diff options
Diffstat (limited to 'website/content/en/status/report-2022-01-2022-03/dpaa2.adoc')
| -rw-r--r-- | website/content/en/status/report-2022-01-2022-03/dpaa2.adoc | 32 |
1 files changed, 32 insertions, 0 deletions
diff --git a/website/content/en/status/report-2022-01-2022-03/dpaa2.adoc b/website/content/en/status/report-2022-01-2022-03/dpaa2.adoc new file mode 100644 index 0000000000..b7fdc16a03 --- /dev/null +++ b/website/content/en/status/report-2022-01-2022-03/dpaa2.adoc @@ -0,0 +1,32 @@ +=== NXP DPAA2 support + +Links + +link:https://github.com/mcusim/freebsd-src/tree/lx2160acex7-dev/sys/dev/dpaa2[Development] URL: https://github.com/mcusim/freebsd-src/tree/lx2160acex7-dev/sys/dev/dpaa2[https://github.com/mcusim/freebsd-src/tree/lx2160acex7-dev/sys/dev/dpaa2] + + +Contact: Dmitry Salychev <dsl@mcusim.org> + +Contact: Bjoern A. Zeeb <bz@FreeBSD.org> + +Some of the NXP SoCs (LX2160A, LS1088A) are shipped with link:https://www.nxp.com/design/qoriq-developer-resources/second-generation-data-path-acceleration-architecture-dpaa2:DPAA2[DPAA2], the second generation of the data path acceleration architecture. +It allows to dynamically configure and wire packet processing "objects" (like DPNI for a network interface, DPMAC for media access controller, etc.) together to form a network-on-a-chip. + +link:https://solidrun.atlassian.net/wiki/spaces/developer/pages/197494288/HoneyComb+LX2+ClearFog+CX+LX2+Quick+Start+Guide[Honeycomb LX2], as one of the powerful ARM64 boards from SolidRun, is also built around NXP LX2160A and lacks on-board GbE support in FreeBSD. +These drivers are about to change it. + +It's still work in progress and a rough implementation of the drivers for some DPAA2 objects is available at the moment. +A new network interface (dpni) can be used on Honeycomb for testing. + +TODO: + +* Update dpaa2_mc_fdt driver for the link:https://www.crowdsupply.com/traverse-technologies/ten64[Traverse Ten64] board + +* Review and fix DPAA2 resources allocation to support as many DPNIs as possible + +* Free driver resources properly + +* Implement CPU affinity for DPIOs and DPNIs + +* Implement cached memory-backed software portals + +* Profile and mitigate bottlenecks + +Sponsor: Bare Enthusiasm :) |