Can "mcontrol execute address before" trigger ever fire on an ECALL instruction?

[Silabs-ArjanB]

Hi,

The "Version 1.0, Revised 2025-02-21: Ratified" version of the RISC-V Debug Specification states the following about the mcontrol timing field:

0 (before): The action for this trigger will be taken just before the instruction that triggered it is retired, but after all preceding instructions are retired. xepc or dpc (depending on action) must be set to the virtual address of the instruction that matched.

Certain instructions (e.g., ECALL) can never retire as per definition in the RISC-V Privileged specification. Therefore, if we take the above literally, an "mcontrol execute address before" trigger that matches the execute address of an ECALL instruction will never fire.

Is that really the intention of the above definition of before timing? Should it maybe be reworded to something like "just before the instruction that triggered it is retired or traps?

Best regards,
Arjan

[pdonahue-ventana]

The Priority section lists mcontrol/mcontrol6 execute address before as higher priority than environment call. That table clearly conveys the intention whereas the prose that you quote may have received a little less scrutiny with respect to cases like the one you bring up. I'd have to spend some time to consider whether "retired or traps" is 100% correct with respect to all cases: fetch-related traps, various decode-related traps like ecall, load-related traps on load/store address/data before, etc.

This can stay open until we're ready to make a batch of clarifications to the spec. In the meantime, rely on the priority table.

[Silabs-ArjanB]

Hi @pdonahue-ventana , thank you for your quick answer. You can close the ticket when you want. Best regards, Arjan