Hi Cody, Can you please refer our SDK document? It should have clear instructions. We can sync up if you still have any questions. Regards, Hiral ________________________________ From: Cody Cheng Sent: Friday, December 13, 2024 11:47 AM To: Gnanesh Kambalu Palanethra ; Hiral Shah Cc: JogaRao Nartu ; Bharath Rajendra ; Patrick Robb ; ci@dpdk.org Subject: [EXTERNAL] CN10K Crypto Test Issue Hi Gnanesh & Hiral, My name is Cody Cheng, I'm a tester at the University of New Hampshire DPDK Community Test Lab. We are hosting a CN10K board here which is currently running some ethernet device tests on DPDK, and we would like to extend Hi Gnanesh & Hiral, My name is Cody Cheng, I'm a tester at the University of New Hampshire DPDK Community Test Lab. We are hosting a CN10K board here which is currently running some ethernet device tests on DPDK, and we would like to extend our testing to include crypto device testing. Gnanesh has written a patch which adds testcases to the DPDK Test Suite which should allow us to do so. However, I am running into difficulties with the crypto devices I create in DPDK from the CN10K board. I would appreciate it if I can sync with one of you next week so that I can show our current configuration, and run through the DPDK crypto device setup process and CN10K autotest (not passing currently). I am guessing there is some error in our configuration. My coworker Patrick Robb has mentioned he met with Hiral previously and it was a big help for understanding how to flash the correct firmware, build the SDK and tftpboot it, chroot to Ubuntu etc. I hope we can do something similar to clear up the confusion regarding crypto devs. What timezones are you in? I would be happy to schedule a Zoom call. Otherwise, I will share some of the system info and the process I have run through below, which might begin to give you an idea regarding our status: ======================= Marvell CN10k Boot Stub ======================= Firmware Version: 2024-12-07 02:04:42 EBF Version: 12.24.11, Branch: /MarvellSDK/base-SDK12.24.11/cn10ka-release-output/build/marvell-external-fw-SDK12.24.11/firmware/ebf, Built: Sat, 07 Dec 2024 02:02:27 +0000 Board Model: crb106 Board Revision: r1p1 Board Serial: Chip: 0xb9 Pass A1 SKU: MV-CN10624-A1-AAP LLC: 49152 KB Boot: SPI1_CS0,EMMC_CS0, using SPI1_CS0 AVS: Enabled I am setting up 1 crypto virtual function using the commands here: https://urldefense.proofpoint.com/v2/url?u=https-3A__doc.dpdk.org_guides_cryptodevs_cnxk.html-23initialization&d=DwIBaQ&c=nKjWec2b6R0mOyPaz7xtfQ&r=vGy6A_Vxl0vuken84vHnqSHz1sugoMRzgsw2uuNRAQ4&m=dN2bwgTF6BzuBPsPbKfZfmAua-foxlSSnMAoU8EjwCE3XN40flneiz1d7Mwl83iN&s=_d4kcuFSGcAGResAexjf_pFNDJ0Szm68pM2tw5BFzUM&e= So afterwards I am left with 1 VF bound to vfio-pci at 0002:10.00.1 At this stage, according to the docs I should be able to launch dpdk-test and run the cn10k symmetrical crypto autotest, using the commands below: ``` ./dpdk-test RTE>>cryptodev_cn10k_autotest ``` However, the auto tests fail and fall into an error loop which I have attached the logs of in this email. Here is the EAL output from the logs: EAL: Detected CPU lcores: 24 EAL: Detected NUMA nodes: 1 EAL: Detected static linkage of DPDK EAL: Multi-process socket /var/run/dpdk/rte/mp_socket EAL: Selected IOVA mode 'VA' EAL: VFIO support initialized EAL: Using IOMMU type 1 (Type 1) CNXK: RoC Model: cn10ka_a1 (HW_PLATFORM) CRYPTODEV: Creating cryptodev 0002:20:00.1 CRYPTODEV: Initialisation parameters - name: 0002:20:00.1,socket id: 0, max queue pairs: 0 APP: HPET is not enabled, using TSC as default timer In the EAL output, max queue pairs is 0 even though in the docs, it says the Maximum queue pairs limit is set to a default of 63. Could this be related to the issue? Also, here is my kernel cmdline parameters: console=ttyAMA0,115200n8 earlycon=p1011,0x87e028000000 maxcpus=24 rootwait root=/dev/nvme0n1p1 rw coherent_pool=16M default_hugepagesz=512M hugepagesz=512M hugepages=8 Does this look correct? I have also tried setting `iommu.passthrough=1` in the boot arguments but that resulted in the same dpdk-test failure. Best Regards, Cody Cheng