Equations for triggering an interrupt are not clear when an IOPMP instance hits no entry or multiple non-priority entries

[tyshyu]

Hi all,

The latest commits (3369da4 7a76c5d) add some equations for describing how to trigger an interrupt in Section 2.7 (chapter2.adoc):

interrupt_on_read = ERR_CFG.ie && !ENTRY_CFG.sire
interrupt_on_write = ERR_CFG.ie && !ENTRY_CFG.siwe
interrupt_on_execution = ERR_CFG.ie && !ENTRY_CFG.sixe

The aforementioned equations are not clear if an IOPMP instance has no hit and multiple hits on non-priority entries.

For a read violation, would it be like the following equation?

interrupt_on_read =
ERR_CFG.ie && (no_hit || 
  ! (
          (entry_hit[0] && ENTRY_CFG(0).sire)
      ||  (entry_hit[1] && ENTRY_CFG(1).sire)
      ||  (entry_hit[2] && ENTRY_CFG(2).sire)
      ...
      ||  (entry_hit[N-1] && ENTRY_CFG(N-1).sire)
    )
)

no_hit means an IOPMP instance hits no entry. entry_hit[i] means an IOPMP instance hits or partially hits entry i. N means number of implemented IOPMP entries.

The equation means any entry is hit and its sire bit is 1, the interrupt will be suppressed.

[channingt]

For non-prioritized entries:

1. if an entry is hit, then no interrupt generated;
2. if no entry is hit, then generate interrupt.

My understanding is, partial hit is considered the entry no hit. @paul-andes , what's your thought? We can discuss on the meeting.

[paul-andes]

A partial hit of an entry is not considered a hit to the entry.

For more details on non-priority entries, 1) if a transaction hits an entry that grants enough permission, the access succeeds. 2) if the transaction hits an entry that doesn't grant enough permission but suppresses the corresponding interrupt/bus error, the access fails but NO interrupt/bus-error is generated. 3) if no transaction grants enough permission, it is considered a "no hit."

[paul-andes]

The discussion will keep on in the following link: https://lists.riscv.org/g/tech-iopmp/message/293