Reviewing use of `currentlyEnabled` vs `hartSupports`

[pmundkur]

• First, some currentlyEnabled clauses seem inaccurate: e.g., currentlEnabled(Ext_Sv39) doesn't capture whether satp[Mode] is actually set to Sv39.
• Secondly, legalization of CSRs like menvcfg should be checking hartsupports instead of currentlyEnabled

When making this change, the use of currentlyEnabled should be carefully reviewed to check downstream consequences.

[pmundkur]

See #1626 (review)

[KotorinMinami]

Perhaps I'm missing something, but why?

currentlyEnabled(Ext_Sv39) doesn't capture whether satp[Mode] is actually set to Sv39

the currentlyEnabled(Ext_Sv39) actually guard whether satp[Mode] is actually set to Sv39 in legalize_satp I think

    match satpMode_of_bits(arch, s[Mode]) {
      None()  => prev_value,
      Some(Sv_mode) =>  match Sv_mode {
        Bare if currentlyEnabled(Ext_Svbare) => s.bits,
        Sv39 if currentlyEnabled(Ext_Sv39) => s.bits,
        Sv48 if currentlyEnabled(Ext_Sv48) => s.bits,
        Sv57 if currentlyEnabled(Ext_Sv57) => s.bits,
        _ => prev_value,
      }

[pmundkur]

Good catch. We need to properly define what currentlyEnabled means for each extension. Taken literally, it is currently misleading.