Write zero to tdata1 for mcontrol6 trigger

[Qin0803]

In RV debug spec, it describes that,

Writing 0 to this register must result in a trigger that is disabled. If this trigger supports multiple types, then the hardware should disable it by changing type to 15

In specific implementation, for an address trigger of mcontrol6, if zero is written to the trigger's tdata1 register, whether to allow tdata1.type to be written to 15 , indicates that the trigger is disabled, and other fields in tdata1 remain original values except for the tdata1.type field?

[Qin0803]

Hi, @pdonahue-ventana , can you help answer the question above? Thanks

[pdonahue-ventana]

It's not clear to me if there are any legal/WARL requirements for the other fields when tdata1.type=15 other than the fact that hardware ignores the other fields. That format is not defined in the same way that mcontrol6, itrigger, icount, etc. are defined.

I think that most implementations have triggers with hardwired types (e.g. tselect=0 has hardwired type=6, tselect=1 has hardwired type=3, tselect=2 has hardwired type=4, etc.). The existence of type=15 shows that there was some thought about multi-type triggers but it might not be as developed as everything else.

If I were implementing a TM then I would make tdata1[XLEN-6:0]=0 for type=15 just to be safe. However, an argument can be made that the other fields can be implementation-defined. Hardware is required to ignore those bits so it's not likely to matter.