October 26, 2023 ##################################################################### Attendees 1. Patrick Robb 2. Paul Szczepanek 3. Aaron Conole ##################################################################### Agenda 1. General Announcements 2. CI Status 3. DTS Improvements & Test Development 4. Any other business ##################################################################### Minutes ===================================================================== General Announcements * Patrick will share out the 2023 Community Lab retrospective today, along with a last call on requests to be included in the work item voting sheet: https://docs.google.com/spreadsheets/d/1DL1t_h-JOz1tTcO7hc8EpJnUQsHyj8YxJqx-JxrvKz0/edit?usp=sharing * There is a joint tech board / gov board meeting next Wednesday on the topic of extending the scope/stack of DPDK. ===================================================================== CI Status --------------------------------------------------------------------- UNH-IOL Community Lab * Arm64 unit testing was re-enabled after David’s patch cleaning up ethdev shared memory on port cleanup reached main and all next maintainers rebased * Intel 8970 QAT Accelerator card: This is still blocked, but UNH will ping Dharmik etc. on how to proceed * TS-Factory: * We are pulling a fresh version of the ts-factory framework with updates (from Andrew) to allow manual runs to produce all salient artifacts, then sending them to be published on Oktet Lab’s instance * After, we’ll begin on setting up the Jenkins automation and hosting our own Bublik instance for publishing results * As a reminder our test matrix for ts-factory will be as follows: * X86_64: Intel XL710, ConnectX-cx5 * Arm64: Intel XL710, ConnectX-cx5 * Apply patchset script: * Need to pick this back up, but the draft version “works” and all that is remaining is implementing the depends-on support, and performing some testing on it before bringing it to production. * Marvell Octeon board: They have granted me access to the CN106XX board documentation, and I have some (very basic) info from them about how they utilize DTS, although I hope to get more info. However, I’m putting together a parts purchase list now (for all components to facilitate the bare board) as I am now satisfied that using the hardware in the lab will be feasible. --------------------------------------------------------------------- Intel Lab * None --------------------------------------------------------------------- Loongarch Lab * None --------------------------------------------------------------------- Github Actions * Should be going live within a week for the email based retesting framework. Aaron needs to sync with Michael on this. ===================================================================== DTS Improvements & Test Development * Scatter: Jeremy’s rewrite of the testsuite in the new DTS framework is working. * The one interesting thing is that as it is written in old dts, topology=loop is being used as a testpmd arg, basically just meaning that the ingress port on the DUT becomes also the egress port. This does not align with the given addresses assignment function implemented in the TestSuite function, so he overrode this behavior accordingly. It’s a bit ugly to do this, so probably what is best to do is to have testpmd forward from port 0 to 1, or 1 to 0, with topology=paired. The only important thing to do with this test is to set the rx ring mbuf size < frame size, and then interrogate the packets the tester receives back (showing that you are receiving scattered segments). Testpmd forwarding approach is unrelated to validating scatter functionality. * However, it does highlight that I should sync with Jeremy (and hopefully the entire DTS WG, to see how to best support different topologies in DTS, and how to support full-duplex traffic. * To that end, I would like to revive the DTS WG meetings (seperate from CI meetings) because A. Jeremy can’t make the CI meetings, and B. we will possibly be adding an additional DTS developer to the community lab * What time could this work at? 10 AM Wed? ===================================================================== Any other business * Next meeting is Nov 9, 2023