[qemu] Earlgrey 1.0.0 `lc_ctrl` updates

[moidx]

Description

Update lc_ctrl implementation to support:

• JTAG support
• Support lifecycle transitions via JTAG and software

[luismarques]

Life cycle transitions triggered through software are working. JTAG supported needs to be ported.

[luismarques]

Life cycle transitions triggered through software are working. JTAG supported needs to be ported.

That comment wasn't correct. At the time, the transitions were only working due to a workaround where the LC tokens weren't being properly checked. Since then, we implemented the OTP secret partition descrambling (and fixed the PRESENT descrambling algorithm in QEMU), so this is now really working without a workaround.

I have a hacky implementation of JTAG support for the LC Ctrl, using the chardev originally intended for the RV_DM TAP, and instantiating the DTM and TL-UL debug bridge at the top level (like Darjeeling does). With it, I was able to run a manufacturing test that performs a complete life cycle transition over JTAG. I will convert that implementation into the final proper one, where those are blocks are instantiated inside the LC Ctrl block, like it's done in Earlgrey.

[AlexJones0]

While working on ROM_EXT boot service functionality, one issue encountered with the lc_ctrl was that the HW_CFG was being loaded from OTP incorrectly: lowRISC/qemu#214. It also turned out that Earlgrey was disabling Ibex's lc_ctrl integration by default: lowRISC/qemu#273.

We also debugged an issue where disconnecting from the LC_CTRL (or any) JTAG would quit QEMU, killing the VM completely, caused by a default property on the QEMU TAP Ctrl interface. THis was fixed in #28601.

[luismarques]

The QEMU LC Ctrl JTAG implementation PR is going through the last round of reviews and should be merged shortly. The required OTT changes to support that JTAG connection have also been published. With those combined changes, manufacturing tests doing lifecycle transitions still pass (compared with the original, hacky implementations).