9. DPAA Poll Mode Driver

9.1.1. DPAA Overview

Reference: FSL DPAA Architecture.

The QorIQ Data Path Acceleration Architecture (DPAA) is a set of hardware components on specific QorIQ series multicore processors. This architecture provides the infrastructure to support simplified sharing of networking interfaces and accelerators by multiple CPU cores, and the accelerators themselves.

DPAA includes:

• Cores
• Network and packet I/O
• Hardware offload accelerators
• Infrastructure required to facilitate flow of packets between the components above

Infrastructure components are:

• The Queue Manager (QMan) is a hardware accelerator that manages frame queues. It allows CPUs and other accelerators connected to the SoC datapath to enqueue and dequeue ethernet frames, thus providing the infrastructure for data exchange among CPUs and datapath accelerators.
• The Buffer Manager (BMan) is a hardware buffer pool management block that allows software and accelerators on the datapath to acquire and release buffers in order to build frames.

Hardware accelerators are:

• SEC - Cryptographic accelerator
• PME - Pattern matching engine

The Network and packet I/O component:

• The Frame Manager (FMan) is a key component in the DPAA and makes use of the DPAA infrastructure (QMan and BMan). FMan is responsible for packet distribution and policing. Each frame can be parsed, classified and results may be attached to the frame. This meta data can be used to select particular QMan queue, which the packet is forwarded to.

9.2.1. DPAA Bus driver

The DPAA bus driver is a rte_bus driver which scans the platform like bus. Key functions include:

• Scanning and parsing the various objects and adding them to their respective device list.
• Performing probe for available drivers against each scanned device
• Creating necessary ethernet instance before passing control to the PMD

9.2.2. DPAA NIC Driver (PMD)

DPAA PMD is traditional DPDK PMD which provides necessary interface between RTE framework and DPAA internal components/drivers.

• Once devices have been identified by DPAA Bus, each device is associated with the PMD
• PMD is responsible for implementing necessary glue layer between RTE APIs and lower level QMan and FMan blocks. The Ethernet driver is bound to a FMAN port and implements the interfaces needed to connect the DPAA network interface to the network stack. Each FMAN Port corresponds to a DPDK network interface.

9.2.3. DPAA Mempool Driver

DPAA has a hardware offloaded buffer pool manager, called BMan, or Buffer Manager.

• Using standard Mempools operations RTE API, the mempool driver interfaces with RTE to service each mempool creation, deletion, buffer allocation and deallocation requests.
• Each FMAN instance has a BMan pool attached to it during initialization. Each Tx frame can be automatically released by hardware, if allocated from this pool.

9.5.1. Config File Options

The following options can be modified in the config file. Please note that enabling debugging options may affect system performance.

• CONFIG_RTE_LIBRTE_DPAA_BUS (default n)

  By default it is enabled only for defconfig_arm64-dpaa-* config. Toggle compilation of the librte_bus_dpaa driver.

• CONFIG_RTE_LIBRTE_DPAA_PMD (default n)

  By default it is enabled only for defconfig_arm64-dpaa-* config. Toggle compilation of the librte_pmd_dpaa driver.

• CONFIG_RTE_LIBRTE_DPAA_DEBUG_DRIVER (default n)

  Toggles display of bus configurations and enables a debugging queue to fetch error (Rx/Tx) packets to driver. By default, packets with errors (like wrong checksum) are dropped by the hardware.

• CONFIG_RTE_LIBRTE_DPAA_HWDEBUG (default n)

  Enables debugging of the Queue and Buffer Manager layer which interacts with the DPAA hardware.

• CONFIG_RTE_MBUF_DEFAULT_MEMPOOL_OPS (default dpaa)

  This is not a DPAA specific configuration - it is a generic RTE config. For optimal performance and hardware utilization, it is expected that DPAA Mempool driver is used for mempools. For that, this configuration needs to enabled.

9.5.2. Environment Variables

DPAA drivers uses the following environment variables to configure its state during application initialization:

• DPAA_NUM_RX_QUEUES (default 1)

  This defines the number of Rx queues configured for an application, per port. Hardware would distribute across these many number of queues on Rx of packets. In case the application is configured to use lesser number of queues than configured above, it might result in packet loss (because of distribution).

• DPAA_PUSH_QUEUES_NUMBER (default 4)

  This defines the number of High performance queues to be used for ethdev Rx. These queues use one private HW portal per queue configured, so they are limited in the system. The first configured ethdev queues will be automatically be assigned from the these high perf PUSH queues. Any queue configuration beyond that will be standard Rx queues. The application can choose to change their number if HW portals are limited. The valid values are from ‘0’ to ‘4’. The valuse shall be set to ‘0’ if the application want to use eventdev with DPAA device.

9.7.1. Platform Requirement

DPAA drivers for DPDK can only work on NXP SoCs as listed in the Supported DPAA SoCs.

9.7.2. Maximum packet length

The DPAA SoC family support a maximum of a 10240 jumbo frame. The value is fixed and cannot be changed. So, even when the rxmode.max_rx_pkt_len member of struct rte_eth_conf is set to a value lower than 10240, frames up to 10240 bytes can still reach the host interface.

9.7.3. Multiprocess Support

Current version of DPAA driver doesn’t support multi-process applications where I/O is performed using secondary processes. This feature would be implemented in subsequent versions.