2. Marvell cnxk SSO Eventdev Driver
The SSO PMD (librte_event_cnxk) and provides poll mode eventdev driver support for the inbuilt event device found in the Marvell OCTEON cnxk SoC family.
More information about OCTEON cnxk SoC can be found at Marvell Official Website.
2.1. Supported OCTEON cnxk SoCs
- CN9XX
- CN10XX
2.2. Features
Features of the OCTEON cnxk SSO PMD are:
- 256 Event queues
- 26 (dual) and 52 (single) Event ports on CN9XX
- 52 Event ports on CN10XX
- HW event scheduler
- Supports 1M flows per event queue
- Flow based event pipelining
- Flow pinning support in flow based event pipelining
- Queue based event pipelining
- Supports ATOMIC, ORDERED, PARALLEL schedule types per flow
- Event scheduling QoS based on event queue priority
- Open system with configurable amount of outstanding events limited only by DRAM
- HW accelerated dequeue timeout support to enable power management
- HW managed event timers support through TIM, with high precision and time granularity of 2.5us on CN9K and 1us on CN10K.
- Up to 256 TIM rings a.k.a event timer adapters.
- Up to 8 rings traversed in parallel.
- HW managed packets enqueued from ethdev to eventdev exposed through event eth RX adapter.
- N:1 ethernet device Rx queue to Event queue mapping.
- Lockfree Tx from event eth Tx adapter using
RTE_ETH_TX_OFFLOAD_MT_LOCKFREE
capability while maintaining receive packet order. - Full Rx/Tx offload support defined through ethdev queue configuration.
- HW managed event vectorization on CN10K for packets enqueued from ethdev to eventdev configurable per each Rx queue in Rx adapter.
- Event vector transmission via Tx adapter.
- Up to 2 event link profiles.
2.3. Prerequisites and Compilation procedure
See Marvell cnxk platform guide for setup information.
2.4. Runtime Config Options
Maximum number of in-flight events
(default8192
)In Marvell OCTEON cnxk the max number of in-flight events are only limited by DRAM size, the
xae_cnt
devargs parameter is introduced to provide upper limit for in-flight events.For example:
-a 0002:0e:00.0,xae_cnt=16384
CN9K Getwork mode
CN9K
single_ws
devargs parameter is introduced to select single workslot mode in SSO and disable the default dual workslot mode.For example:
-a 0002:0e:00.0,single_ws=1
CN10K Getwork mode
CN10K supports three getwork prefetch modes no prefetch[0], prefetch immediately[1] and delayed prefetch on forward progress event[2]. The default getwork mode is 2.
For example:
-a 0002:0e:00.0,gw_mode=1
Event Group QoS support
SSO GGRPs i.e. queue uses DRAM & SRAM buffers to hold in-flight events. By default the buffers are assigned to the SSO GGRPs to satisfy minimum HW requirements. SSO is free to assign the remaining buffers to GGRPs based on a preconfigured threshold. We can control the QoS of SSO GGRP by modifying the above mentioned thresholds. GGRPs that have higher importance can be assigned higher thresholds than the rest. The dictionary format is as follows [Qx-TAQ-IAQ][Qz-TAQ-IAQ] expressed in percentages, 0 represents default.
For example:
-a 0002:0e:00.0,qos=[1-50-50]
CN10K WQE stashing support
CN10K supports stashing the scheduled WQE carried by rte_event to the cores L2 Dcache. The number of cache lines to be stashed and the offset is configurable per HWGRP i.e. event queue. The dictionary format is as follows [Qx|stash_offset|stash_length] here the stash offset can be a negative integer. By default, stashing is enabled on queues which have been connected to Rx adapter. Both MBUF and NIX_RX_WQE_HDR + NIX_RX_PARSE_S are stashed.
For example:
For stashing mbuf on queue 0 and mbuf + headroom on queue 1 -a 0002:0e:00.0,stash="[0|-1|1][1|-1|2]"
Force Rx Back pressure
Force Rx back pressure when same mempool is used across ethernet device connected to event device.
For example:
-a 0002:0e:00.0,force_rx_bp=1
TIM disable NPA
By default chunks are allocated from NPA then TIM can automatically free them when traversing the list of chunks. The
tim_disable_npa
devargs parameter disables NPA and uses software mempool to manage chunksFor example:
-a 0002:0e:00.0,tim_disable_npa=1
TIM modify chunk slots
The
tim_chnk_slots
devargs can be used to modify number of chunk slots. Chunks are used to store event timers, a chunk can be visualised as an array where the last element points to the next chunk and rest of them are used to store events. TIM traverses the list of chunks and enqueues the event timers to SSO. The default value is 255 and the max value is 4095.For example:
-a 0002:0e:00.0,tim_chnk_slots=1023
TIM enable arm/cancel statistics
The
tim_stats_ena
devargs can be used to enable arm and cancel stats of event timer adapter.For example:
-a 0002:0e:00.0,tim_stats_ena=1
TIM limit max rings reserved
The
tim_rings_lmt
devargs can be used to limit the max number of TIM rings i.e. event timer adapter reserved on probe. Since, TIM rings are HW resources we can avoid starving other applications by not grabbing all the rings.For example:
-a 0002:0e:00.0,tim_rings_lmt=5
TIM ring control internal parameters
When using multiple TIM rings the
tim_ring_ctl
devargs can be used to control each TIM rings internal parameters uniquely. The following dict format is expected [ring-chnk_slots-disable_npa-stats_ena]. 0 represents default values.For Example:
-a 0002:0e:00.0,tim_ring_ctl=[2-1023-1-0]
TIM external clock frequency
The
tim_eclk_freq
devagrs can be used to pass external clock frequencies when external clock source is selected.External clock frequencies are mapped as follows:
RTE_EVENT_TIMER_ADAPTER_EXT_CLK0 = TIM_CLK_SRC_10NS, RTE_EVENT_TIMER_ADAPTER_EXT_CLK1 = TIM_CLK_SRC_GPIO, RTE_EVENT_TIMER_ADAPTER_EXT_CLK2 = TIM_CLK_SRC_PTP, RTE_EVENT_TIMER_ADAPTER_EXT_CLK3 = TIM_CLK_SRC_SYNCE
The order of frequencies supplied to device args should be GPIO-PTP-SYNCE.
For Example:
-a 0002:0e:00.0,tim_eclk_freq=122880000-1000000000-0
2.5. Debugging Options
# | Component | EAL log command |
---|---|---|
1 | SSO | –log-level=’pmd.common.cnxk.event,8’ |
2 | TIM | –log-level=’pmd.common.cnxk.timer,8’ |
2.6. Limitations
2.6.1. Rx adapter support
Using the same mempool for all the ethernet device ports connected to event device would cause back pressure to be asserted only on the first ethernet device. Back pressure is automatically disabled when using same mempool for all the ethernet devices connected to event device to override this applications can use force_rx_bp=1 device arguments. Using unique mempool per each ethernet device is recommended when they are connected to event device.