Files
mlkem-sync/sync_rtl/sha3/sha3_top.v
FallenSigh 106b2925a8 feat(sha3): multi-block SHA3-256 absorb for H(ek); KeyGen golden vectors
Stage 0+1 of mlkem_top KeyGen integration:
- sha3_top: add multi-block SHA3-256 absorb FSM (mb_en/mb_block_i/mb_valid_i/
  mb_last_i/mb_ready_o). Caller pre-pads final block; module does pure absorb
  loop (state^=block; Keccak-p). Single-block G/H/J paths bit-identical when
  mb_en=0. Sticky digest register holds output until consumer acks.
- tb_sha3_mb_xsim: self-checking TB streams 800B ek (6 blocks) -> H(ek),
  verified == hashlib.sha3_256. Proper valid/ready handshake (no force).
- Existing G/H/J TBs (xsim + Verilator) tie off mb_* ports; both frameworks
  regress clean (Verilator 25/25, XSIM G/H/J + keccak + 7-vec + multiblock).
- test_framework/modules/mlkem_keygen/golden: full 256-coeff per-stage
  intermediates (rho/sigma, A_hat, s/e, s_hat/e_hat, t_hat, ek, dk_pke) for
  KAT count=0..4, dumped by ml-kem-r and self-verified against NIST KAT.
2026-06-27 23:37:23 +08:00

216 lines
8.7 KiB
Verilog

// sha3_top.v - SHA3/SHAKE top wrapper with valid/ready interface
//
// Implements SHA3-512 (G), SHA3-256 (H), SHAKE-256 (J) over a single
// Keccak-f[1600] core. Supports single-block absorption, plus a separate
// multi-block SHA3-256 streaming absorb path for long inputs (e.g. H(ek)).
//
// Single-block modes (mb_en=0):
// 00 = G (SHA3-512): rate=576, suffix=01, msg_len=264, out=512
// 01 = H (SHA3-256): rate=1088, suffix=01, msg_len=256, out=256
// 10 = J (SHAKE-256): rate=1088, suffix=1111,msg_len=512,out=256
//
// Multi-block SHA3-256 (mb_en=1):
// Streams pre-padded 1088-bit rate blocks. The CALLER applies SHA3-256
// padding (suffix 0x06 ... 0x80) to the final block so this module only
// does the absorb loop: state ^= block; state = Keccak-p(state); repeat.
// The single-block paths above are bit-identical when mb_en=0.
//
// Interface (single-block):
// mode[1:0] - 00=G, 01=H, 10=J
// data_i - 512-bit message input
// valid_i / ready_o / hash_o[511:0] / valid_o / ready_i
//
// Interface (multi-block SHA3-256, active when mb_en=1):
// mb_en - 1 selects the multi-block absorb path
// mb_block_i - one pre-padded 1088-bit (136-byte) rate block, byte 0 in [7:0]
// mb_valid_i - block valid
// mb_ready_o - module can accept a block
// mb_last_i - asserted with the final (already-padded) block
// result reuses hash_o[255:0] / valid_o / ready_i
module sha3_top (
input clk,
input rst_n,
input [1:0] mode,
input [511:0] data_i,
input valid_i,
output ready_o,
output [511:0] hash_o,
output valid_o,
input ready_i,
// --- multi-block SHA3-256 absorb (tie mb_en=0 to disable) ---
input mb_en,
input [1087:0] mb_block_i,
input mb_valid_i,
input mb_last_i,
output mb_ready_o
);
// ================================================================
// FSM state encoding
// ================================================================
localparam ST_IDLE = 2'd0;
localparam ST_PERMUTE = 2'd1;
localparam ST_SQUEEZE = 2'd2;
reg [1:0] state_r, state_next;
// ================================================================
// Absorb state: message || suffix || pad10*1 into rate bits
// Built combinationally from data_i and mode.
//
// G: padded_block = {1, 308'b0, 1, 2'b01, data_i[263:0]}
// absorb_state = {1024'b0, padded_block_576}
//
// H: padded_block = {1, 828'b0, 1, 2'b01, data_i[255:0]}
// absorb_state = {512'b0, padded_block_1088}
//
// J: padded_block = {1, 570'b0, 1, 4'b1111, data_i[511:0]}
// absorb_state = {512'b0, padded_block_1088}
// ================================================================
wire [575:0] g_pad;
wire [1087:0] h_pad;
wire [1087:0] j_pad;
assign g_pad = {1'b1, {308{1'b0}}, 1'b1, 2'b10, data_i[263:0]};
assign h_pad = {1'b1, {828{1'b0}}, 1'b1, 2'b10, data_i[255:0]};
// J: SHAKE suffix is "1111" — all ones, order irrelevant
assign j_pad = {1'b1, {570{1'b0}}, 1'b1, 4'b1111, data_i[511:0]};
wire [1599:0] absorb_state;
assign absorb_state = (mode == 2'b00) ? {{(1600-576){1'b0}}, g_pad} :
(mode == 2'b01) ? {{(1600-1088){1'b0}}, h_pad} :
(mode == 2'b10) ? {{(1600-1088){1'b0}}, j_pad} :
1600'd0;
// ================================================================
// Multi-block SHA3-256 absorb FSM (active only when mb_en=1)
//
// Running sponge state mb_state_r (init 0). For each pre-padded
// 1088-bit rate block: mb_state_r ^= block; mb_state_r = Keccak-p(...).
// After the last block, squeeze 256 bits. The caller pads the final
// block (SHA3-256 suffix 0x06 ... 0x80), so this FSM is pure absorb.
// ================================================================
localparam MB_IDLE = 2'd0; // ready for a block (or first block)
localparam MB_PERMUTE = 2'd1; // keccak running on xored state
localparam MB_DONE = 2'd2; // squeeze: present 256-bit digest
reg [1:0] mb_state, mb_state_next;
reg [1599:0] mb_state_r; // running sponge state
reg mb_last_r; // captured last-block flag
reg [255:0] mb_digest_r; // latched 256-bit digest (sticky in MB_DONE)
// XOR the incoming block into the low 1088 bits (rate) of the state.
wire [1599:0] mb_xored;
assign mb_xored = mb_state_r ^ {{(1600-1088){1'b0}}, mb_block_i};
// Accept a block only in MB_IDLE while enabled.
assign mb_ready_o = mb_en && (mb_state == MB_IDLE);
wire mb_accept = mb_en && (mb_state == MB_IDLE) && mb_valid_i;
wire mb_kc_valid = mb_accept; // start keccak on the accept cycle
always @(*) begin
mb_state_next = mb_state;
case (mb_state)
MB_IDLE: if (mb_accept) mb_state_next = MB_PERMUTE;
MB_PERMUTE: if (kc_valid_o) mb_state_next = mb_last_r ? MB_DONE : MB_IDLE;
MB_DONE: if (ready_i) mb_state_next = MB_IDLE;
default: mb_state_next = MB_IDLE;
endcase
end
// ================================================================
// Keccak core
// ================================================================
wire kc_valid_i;
/* verilator lint_off UNUSEDSIGNAL */
wire [1599:0] kc_state_o;
wire kc_ready_o;
/* verilator lint_on UNUSEDSIGNAL */
wire kc_valid_o;
// Keccak input: multi-block xored state when mb_en, else single-block absorb.
wire [1599:0] kc_state_i_mux;
assign kc_state_i_mux = mb_en ? mb_xored : absorb_state;
keccak_core #(.ROUNDS(24)) u_keccak (
.clk (clk),
.rst_n (rst_n),
.state_i (kc_state_i_mux),
.valid_i (kc_valid_i),
.ready_o (kc_ready_o), // unused but must connect
.state_o (kc_state_o),
.valid_o (kc_valid_o),
.ready_i (1'b1) // always accept output
);
// ================================================================
// FSM combinational logic
// ================================================================
assign ready_o = !mb_en && (state_r == ST_IDLE);
// kc_valid_i: single-block start (state_next==PERMUTE) OR multi-block accept.
assign kc_valid_i = mb_en ? mb_kc_valid : (state_next == ST_PERMUTE);
always @(*) begin
state_next = state_r;
case (state_r)
ST_IDLE: if (valid_i && ready_o) state_next = ST_PERMUTE;
ST_PERMUTE: if (kc_valid_o) state_next = ST_SQUEEZE;
ST_SQUEEZE: if (valid_o && ready_i) state_next = ST_IDLE;
default: state_next = ST_IDLE;
endcase
end
// ================================================================
// Output
// ================================================================
// Register for squeezed output (only 512 bits needed)
reg [511:0] squeezed_state_r;
// valid_o / hash_o serve both paths, selected by mb_en.
assign valid_o = mb_en ? (mb_state == MB_DONE) : (state_r == ST_SQUEEZE);
assign hash_o = mb_en ? {256'b0, mb_digest_r} : squeezed_state_r;
// ================================================================
// Sequential logic
// ================================================================
always @(posedge clk or negedge rst_n) begin
if (!rst_n) begin
state_r <= ST_IDLE;
squeezed_state_r <= 512'd0;
mb_state <= MB_IDLE;
mb_state_r <= 1600'd0;
mb_last_r <= 1'b0;
mb_digest_r <= 256'd0;
end else begin
state_r <= state_next;
mb_state <= mb_state_next;
// --- single-block: latch squeezed output ---
if (state_r == ST_PERMUTE && kc_valid_o) begin
squeezed_state_r <= kc_state_o[511:0];
end
// --- multi-block: capture last flag on accept ---
if (mb_accept) begin
mb_last_r <= mb_last_i;
end
// --- multi-block: latch permuted state when keccak finishes ---
if (mb_state == MB_PERMUTE && kc_valid_o) begin
mb_state_r <= kc_state_o;
// On the final block, latch the 256-bit digest (sticky for MB_DONE).
if (mb_last_r) mb_digest_r <= kc_state_o[255:0];
end
// --- multi-block: clear running state after digest consumed ---
if (mb_state == MB_DONE && ready_i) begin
mb_state_r <= 1600'd0;
end
end
end
endmodule