Implement progbuf and attempt to halt/resume using existing trap logic (very broken)

bin/hw_debug_test.py

@@ -38,7 +38,13 @@ random_order = False
 
 def prog_buff_test(cvw):
     cvw.halt()
-    cvw.write_dmi("0x20", "0x7ee7beef")
+    cvw.write_dmi("0x20", "0x00100073")
+    #cvw.write_dmi("0x21", "0xf00daedd")
+    #cvw.write_dmi("0x22", "0xdaede105")
+    #cvw.write_dmi("0x23", "0x00100073") # ebreak
+
+    cvw.write_data("DPC", "0x00002000") # Progbuf addr
+    cvw.resume()
     print()
 
 
@@ -46,10 +52,13 @@ def flow_control_test(cvw):
     #time.sleep(70)  # wait for OpenSBI
 
     cvw.halt()
-    cvw.read_data("DCSR")
+    time.sleep(1)
+    #cvw.read_data("DCSR")
     for _ in range(50):
-        cvw.step()
+        cvw.resume()
-        print(cvw.read_data("PCM"))
+        cvw.halt()
+        #cvw.step()
+        #print(cvw.read_data("PCM"))
     cvw.resume()
 
 
@@ -154,11 +163,11 @@ def random_hex(reg_name):
 
 
 with OpenOCD() as cvw:
-    print(cvw.read_dmi("0x16"))
+    #cvw.trst()
-    cvw.trst()
     cvw.reset_dm()
+    time.sleep(1)
     cvw.reset_hart()
-    quit()
+    time.sleep(1)
     #register_rw_test(cvw)
-    flow_control_test(cvw)
+    #flow_control_test(cvw)
-    #prog_buff_test(cvw)
+    prog_buff_test(cvw)

bin/openocd_tcl_wrapper.py

@@ -114,6 +114,12 @@ class OpenOCD:
             raise Exception("Error: Hart failed to reset")
         self.write_dmi("0x10", "0x10000001")  # ack HaveReset
 
+    def write_progbuf(self, data):
+        #TODO query progbuf size and error is len(data) is greater
+        baseaddr = 0x20
+        for idx, instr in enumerate(data):
+            self.write_dmi(hex(baseaddr+idx), instr)
+
     def set_haltonreset(self):
         self.write_dmi("0x10", "0x9")
 
@@ -139,7 +145,7 @@ class OpenOCD:
             dcsr |= 0x4
             self.write_data("DCSR", hex(dcsr))
         # Resume once
-        self.resume()
+        self.write_dmi("0x10", "0x40000001")
         # Unset step bit
         dcsr &= ~0x4
         self.write_data("DCSR", hex(dcsr))

config/rv64gc/config.vh

@@ -186,6 +186,9 @@ localparam logic [63:0] SDC_RANGE        = 64'h0000007F;
 localparam logic SPI_SUPPORTED = 1;
 localparam logic [63:0] SPI_BASE         = 64'h10040000;
 localparam logic [63:0] SPI_RANGE        = 64'h00000FFF;
+// Debug program buffer support is enabled with DEBUG_SUPPORTED
+localparam logic [63:0] PROGBUF_BASE     = 64'h00002000;
+localparam logic [63:0] PROGBUF_RANGE    = 64'h0000000F;
 
 // Bus Interface width
 localparam AHBW = (XLEN);

config/shared/debug.vh

@@ -48,6 +48,21 @@
 `define ABSTRACTAUTO `ADDR_WIDTH'h18
 `define NEXTDM       `ADDR_WIDTH'h1d
 `define PROGBUF0     `ADDR_WIDTH'h20
+`define PROGBUF1     `ADDR_WIDTH'h21
+`define PROGBUF2     `ADDR_WIDTH'h22
+`define PROGBUF3     `ADDR_WIDTH'h23
+`define PROGBUF4     `ADDR_WIDTH'h24
+`define PROGBUF5     `ADDR_WIDTH'h25
+`define PROGBUF6     `ADDR_WIDTH'h26
+`define PROGBUF7     `ADDR_WIDTH'h27
+`define PROGBUF8     `ADDR_WIDTH'h28
+`define PROGBUF9     `ADDR_WIDTH'h29
+`define PROGBUFA     `ADDR_WIDTH'h2A
+`define PROGBUFB     `ADDR_WIDTH'h2B
+`define PROGBUFC     `ADDR_WIDTH'h2C
+`define PROGBUFD     `ADDR_WIDTH'h2D
+`define PROGBUFE     `ADDR_WIDTH'h2E
+`define PROGBUFF     `ADDR_WIDTH'h2F
 //`define dmcs2        `ADDR_WIDTH'h32
 `define SBCS         `ADDR_WIDTH'h38
 

config/shared/parameter-defs.vh

@@ -85,6 +85,8 @@ localparam cvw_t P = '{
   SPI_SUPPORTED :        SPI_SUPPORTED,
   SPI_BASE :        SPI_BASE,
   SPI_RANGE :        SPI_RANGE,
+  PROGBUF_BASE : PROGBUF_BASE,
+  PROGBUF_RANGE : PROGBUF_RANGE,
   GPIO_LOOPBACK_TEST :        GPIO_LOOPBACK_TEST,
   SPI_LOOPBACK_TEST :        SPI_LOOPBACK_TEST,
   UART_PRESCALE :        UART_PRESCALE ,

src/cvw.sv

@@ -140,6 +140,9 @@ typedef struct packed {
   logic         SPI_SUPPORTED;
   logic [63:0]  SPI_BASE;
   logic [63:0]  SPI_RANGE;
+  // Debug program buffer support is enabled with DEBUG_SUPPORTED
+  logic [63:0]  PROGBUF_BASE;
+  logic [63:0]  PROGBUF_RANGE;
 
   // Test modes
 

src/debug/dm.sv

@@ -74,12 +74,14 @@ module dm import cvw::*; #(parameter cvw_t P) (
   output logic            DebugRegUpdate, // writes values from scan register after scanning in			
   
   // Program Buffer
+  output logic [3:0]      ProgBufAddr,
   output logic            ProgBuffScanEn,
-  output logic            ProgBuffScanOut,
   output logic            ExecProgBuff
 );
   `include "debug.vh"
 
+  localparam PROGBUF_SIZE = (P.PROGBUF_RANGE+1)/4;
+
   // DMI Signals
   logic                   ReqReady;
   logic                   ReqValid;
@@ -93,7 +95,7 @@ module dm import cvw::*; #(parameter cvw_t P) (
 
   // JTAG ID for Wally:  
   // Version [31:28] = 0x1 : 0001
-  // PartNumber [27:12] = 0x2A : 00000000_00101010
+  // PartNumber [27:12] = 0x2A : Wally (00000000_00101010)
   // JEDEC number [11:1] = 0x602 : Bank 13 (1100) Open HW Group (0000010) 
   // [0] = 1
   localparam JTAG_DEVICE_ID = 32'h1002AC05; 
@@ -151,7 +153,7 @@ module dm import cvw::*; #(parameter cvw_t P) (
   // DMStatus
   const logic        NdmResetPending = 0;
   const logic        StickyUnavail = 0;
-  const logic        ImpEBreak = 1;
+  const logic        ImpEBreak = 0;
   logic              AllHaveReset;
   logic              AnyHaveReset;
   logic              AllResumeAck;
@@ -170,7 +172,7 @@ module dm import cvw::*; #(parameter cvw_t P) (
   const logic        ConfStrPtrValid = 0; // Used with SysBusAccess
   const logic [3:0]  Version = 3;    // DM Version
   // AbstractCS
-  const logic [4:0]  ProgBufSize = 1;
+  const logic [4:0]  ProgBufSize = PROGBUF_SIZE;
   logic              Busy;
   const logic        RelaxedPriv = 1;
   logic [2:0]        CmdErr;
@@ -248,12 +250,12 @@ module dm import cvw::*; #(parameter cvw_t P) (
               {`OP_READ,`COMMAND}                           : State <= READ_ZERO;
               {`OP_WRITE,`SBCS}                             : State <= READ_ZERO;
               {`OP_READ,`SBCS}                              : State <= R_SYSBUSCS;
-              {`OP_WRITE,`PROGBUF0}                    : State <= W_PROGBUF;
+              [{`OP_WRITE,`PROGBUF0}:{`OP_WRITE,`PROGBUF3}] : State <= W_PROGBUF; // TODO: update decode range dynamically using PROGBUF_RANGE
-              {`OP_READ,`PROGBUF0},
+              [{`OP_READ,`PROGBUF0}:{`OP_READ,`PROGBUFF}],
               {2'bx,`HARTINFO},
               {2'bx,`ABSTRACTAUTO},
               {2'bx,`NEXTDM}                                : State <= READ_ZERO;
-              default                                  : State <= READ_ZERO;//INVALID;
+              default                                       : State <= INVALID;
             endcase
         end
 
@@ -369,8 +371,10 @@ module dm import cvw::*; #(parameter cvw_t P) (
         W_PROGBUF : begin
           if (Busy)
             CmdErr <= ~|CmdErr ? `CMDERR_BUSY : CmdErr;
-          else
+          else begin
             NewAcState <= PROGBUFF_WRITE;
+            ProgBufAddr <= ReqAddress[$clog2(PROGBUF_SIZE)-1:0];
+          end
           RspOP <= `OP_SUCCESS;
           State <= ACK;
         end
@@ -441,7 +445,6 @@ module dm import cvw::*; #(parameter cvw_t P) (
 
   // Program Buffer
   assign ProgBuffScanEn = (AcState == PROGBUFF_WRITE);
-  assign ProgBuffScanOut = ScanReg[0];
 
   // Scan Chain
   assign DebugScanOut = ScanReg[0];

src/debug/dmc.sv

@@ -27,12 +27,28 @@
 // and limitations under the License.
 ////////////////////////////////////////////////////////////////////////////////////////////////
 
-// Note: This module controls all of the per-hart debug state.
+// On HaltReq/eBreak:
-// In a multihart system, this module should be instantiated under wallypipelinedcore
+// store value of NextPC in DPC
+// trigger trap (flush pipe)
+// stall pipe
+
+// On Step:
+// unstall pipe until instruction receaches M stage
+// goto: HaltReq/eBreak
+
+// On exec_progbuf
+// change NextPC to progbuf_address (using DPC?)
+// goto: resume (implicic ebreak will return to debug mode)
+
+// On Resume:
+// update NextPC from DPC
+// Unstall pipe
 
 module dmc (
   input  logic       clk, reset,
   input  logic       Step,
+  input  logic       ebreakM,      // ebreak instruction
+  input  logic       ebreakEn,     // DCSR: enter debug mode on ebreak
   input  logic       HaltReq,      // Initiates core halt
   input  logic       ResumeReq,    // Initiates core resume
   input  logic       HaltOnReset,  // Halts core immediately on hart reset
@@ -46,14 +62,14 @@ module dmc (
 
   output logic       EnterDebugMode, // Store PCNextF in DPC when entering Debug Mode
   output logic       ExitDebugMode,  // Updates PCNextF with the current value of DPC
-  output logic       ForceNOP        // Fills the pipeline with NOP
+  output logic       ForceBreakPoint // Causes artificial ebreak that puts core in debug mode
 );
   `include "debug.vh"
 
-  enum logic [1:0] {RUNNING, FLUSH, HALTED, RESUME} State;
+   enum logic [2:0] {RUNNING, HALTED, RESUME, STEP, PROGBUF} State;
 
-  localparam NOP_CYCLE_DURATION = 0;
+  localparam E2M_CYCLE_COUNT = 3;
-  logic [$clog2(NOP_CYCLE_DURATION+1)-1:0] Counter;
+  logic [$clog2(E2M_CYCLE_COUNT+1)-1:0] Counter;
 
   always_ff @(posedge clk) begin
     if (reset)
@@ -62,12 +78,13 @@ module dmc (
       HaveReset <= 0;
   end
 
-  assign DebugMode = (State != RUNNING);
+  assign ForceBreakPoint = (State == RUNNING) & HaltReq | (State == STEP) & ~|Counter;
-  assign DebugStall = (State == HALTED);
 
-  assign EnterDebugMode = (State == FLUSH) & ~|Counter;
+  assign DebugMode = (State != RUNNING);
+  assign DebugStall = (State == HALTED) | (State == RESUME);
+
+  assign EnterDebugMode = (State == RUNNING) & (ebreakM & ebreakEn) | ForceBreakPoint;
   assign ExitDebugMode = (State == HALTED) & ResumeReq;
-  assign ForceNOP = (State == FLUSH);
 
   always_ff @(posedge clk) begin
     if (reset) begin
@@ -77,33 +94,37 @@ module dmc (
       case (State)
         RUNNING : begin
           if (HaltReq) begin
-            Counter <= NOP_CYCLE_DURATION;
-            State <= FLUSH;
-            DebugCause <= `CAUSE_HALTREQ;
-          end 
-          //else if (eBreak) TODO: halt on ebreak if DCSR bit is set
-          // DebugCause <= `CAUSE_EBREAK;
-        end
-
-        // fill the pipe with NOP before halting
-        FLUSH : begin
-          if (~|Counter)
             State <= HALTED;
-          else
+            DebugCause <= `CAUSE_HALTREQ;
-            Counter <= Counter - 1;
+          end else if (ebreakM & ebreakEn) begin
+            State <= HALTED;
+            DebugCause <= `CAUSE_EBREAK;
+          end
         end
 
         HALTED : begin
-          if (ResumeReq) begin
+          if (ResumeReq)
+            State <= RESUME;
+        end
+
+        // Wait a cycle to load PCF from DPC before resuming
+        // TODO: test without resume stage
+        RESUME : begin
           if (Step) begin
-              Counter <= NOP_CYCLE_DURATION;
+            Counter <= E2M_CYCLE_COUNT;
-              State <= FLUSH;
+            State <= STEP;
-              DebugCause <= `CAUSE_STEP;
           end else begin
             State <= RUNNING;
             ResumeAck <= 1;
           end
         end
+
+        STEP : begin
+          if (~|Counter) begin
+            DebugCause <= `CAUSE_STEP;
+            State <= HALTED;
+          end else
+            Counter <= Counter - 1;
         end
       endcase
     end

src/hazard/hazard.sv

@@ -108,9 +108,8 @@ module hazard import cvw::*;  #(parameter cvw_t P) (
   assign LatestUnstalledW = ~StallW & StallM;
   
   // Each stage flushes if the previous stage is the last one stalled (for cause) or the system has reason to flush
-  // Do not flush if halted for Debug
+  assign FlushD = (LatestUnstalledD | FlushDCause);
-  assign FlushD = ~DebugStall & (LatestUnstalledD | FlushDCause);
+  assign FlushE = (LatestUnstalledE | FlushECause);
-  assign FlushE = ~DebugStall & (LatestUnstalledE | FlushECause);
+  assign FlushM = (LatestUnstalledM | FlushMCause);
-  assign FlushM = ~DebugStall & (LatestUnstalledM | FlushMCause);
+  assign FlushW = (LatestUnstalledW | FlushWCause);
-  assign FlushW = ~DebugStall & (LatestUnstalledW | FlushWCause);
 endmodule

src/ifu/ifu.sv

@@ -98,11 +98,13 @@ module ifu import cvw::*;  #(parameter cvw_t P) (
   output logic                 ICacheAccess,                             // Report I$ read to performance counters
   output logic                 ICacheMiss,                               // Report I$ miss to performance counters
   // Debug Mode logic
+  output logic [P.XLEN-1:0]    PCNextF,                                  // Next PCF, selected from Branch predictor, Privilege, or PC+2/4
   input  logic                 ExitDebugMode,
   input  logic [P.XLEN-1:0]    DPC,
-  output logic [P.XLEN-1:0]    PCNextF,                                  // Next PCF, selected from Branch predictor, Privilege, or PC+2/4
+  input  logic                 ProgBuffScanEn,
-  input  logic                 ForceNOP,
   // Debug scan chain
+  input  logic [3:0]           ProgBufAddr,
+  input  logic                 ProgBufScanIn,
   input  logic                 DebugScanEn,
   input  logic                 DebugScanIn,
   output logic                 DebugScanOut
@@ -111,7 +113,6 @@ module ifu import cvw::*;  #(parameter cvw_t P) (
   localparam [31:0]            nop = 32'h00000013;                       // instruction for NOP
   localparam            LINELEN = P.ICACHE_SUPPORTED ? P.ICACHE_LINELENINBITS : P.XLEN;
 
-  logic [P.XLEN-1:0]           PCNextFM;                                 // (muxed for debug) Next PCF, selected from Branch predictor, Privilege, or PC+2/4
   logic [P.XLEN-1:0]           PC1NextF;                                 // Branch predictor next PCF
   logic [P.XLEN-1:0]           PC2NextF;                                 // Selected PC between branch prediction and next valid PC if CSRWriteFence
   logic [P.XLEN-1:0]           UnalignedPCNextF;                         // The next PCF, but not aligned to 2 bytes. 
@@ -128,6 +129,7 @@ module ifu import cvw::*;  #(parameter cvw_t P) (
   logic [31:0]                 IROMInstrF;                               // Instruction from the IROM
   logic [31:0]                 ICacheInstrF;                             // Instruction from the I$
   logic [31:0]                 InstrRawF;                                // Instruction from the IROM, I$, or bus
+  logic [31:0]                 ProgBufInstrF;                            // Instruction from the ProgBuf
   logic                        CompressedF, CompressedE;                 // The fetched instruction is compressed
   logic [31:0]                 PostSpillInstrRawF;                       // Fetch instruction after merge two halves of spill
   logic [31:0]                 InstrRawD;                                // Non-decompressed instruction in the Decode stage
@@ -146,6 +148,7 @@ module ifu import cvw::*;  #(parameter cvw_t P) (
   logic                        BusCommittedF;                            // Bus memory operation in flight, delay interrupts
   logic                        CacheCommittedF;                          // I$ memory operation started, delay interrupts
   logic                        SelIROM;                                  // PMA indicates instruction address is in the IROM
+  logic                        SelProgBuf;                               // PMA indicates instruction address is in Program Buffer
   logic [15:0]                 InstrRawE, InstrRawM;
   logic [LINELEN-1:0]          FetchBuffer;
   logic [31:0]                 ShiftUncachedInstr;
@@ -196,7 +199,7 @@ module ifu import cvw::*;  #(parameter cvw_t P) (
          .TLBFlush,
          .PhysicalAddress(PCPF),
          .TLBMiss(ITLBMissF),
-         .Cacheable(CacheableF), .Idempotent(), .SelTIM(SelIROM),
+         .Cacheable(CacheableF), .Idempotent(), .SelTIM(SelIROM), .SelProgBuf,
          .InstrAccessFaultF, .LoadAccessFaultM(), .StoreAmoAccessFaultM(),
          .InstrPageFaultF, .LoadPageFaultM(), .StoreAmoPageFaultM(),
          .LoadMisalignedFaultM(), .StoreAmoMisalignedFaultM(),
@@ -209,6 +212,7 @@ module ifu import cvw::*;  #(parameter cvw_t P) (
     assign PCPF = PCFExt[P.PA_BITS-1:0];
     assign CacheableF = 1'b1;
     assign SelIROM = '0;
+    assign SelProgBuf = '0;
   end
 
   ////////////////////////////////////////////////////////////////////////////////////////////////
@@ -218,7 +222,7 @@ module ifu import cvw::*;  #(parameter cvw_t P) (
   // CommittedM tells the CPU's privileged unit the current instruction
   // in the memory stage is a memory operaton and that memory operation is either completed
   // or is partially executed. Partially completed memory operations need to prevent an interrupts.
-  // There is not a clean way to restore back to a partial executed instruction.  CommiteedM will
+  // There is not a clean way to restore back to a partial executed instruction.  CommittedM will
   // delay the interrupt until the LSU is in a clean state.
   assign CommittedF = CacheCommittedF | BusCommittedF;
 
@@ -314,7 +318,14 @@ module ifu import cvw::*;  #(parameter cvw_t P) (
   assign IFUStallF = IFUCacheBusStallF | SelSpillNextF;
   assign GatedStallD = StallD & ~SelSpillNextF;
   
+  if (P.DEBUG_SUPPORTED) begin
+    logic [31:0] PostSpillInstrRawFM;
+    progbuf #(P) progbuf(.clk, .reset, .Addr(PCNextF[3:0]), .ProgBufInstrF, .ScanAddr(ProgBufAddr), .Scan(ProgBuffScanEn), .ScanIn(ProgBufScanIn));
+    assign PostSpillInstrRawFM = SelProgBuf ? ProgBufInstrF : PostSpillInstrRawF;
+    flopenl #(32) AlignedInstrRawDFlop(clk, reset | FlushD, ~StallD, PostSpillInstrRawFM, nop, InstrRawD);
+  end else begin
     flopenl #(32) AlignedInstrRawDFlop(clk, reset | FlushD, ~StallD, PostSpillInstrRawF, nop, InstrRawD);
+  end
 
   ////////////////////////////////////////////////////////////////////////////////////////////////
   // PCNextF logic
@@ -324,10 +335,11 @@ module ifu import cvw::*;  #(parameter cvw_t P) (
     mux2 #(P.XLEN) pcmux2(.d0(PC1NextF), .d1(NextValidPCE), .s(CSRWriteFenceM),.y(PC2NextF));
   else assign PC2NextF = PC1NextF;
 
+
   mux3 #(P.XLEN) pcmux3(PC2NextF, EPCM, TrapVectorM, {TrapM, RetM}, UnalignedPCNextF);
+
   if (P.DEBUG_SUPPORTED) begin
-    mux2 #(P.XLEN) pcresetmux({UnalignedPCNextF[P.XLEN-1:1], 1'b0}, P.RESET_VECTOR[P.XLEN-1:0], reset, PCNextFM);
+    mux3 #(P.XLEN) pcresetmux(.d2(P.RESET_VECTOR[P.XLEN-1:0]), .d1(DPC), .d0({UnalignedPCNextF[P.XLEN-1:1], 1'b0}), .s({reset,ExitDebugMode}), .y(PCNextF));
-    assign PCNextF = ExitDebugMode ? DPC : PCNextFM;
     flopen #(P.XLEN) pcreg(clk, ~StallF | reset | ExitDebugMode, PCNextF, PCF);
   end else begin
     mux2 #(P.XLEN) pcresetmux({UnalignedPCNextF[P.XLEN-1:1], 1'b0}, P.RESET_VECTOR[P.XLEN-1:0], reset, PCNextF);

src/ifu/progbuf.sv

@@ -0,0 +1,69 @@
+///////////////////////////////////////////
+// progbuf.sv
+//
+// Written: matthew.n.otto@okstate.edu
+// Created: 18 June 2024
+//
+// Purpose: Holds small programs to be executed in debug mode
+//          This module acts like a small ROM except it can be written by serial Scanning via the Debug Module
+//
+// A component of the CORE-V-WALLY configurable RISC-V project.
+// https://github.com/openhwgroup/cvw
+// 
+// Copyright (C) 2021-24 Harvey Mudd College & Oklahoma State University
+//
+// SPDX-License-Identifier: Apache-2.0 WITH SHL-2.1
+//
+// Licensed under the Solderpad Hardware License v 2.1 (the “License”); you may not use this file 
+// except in compliance with the License, or, at your option, the Apache License Version 2.0. You 
+// may obtain a copy of the License at
+//
+// https://solderpad.org/licenses/SHL-2.1/
+//
+// Unless required by applicable law or agreed to in writing, any work distributed under the 
+// License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, 
+// either express or implied. See the License for the specific language governing permissions 
+// and limitations under the License.
+////////////////////////////////////////////////////////////////////////////////////////////////
+
+module progbuf import cvw::*;  #(parameter cvw_t P) (
+    input  logic        clk, reset,
+    input  logic [3:0]  Addr,
+    output logic [31:0] ProgBufInstrF,
+
+    input  logic [3:0]  ScanAddr,
+    input  logic        Scan,
+    input  logic        ScanIn
+);
+
+  localparam PROGBUF_SIZE = (P.PROGBUF_RANGE+1)/4;
+  localparam ADDR_WIDTH = $clog2(PROGBUF_SIZE);
+  
+  bit [31:0] RAM [PROGBUF_SIZE-1:0];
+
+  logic EnPrevClk;
+  logic WriteProgBuf;
+  logic [32:0] WriteData;
+  logic [ADDR_WIDTH-1:0] AddrM;
+
+  flopr #(1) Scanenhist (.clk, .reset, .d(Scan), .q(EnPrevClk));
+  assign WriteProgBuf = ~Scan & EnPrevClk;
+
+  assign WriteData[32] = ScanIn;
+  genvar i;
+  for (i=0; i<32; i=i+1) begin
+    flopenr #(1) Scanreg (.clk, .reset, .en(Scan), .d(WriteData[i+1]), .q(WriteData[i]));
+  end
+
+  assign AddrM = WriteProgBuf ? ScanAddr[ADDR_WIDTH-1:0] : Addr[ADDR_WIDTH-1:0];
+
+  always_ff @(posedge clk) begin
+    if (WriteProgBuf)
+      RAM[AddrM] <= WriteData;
+    if (reset)
+      ProgBufInstrF <= 0;
+    else
+      ProgBufInstrF <= RAM[AddrM];
+  end
+
+endmodule

src/mmu/adrdecs.sv

@@ -33,7 +33,7 @@ module adrdecs import cvw::*;  #(parameter cvw_t P) (
   input  logic [P.PA_BITS-1:0] PhysicalAddress,
   input  logic                 AccessRW, AccessRX, AccessRWXC,
   input  logic [1:0]           Size,
-  output logic [11:0]          SelRegions
+  output logic [12:0]          SelRegions
 );
 
   localparam logic [3:0]       SUPPORTED_SIZE = (P.LLEN == 32 ? 4'b0111 : 4'b1111);
@@ -49,6 +49,7 @@ module adrdecs import cvw::*;  #(parameter cvw_t P) (
   adrdec #(P.PA_BITS) plicdec(PhysicalAddress, P.PLIC_BASE[P.PA_BITS-1:0], P.PLIC_RANGE[P.PA_BITS-1:0], P.PLIC_SUPPORTED, AccessRW, Size, 4'b0100, SelRegions[9]);
   adrdec #(P.PA_BITS) sdcdec(PhysicalAddress, P.SDC_BASE[P.PA_BITS-1:0], P.SDC_RANGE[P.PA_BITS-1:0], P.SDC_SUPPORTED, AccessRW, Size, SUPPORTED_SIZE & 4'b1100, SelRegions[10]); 
   adrdec #(P.PA_BITS) spidec(PhysicalAddress, P.SPI_BASE[P.PA_BITS-1:0], P.SPI_RANGE[P.PA_BITS-1:0], P.SPI_SUPPORTED, AccessRW, Size, 4'b0100, SelRegions[11]);
+  adrdec #(P.PA_BITS) progbufdec(PhysicalAddress, P.PROGBUF_BASE[P.PA_BITS-1:0], P.PROGBUF_RANGE[P.PA_BITS-1:0], P.DEBUG_SUPPORTED, AccessRX, Size, SUPPORTED_SIZE, SelRegions[12]);
 
   assign SelRegions[0] = ~|(SelRegions[11:1]); // none of the regions are selected
 endmodule

src/mmu/mmu.sv

@@ -50,6 +50,7 @@ module mmu import cvw::*;  #(parameter cvw_t P,
   output logic                 Cacheable,          // PMA indicates memory address is cachable
   output logic                 Idempotent,         // PMA indicates memory address is idempotent
   output logic                 SelTIM,             // Select a tightly integrated memory
+  output logic                 SelProgBuf,         // Select ProgBuf
   // Faults
   output logic                 InstrAccessFaultF, LoadAccessFaultM, StoreAmoAccessFaultM, // access fault sources
   output logic                 InstrPageFaultF, LoadPageFaultM, StoreAmoPageFaultM,       // page fault sources
@@ -112,7 +113,7 @@ module mmu import cvw::*;  #(parameter cvw_t P,
 
   pmachecker #(P) pmachecker(.PhysicalAddress, .Size, .CMOpM, 
     .AtomicAccessM, .ExecuteAccessF, .WriteAccessM, .ReadAccessM, .PBMemoryType,
-    .Cacheable, .Idempotent, .SelTIM, 
+    .Cacheable, .Idempotent, .SelTIM, .SelProgBuf,
     .PMAInstrAccessFaultF, .PMALoadAccessFaultM, .PMAStoreAmoAccessFaultM);
  
   if (P.PMP_ENTRIES > 0) begin : pmp

src/mmu/pmachecker.sv

@@ -38,7 +38,7 @@ module pmachecker import cvw::*;  #(parameter cvw_t P) (
   input  logic                 WriteAccessM,   // Write access 
   input  logic                 ReadAccessM,    // Read access
   input  logic [1:0]           PBMemoryType,     // PBMT field of PTE during TLB hit, or 00 otherwise
-  output logic                 Cacheable, Idempotent, SelTIM,
+  output logic                 Cacheable, Idempotent, SelTIM, SelProgBuf,
   output logic                 PMAInstrAccessFaultF,
   output logic                 PMALoadAccessFaultM,
   output logic                 PMAStoreAmoAccessFaultM
@@ -46,7 +46,7 @@ module pmachecker import cvw::*;  #(parameter cvw_t P) (
 
   logic                        PMAAccessFault;
   logic                        AccessRW, AccessRWXC, AccessRX;
-  logic [11:0]                 SelRegions;
+  logic [12:0]                 SelRegions;
   logic                        AtomicAllowed;
   logic                        CacheableRegion, IdempotentRegion;
 
@@ -72,6 +72,9 @@ module pmachecker import cvw::*;  #(parameter cvw_t P) (
   // Check if tightly integrated memories are selected
   assign SelTIM = SelRegions[1] | SelRegions[2]; // exclusion-tag: unused-tim
 
+  // Debug program buffer
+  assign SelProgBuf = SelRegions[12];
+
   // Detect access faults
   assign PMAAccessFault          = SelRegions[0] & AccessRWXC | AtomicAccessM & ~AtomicAllowed;  
   assign PMAInstrAccessFaultF    = ExecuteAccessF & PMAAccessFault;

src/privileged/csr.sv

@@ -96,9 +96,9 @@ module csr import cvw::*;  #(parameter cvw_t P) (
   // Debug Mode output
   input  logic                     DebugMode,
   input  logic [2:0]               DebugCause,
+  output logic                     ebreakEn,
   output logic                     Step,
   output logic [P.XLEN-1:0]        DPC,
-  input  logic [P.XLEN-1:0]        PCNextF,
   input  logic                     EnterDebugMode,
   // Debug scan chain
   input  logic                     DebugSel,
@@ -215,10 +215,10 @@ module csr import cvw::*;  #(parameter cvw_t P) (
   ///////////////////////////////////////////
 
   assign CSRAdrM = InstrM[31:20];
-  assign UnalignedNextEPCM = TrapM ? PCM : CSRWriteValDM;
+  assign UnalignedNextEPCM = TrapM ? PCM : CSRWriteValM;
   assign NextEPCM = P.ZCA_SUPPORTED ? {UnalignedNextEPCM[P.XLEN-1:1], 1'b0} : {UnalignedNextEPCM[P.XLEN-1:2], 2'b00}; // 3.1.15 alignment
-  assign NextCauseM = TrapM ? {InterruptM, CauseM}: {CSRWriteValDM[P.XLEN-1], CSRWriteValDM[3:0]};
+  assign NextCauseM = TrapM ? {InterruptM, CauseM}: {CSRWriteValM[P.XLEN-1], CSRWriteValM[3:0]};
-  assign NextMtvalM = TrapM ? NextFaultMtvalM : CSRWriteValDM;
+  assign NextMtvalM = TrapM ? NextFaultMtvalM : CSRWriteValM;
   assign UngatedCSRMWriteM = CSRWriteDM & (PrivilegeModeW == P.M_MODE);
   assign CSRMWriteM = UngatedCSRMWriteM & InstrValidNotFlushedM;
   assign CSRSWriteM = CSRWriteDM & (|PrivilegeModeW) & InstrValidNotFlushedM;
@@ -305,7 +305,7 @@ module csr import cvw::*;  #(parameter cvw_t P) (
   if (P.DEBUG_SUPPORTED) begin:csrd
     csrd #(P) csrd(.clk, .reset, .DebugMode, .PrivilegeModeW,
     .CSRWriteDM, .CSRAdrM(CSRAdrDM), .CSRWriteValM(CSRWriteValDM), .CSRDReadValM, .IllegalCSRDAccessM,
-    .DebugCause, .Step, .DPC, .PCNextF, .EnterDebugMode);
+    .DebugCause, .ebreakEn, .Step, .DPC, .PCM, .EnterDebugMode);
   end else begin
     assign CSRDReadValM = '0;
     assign IllegalCSRDAccessM = 1'b1; // Debug isn't supported

src/privileged/csrd.sv

@@ -35,12 +35,12 @@ module csrd import cvw::*;  #(parameter cvw_t P) (
   input  logic [P.XLEN-1:0] CSRWriteValM,
   output logic [P.XLEN-1:0] CSRDReadValM,
   output logic              IllegalCSRDAccessM,
-
+  input  logic [P.XLEN-1:0] PCM,
+  input  logic              EnterDebugMode,
   input  logic [2:0]        DebugCause,
+  output logic              ebreakEn,
   output logic              Step,
-  output logic [P.XLEN-1:0] DPC,
+  output logic [P.XLEN-1:0] DPC
-  input  logic [P.XLEN-1:0] PCNextF,
-  input  logic              EnterDebugMode
 );
   `include "debug.vh"
 
@@ -58,8 +58,8 @@ module csrd import cvw::*;  #(parameter cvw_t P) (
   const logic       ebreakVS = 0;
   const logic       ebreakVU = 0;
   logic             ebreakM;
-  logic             ebreakS;
+  const logic       ebreakS = 0;
-  logic             ebreakU;
+  const logic       ebreakU = 0;
   const logic       StepIE = 0;
   const logic       StopCount = 0;
   const logic       StopTime = 0;
@@ -70,6 +70,7 @@ module csrd import cvw::*;  #(parameter cvw_t P) (
   logic [1:0]       Prv;
 
   
+  assign ebreakEn = ebreakM; // Only support ebreak from M mode
   assign CSRDWriteM = CSRWriteDM & (PrivilegeModeW == P.M_MODE) & DebugMode;
 
   assign WriteDCSRM = CSRDWriteM & (CSRAdrM == DCSR_ADDR);
@@ -85,15 +86,13 @@ module csrd import cvw::*;  #(parameter cvw_t P) (
     end
   end
 
-  flopenr #(4) DCSRreg (clk, reset, WriteDCSRM, 
+  flopenr #(4) DCSRreg (clk, reset, WriteDCSRM, {CSRWriteValM[`EBREAKM], CSRWriteValM[`STEP]}, {ebreakM, Step});
-    {CSRWriteValM[`EBREAKM], CSRWriteValM[`EBREAKS], CSRWriteValM[`EBREAKU], CSRWriteValM[`STEP]}, 
-    {ebreakM, ebreakS, ebreakU, Step});
 
   assign DCSR = {DebugVer, 10'b0, ebreakVS, ebreakVU, ebreakM, 1'b0, ebreakS, ebreakU, StepIE,
                       StopCount, StopTime, Cause, V, MPrvEn, NMIP, Step, Prv};
 
-  assign DPCWriteVal = EnterDebugMode ? PCNextF : CSRWriteValM;
+  assign DPCWriteVal = EnterDebugMode ? PCM : CSRWriteValM;
-  flopenr #(P.XLEN) DPCreg (clk, reset, WriteDPCM | EnterDebugMode, DPCWriteVal, DPC);
+  flopenr #(P.XLEN) DPCreg (clk, reset, WriteDPCM | EnterDebugMode, DPCWriteVal, DPC); // TODO: reset to something sane (0x80000000?)
 
   always_comb begin
     CSRDReadValM = '0;

src/privileged/privdec.sv

@@ -31,6 +31,7 @@
 module privdec import cvw::*;  #(parameter cvw_t P) (
   input  logic         clk, reset,
   input  logic         StallW, FlushW,
+  input  logic         ForceBreakPoint,                             // Debug Module initiated break to debug mode
   input  logic [31:15] InstrM,                              // privileged instruction function field
   input  logic         PrivilegedM,                         // is this a privileged instruction (from IEU controller)
   input  logic         IllegalIEUFPUInstrM,                 // Not a legal IEU instruction
@@ -40,13 +41,14 @@ module privdec import cvw::*;  #(parameter cvw_t P) (
   output logic         IllegalInstrFaultM,                  // Illegal instruction
   output logic         EcallFaultM, BreakpointFaultM,       // Ecall or breakpoint; must retire, so don't flush it when the trap occurs
   output logic         sretM, mretM, RetM,                  // return instructions
-  output logic         wfiM, wfiW, sfencevmaM               // wfi / sfence.vma / sinval.vma instructions
+  output logic         wfiM, wfiW, sfencevmaM,              // wfi / sfence.vma / sinval.vma instructions
+  output logic         ebreakM                              // ebreak instruction
 );
 
   logic                rs1zeroM;                            // rs1 field = 0
   logic                IllegalPrivilegedInstrM;             // privileged instruction isn't a legal one or in legal mode
   logic                WFITimeoutM;                         // WFI reaches timeout threshold
-  logic                ebreakM, ecallM;                     // ebreak / ecall instructions
+  logic                ecallM;                              // ecall instructions
   logic                sinvalvmaM;                          // sinval.vma
   logic                sfencewinvalM, sfenceinvalirM;       // sfence.w.inval, sfence.inval.ir
   logic                invalM;                              // any of the svinval instructions
@@ -94,7 +96,10 @@ module privdec import cvw::*;  #(parameter cvw_t P) (
   // Extract exceptions by name and handle them 
   ///////////////////////////////////////////
 
-  assign BreakpointFaultM = ebreakM; // could have other causes from a debugger
+  if (P.DEBUG_SUPPORTED)
+    assign BreakpointFaultM = ebreakM | ForceBreakPoint;
+  else
+    assign BreakpointFaultM = ebreakM;
   assign EcallFaultM = ecallM;
 
   ///////////////////////////////////////////

src/privileged/privileged.sv

@@ -97,12 +97,14 @@ module privileged import cvw::*;  #(parameter cvw_t P) (
   output logic              BigEndianM,                                     // Use big endian in current privilege mode
   // Fault outputs                                                         
   output logic              wfiM, IntPendingM,                              // Stall in Memory stage for WFI until interrupt pending or timeout
+  output logic              ebreakM,                                        // Notifies DM to enter debug mode
   // Debuge Mode
+  output logic              ebreakEn,
+  input  logic              ForceBreakPoint,
   input  logic              DebugMode,
   input  logic [2:0]        DebugCause,
   output logic              Step,
   output logic [P.XLEN-1:0] DPC,
-  input  logic [P.XLEN-1:0] PCNextF,
   input  logic              EnterDebugMode,
   // Debug scan chain
   input  logic              DebugSel,
@@ -142,9 +144,9 @@ module privileged import cvw::*;  #(parameter cvw_t P) (
 
   // decode privileged instructions
   privdec #(P) pmd(.clk, .reset, .StallW, .FlushW, .InstrM(InstrM[31:15]), 
-    .PrivilegedM, .IllegalIEUFPUInstrM, .IllegalCSRAccessM, 
+    .PrivilegedM, .IllegalIEUFPUInstrM, .IllegalCSRAccessM, .ForceBreakPoint,
     .PrivilegeModeW, .STATUS_TSR, .STATUS_TVM, .STATUS_TW, .IllegalInstrFaultM, 
-    .EcallFaultM, .BreakpointFaultM, .sretM, .mretM, .RetM, .wfiM, .wfiW, .sfencevmaM);
+    .EcallFaultM, .BreakpointFaultM, .sretM, .mretM, .RetM, .wfiM, .wfiW, .sfencevmaM, .ebreakM);
 
   // Control and Status Registers
   csr #(P) csr(.clk, .reset, .FlushM, .FlushW, .StallE, .StallM, .StallW,
@@ -163,7 +165,7 @@ module privileged import cvw::*;  #(parameter cvw_t P) (
     .SetFflagsM, .FRM_REGW, .ENVCFG_CBE, .ENVCFG_PBMTE, .ENVCFG_ADUE,
     .EPCM, .TrapVectorM,
     .CSRReadValW, .IllegalCSRAccessM, .BigEndianM,
-    .DebugMode, .DebugCause, .Step, .DPC, .PCNextF, .EnterDebugMode,
+    .DebugMode, .DebugCause, .ebreakEn, .Step, .DPC, .EnterDebugMode,
     .DebugSel, .DebugRegAddr, .DebugCapture, .DebugRegUpdate, .DebugScanEn, .DebugScanIn, .DebugScanOut);
 
   // pipeline early-arriving trap sources

src/wally/wallypipelinedcore.sv

@@ -53,6 +53,7 @@ module wallypipelinedcore import cvw::*; #(parameter cvw_t P) (
    output logic                  ResumeAck,
    output logic                  HaveReset,
    output logic                  DebugStall,
+   input  logic                  ExecProgBuff,
    // Debug scan chain
    input  logic                  DebugScanEn,    // puts scannable flops into scan mode
    output logic                  DebugScanOut,   // (misc) scan chain data out
@@ -66,13 +67,16 @@ module wallypipelinedcore import cvw::*; #(parameter cvw_t P) (
    input  logic                  CSRSel,         // selects CSR scan chain
    input  logic [11:0]           DebugRegAddr,   // address for scanable regfiles (GPR, FPR, CSR)
    input  logic                  DebugCapture,   // latches values into scan register before scanning out
-   input  logic                  DebugRegUpdate  // writes values from scan register after scanning in							       							       
+   input  logic                  DebugRegUpdate,  // writes values from scan register after scanning in	
+   input  logic [3:0]            ProgBufAddr,
+   input  logic                  ProgBuffScanEn
 );
 
   logic                          StallF, StallD, StallE, StallM, StallW;
   logic                          FlushD, FlushE, FlushM, FlushW;
   logic                          TrapM, RetM;
   logic                          DebugMode, Step;
+  logic                          ebreakEn;
 
   //  signals that must connect through DP
   logic                          IntDivE, W64E;
@@ -190,13 +194,14 @@ module wallypipelinedcore import cvw::*; #(parameter cvw_t P) (
   logic                          BranchD, BranchE, JumpD, JumpE;
   logic                          DCacheStallM, ICacheStallF;
   logic                          wfiM, IntPendingM;
+  logic                          ebreakM;
 
   // Debug mode logic
   logic [P.XLEN-1:0]             DPC, PCNextF;
   logic                          ExitDebugMode;
   logic                          EnterDebugMode;
   logic [2:0]                    DebugCause;
-  logic                          ForceNOP;
+  logic                          ForceBreakPoint;
   // Debug register scan chain interconnects
   logic [2:0]                    DebugScanReg;
 
@@ -222,7 +227,7 @@ module wallypipelinedcore import cvw::*; #(parameter cvw_t P) (
     .STATUS_MPP, .ENVCFG_PBMTE, .ENVCFG_ADUE, .ITLBWriteF, .sfencevmaM, .ITLBMissF,
     // pmp/pma (inside mmu) signals. 
     .PMPCFG_ARRAY_REGW,  .PMPADDR_ARRAY_REGW, .InstrAccessFaultF, .InstrUpdateDAF,
-    .ExitDebugMode, .DPC, .PCNextF, .ForceNOP,
+    .ExitDebugMode, .DPC, .PCNextF, .ProgBuffScanEn, .ProgBufAddr, .ProgBufScanIn(DebugScanIn),
     .DebugScanEn(DebugScanEn & MiscSel), .DebugScanIn(DebugScanReg[0]), .DebugScanOut(DebugScanReg[1]));
     
   // integer execution unit: integer register file, datapath and controller
@@ -320,9 +325,9 @@ module wallypipelinedcore import cvw::*; #(parameter cvw_t P) (
   if (P.DEBUG_SUPPORTED) begin
     dmc debugcontrol(
       .clk, .reset,
-      .Step, .HaltReq, .ResumeReq, .HaltOnReset, .AckHaveReset,
+      .Step, .ebreakM, .ebreakEn, .HaltReq, .ResumeReq, .HaltOnReset, .AckHaveReset,
       .ResumeAck, .HaveReset, .DebugMode, .DebugCause, .DebugStall,
-      .EnterDebugMode, .ExitDebugMode, .ForceNOP);
+      .EnterDebugMode, .ExitDebugMode, .ForceBreakPoint);
   end else begin
     assign DebugStall = 1'b0;
   end
@@ -349,8 +354,8 @@ module wallypipelinedcore import cvw::*; #(parameter cvw_t P) (
       .PrivilegeModeW, .SATP_REGW,
       .STATUS_MXR, .STATUS_SUM, .STATUS_MPRV, .STATUS_MPP, .STATUS_FS, 
       .PMPCFG_ARRAY_REGW, .PMPADDR_ARRAY_REGW, 
-      .FRM_REGW, .ENVCFG_CBE, .ENVCFG_PBMTE, .ENVCFG_ADUE, .wfiM, .IntPendingM, .BigEndianM,
+      .FRM_REGW, .ENVCFG_CBE, .ENVCFG_PBMTE, .ENVCFG_ADUE, .wfiM, .IntPendingM, .BigEndianM, .ebreakM,
-      .DebugMode, .DebugCause, .Step, .DPC, .PCNextF, .EnterDebugMode,
+      .ebreakEn, .ForceBreakPoint, .DebugMode, .DebugCause, .Step, .DPC, .EnterDebugMode,
       .DebugSel(CSRSel), .DebugRegAddr, .DebugCapture, .DebugRegUpdate, .DebugScanEn(DebugScanEn & CSRSel), .DebugScanIn, .DebugScanOut(CSRScanOut));
     if (P.DEBUG_SUPPORTED) begin
       flopenrs #(1) scantrapm (.clk, .reset, .en(DebugCapture), .d(TrapM), .q(), .scan(DebugScanEn), .scanin(DebugScanIn), .scanout(DebugScanReg[0]));

src/wally/wallypipelinedsoc.sv

@@ -82,6 +82,7 @@ module wallypipelinedsoc import cvw::*; #(parameter cvw_t P)  (
   logic                        ResumeAck;
   logic                        HaveReset;
   logic                        DebugStall;
+  logic                        ExecProgBuff;
   // Debug Module signals
   logic                        DebugScanEn;
   logic                        DebugScanIn;
@@ -96,9 +97,8 @@ module wallypipelinedsoc import cvw::*; #(parameter cvw_t P)  (
   logic [11:0]                 DebugRegAddr;
   logic                        DebugCapture;
   logic                        DebugRegUpdate;
+  logic [3:0]                  ProgBufAddr;
   logic                        ProgBuffScanEn;
-  logic                        ProgBuffScanOut;
-  logic                        ExecProgBuff;
 
   // synchronize reset to SOC clock domain
   synchronizer resetsync(.clk, .d(reset_ext), .q(reset));
@@ -108,9 +108,10 @@ module wallypipelinedsoc import cvw::*; #(parameter cvw_t P)  (
     .MTimerInt, .MExtInt, .SExtInt, .MSwInt, .MTIME_CLINT,
     .HRDATA, .HREADY, .HRESP, .HCLK, .HRESETn, .HADDR, .HWDATA, .HWSTRB,
     .HWRITE, .HSIZE, .HBURST, .HPROT, .HTRANS, .HMASTLOCK,
-    .HaltReq, .ResumeReq, .HaltOnReset, .AckHaveReset, .ResumeAck, .HaveReset, .DebugStall,
+    .HaltReq, .ResumeReq, .HaltOnReset, .AckHaveReset, .ResumeAck, .HaveReset, .DebugStall, .ExecProgBuff,
     .DebugScanEn, .DebugScanOut(DebugScanIn), .GPRScanOut(GPRScanIn), .FPRScanOut(FPRScanIn), .CSRScanOut(CSRScanIn), 
-    .DebugScanIn(DebugScanOut), .MiscSel, .GPRSel, .FPRSel, .CSRSel, .DebugRegAddr, .DebugCapture, .DebugRegUpdate);
+    .DebugScanIn(DebugScanOut), .MiscSel, .GPRSel, .FPRSel, .CSRSel, .DebugRegAddr, .DebugCapture, .DebugRegUpdate,
+    .ProgBufAddr, .ProgBuffScanEn);
 
   // instantiate uncore if a bus interface exists
   if (P.BUS_SUPPORTED) begin : uncoregen // Hack to work around Verilator bug https://github.com/verilator/verilator/issues/4769
@@ -130,7 +131,7 @@ module wallypipelinedsoc import cvw::*; #(parameter cvw_t P)  (
       .HaltReq, .ResumeReq, .HaltOnReset, .AckHaveReset, .ResumeAck, .HaveReset, .DebugStall,
       .DebugScanEn, .DebugScanIn, .GPRScanIn, .FPRScanIn, .CSRScanIn, .DebugScanOut,
       .MiscSel, .GPRSel, .FPRSel, .CSRSel, .RegAddr(DebugRegAddr), .DebugCapture, .DebugRegUpdate,
-      .ProgBuffScanEn, .ProgBuffScanOut, .ExecProgBuff);
+      .ProgBufAddr, .ProgBuffScanEn, .ExecProgBuff);
   end
 
 endmodule