CSR Bit Fields

RISC-V pmpaddr16 CSR Register

Address 0x3C0Privilege MachineAccess RW / MXLEN: RV32 encodes PA[33:2]; RV64 [53:0] encodes PA[55:2] and [63:54] is zeroMachine physical memory protection CSRs

pmpaddr16 (0x3C0) is the WARL address register for machine PMP entry 16 when 64 PMP entries are implemented; it encodes physical address bits 33:2 on RV32 and 55:2 on RV64.

Field Map

Understand pmpaddr16 By Bit Fields

2 key fields
63:54

0

WARL/reads zero

RV64: the upper ten bits are zero in the standard pmpaddr format; this segment is part of the WARL format.

0 (bits 63:54) — RV64: the upper ten bits are zero in the standard pmpaddr format; this segment is part of the WARL format.

What This Field Controls

  • - RV64: the upper ten bits are zero in the standard pmpaddr format; this segment is part of the WARL format.

Common Values

This field is better understood together with surrounding context than as a fixed memorized enumeration.

Open Official Manual
53:0

address[55:2]

WARL/may be RO0

RV64: encodes bits 55:2 of a 56-bit physical address. Unimplemented physical-address bits make the field WARL; every PMP CSR field may be implemented as read-only zero.

address[55:2] (bits 53:0) — RV64: encodes bits 55:2 of a 56-bit physical address. Unimplemented physical-address bits make the field WARL; every PMP CSR field may be implemented as read-only zero.

What This Field Controls

  • - RV64: encodes bits 55:2 of a 56-bit physical address. Unimplemented physical-address bits make the field WARL; every PMP CSR field may be implemented as read-only zero.

Common Values

This field is better understood together with surrounding context than as a fixed memorized enumeration.

Open Official Manual
Official Basis & Search Notes

pmpaddr16 is the MXLEN-bit WARL address encoding for entry 16 when 64 PMP entries are implemented. pmp16cfg.A determines its meaning; TOR uses pmpaddr15 as the lower bound.

RV32 pmpaddr encodes PA[33:2]; RV64 [53:0] encodes PA[55:2], with [63:54] zero.
TOR entry 16 uses pmpaddr15 as its lower bound and pmpaddr16 as its upper bound; NAPOT uses low pmpaddr16 bits to encode a naturally aligned power-of-two region of at least 8 bytes.
Platform grain changes low-bit readback; pmp16cfg.L or a locked TOR entry 17 can cause writes to pmpaddr16 to be ignored. Smepmp MML=1 changes L's permission meaning, but the relevant L=1 rules remain locked unless RLB=1 temporarily bypasses locking.

What To Check First When Reading This CSR

  • - PMP is optional; implementations may provide 0, 16, or 64 entries, with the lowest-numbered entries implemented first. Therefore, pmpaddr16 exists only when 64 entries are implemented; every PMP CSR field is WARL and may be read-only zero.
  • - pmpaddr16 is accessible only to M-mode. Its readback depends on pmp16cfg.A and the platform-wide PMP grain G.
  • - With G>=2 and A=NAPOT, pmpaddr16[G-2:0] reads as ones; with G>=1 and A=OFF/TOR, pmpaddr16[G-1:0] reads as zeros, but those bits do not affect TOR matching. Changing pmp16cfg.A[1] changes readback, not the underlying stored value.
  • - Reset specifies defaults only for writable PMP A/L fields (subject to platform overrides); pmpaddr16's reset value is unspecified, but no WARL field contains an illegal value.

Risk Checks Before Writing

  • - Writes to pmpaddr16 are ignored when pmp16cfg.L=1; pmpaddr16 is also unwritable as entry 17's lower bound when that entry is locked with pmp17cfg.A=TOR. Without Smepmp, locking persists until hart reset. With Smepmp MML=1, L changes permission meaning, but the L=1 M-mode-only or locked Shared-Region rules it defines remain locked unless mseccfg.RLB=1 temporarily bypasses locking.
  • - pmpaddr16 is WARL; read it back after writing and interpret the masked value using the current pmp16cfg.A and grain.
  • - With pmp16cfg.A=TOR, entry 16 matches pmpaddr15 <= y < pmpaddr16; if the lower bound is not less than the upper bound, it matches no addresses. NA4 is not selectable when G>=1.

Put It Back Into A Real Flow

1

Read pmp16cfg.A to determine whether entry 16 uses OFF, TOR, NA4, or NAPOT.

2

Write pmpaddr16 for the selected mode; TOR also requires lower bound pmpaddr15, while NAPOT encodes range size in low bits.

3

Read back pmpaddr16 and the corresponding pmpcfg CSR to confirm the effective WARL/grain encoding; set pmp16cfg.L only after final verification.

FAQ

Where do pmpaddr16's TOR bounds come from?

With pmp16cfg.A=TOR, pmpaddr15 supplies the lower bound and pmpaddr16 the upper bound, so entry 16 matches the half-open range pmpaddr15 <= y < pmpaddr16; it matches no addresses when the lower bound is not less than the upper bound.

Why can pmpaddr16 read back differently from the written value?

pmpaddr16 is WARL, and platform grain plus pmp16cfg.A[1] masks low-bit readback: NAPOT can read ones while OFF/TOR reads zeros. Changing A[1] does not alter the underlying stored value.