[On-prem U250] Host reboots after `firesim runworkload`. 1.18.0 regression?

[caizixian]

[On-prem U250] Host reboots after `firesim runworkload`. 1.18.0 regression?

[caizixian]

1.17.1 works fine.

[caizixian]

Possibly related to #1692

[RealJustinNi]

Solved like #1695

[caizixian]

@RealJustinNi thanks for the link.

In my case, I'm following the getting started guide, and didn't elaborate any design myself. The bitstream flashed is downloaded by FireSim

firesim/deploy/sample-backup-configs/sample_config_hwdb.yaml

Line 66 in 535dcdc

alveo_u250_firesim_rocket_singlecore_no_nic:

(1.18.0). The memory configuration file is from the same tarball. So I didn't think that reprogramming the memory device is necessary.

Regardless of the above, this still seems to be a regression where the same steps work find on a fresh 1.17.1 checkout.

[RealJustinNi]

@caizixian Hello, we have also recently encountered similar issues when running Firesim 1.18.0 and 1.17.1 workloads on the U200 platform about one week ago. Our bitstream is alveo_u200_firesim_rocket_singlecore_no_nic. Interestingly, under version 1.17.1, both firesim infrasetup and firesim runworkload run correctly, and Linux boots up without any problems. However, after successfully executing firesim infrasetup in version 1.18.0, runworkload results in a host freeze issue.

We ultimately resolved the issue by recompiling a buildstream and then proceeding with the FPGA re-programming. After looking at another issue you mentioned, we discovered that it was indeed an issue with the version pointer. Thank you very much!