAD] XLIO DETAILS: TCP timestamp option 0 [XLIO_TCP_TIMESTAMP_OPTION] XLIO DETAILS: TCP nodelay 0 [XLIO_TCP_NODELAY] XLIO DETAILS: TCP quickack 0 [XLIO_TCP_QUICKACK] XLIO DETAILS: Exception handling mode -1(just log debug message) [XLIO_EXCEPTION_HANDLING] XLIO DETAILS: Avoid sys-calls on tcp fd Disabled [XLIO_AVOID_SYS_CALLS_ON_TCP_FD] XLIO DETAILS: Allow privileged sock opt Enabled [XLIO_ALLOW_PRIVILEGED_SOCK_OPT] XLIO DETAILS: Delay after join (msec) 0 [XLIO_WAIT_AFTER_JOIN_MSEC] XLIO DETAILS: Internal Thread Affinity -1 [XLIO_INTERNAL_THREAD_AFFINITY] XLIO DETAILS: Internal Thread Cpuset [XLIO_INTERNAL_THREAD_CPUSET] XLIO DETAILS: Internal Thread Arm CQ Disabled [XLIO_INTERNAL_THREAD_ARM_CQ] XLIO DETAILS: Internal Thread TCP Handling 0 (deferred) [XLIO_INTERNAL_THREAD_TCP_TIMER_HANDLING] XLIO DETAILS: Thread mode Multi spin lock [XLIO_THREAD_MODE] XLIO INFO : Buffer batching mode 0 (No batching buffers) [XLIO_BUFFER_BATCHING_MODE] XLIO DETAILS: Mem Allocate type 2 (Huge Pages) [XLIO_MEM_ALLOC_TYPE] XLIO DETAILS: Num of UC ARPs 3 [XLIO_NEIGH_UC_ARP_QUATA] XLIO DETAILS: UC ARP delay (msec) 10000 [XLIO_NEIGH_UC_ARP_DELAY_MSEC] XLIO DETAILS: Num of neigh restart retries 1 [XLIO_NEIGH_NUM_ERR_RETRIES] XLIO DETAILS: TSO support Auto [XLIO_TSO] XLIO INFO : LRO support Disabled [XLIO_LRO] XLIO DETAILS: BF (Blue Flame) Enabled [XLIO_BF] XLIO DETAILS: UTLS RX support Disabled [XLIO_UTLS_RX] XLIO DETAILS: UTLS TX support Enabled [XLIO_UTLS_TX] XLIO DETAILS: UTLS high watermark DEK cache size 1024 [XLIO_UTLS_HIGH_WMARK_DEK_CACHE_SIZE] XLIO DETAILS: UTLS low watermark DEK cache size 512 [XLIO_UTLS_LOW_WMARK_DEK_CACHE_SIZE] XLIO DETAILS: Src port stirde 2 [XLIO_SRC_PORT_STRIDE] XLIO DETAILS: Number of Nginx workers 0 [XLIO_NGINX_WORKERS_NUM] XLIO DETAILS: Size of UDP socket pool 0 [XLIO_NGINX_UDP_POOL_SIZE] XLIO DETAILS: Max RX reuse buffs UDP pool 0 [XLIO_NGINX_UDP_POOL_REUSE_BUFFS] XLIO DETAILS: fork() support Enabled [XLIO_FORK] XLIO DETAILS: close on dup2() Enabled [XLIO_CLOSE_ON_DUP2] XLIO DETAILS: MTU 0 (follow actual MTU) [XLIO_MTU] XLIO DETAILS: MSS 0 (follow XLIO_MTU) [XLIO_MSS] XLIO DETAILS: TCP CC Algorithm 0 (LWIP) [XLIO_TCP_CC_ALGO] XLIO DETAILS: TCP abort on close Disabled [XLIO_TCP_ABORT_ON_CLOSE] XLIO DETAILS: Polling Rx on Tx TCP Disabled [XLIO_RX_POLL_ON_TX_TCP] XLIO DETAILS: RX CQ wait control Disabled [XLIO_RX_CQ_WAIT_CTRL] XLIO DETAILS: Trig dummy send getsockname() Disabled [XLIO_TRIGGER_DUMMY_SEND_GETSOCKNAME] XLIO INFO : Skip CQ polling in rx Epoll Only [XLIO_SKIP_POLL_IN_RX] XLIO INFO : ---------------------------------------------------------------------------

Configuration Parameters Values

The following table lists the XLIO configuration parameters and their possible values.

XLIO Configuration Parameter

Description and Examples


PANIC = 0 – Panic level logging.

This trace level causes fatal behavior and halts the application, typicallXLIO_TX_NONBLOCKED_EAGAINSy caused by memory allocation problems. PANIC level is rarely used.

ERROR = 1 – Runtime errors in XLIO.

Typically, this trace level assists you to identify internal logic errors, such as errors from underlying OS or InfiniBand verb calls, and internal double mapping/unmapping of objects.

WARN = WARNING = 2– Runtime warning that does not disrupt the application workflow.

A warning may indicate problems in the setup or in the overall setup configuration. For example, address resolution failures (due to an incorrect routing setup configuration), corrupted IP packets in the receive path, or unsupported functions requested by the user application.

INFO = INFORMATION = 3– General information passed to the application user.

This trace level includes configuration logging or general information to assist you with better use of the XLIO library.

DETAILS – Greater general information passed to the user of the application.

This trace level includes printing of all environment variables of XLIO at start up.

DEBUG = 4 – High-level insight to the operations performed in XLIO.

All socket API calls are logged in this logging level, and internal high-level control channels log their activity.

FINE = FUNC = 5 – Low-level runtime logging of activity.

This logging level includes basic Tx and Rx logging in the fast path. Note that using this setting lowers application performance. We recommend that you use this level with the XLIO_LOG_FILE parameter.

FINER = FUNC_ALL = 6 – Very low-level runtime logging of activity. This logging level drastically lowers application performance. We recommend that you use this level with the XLIO_LOG_FILE parameter.


Provides additional logging details on each log line.

0 = Basic log line

1 = With ThreadId

2 = With ProcessId and ThreadId

3 = With Time, ProcessId, and ThreadId (Time is the amount of milliseconds from the start of the process)

Default: 0

For XLIO_TRACELEVEL >= 4, this value defaults to 2.


Redirects all XLIO logging to a specific user-defined file.

This is very useful when raising the XLIO_TRACELEVEL.

The XLIO replaces a single '%d' appearing in the log file name with the pid of the process loaded with XLIO. This can help when running multiple instances of XLIO, each with its own log file name.

Example: XLIO_LOG_FILE=/tmp/xlio_log.txt


Sets the full path to the XLIO configuration file.

Example: XLIO_CONFIG_FILE=/tmp/libxlio.conf

Default: /etc/libxlio.conf


Uses a color scheme when logging; red for errors and warnings, and dim for very low level debugs.

XLIO_LOG_COLORS is automatically disabled when logging is done directly to a non-terminal device (for example, when XLIO_LOG_FILE is configured).

Default: 1 (Enabled)


Calculates the XLIO CPU usage during polling hardware loops. This information is available through XLIO stats utility.

Default: 0 (Disabled)


Specifies a group of rules from libxlio.conf for XLIO to apply.

Example: XLIO_APPLICATION_ID=iperf_server

Default: XLIO_DEFAULT_APPLICATION_ID (match only the '*' group rule)


When enabled, the XLIO handler is called when an interrupt signal is sent to the process.

XLIO also calls the application's handler, if it exists.

Range: 0 to 1

Default: 0 (Enabled)


When enabled, a print backtrace is performed, if a segmentation fault occurs.

Range: 0 to 1

Default: 1 (Disabled)


Maximum number of sockets monitored by the XLIO statistics mechanism.

Range: 0 to 1024

Default: 100


Redirects socket statistics to a specific user-defined file.

XLIO dumps each socket's statistics into a file when closing the socket.

Example: XLIO_STATS_FILE=/tmp/stats


Sets the directory path for XLIO to create the shared memory files for xlio_stats.

If this value is set to an empty string: “ “, no shared memory files are created.

Default: /tmp/


Sets the directory path for XLIO to write files used by xliod.

Default value is /tmp/xlio

Note: when used xliod must be run with --notify-dir directing the same folder.


Limits the number of TCP SYN packets that XLIO handles per second for each listen socket.

Example: by setting this value to 10, the maximal number of TCP connection accepted by XLIO per second for each listen socket will be 10.

Set this value to 0 for XLIO to handle an unlimited number of TCP SYN packets per second for each listen socket.

Value range is 0 to 100000.

Default value is 0 (no limit)


Number of TCP LWIP segments allocation for each XLIO process.

Default: 1000000


Number of global Tx data buffer elements allocation.

Default: 200000


Number of Work Request Elements allocated in all transmit QP's. The number of QP's can change according to the number of network offloaded interfaces.

Default: 3000

The size of the Tx buffers is determined by the XLIO_MTU parameter value (see below).

If this value is raised, the packet rate peaking can be better sustained; however, this increases memory usage. A smaller number of data buffers gives a smaller memory footprint, but may not sustain peaks in the data rate.


Controls the number of aggregated Work Requests Elements before receiving a completion signal (CQ entry) from the hardware. Previously this number was hard coded as 64.

The new update allows a better control of the jitter encountered in the Tx completion handling.

Valid value range: 1-64

Default: 64


Max send inline data set for QP.

Data copied into the INLINE space is at least 32 bytes of headers and the rest can be user datagram payload.

XLIO_TX_MAX_INLINE=0 disables INLINEing on the TX transmit path. In older releases this parameter was called XLIO_MAX_INLINE.

Default: 220


Sets the initial value used internally by the XLIO to control multicast loopback packet behavior during transmission. An application that calls setsockopt() with IP_MULTICAST_LOOP overwrites the initial value set by this parameter.

Range: 0 - Disabled, 1 - Enabled

Default: 1


Returns value 'OK' on all send operations performed on a non-blocked UDP socket. This is the OS default behavior. The datagram sent is silently dropped inside the XLIO or the network stack.

When set to Enabled (set to 1), XLIO returns with error EAGAIN in case it could not perform the send operation, and the datagram was dropped.

In both cases, a dropped Tx statistical counter is incremented.

Default: 0 (Disabled)


Accelerates an offloaded send operation by optimizing the cache. Different values give an optimized send rate on different machines. We recommend that you adjust this parameter to your specific hardware.

Range: 0 to MTU size

Disable with a value of 0

Default: 256 bytes


The number of buffers fetched from the ring pool by a socket at once. Higher number for less ring accesses to fetch buffers. Lower number for less memory consumption by a socket.

Min value: 1

Default value: 16


The number of Rx data buffer elements allocated for the processes. These data buffers are used by all QPs on all HCAs, as determined by the XLIO_QP_LOGIC.

Default: 200000


The number of Work Request Elements allocated in all received QPs.

Default: 16000


Number of Work Request Elements and RX buffers to batch before recycling.

Batching decreases the latency mean, but might increase latency STD.

Valid value range: 1-1024

Default: 1024


The minimum value in bytes used per socket by the XLIO when applications call to setsockopt(SO_RCVBUF).

If the application tries to set a smaller value than configured in XLIO_RX_BYTES_MIN, XLIO forces this minimum limit value on the socket.

XLIO offloaded sockets receive the maximum amount of ready bytes. If the application does not drain sockets and the byte limit is reached, newly received datagrams are dropped.

The application's socket usage of current, max,dropped bytes and packet counters, can be monitored using xlio_stats.

Default: 65536


The number of times to unsuccessfully poll an Rx for XLIO packets before going to sleep.

Range: -1, 0 … 100,000,000

Default: 100,000

This value can be reduced to lower the load on the CPU. However, the price paid for this is that the Rx latency is expected to increase.

Recommended values:

  • 10000 – when CPU usage is not critical and Rx path latency is critical.

  • 0 – when CPU usage is critical, while Rx path latency is not.

  • -1 – causes infinite polling.

Once the XLIO has gone to sleep, if it is in blocked mode, it waits for an interrupt; if it is in non-blocked mode, it returns -1.

This Rx polling is performed when the application is working with direct blocked calls to read(), recv(), recvfrom(), and recvmsg().

When the Rx path has successful poll hits, the latency improves dramatically. However, this causes increased CPU utilization. For more information, see Debugging, Troubleshooting, and Monitoring.


XLIO maps all UDP sockets as potential Offloaded-capable. Only after ADD_MEMBERSHIP is set, the offload starts working and the CQ polling starts XLIO.

This parameter controls the polling count during this transition phase where the socket is a UDP unicast socket and no multicast addresses were added to it.

Once the first ADD_MEMBERSHIP is called, the XLIO_RX_POLL (above) takes effect.

Value range is similar to the XLIO_RX_POLL (above).

Default: 0


Defines the ratio between XLIO CQ poll and OS FD poll.

This will result in a single poll of the not-offloaded sockets every XLIO_RX_UDP_POLL_OS_RATIO offloaded socket (CQ) polls. No matter if the CQ poll was a hit or miss. No matter if the socket is blocking or non-blocking.

When disabled, only offloaded sockets are polled.

This parameter replaces the two old parameters:



Disable with 0

Default: 10


Defines the size of UDP socket pool for NGINX.

For any value different that 0 - close() socket will not destroy the socket, but will place it in a pool for the next socket UDP creation.

Disable with 0

Default: 0


Defines timestamp conversion method.

The value of XLIO_HW_TS_CONVERSION is determined by all devices, that is, if the hardware of one device does not support the conversion, then it will be disabled for the other devices.

Currently only UDP RX flow is supported.

Options = [0,1,2,3,4]:

  • 0 – Disabled

  • 1 – Raw-HW time

    Only convert the timestamp to seconds.nano_seconds time units (or disable if hardware does not supports).

  • 2 – Best possible – Raw-HW or system time.

    Sync to system time, then Raw hardware time. Disable if none of them are supported by hardware.

  • 3 – Sync to system time

    Convert the timestamp to seconds.nano_seconds time units. Comparable to UDP receive software timestamp. Disable if hardware does not support.

  • 4 – PTP Sync

    Convert the timestamp to seconds.nano_seconds time units. In case it is not supported – will apply option 3 (or disable if hardware does not support).

  • 5 - RTC Sync

    Convert the time stamp to seconds.nano_seconds time units. In case it is not supported - will apply option 3 (or disable if hardware does not support it)

Default value: 3 (Sync to system time)


When an application is running with multiple threads on a limited number of cores, there is a need for each thread polling inside XLIO (read, readv, recv, and recvfrom) to yield the CPU to another polling thread so as not to starve them from processing incoming packets.

Default: 0 (Disabled)


The size of the receive buffer to prefetch into the cache while processing ingress packets.

The default is a single cache line of 64 bytes which should be at least 32 bytes to cover the IP+UDP headers and a small part of the user payload.

Increasing this size can help improve performance for larger user payloads.

Range: 32 bytes to MTU size

Default: 256 bytes


Socket's receive path CQ drain logic rate control.

When disabled (default), the socket's receive path attempts to return a ready packet from the socket's receive ready packet queue. If the ready receive packet queue is empty, the socket checks the CQ for ready completions for processing.

When enabled, even if the socket's receive ready packet queue is not empty, this parameter checks the CQ for ready completions for processing. This CQ polling rate is controlled in nanosecond resolution to prevent CPU consumption due to over CQ polling. This enables improved 'real-time' monitoring of the socket ready packet queue.

Recommended value is 100-5000 (nsec)

Default: 0 (Disabled)


Enables TCP RX polling during TXP TX operation for faster TCP ACK reception

Default: 0 (Disabled)


Controls the number of TCP streams to perform GRO (generic receive offload) simultaneously.

Disable GRO with a value of 0.

Default: 32


Uses only 3 tuple rules for TCP, instead of using 5 tuple rules.

This can improve performance for a server with a listen socket which accepts many connections from the same source IP.

Enable with a value of 1.

Default: 0 (Disabled)


This parameter is relevant in case the application uses connected UDP sockets. 3 tuple rules are used in hardware flow steering rule when the parameter is enabled, and in 5 tuple flow steering rule when it is disabled. Enabling this option can reduce hardware flow steering resources. However, when it is disabled, the application might see benefits in latency and cycles per packet. Default: 1 (Enable)


Uses only L2 rules for Ethernet Multicast.

All loopback traffic will be handled by XLIO instead of OS.

Enable with a value of 1.

Default: 0 (Disabled)


Enables and disables Striding Receive Queues.

Each WQE in a Striding RQ may receive several packets. Thus, the WQE buffer size is controlled by XLIO_STRQ_NUM_STRIDES x XLIO_STRQ_STRIDE_SIZE_BYTES

Values: on, off

Default: on (Enabled)


The number of strides in each receive WQE. Must be power of two and in range [512 - 65536].

Default: 16384


The size, in bytes, of each stride in a receive WQE. Must be power of two and in range [64 - 8192].

Default: 512


The initial number of stride objects in the strides pool. Each received packet is represented by a stride object. Each stride object points to a portion of a buffer allocated for a receive WQE. When the pool runs out of stride objects it expands by another portion of this value.

Default: 262144


Number of spare stride objects a CQ holds to allow faster allocation of a stride object when a packet arrives.

Default: 16384


The duration in micro-seconds (usec) in which to poll the hardware on Rx path before blocking for an interrupt (when waiting and also when calling select(), poll(), or epoll_wait()).

Range: -1, 0 … 100,000,000

Default: 100,000

When the selected path has successfully received poll hits, the latency improves dramatically. However, this comes at the expense of CPU utilization. For more information, see Debugging, Troubleshooting, and Monitoring.


This enables polling the OS file descriptors while the user thread calls select(), poll(), or epoll_wait(), and XLIO is busy in the offloaded socket polling loop. This results in a single poll of the non-offloaded sockets every XLIO_SELECT_POLL_RATIO offloaded socket (CQ) polls.

When disabled, only offloaded sockets are polled.

(See XLIO_SELECT_POLL for more information.)

Disable with 0

Default: 10


In select(), poll(), or epoll_wait()forces the XLIO to check the non-offloaded sockets even though an offloaded socket has a ready packet that was found while polling.

Range: 0 … 10,000

Default: 4


The maximum size of the array while polling the CQs in the XLIO.

Default: 16


Internal XLIO thread safety which checks that the CQ is drained at least once every N milliseconds. This mechanism allows XLIO to progress the TCP stack even when the application does not access its socket (so it does not provide a context to XLIO). If the CQ was already drained by the application receive socket API calls, this thread goes back to sleep without any processing.

Disable with 0

Default: 10 milliseconds


Each time the XLIO's internal thread starts its CQ draining, it stops when it reaches this maximum value.

The application is not limited by this value in the number of CQ elements that it can ProcessId from calling any of the receive path socket APIs.

Default: 2048


Enable CQ interrupt moderation.

Default: 1 (Enabled)


Number of packets to hold before generating interrupt.

Default: 48


Period in microseconds for holding the packet before generating interrupt.

Default: 50


Maximum count value to use in the adaptive interrupt moderation algorithm.

Default: 560


Maximum period value to use in the adaptive interrupt moderation algorithm.

Default: 250


Frequency of interrupt moderation adaptation.

Interval in milliseconds between adaptation attempts.

Use value of 0 to disable adaptive interrupt moderation.

Default: 250


Desired interrupts rate per second for each ring (CQ).

The count and period parameters for CQ moderation will change automatically to achieve the desired interrupt rate for the current traffic rate.

Default: 5000


If disabled (default), the CQ does not try to compensate for each poll on the receive path. It uses a "debt" to remember how many WRE are missing from each QP, so that it can fill it when buffers become available.

If enabled, CQ tries to compensate QP for each polled receive completion. If there is a shortage of buffers, it reposts a recently completed buffer. This causes a packet drop, and is monitored in xlio_stats.

Default: 1 (Enabled)


The number of spare receive buffer CQ holds that can be allowed for filling up QP while full receive buffers are being processed inside XLIO.

Default: 256 buffers


Creates all sockets as offloaded/not-offloaded by default.

  • 1 is used for offloaded

  • 0 is used for not-offloaded

Default: 1 (Enabled)


Control XLIO internal thread wakeup timer resolution (in milliseconds).

Default: 10 (milliseconds)


Controls XLIO internal TCP timer resolution (fast timer) (in milliseconds). Minimum value is the internal thread wakeup timer resolution (XLIO_TIMER_RESOLUTION_MSEC).

Default: 100 (milliseconds)


Select which TCP control flows are done in the internal thread.

This feature should be kept disabled if using blocking poll/select (epoll is OK).

  • Use value of 'disable' to disable.

  • Use value of 'delegate' to handle TCP timers in application context threads.

    • In this mode the socket must be handled by the same thread from the time of its creation to the time of its destruction. Otherwise, it may lead to an unexpected behaviour.

  • Use value of 'with_wakeup' for waking up the thread when there is work to do.

  • Use value of 'no_wakeup' for waiting for thread timer to expire.

Default value is disabled.


Currently, LWIP is not supporting RTTM and PAWS mechanisms.

See RFC1323 for info.

  • Use value of 0 to disable (enabling causing a slight performance degradation of ~50-100 nano sec per half round trip).

  • Use value of 1 for enable.

  • Use value of 2 for OS follow up.

Default: 0 (disabled)


If set, it disables the Nagle algorithm option for each TCP socket during initialization. Meaning that TCP segments are always sent as soon as possible, even if there is only a small amount of data.

For more information on TCP_NODELAY flag refer to TCP manual page.

Valid Values are:

  • 0 to disable.

  • 1 to enable (default)


If set, it disables the delayed acknowledge ability. Meaning that TCP will respond after every packet.

For more information on TCP_QUICKACK flag refer to TCP manual page.

Valid Values are:

  • 0 to disable.

  • 1 to enable (default)


Handles missing support or error cases in Socket API or functionality by XLIO.

It quickly identifies XLIO unsupported Socket API or features.

  • Use value of -1 to handle DEBUG severity

  • Use value of 0 to log DEBUG message and try recovering via Kernel network stack (un-offloading the socket)

  • Use value of 1 to log ERROR message and try recovering via Kernel network stack (un-offloading the socket)

  • Use value of 2 to log ERROR message and return API respectful error code

  • Use value of 3 to log ERROR message and abort application (throw xlio_error exception).

Default: -1


For TCP fd, avoid system calls for the supported options of: ioctl, fcntl, getsockopt, setsockopt.

Non-supported options will go to OS.

To activate, use XLIO_AVOID_SYS_CALLS_ON_TCP_FD=1.

Default: 0 (disabled)


By default XLIO is ready for multi-threaded applications, meaning it is thread-safe.

If the user application is single threaded, use this configuration parameter to help eliminate XLIO locks and improve performance.


  • 0 – Single-threaded application

  • 1 – Multi-threaded application with spin lock

  • 2 – Multi-threaded application with mutex lock

  • 3 – Multi-threaded application with more threads than cores using spin lock

Default: 1 (Multi with spin lock)


XLIO will try to allocate data buffers as configured:

  • 0 or "ANON" – using malloc

  • 2 or "HUGE" - using huge pages.

XLIO also overrides rdma-core parameters MLX_QP_ALLOC_TYPE and MLX_CQ_ALLOC_TYPE


XLIO parameter | rdma-core parameter



HUGE | ALL (PREFER_CONTIG for >32MB hugepage)

Default value is 2 (Huge pages)


Controls XLIO fork support. Setting this flag on will cause XLIO to call ibv_fork_init() function. ibv_fork_init() initializes libibverbs's data structures to handle fork() function calls correctly and avoid data corruption.

If ibv_fork_init() is not called or returns a non-zero status, then libibverbs data structures are not fork()-safe and the effect of an application calling fork() is undefined.

ibv_fork_init() works on Linux kernels 2.6.17 and later, which support the MADV_DONTFORK flag for madvise().

XLIO allocates huge pages by default.

For limitations of using fork() with XLIO, please refer to the Release Notes.

Default: 1 (Enabled)


Size of each Rx and Tx data buffer (Maximum Transfer Unit).

This value sets the fragmentation size of the packets sent by the XLIO library.

  • If XLIO_MTU is 0, then for each interface XLIO will follow the actual MTU

  • If XLIO_MTU is greater than 0, then this MTU value is applicable to all interfaces regardless of their actual MTU

Default: 0 (following interface actual MTU)


Defines the max TCP payload size that can be sent without IP fragmentation.

Value of 0 will set XLIO's TCP MSS to be aligned with XLIO_MTU configuration (leaving 40 bytes of room for IP + TCP headers; "TCP MSS = XLIO_MTU - 40").

Other XLIO_MSS values will force XLIO's TCP MSS to that specific value.

Default: 0 (following XLIO_MTU)


When this parameter is enabled, XLIO handles the duplicated file descriptor (oldfd), as if it is closed (clear internal data structures) and only then forwards the call to the OS.

This is, in effect, a very rudimentary dup2 support. It supports only the case where dup2 is used to close file descriptors.

Default: 1 (Enabled)


Controls which CPU core(s) the XLIO internal thread is serviced on. The CPU set should be provided as either a hexadecimal value that represents a bitmask or as a comma delimited of values (ranges are ok). Both the bitmask and comma delimited list methods are identical to what is supported by the taskset command. See the man page on taskset for additional information.

The -1 value disables the Internal Thread Affinity setting by XLIO.

Bitmask examples:

0x00000001 – Run on processor 0

0x00000007 – Run on processors 1,2, and 3

Comma delimited examples:

0,4,8 – Run on processors 0,4, and 8

0,1,7-10 – Run on processors 0,1,7,8,9 and 10

Default: -1.


Selects a CPUSET for XLIO internal thread (For further information, see man page of cpuset).

The value is either the path to the CPUSET (for example: /dev/cpuset/my_set), or an empty string to run it on the same CPUSET the process runs on.


Wakes up the internal thread for each packet that the CQ receives.

Polls and processes the packet and brings it to the socket layer.

This can minimize latency for a busy application that is not available to receive the packet when it arrives.

However, this might decrease performance for high pps rate applications.

Default: 0 (Disabled)


Selects the internal thread policy when handling TCP timers.

Use value of 0 for deferred handling. The internal thread will not handle TCP timers upon timer expiration (once every 100ms) in order to let application threads handling it first.

Use value of 1 for immediate handling. The internal thread will try locking and handling TCP timers upon timer expiration (once every 100ms). Application threads may be blocked till internal thread finishes handling TCP timers

Default value is 0 (deferred handling)


This parameter indicates the time of delay the first packet is send after receiving the multicast JOINED event from the SM.

This is helpful to overcome loss of first few packets of an outgoing stream due to SM lengthy handling of MFT configuration on the switch chips.

Default: 0 (milli-sec)


XLIO will send UC ARP in case neigh state is NUD_STALE.

If that neigh state is still NUD_STALE XLIO will try

XLIO_NEIGH_UC_ARP_QUATA retries to send UC ARP again and then will send BC ARP.

Default: 3


This parameter indicates number of msec to wait between every UC ARP.

Default: 10000


Indicates number of retries to restart NEIGH state machine if NEIGH receives ERROR event.

Default: 1


Enables/disables BlueFlame usage of the card.

Default: 1 (Enabled)


With Segmentation Offload, or TCP Large Send, TCP can pass a buffer to be transmitted that is bigger than the maximum transmission unit (MTU) supported by the medium. Intelligent adapters implement large sends by using the prototype TCP and IP headers of the incoming send buffer to carve out segments of required size. Copying the prototype header and options, then calculating the sequence number and checksum fields creates TCP segment headers.

Expected benefits: Throughput increase and CPU unload.

Default value: auto


Depends on ethtool setting and adapter ability.

See ethtool -k <eth0> | grep tcp-segmentation-offload


Enabled in case adapter supports it




Large receive offload (LRO) is a technique for increasing inbound throughput of high-bandwidth network connections by reducing central processing unit (CPU) overhead. It works by aggregating multiple incoming packets from a single stream into a larger buffer before they are passed higher up the networking stack, thus reducing the number of packets that must be processed.

Default value: auto


Depends on ethtool setting and adapter ability.

See ethtool -k <eth0> | grep large-receive-offload


Enabled in case adapter supports it




This parameter triggers dummy packet sent from getsockname() to warm up the caches.

For more information see section "Dummy Send" to Improve Low Message Rate Latency.

Default: 0 (Disable)


When this parameter is enabled, XLIO offloads TLS TX path through kTLS API if possible.

Default: 1 (Enabled)


When this parameter is enabled, XLIO offloads TLS RX path through kTLS API if possible.

Default: 1 (Enabled)


XLIO_SPEC sets all the required configuration parameters of XLIO. Usually, no additional configuration is required.

Example #1: XLIO_SPEC=latency (XLIO predefined specification profile for latency: Latency profile spec – optimized latency on all use cases. System is tuned to keep balance between Kernel and XLIO. Note: It may not reach the maximum bandwidth)

Example #2: XLIO_SPEC=multi_ring_latency (XLIO predefined specification profile for Multi ring latency – optimized for use cases that are keen on latency where two applications communicate using send-only and receive-only TCP sockets)


This number must be equal to ‘worker_processes’ attribute of nginx configuration file.

Default: 0 (Disable)



Ring allocation logic is used to separate the traffic into different rings.

By default, all sockets use the same ring for both RX and TX over the same interface. For different interfaces, different rings are used, even when specifying the logic to be per socket or thread.

Using different rings is useful when tuning for a multi-threaded application and aiming for HW resource separation.


This feature might decrease performance for applications which their main processing loop is based on select() and/or poll().

The logic options are:

  • 0 – Ring per interface

  • 1 – Ring per IP address (using IP address)

  • 10 – Ring per socket (using socket ID as separator)

  • 20 – Ring per thread (using the ID of the thread in which the socket was created)

  • 30 – Ring per core (using CPU ID)

  • 31 – Ring per core - attach threads: attach each thread to a CPU core

Default: 0


Enable/Disable flow-tag. Flow Tags improve RX performance for 5T rules.

Default: 0 (Flow tag enabled)


Allow TCP socket to skip CQ polling in Rx socket call.

0 - Disabled

1 - Skip always

2 - Skip only if this socket was added to epoll before

Default: 0 (Disabled)


Enable the Spcketxtreme API. The potential benefits of SocketXtreme are elimination of the copy operations, higher throughput, and lower latency.


Adding a threshold parameter will trigger TCP_NODELAY only if the first segment length is larger than the threshold.

This skips the TCP_NODELAY for small segments and allows some aggregation.

Only the first segment is checked because the current behavior triggers output regardless of TCP_NODELAY when an unsent list contains at least two segments.

The value is in bytes.

Default: 0


Postpone the close(2) syscall to the socket destructor to prevent the kernel from freeing the resources before RFS destruction. This variable may lead to unexpected EADDRINUSE and excessive fd consumption in the application.

0 - Disabled

1 - Enable

Default: 0 (Disabled)


The number of TCP segments fetched from the segments pool by a ring at once.

Min value: 1

Default: 1024


Control locking type mechanism for some specific flows.

Note that usage of Mutex might increase latency.

0 - Spin

1 - Mutex

Default: 0 (Spin)


Turn on/off interrupt distribution for NGINX.

Default: false

For more information see section 11.3


This number must be equal to ‘worker_processes’ attribute of nginx configuration file.

Default: 0 (Disable)


Pre-allocated memory limit for buffers. Note that the limit doesn't include

dynamic memory allocation and XLIO memory consumption can exceed the limit.

Default value is 2GB


Memory limit for allocations done by external allocator.


Request XLIO to allocate HugePages of pecified size.

The value should be specified as log2 of the size.

Defualt: 0 (auto detect optimal size)


Print a report in human readable format in case of runtime errors. The report is

printed during termination phase. It can be missed if the process is killed with

SIGKILL signal.

Default: 0 (Disabled)

Beta Level Features Configuration Parameters

The following table lists configuration parameters and their possible values for new XLIO Beta level features. The parameters below are disabled by default.

These XLIO features are still experimental and subject to changes. They can help improve performance of multithread applications.

We recommend altering these parameters in a controlled environment until reaching the best performance tuning.

XLIO Configuration Parameter

Description and Examples



Ring migration ratio is used with the "ring per thread" logic in order to decide when it is beneficial to replace the socket's ring with the ring allocated for the current thread.

Each XLIO_RING_MIGRATION_RATIO iteration (of accessing the ring), the current thread ID is checked to see whether the ring matches the current thread.

If not, ring migration is considered. If the ring continues to be accessed from the same thread for a certain iteration, the socket is migrated to this thread ring.

Use a value of -1 in order to disable migration.

Default: -1


Limits the number of rings that can be allocated per interface.

For example, in ring allocation per socket logic, if the number of sockets using the same interface is larger than the limit, several sockets will share the same ring.


XLIO_RX_BUFS might need to be adjusted in order to have enough buffers for all rings in the system. Each ring consumes XLIO_RX_WRE buffers.

Use a value of 0 for an unlimited number of rings.

Default: 0 (no limit)


XLIO can use the on-device-memory to store the egress packet if it does not fit into the BF inline buffer. This improves application egress latency by reducing the PCI transactions.

Using XLIO_RING_DEV_MEM_TX, enables the user to set the amount of the on-device-memory buffer allocated for each TX ring.

The total size of the on-device-memory is limited to 256k for a single port HCA and to 128k for dual port HCA.

Default value is 0


TCP congestion control algorithm.

The default algorithm coming with LWIP is a variation of Reno/New-Reno.

The new Cubic algorithm was adapted from FreeBsd implementation.

Use value of 0 for LWIP algorithm.

Use value of 1 for the Cubic algorithm.

Use value of 2 in order to disable the congestion algorithm.

Default: 0 (LWIP).

XLIO can be loaded using Dynamically Loaded (DL) libraries. These libraries are not automatically loaded at program link time or start-up as with LD_PRELOAD. Instead, there is an API for opening a library, looking up symbols, handling errors, and closing the library.

The example below demonstrates how to load socket() function. Similarly, users should load all other network-related functions as declared in sock-redirect.h:


#include <stdlib.h> #include <stdio.h> #include <dlfcn.h> #include <arpa/inet.h> #include <sys/socket.h>   typedef int (*socket_fptr_t) (int __domain, int __type, int __protocol);   int main(int argc, const char** argv) { void* lib_handle; socket_fptr_t xlio_socket; int fd;   lib_handle = dlopen("libxlio.so", RTLD_LAZY); if (!lib_handle) { printf("FAILED to load libxlio.so\n"); exit(1); }   xlio_socket = (socket_fptr_t)dlsym(lib_handle, "socket"); if (xlio_socket == NULL) { printf("FAILED to load socket()\n"); exit(1); }   fd = xlio_socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP); if (fd < 0) { printf("FAILED open socket()\n"); exit(1); }   printf("socket creation succeeded fd = %d\n", fd); close(fd); dlclose(lib_handle); return 0; }

For more information, please refer to dlopen man page.

For a complete example that includes all the necessary functions, see sockperf’s xlio-redirect.h and xlio_socket-redirect.cpp files.

© Copyright 2024, NVIDIA. Last updated on May 6, 2024.