IceSugarPro (Lattice ECP5-25) broken since recent commit 0bc25944692d508a44878005ca5f292f5a7f9180

[robinsonb5]

The recent commit 0bc2594 has broken the ability of openFPGALoader to program the icesugarpro FPGA board with Lattice ECP5-25. It goes through the motions of programming, but shows an error message afterwards, and the FPGA doesn't start.

Prior to this commit it was working fine.

The board has a CMSIS-DAP programmer built in, which isn't automatically recognised by openFPGALoader but can be used if the USB vendor and product IDs are specified.

I'm attempting to configure the SRAM in the FPGA with the following command:

~/FPGA/openFPGALoader/openFPGALoader --verbose-level=2 -c cmsisdap --vid 0x1d50 --pid 0x602b impl1/tg68test_impl1.bit

The debug-verbose output is as follows:

empty
Cable VID overridden
Cable PID overridden
Found 1 compatible device:
0x1d50 0x602b 0x3 DAPLink CMSIS-DAP
VID: 1d50
PID: 602b
serial number : 0710000100420043380000184e514b43a5a5a5a597969908
firmware version : 0255
hardware capabilities : 13
SWO trace buffer size : NA
max packet cnt : 04
max packet size : 64
Hardware cap 00 01 13
Raw IDCODE:

• 0 -> 0x41111043
• 1 -> 0xffffffff
  Fetched TDI, end-of-chain
  found 1 devices
  index 0:
  idcode 0x41111043
  manufacturer lattice
  family ECP5
  model LFE5U-25
  irlength 8
  File type : bit
  Open file: DONE
  b3bdffff
  Parse file: DONE
  bitstream header infos
  Architecture: sa5p00
  Bits: 4476704
  Bitstream CRC: 0x6C7C
  Bitstream Status: Final Version 10.27
  Cols: 592
  Date: Tue Mar 3 17:43:40 2026
  Design name: tg68test_impl1.ncd
  Part: LFE5U-25F-6CABGA256
  Readback: Off
  Rows: 7562
  Security: Off
  Version: Diamond (64-bit) 3.11.0.396.4
  idcode: 41111043
  IDCode : 41111043
  displayReadReg
  Config Target Selection : 0
  BSE Error Code
  Abort ERR
  Enable configuration: DONE
  SRAM erase: DONE
  Loading: [==================================================] 100.00%
  Fail
  displayReadReg
  Config Target Selection : 0
  JTAG Active
  ISC Enable
  Write Enable
  Read Enable
  BSE Error Code
  Abort ERR
  Error: Failed to program FPGA: std::exception

I have an icepi-zero and a ULX3S which I can test with, too, if it will help.

[trabucayre]

Thanks.
I will check why this fails.
With a classic JTAG interface loading and writing bitstream is perfectly fine. Before pushing this commit I have performed tests with multiple boards via this link, but missed protocol like CMSIS-DAP.

[trabucayre]

I have tested with an colorlight i5 (cmsis-DAP interface too) without issue: load and flash are working.
But I have used trellis toolchain instead of diamond: maybe something wrong here.

[robinsonb5]

Thanks for giving this your attention - I'll try a trellis bitstream on IceSugarPro, and a Diamond bitstream on one of the other boards and report back.

[trabucayre]

I have produces a bitstream for colorlight i5 with diamond. The configuration data after parsing contains exactly the same header sequence: 0xff 0xff 0xbd 0xb3 ....
And with bitstream produces by diamond I have the same issue...
But If I can't find why this fails after the mentioned commit I think I will revert it to avoid this issue.

[trabucayre]

Configuration data before and after differs only by one Dummy Byte but I have reverted previous commit with 0bc2594

Needs to check CMSIS-DAP maybe something is wrong/lost during process

[robinsonb5]

Interesting - the plot thickens...

I've just tested (SRAM, not flash, and before you reverted the suspect commit) with a Trellis bitstream and a Diamond bitstream on both IceSugarPro and IcePi-Zero, and also a Trellis bitstream on ulx3s (85k variant) - and I'm getting the same error with every combination, so whatever it is, it's not specific to CMSIS-DAP.

I've attached a few bitstreams, just in case there's something different about the ones I'm generating compared with yours.

I'm on Linux, by the way - the behaviour is the same on two different versions of Linux Mint, one based on Ubuntu 20.04, the other on 22.04.

ecp5_bitstreams.zip

[trabucayre]

With an up to data repository including last commit?

[robinsonb5]

Those tests were performed before your most recent commit which reverts the change. (We posted those comments within seconds of each other!)

With the most recent commit applied the problem goes away, as expected.

As I say, the surprise is that on my systems the problem seemed to affect all ECP5-based boards, not just the icesugarpro with its CMSIS-DAP interface.

[trabucayre]

It's really weird. Before pushing this commit I have tested using more than one distribution (debian, gentoo, ...) and macOS without having issue.
I must admit I have to real idea why this fails.
I think I have to create a minor release to have a stable tag with the fix applied.

[robinsonb5]

Very strange - not to mention intriguing!

Is there something different about my bitstreams compared to yours? Can you reproduce the issue using the bitstreams I attached?