Where to upskill for working with HSS

[LucienMorey]

I am trying to bring up a me-mp1-250-si-d3en made by Enclustra FPGA Solutions, but I am not having much luck. They provide a forked version of HSS that is older than the current release. After deploying the generated elf file in my bitstream over jtag I get two different errors seemingly at random. These are the logs:

Error 1

HSS: decompressing from eNVM to L2 Scratch ... Passed
[0.43629] Switching to UART 0 ...
DDR training ... Passed ( 5431 ms)
[5.484366] PolarFire(R) SoC Hart Software Services (HSS) - version 0.99.33-dev-build
MPFS HAL version 2.0.101 / DDR Driver version 0.4.018 / BOARD=enclustra-mercury-mp1
(c) Copyright 2017-2022 Microchip FPGA Embedded Systems Solutions.

incorporating OpenSBI - version 1.0
(c) Copyright 2019-2022 Western Digital Corporation.

[5.517784] Build ID: 0e4d634c8fd70316fecc258870cb365b7c0a0dc8
[5.525327] Built with the following tools:
 - riscv64-unknown-elf-gcc (xPack GNU RISC-V Embedded GCC (Microsemi SoftConsole build), 64-bit) 8.3.0
 - GNU ld (xPack GNU RISC-V Embedded GCC (Microsemi SoftConsole build), 64-bit) 2.32

[5.549388] NOTICE: Running from L2 Scratchpad

[5.555594] Serial Number:
b120f4235a931116166d7006f6ec5a2b00000000000000000000000000000000000000000000000000000000000000000000
[5.569534] Segment Configuration:
        Cached: SEG0_1: offset 0x1000000000, physical DDR 0x00000000
[5.581087] L2 Cache Configuration:
    L2-Scratchpad:  4 ways (512 KiB)
         L2-Cache:  8 ways (1024 KiB)
           L2-LIM:  4 ways (512 KiB)
[5.597032] DDR-Lo size is    0 MiB
[5.601711] DDR-Hi size is 2048 MiB
[6.106461] ETH PHY found at MDIO address 0
[6.111938] ETH PHY found at MDIO address 3
[6.118136] Initializing memory offset 0x1000000000 size 0x80000000
[27.934882] Attempting to select SDCARD ... Passed
Press a key to enter CLI, ESC to skip
Timeout in 1 second
..
[29.156019] CLI boot interrupt timeout
[29.161080] Initializing IPI Queues (6056 bytes @ 0xa02b458)...
[29.168527] Initializing PMPs
[29.172728] Initializing Boot Image ...
[29.177884] Trying to boot via MMC ...
[29.182944] Preparing to copy from MMC to DDR ...
[29.194940] Validated GPT Header ...
[29.199888] Unable to read block for LBA 2 (partition entry 0)
[29.207526] GPT_ValidatePartitionEntries() failed
[29.213923] GPT_FindBootSectorIndex() failed
[29.219843] GPT_PartitionIdToLBAOffset() failed - using offset 0
[29.227673] Attempting to read image header (1632 bytes) ...
[29.240267] HSS_MMC_ReadBlock() failed
[29.245614] HSS_BootInit() returned 0
Init failed, press a key to prevent restart
Timeout in 5 seconds
.....

Error 2

HSS: decompressing from eNVM to L2 Scratch ... Passed
[0.43634] Switching to UART 0 ...
DDR training ... Passed ( 5839 ms)
[5.892687] PolarFire(R) SoC Hart Software Services (HSS) - version 0.99.33-dev-build
MPFS HAL version 2.0.101 / DDR Driver version 0.4.018 / BOARD=enclustra-mercury-mp1
(c) Copyright 2017-2022 Microchip FPGA Embedded Systems Solutions.

incorporating OpenSBI - version 1.0
(c) Copyright 2019-2022 Western Digital Corporation.

[5.926105] Build ID: 0e4d634c8fd70316fecc258870cb365b7c0a0dc8
[5.933648] Built with the following tools:
 - riscv64-unknown-elf-gcc (xPack GNU RISC-V Embedded GCC (Microsemi SoftConsole build), 64-bit) 8.3.0
 - GNU ld (xPack GNU RISC-V Embedded GCC (Microsemi SoftConsole build), 64-bit) 2.32

[5.957709] NOTICE: Running from L2 Scratchpad

[5.963915] Serial Number:
b120f4235a931116166d7006f6ec5a2b00000000000000000000000000000000000000000000000000000000000000000000
[5.977855] Segment Configuration:
        Cached: SEG0_1: offset 0x1000000000, physical DDR 0x00000000
[5.989408] L2 Cache Configuration:
    L2-Scratchpad:  4 ways (512 KiB)
         L2-Cache:  8 ways (1024 KiB)
           L2-LIM:  4 ways (512 KiB)
[6.05354] DDR-Lo size is    0 MiB
[6.09937] DDR-Hi size is 2048 MiB
[6.514592] ETH PHY found at MDIO address 0
[6.520069] ETH PHY found at MDIO address 3
[6.526266] Initializing memory offset 0x1000000000 size 0x80000000
[28.343012] Attempting to select SDCARD ... Failed
[28.632742] Attempting to select eMMC ... Passed
Press a key to enter CLI, ESC to skip
Timeout in 1 second
..
[29.925680] CLI boot interrupt timeout
[29.930741] Initializing IPI Queues (6056 bytes @ 0xa02b458)...
[29.938188] Initializing PMPs
[29.942389] Initializing Boot Image ...
[29.947545] Trying to boot via MMC ...
[29.952606] Preparing to copy from MMC to DDR ...
[29.960010] GPT_PartitionIdToLBAOffset() failed - using offset 0
[29.967839] Attempting to read image header (1632 bytes) ...
[29.975208] magic is d78c9a49 vs expected b007c0de or c08b8355
[29.982847] HSS_Boot_VerifyMagic() failed
[29.988480] HSS_BootInit() returned 0
Init failed, press a key to prevent restart
Timeout in 5 seconds
.....

As far as I can tell, the errors thrown are related to how HSS tries to unpack a bootable image. Enclustra doesn't provide substantial documentation for dealing with issues and is in a different timezone than me, so any support I get has an irritating turnaround. I have been looking around here at the source and haven't found the thing that makes it all click in the docs repo or the hart-software-services repository .

I have also tried rebasing their board packaging on the 2024.6 release of HSS. This has made things more opaque because more of the bringup process is being run as a service. As best as I can tell from the documentation, things now all run independently of the specification in hss_board_init.c and hss_registry.c, but any debug prints are no longer included in the console.

Can anyone point me in the right direction so I can try to get back to a baseline?