12. CNXK Poll Mode driver

The CNXK ETHDEV PMD (librte_net_cnxk) provides poll mode ethdev driver support for the inbuilt network device found in Marvell OCTEON CN9K/CN10K SoC family as well as for their virtual functions (VF) in SR-IOV context.

More information can be found at Marvell Official Website.

12.1. Features

Features of the CNXK Ethdev PMD are:

Packet type information
Promiscuous mode
Jumbo frames
SR-IOV VF
Lock-free Tx queue
Multiple queues for TX and RX
Receiver Side Scaling (RSS)
MAC filtering
Generic flow API
Inner and Outer Checksum offload
Port hardware statistics
Link state information
Link flow control
MTU update
Scatter-Gather IO support
Vector Poll mode driver
Debug utilities - Context dump and error interrupt support
Support Rx interrupt
Inline IPsec processing support
Ingress meter support
Queue based priority flow control support
Port representors
Represented port pattern matching and action
Port representor pattern matching and action

12.2. Prerequisites

See Marvell cnxk platform guide for setup information.

12.3. Driver compilation and testing

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

Running testpmd:

Follow instructions available in the document compiling and testing a PMD for a NIC to run testpmd.

Example output:

./<build_dir>/app/dpdk-testpmd -c 0xc -a 0002:02:00.0 -- --portmask=0x1 --nb-cores=1 --port-topology=loop --rxq=1 --txq=1
EAL: Detected 4 lcore(s)
EAL: Detected 1 NUMA nodes
EAL: Multi-process socket /var/run/dpdk/rte/mp_socket
EAL: Selected IOVA mode 'VA'
EAL: No available hugepages reported in hugepages-16777216kB
EAL: No available hugepages reported in hugepages-2048kB
EAL: Probing VFIO support...
EAL: VFIO support initialized
EAL:   using IOMMU type 1 (Type 1)
[ 2003.202721] vfio-pci 0002:02:00.0: vfio_cap_init: hiding cap 0x14@0x98
EAL: Probe PCI driver: net_cn10k (177d:a063) device: 0002:02:00.0 (socket 0)
PMD: RoC Model: cn10k
testpmd: create a new mbuf pool <mb_pool_0>: n=155456, size=2176, socket=0
testpmd: preferred mempool ops selected: cn10k_mempool_ops
Configuring Port 0 (socket 0)
PMD: Port 0: Link Up - speed 25000 Mbps - full-duplex

Port 0: link state change event
Port 0: 96:D4:99:72:A5:BF
Checking link statuses...
Done
No commandline core given, start packet forwarding
io packet forwarding - ports=1 - cores=1 - streams=1 - NUMA support enabled, MP allocation mode: native
Logical Core 3 (socket 0) forwards packets on 1 streams:
  RX P=0/Q=0 (socket 0) -> TX P=0/Q=0 (socket 0) peer=02:00:00:00:00:00

  io packet forwarding packets/burst=32
  nb forwarding cores=1 - nb forwarding ports=1
  port 0: RX queue number: 1 Tx queue number: 1
    Rx offloads=0x0 Tx offloads=0x10000
    RX queue: 0
      RX desc=4096 - RX free threshold=0
      RX threshold registers: pthresh=0 hthresh=0  wthresh=0
      RX Offloads=0x0
    TX queue: 0
      TX desc=512 - TX free threshold=0
      TX threshold registers: pthresh=0 hthresh=0  wthresh=0
      TX offloads=0x0 - TX RS bit threshold=0
Press enter to exit

12.4. Runtime Config Options

Rx&Tx scalar mode enable (default 0)

PMD supports both scalar and vector mode, it may be selected at runtime using scalar_enable devargs parameter.
RSS reta size (default 64)
RSS redirection table size may be configured during runtime using reta_size devargs parameter.

For example:
-a 0002:02:00.0,reta_size=256
With the above configuration, reta table of size 256 is populated.
Flow priority levels (default 3)
RTE Flow priority levels can be configured during runtime using flow_max_priority devargs parameter.

For example:
-a 0002:02:00.0,flow_max_priority=10
With the above configuration, priority level was set to 10 (0-9). Max priority level supported is 32.
Reserve Flow entries (default 8)
RTE flow entries can be pre allocated and the size of pre allocation can be selected runtime using flow_prealloc_size devargs parameter.

For example:
-a 0002:02:00.0,flow_prealloc_size=4
With the above configuration, pre alloc size was set to 4. Max pre alloc size supported is 32.
Max SQB buffer count (default 512)
Send queue descriptor buffer count may be limited during runtime using max_sqb_count devargs parameter.

For example:
-a 0002:02:00.0,max_sqb_count=64
With the above configuration, each send queue’s descriptor buffer count is limited to a maximum of 64 buffers.
SQB slack count (default 12)
Send queue descriptor slack count added to SQB count when a Tx queue is created, can be set using sqb_slack devargs parameter.

For example:
-a 0002:02:00.0,sqb_slack=32
With the above configuration, each send queue’s descriptor buffer count will be increased by 32, while keeping the queue limit to default configuration.
Switch header enable (default none)
A port can be configured to a specific switch header type by using switch_header devargs parameter.

For example:
-a 0002:02:00.0,switch_header="higig2"
With the above configuration, higig2 will be enabled on that port and the traffic on this port should be higig2 traffic only. Supported switch header types are “chlen24b”, “chlen90b”, “dsa”, “exdsa”, “higig2”, “vlan_exdsa” and “pre_l2”.
Flow pre_l2 info (default 0x0/0x0/0x0)
pre_l2 headers are custom headers placed before the ethernet header. For parsing custom pre_l2 headers, an offset, mask within the offset and shift direction has to be provided within the custom header that holds the size of the custom header. This is valid only with switch header pre_l2. Maximum supported offset range is 0 to 255 and mask range is 1 to 255 and shift direction, 0: left shift, 1: right shift. Info format will be “offset/mask/shift direction”. All parameters has to be in hexadecimal format and mask should be contiguous. Info can be configured using flow_pre_l2_info devargs parameter.

For example:
-a 0002:02:00.0,switch_header="pre_l2",flow_pre_l2_info=0x2/0x7e/0x1
With the above configuration, custom pre_l2 header will be enabled on that port and size of the header is placed at byte offset 0x2 in the packet with mask 0x7e and right shift will be used to get the size. That is, size will be (pkt[0x2] & 0x7e) >> shift count. Shift count will be calculated based on mask and shift direction. For example, if mask is 0x7c and shift direction is 1 (i.e., right shift) then the shift count will be 2, that is, absolute position of the rightmost set bit. If the mask is 0x7c and shift direction is 0 (i.e., left shift) then the shift count will be 1, that is, (8 - n), where n is the absolute position of leftmost set bit.
RSS tag as XOR (default 0)
The HW gives two options to configure the RSS adder i.e
- rss_adder<7:0> = flow_tag<7:0> ^ flow_tag<15:8> ^ flow_tag<23:16> ^ flow_tag<31:24>
- rss_adder<7:0> = flow_tag<7:0>
Latter one aligns with standard NIC behavior vs former one is a legacy RSS adder scheme used in OCTEON 9 products.

By default, the driver runs in the latter mode. Setting this flag to 1 to select the legacy mode.

For example to select the legacy mode(RSS tag adder as XOR):
-a 0002:02:00.0,tag_as_xor=1
Min SPI for inbound inline IPsec (default 0)
Min SPI supported for inbound inline IPsec processing can be specified by ipsec_in_min_spi devargs parameter.

For example:
-a 0002:02:00.0,ipsec_in_min_spi=6
With the above configuration, application can enable inline IPsec processing for inbound SA with min SPI of 6.
Max SPI for inbound inline IPsec (default 255)
Max SPI supported for inbound inline IPsec processing can be specified by ipsec_in_max_spi devargs parameter.

For example:
-a 0002:02:00.0,ipsec_in_max_spi=128
With the above configuration, application can enable inline IPsec processing with max SPI of 128.
Max SA's for outbound inline IPsec (default 4096)
Max number of SA’s supported for outbound inline IPsec processing can be specified by ipsec_out_max_sa devargs parameter.

For example:
-a 0002:02:00.0,ipsec_out_max_sa=128
With the above configuration, application can enable inline IPsec processing for 128 outbound SAs.
Enable custom SA action (default 0)
Custom SA action can be enabled by specifying custom_sa_act devargs parameter.

For example:
-a 0002:02:00.0,custom_sa_act=1
With the above configuration, application can enable custom SA action. This configuration allows the potential for a MCAM entry to match many SAs, rather than only match a single SA. For cnxk device sa_index will be calculated based on SPI value. So, it will be 1 to 1 mapping. By enabling this devargs and setting a MCAM rule, will allow application to configure the sa_index as part of session create. And later original SPI value can be updated using session update. For example, application can set sa_index as 0 using session create as SPI value and later can update the original SPI value (for example 0x10000001) using session update. And create a flow rule with security action and algorithm as RTE_PMD_CNXK_SEC_ACTION_ALG0 and sa_hi as 0x1000 and sa_lo as 0x0001.

Outbound CPT LF queue size (default 8200)

Size of Outbound CPT LF queue in number of descriptors can be specified by outb_nb_desc devargs parameter.

For example:
  -a 0002:02:00.0,outb_nb_desc=16384

With the above configuration, Outbound CPT LF will be created to accommodate
at max 16384 descriptors at any given time.

Outbound CPT LF count (default 1)
Number of CPT LF’s to attach for Outbound processing can be specified by outb_nb_crypto_qs devargs parameter.

For example:
-a 0002:02:00.0,outb_nb_crypto_qs=2
With the above configuration, two CPT LF’s are setup and distributed among all the Tx queues for outbound processing.
Disable using inline ipsec device for inbound (default 0)
In CN10K, in event mode, driver can work in two modes,
1. Inbound encrypted traffic received by probed ipsec inline device while plain traffic post decryption is received by ethdev.
2. Both Inbound encrypted traffic and plain traffic post decryption are received by ethdev.
By default event mode works using inline device i.e mode 1. This behaviour can be changed to pick mode 2 by using no_inl_dev devargs parameter.

For example:
-a 0002:02:00.0,no_inl_dev=1 -a 0002:03:00.0,no_inl_dev=1
With the above configuration, inbound encrypted traffic from both the ports is received by ipsec inline device.
Inline IPsec device channel and mask (default none)
Set channel and channel mask configuration for the inline IPSec device. This will be used when creating flow rules with RTE_FLOW_ACTION_TYPE_SECURITY action.

By default, RTE Flow API sets the channel number of the port on which the rule is created in the MCAM entry and matches it exactly. This behaviour can be modified using the inl_cpt_channel devargs parameter.

For example:
-a 0002:1d:00.0,inl_cpt_channel=0x100/0xf00
With the above configuration, RTE Flow rules API will set the channel and channel mask as 0x100 and 0xF00 in the MCAM entries of the flow rules created with RTE_FLOW_ACTION_TYPE_SECURITY action. Since channel number is set with this custom mask, inbound encrypted traffic from all ports with matching channel number pattern will be directed to the inline IPSec device.
Inline IPsec device flow rules (default none)
For inline IPsec device, reserve number of rules specified by max_ipsec_rules and use them while installing rules with action as security. Rule priority should be 0. If specified number of rules not available, then only available number of rules will be allocated and used. If application try to insert more than allocated rules, flow creation will fail.

For example:
-a 0002:1d:00.0,max_ipsec_rules=100
With the above configuration, 100 rules will be allocated from 0-99 if available and will be used for rules with action security. If 100 rules are not available, and only 50 are available, then only 50 rules will be allocated and used for flow rule creation. If application try to add more than 50 rules, the flow creation will fail.
SDP device channel and mask (default none)
Set channel and channel mask configuration for the SDP device. This will be used when creating flow rules on the SDP device.

By default, for rules created on the SDP device, the RTE Flow API sets the channel number and mask to cover the entire SDP channel range in the channel field of the MCAM entry. This behaviour can be modified using the sdp_channel_mask devargs parameter.

For example:
-a 0002:1d:00.0,sdp_channel_mask=0x700/0xf00
With the above configuration, RTE Flow rules API will set the channel and channel mask as 0x700 and 0xF00 in the MCAM entries of the flow rules created on the SDP device. This option needs to be used when more than one SDP interface is in use and RTE Flow rules created need to distinguish between traffic from each SDP interface. The channel and mask combination specified should match all the channels(or rings) configured on the SDP interface.
Transmit completion handler (default 0)
When transmit completion handler is enabled, the PMD invokes the callback handler provided by the application for every packet which has external buf attached to mbuf and frees main mbuf, external buffer is provided to applicatoin. Once external buffer is handed over to application, it is application responsibility either to free or reuse external buffer using tx_compl_ena devargs parameter.

For example:
-a 0002:01:00.1,tx_compl_ena=1
Meta buffer size per ethdev port for inline inbound IPsec second pass
Size of meta buffer allocated for inline inbound IPsec second pass per ethdev port can be specified by meta_buf_sz devargs parameter. Default value is computed runtime based on pkt mbuf pools created and in use. This option is for OCTEON CN106-B0/CN103XX SoC family.

For example:
-a 0002:02:00.0,meta_buf_sz=512
With the above configuration, PMD would allocate meta buffers of size 512 for inline inbound IPsec processing second pass.
NPC MCAM Aging poll frequency in seconds (default 10)
Poll frequency for aging control thread can be specified by aging_poll_freq devargs parameter.

For example:
-a 0002:01:00.2,aging_poll_freq=50
With the above configuration, driver would poll for aging flows every 50 seconds.
Rx Inject Enable inbound inline IPsec for second pass (default 0)
Rx packet inject feature for inbound inline IPsec processing can be enabled by rx_inj_ena devargs parameter. This option is for OCTEON CN106-B0/CN103XX SoC family.

For example:
-a 0002:02:00.0,rx_inj_ena=1
With the above configuration, driver would enable packet inject from ARM cores to crypto to process and send back in Rx path.
Disable custom meta aura feature (default 0)
Custom meta aura i.e 1:N meta aura is enabled for second pass traffic by default when inl_cpt_channel devarg is provided. The custom meta aura feature can be disabled by setting devarg custom_meta_aura_dis to 1.

For example:
-a 0002:02:00.0,custom_meta_aura_dis=1
With the above configuration, the driver would disable custom meta aura feature for the device 0002:02:00.0.
Enable custom SA for inbound inline IPsec (default 0)
Custom SA for inbound inline IPsec can be enabled by specifying custom_inb_sa devargs parameter. This option needs to be given to both ethdev and inline device.

For example:
-a 0002:02:00.0,custom_inb_sa=1
With the above configuration, inline inbound IPsec post-processing should be done by the application.

Note

Above devarg parameters are configurable per device, user needs to pass the parameters to all the PCIe devices if application requires to configure on all the ethdev ports.

12.5. Limitations

12.5.1. `mempool_cnxk` external mempool handler dependency

The OCTEON CN9K/CN10K SoC family NIC has inbuilt HW assisted external mempool manager. net_cnxk PMD only works with mempool_cnxk mempool handler as it is performance wise most effective way for packet allocation and Tx buffer recycling on OCTEON 9 SoC platform.

12.5.2. `mempool_cnxk` rte_mempool cache sizes for CN10K

The OCTEON CN10K SoC Family supports asynchronous batch allocation of objects from an NPA pool. In the CNXK mempool driver, asynchronous batch allocation is enabled when local caches are enabled. This asynchronous batch allocation will be using an additional local async buffer whose size will be equal to RTE_ALIGN_CEIL(rte_mempool->cache_size, 16). This can result in additional objects being cached locally. While creating an rte_mempool using mempool_cnxk driver for OCTEON CN10K, this must be taken into consideration and the local cache sizes should be adjusted accordingly so that starvation does not happen.

For Eg: If the cache_size passed into rte_mempool_create is 8, then the max objects than can get cached locally on a core would be the sum of max objects in the local cache + max objects in the async buffer i.e 8 + RTE_ALIGN_CEIL(8, 16) = 24.

12.5.3. CRC stripping

The OCTEON CN9K/CN10K SoC family NICs strip the CRC for every packet being received by the host interface irrespective of the offload configuration.

12.5.4. RTE flow GRE support

RTE_FLOW_ITEM_TYPE_GRE_KEY works only when checksum and routing bits in the GRE header are equal to 0.

12.5.5. RTE flow action represented_port support

RTE_FLOW_ACTION_TYPE_REPRESENTED_PORT only works between a PF and its VFs.

12.5.6. RTE flow action port_id support

RTE_FLOW_ACTION_TYPE_PORT_ID is only supported between PF and its VFs.

12.5.7. Custom protocols supported in RTE Flow

The RTE_FLOW_ITEM_TYPE_RAW can be used to parse the below custom protocols.

vlan_exdsa and exdsa can be parsed at L2 level.
NGIO can be parsed at L3 level.

For vlan_exdsa and exdsa, the port has to be configured with the respective switch header.

For example:

-a 0002:02:00.0,switch_header="vlan_exdsa"

The below fields of struct rte_flow_item_raw shall be used to specify the pattern.

relative Selects the layer at which parsing is done.
- 0 for exdsa and vlan_exdsa.
- 1 for NGIO.
offset The offset in the header where the pattern should be matched.
length Length of the pattern.
pattern Pattern as a byte string.

Example usage in testpmd:

./dpdk-testpmd -c 3 -w 0002:02:00.0,switch_header=exdsa -- -i \
               --rx-offloads=0x00080000 --rxq 8 --txq 8
testpmd> flow create 0 ingress pattern eth / raw relative is 0 pattern \
       spec ab pattern mask ab offset is 4 / end actions queue index 1 / end

12.5.8. RTE Flow mark item support

RTE_FLOW_ITEM_TYPE_MARK can be used to create ingress flow rules to match packets from CPT(second pass packets). When mark item type is used, it should be the first item in the patterns specification.

12.6. Inline device support for CN10K

CN10K HW provides a misc device Inline device that supports ethernet devices in providing following features.

Aggregate all the inline IPsec inbound traffic from all the CN10K ethernet devices to be processed by the single inline IPSec device. This allows single rte security session to accept traffic from multiple ports.

Support for event generation on outbound inline IPsec processing errors.

Support CN106xx poll mode of operation for inline IPSec inbound processing.

Inline IPsec device is identified by PCI PF vendid:devid 177D:A0F0 or VF 177D:A0F1.

12.6.1. Runtime Config Options for inline device

Min SPI for inbound inline IPsec (default 0)
Min SPI supported for inbound inline IPsec processing can be specified by ipsec_in_min_spi devargs parameter.

For example:
-a 0002:1d:00.0,ipsec_in_min_spi=6
With the above configuration, application can enable inline IPsec processing for inbound SA with min SPI of 6 for traffic aggregated on inline device.
Max SPI for inbound inline IPsec (default 255)
Max SPI supported for inbound inline IPsec processing can be specified by ipsec_in_max_spi devargs parameter.

For example:
-a 0002:1d:00.0,ipsec_in_max_spi=128
With the above configuration, application can enable inline IPsec processing for inbound SA with max SPI of 128 for traffic aggregated on inline device.
Count of meta buffers for inline inbound IPsec second pass
Number of meta buffers allocated for inline inbound IPsec second pass can be specified by nb_meta_bufs devargs parameter. Default value is computed runtime based on pkt mbuf pools created and in use. Number of meta buffers should be at least equal to aggregated number of packet buffers of all packet mbuf pools in use by Inline IPsec enabled ethernet devices.

For example:
-a 0002:1d:00.0,nb_meta_bufs=1024
With the above configuration, PMD would enable inline IPsec processing for inbound with 1024 meta buffers available for second pass.
Meta buffer size for inline inbound IPsec second pass
Size of meta buffer allocated for inline inbound IPsec second pass can be specified by meta_buf_sz devargs parameter. Default value is computed runtime based on pkt mbuf pools created and in use.

For example:
-a 0002:1d:00.0,meta_buf_sz=512
With the above configuration, PMD would allocate meta buffers of size 512 for inline inbound IPsec processing second pass.
Inline Outbound soft expiry poll frequency in usec (default 100)
Soft expiry poll frequency for Inline Outbound sessions can be specified by soft_exp_poll_freq devargs parameter.

For example:
-a 0002:1d:00.0,soft_exp_poll_freq=1000
With the above configuration, driver would poll for soft expiry events every 1000 usec.
Rx Inject Enable inbound inline IPsec for second pass (default 0)
Rx packet inject feature for inbound inline IPsec processing can be enabled by rx_inj_ena devargs parameter with both inline device and ethdev device. This option is for OCTEON CN106-B0/CN103XX SoC family.

For example:
-a 0002:1d:00.0,rx_inj_ena=1
With the above configuration, driver would enable packet inject from ARM cores to crypto to process and send back in Rx path.
Enable custom SA for inbound inline IPsec (default 0)
Custom SA for inbound inline IPsec can be enabled by specifying custom_inb_sa devargs parameter with both inline device and ethdev.

For example:
-a 0002:1d:00.0,custom_inb_sa=1
With the above configuration, inline inbound IPsec post-processing should be done by the application.

12.7. Port Representors

The CNXK driver supports port representor model by adding virtual ethernet ports providing a logical representation in DPDK for physical function (PF) or SR-IOV virtual function (VF) devices for control and monitoring.

Base device or parent device underneath the representor ports is an eswitch device which is not a cnxk ethernet device but has NIC Rx and Tx capabilities. Each representor port is represented by a RQ and SQ pair of this eswitch device.

Implementation supports representors for both physical function and virtual function.

Port representor ethdev instances can be spawned on an as needed basis through configuration parameters passed to the driver of the underlying base device using devargs -a <base PCI BDF>,representor=pf*vf*