Shim 16.1-8 hang after BIOS update on Dell / Alienware systems when chainloading rEFInd

[olifre]

After updating the system BIOS on
Dell / Alienware 18 Area-51 AA18250
from version 2.4.1 to 2.5.0, Shim 16.1-8 gets stuck somewhere during initialization.
Shim 15.8 continues to boot fine on the same system. Note that the system is also affected by #803 (both old and new BIOS version).

I enabled verbosity via:

mokutil --set-verbosity true

which is extremely slow during scrolling on this system after the first screen is filled. Sadly, the system turns off hard during the lengthy dumping after importing mok state for db / mirroring db (first dump of db works, it turns off during the second dump). I am unsure whether that indicates some timeout or memory corruption.

15.8 also has slow output on the system, but the output is less lengthy and the system manages to boot fine even in this verbose mode.

Ideas on how to debug this further are naturally appreciated.

[olifre]

Further info:

• Presumably, the hang during debug output is caused by the problem addressed in Periodically update the UEFI watchdog timer when we're printing to the console... #807
• The full hang is observed using rEFInd as loader, as built against gnu-efi 3. Using a rEFInd built against gnu-efi 4 (patched version by ArchLinux) lets the loader load fine, but as soon as it loads a kernel (again asking shim to verify it), I observe more or less erratic behaviour, in three attempts:
  • system freeze, disk spun down
  • normal boot
  • system freeze with loud fans, blink pattern of power LED indicates catastrophic chipset failure (8 blue, 1 red)

This smells like some memory allocation issue.

[olifre]

I can reproduce similar issues now on a:
Dell Pro 14 Premium PA14250
after UEFI/BIOS updates (from 2.13.4 to 2.14.4).

That machine is now also affected by #803 , but only when chainloading refind from shim .
Chainloading a more lightweight loader, i.e. systemd-boot or just an UKI, I do not get this message.

Chainloading refind, though, refind gets fully stuck when iterating the loaders.

Potentially, those issues are only triggered if the EFI application is larger / does more allocations?