Add DCSR and DPC CSRs (#614)

This PR addresses Issue #570 and adds the remaining missing debug CSRs
- DCSR
- DPC

---------

Signed-off-by: Katherine Hsu <[email protected]>
Co-authored-by: Derek Hower <[email protected]>

Author: neverlandiz

spec/schemas/csr_schema.json

@@ -249,7 +249,7 @@
         "virtual_address": true,
         "$comment": "Conditionally required; see below",
         "priv_mode": {
-          "enum": ["M", "S", "U", "VS"]
+          "enum": ["M", "S", "U", "VS", "D"]
         },
         "length": {
           "description": "Length, in bits, of the CSR. Can either be a 32, 64 or MXLEN, SXLEN, VSXLEN to indicate that it is dependent on the effective XLEN for a given mode. XLEN here refers to the effective XLEN in the current execution mode.",

spec/std/isa/csr/dcsr.yaml

@@ -0,0 +1,276 @@
+# Copyright (c) Katherine Hsu
+# SPDX-License-Identifier: BSD-3-Clause-Clear
+
+# yaml-language-server: $schema=../../schemas/csr_schema.json
+
+$schema: "csr_schema.json#"
+kind: csr
+name: dcsr
+long_name: Debug Control and Status Register
+address: 0x7B0
+priv_mode: D
+length: 32
+description: |
+  Upon entry into Debug Mode, v and prv are updated with the privilege level the hart was previously in,
+  and cause is updated with the reason for Debug Mode entry. Other than these fields and nmip, the
+  other fields of dcsr are only writable by the external debugger.
+
+  Priority of reasons for entering Debug Mode from highest to lowest is shown below.
+  5:: resethaltreq
+  6:: halt group
+  3:: haltreq
+  2:: trigger
+  1:: ebreak
+  4:: step
+
+  For compatibility with old versions of this spec, resethaltreq and
+  haltreq are allowed to be at different positions than shown as long as:
+  1. resethaltreq is higher priority than haltreq
+  2. the relative order of the other four causes is maintained
+
+  This CSR is read/write.
+
+definedBy: Sdext
+fields:
+  DEBUGVER:
+    location: 31-28
+    type: RO
+    description: |
+      0 (none):: There is no debug support.
+      4 (1.0):: Debug support exists as it is described in this document.
+      15 (custom):: There is debug support, but it does not conform to any available version of this spec.
+    reset_value: UNDEFINED_LEGAL
+  EXTCAUSE:
+    location: 26-24
+    type: RO
+    description: |
+      When cause is 7, this optional field contains the value of a more specific halt reason than "other."
+      Otherwise it contains 0.
+
+      0 (critical error):: The hart entered a critical error state, as defined in the Smdbltrp extension.
+
+      All other values are reserved for future versions of this spec, or for use by other RISC-V extensions.
+    reset_value: 0
+  CETRIG:
+    location: 19
+    type: RW
+    definedBy: Smdbltrp
+    description: |
+      This bit is part of Smdbltrp and only exists when that extension is implemented.
+      0 (disabled):: A hart in a critical error state does not enter
+      Debug Mode but instead asserts the critical-error signal to
+      the platform.
+      1 (enabled):: A hart in a critical error state enters Debug
+      Mode instead of asserting the critical-error signal to the
+      platform. Upon such entry into Debug Mode, the cause
+      field is set to 7, and the extcause field is set to 0, indicating
+      a critical error triggered the Debug Mode entry. This cause
+      has the highest priority among all reasons for entering
+      Debug Mode. Resuming from Debug Mode following an
+      entry from the critical error state returns the hart to the
+      critical error state.
+
+      When cetrig is 1, resuming from Debug Mode following an entry due to a critical
+      error will result in an immediate re-entry into Debug Mode due to the critical error.
+      The debugger may resume with cetrig set to 0 to allow the platform defined actions on
+      critical-error signal to occur. Other possible actions include initiating a hart or platform
+      reset using the Debug Module reset control.
+    reset_value: 0
+  PELP:
+    location: 18
+    type: RW
+    definedBy: Zicfilp
+    description: |
+      This bit is part of Zicfilp and only exists when that extension is implemented.
+      0 (NO_LP_EXPECTED):: No landing pad instruction expected.
+      1 (LP_EXPECTED):: A landing pad instruction is expected.
+    reset_value: 0
+  EBREAKVS:
+    location: 17
+    type: RW
+    definedBy: H
+    description: |
+      0 (exception):: ebreak instructions in VS-mode behave as described in the Privileged Spec.
+      1 (debug mode):: ebreak instructions in VS-mode enter Debug Mode.
+      This bit is hardwired to 0 if the hart does not support virtualization mode.
+    reset_value: 0
+  EBREAKVU:
+    location: 16
+    type: RW
+    definedBy: H
+    description: |
+      0 (exception):: ebreak instructions in VU-mode behave as described in the Privileged Spec.
+      1 (debug mode):: ebreak instructions in VU-mode enter Debug Mode.
+      This bit is hardwired to 0 if the hart does not support virtualization mode.
+    reset_value: 0
+  EBREAKM:
+    location: 15
+    type: RW
+    description: |
+      0 (exception):: ebreak instructions in M-mode behave as described in the Privileged Spec.
+      1 (debug mode):: ebreak instructions in M-mode enter Debug Mode.
+    reset_value: 0
+  EBREAKS:
+    location: 13
+    type: RW
+    definedBy: S
+    description: |
+      0 (exception):: ebreak instructions in S-mode behave as described in the Privileged Spec.
+      1 (debug mode):: ebreak instructions in S-mode enter Debug Mode.
+      This bit is hardwired to 0 if the hart does not support S-mode.
+    reset_value: 0
+  EBREAKU:
+    location: 12
+    type: RW
+    definedBy: U
+    description: |
+      0 (exception):: ebreak instructions in U-mode behave as described in the Privileged Spec.
+      1 (debug mode):: ebreak instructions in U-mode enter Debug Mode.
+      This bit is hardwired to 0 if the hart does not support U-mode.
+    reset_value: 0
+  STEPIE:
+    location: 11
+    description: |
+      0 (interrupts disabled):: Interrupts (including NMI) are disabled during single stepping with step set.
+      This value should be supported.
+      1 (interrupts enabled):: Interrupts (including NMI) are enabled during single stepping with step set.
+      Implementations may hard wire this bit to 0. In that case interrupt behavior can be emulated by the
+      debugger. The debugger must not change the value of this bit while the hart is running.
+    type(): |
+      if (DCSR_STEPIE_TYPE == "read-only-0" || DCSR_STEPIE_TYPE == "read-only-1") {
+        return CsrFieldType::RO;
+      }
+      assert(DCSR_STEPIE_TYPE == "rw", "Unhandled DCSR_STEPIE_TYPE value");
+      return CsrFieldType::RW;
+    reset_value(): |
+      if (DCSR_STEPIE_TYPE == "read-only-0") {
+        return 0;
+      } else if (DCSR_STEPIE_TYPE == "read-only-1") {
+        return 1;
+      }
+      assert(DCSR_STEPIE_TYPE == "rw", "Unhandled DCSR_STEPIE_TYPE value");
+      return UNDEFINED_LEGAL;
+  STOPCOUNT:
+    location: 10
+    description: |
+      0 (normal):: Increment counters as usual.
+      1 (freeze):: Don’t increment any hart-local counters while in Debug Mode or on ebreak instructions
+      that cause entry into Debug Mode. These counters include the instret CSR. On single-hart cores cycle
+      should be stopped, but on multi-hart cores it must keep incrementing.
+      An implementation may hardwire this bit to 0 or 1.
+    type(): |
+      if (DCSR_STOPCOUNT_TYPE == "read-only-0" || DCSR_STOPCOUNT_TYPE == "read-only-1") {
+        return CsrFieldType::RO;
+      }
+      assert(DCSR_STOPCOUNT_TYPE == "rw", "Unhandled DCSR_STOPCOUNT_TYPE value");
+      return CsrFieldType::RW;
+    reset_value(): |
+      if (DCSR_STOPCOUNT_TYPE == "read-only-0") {
+        return 0;
+      } else if (DCSR_STOPCOUNT_TYPE == "read-only-1") {
+        return 1;
+      }
+      assert(DCSR_STOPCOUNT_TYPE == "rw", "Unhandled DCSR_STOPCOUNT_TYPE value");
+      return UNDEFINED_LEGAL;
+  STOPTIME:
+    location: 9
+    description: |
+      0 (normal):: time continues to reflect mtime.
+      1 (freeze):: time is frozen at the time that Debug Mode was entered. When leaving Debug Mode,
+      time will reflect the latest value of mtime again.
+      While all harts have stoptime=1 and are in Debug Mode, mtime is allowed to stop incrementing.
+      An implementation may hardwire this bit to 0 or 1.
+    type(): |
+      if (DCSR_STOPTIME_TYPE == "read-only-0" || DCSR_STOPTIME_TYPE == "read-only-1") {
+        return CsrFieldType::RO;
+      }
+      assert(DCSR_STOPTIME_TYPE == "rw", "Unhandled DCSR_STOPTIME_TYPE value");
+      return CsrFieldType::RW;
+    reset_value(): |
+      if (DCSR_STOPTIME_TYPE == "read-only-0") {
+        return 0;
+      } else if (DCSR_STOPTIME_TYPE == "read-only-1") {
+        return 1;
+      }
+      assert(DCSR_STOPTIME_TYPE == "rw", "Unhandled DCSR_STOPTIME_TYPE value");
+      return UNDEFINED_LEGAL;
+  CAUSE:
+    location: 8-6
+    type: RO
+    description: |
+      Explains why Debug Mode was entered.
+      When there are multiple reasons to enter Debug Mode in a
+      single cycle, hardware should set cause to the cause with
+      the highest priority.
+      1 (ebreak):: An ebreak instruction was executed.
+      2 (trigger):: A Trigger Module trigger fired with action=1.
+      3 (haltreq):: The debugger requested entry to Debug Mode
+      using haltreq.
+      4 (step):: The hart single stepped because step was set.
+      5 (resethaltreq):: The hart halted directly out of reset due to
+      resethaltreq It is also acceptable to report 3 when this
+      happens.
+      6 (group):: The hart halted because it’s part of a halt group.
+      Harts may report 3 for this cause instead.
+      7 (other):: The hart halted for a reason other than the ones
+      mentioned above. extcause may contain a more specific
+      reason.
+    reset_value: 0
+  V:
+    location: 5
+    type: RW
+    definedBy: H
+    description: |
+      Extends the prv field with the virtualization mode the hart was operating in
+      when Debug Mode was entered. A debugger can change this value to change the
+      hart’s virtualization mode when exiting Debug Mode. This bit is hardwired to 0 on harts
+      that do not support virtualization mode.
+    reset_value: 0
+  MPRVEN:
+    location: 4
+    description: |
+      0 (disabled):: mprv in mstatus is ignored in Debug Mode.
+      1 (enabled):: mprv in mstatus takes effect in Debug Mode.
+      Implementing this bit is optional. It may be tied to either 0 or 1.
+    type(): |
+      if (DCSR_MPRVEN_TYPE == "read-only-0" || DCSR_MPRVEN_TYPE == "read-only-1") {
+        return CsrFieldType::RO;
+      }
+      assert(DCSR_MPRVEN_TYPE == "rw", "Unhandled DCSR_MPRVEN_TYPE value");
+      return CsrFieldType::RW;
+    reset_value(): |
+      if (DCSR_MPRVEN_TYPE == "read-only-0") {
+        return 0;
+      } else if (DCSR_MPRVEN_TYPE == "read-only-1") {
+        return 1;
+      }
+      assert(DCSR_MPRVEN_TYPE == "rw", "Unhandled DCSR_MPRVEN_TYPE value");
+      return UNDEFINED_LEGAL;
+  NMIP:
+    location: 3
+    type: RO
+    description: |
+      When set, there is a Non-Maskable-Interrupt (NMI) pending for the hart.
+      Since an NMI can indicate a hardware error condition, reliable debugging
+      may no longer be possible once this bit becomes set. This is implementation-dependent.
+    reset_value: 0
+  STEP:
+    location: 2
+    type: RW
+    description: |
+      When set and not in Debug Mode, the hart will only execute a single instruction
+      and then enter Debug Mode.
+      The debugger must not change the value of this bit while the hart is running.
+    reset_value: 0
+  PRV:
+    location: 1-0
+    type: RW
+    description: |
+      Contains the privilege mode the hart was operating in when Debug Mode was entered.
+      A debugger can change this value to change the hart’s privilege mode when exiting
+      Debug Mode.
+
+      Not all privilege modes are supported on all harts. If the encoding written is not
+      supported or the debugger is not allowed to change to it, the hart may change to any
+      supported privilege mode.
+    reset_value: 3

spec/std/isa/csr/dpc.yaml

@@ -0,0 +1,42 @@
+# Copyright (c) Katherine Hsu
+# SPDX-License-Identifier: BSD-3-Clause-Clear
+
+# yaml-language-server: $schema=../../schemas/csr_schema.json
+
+$schema: "csr_schema.json#"
+kind: csr
+name: dpc
+long_name: Debug PC Register
+address: 0x7B1
+priv_mode: D
+length: MXLEN
+description: |
+  Upon entry to debug mode, dpc is updated with the virtual address of the next instruction to be executed.
+
+  Executing the Program Buffer may cause the value of dpc to become UNSPECIFIED. If that is the case,
+  it must be possible to read/write dpc using an abstract command with postexec not set. The debugger
+  must attempt to save dpc between halting and executing a Program Buffer, and then restore dpc before
+  leaving Debug Mode.
+
+  If the Access Register abstract command supports reading dpc while the hart is running, then the
+  value read should be the address of a recently executed instruction.
+
+  If the Access Register abstract command supports writing dpc while the hart is running, then the
+  executing program should jump to the written address shortly after the write occurs.
+
+  The writability of dpc follows the same rules as mepc as defined in the Privileged Spec. In particular,
+  dpc must be able to hold all valid virtual addresses and the writability of the low bits depends on
+  IALIGN.
+
+  When resuming, the hart’s PC is updated to the virtual address stored in dpc. A debugger may write
+  dpc to change where the hart resumes.
+
+  This CSR is read/write.
+definedBy: Sdext
+fields:
+  DPC:
+    location_rv32: 31-0
+    location_rv64: 63-0
+    type: RW
+    description: Debug PC Value
+    reset_value: UNDEFINED_LEGAL

spec/std/isa/ext/Sdext.yaml

@@ -13,4 +13,53 @@ type: privileged
 versions:
   - version: "1.0.0"
     state: ratified
-    ratification_date: null
+    ratification_date: 2025-02
+params:
+  DCSR_MPRVEN_TYPE:
+    schema:
+      type: string
+      enum: [read-only-0, read-only-1, rw]
+    description: |
+      Implementation of dcsr.MPRVEN is optional.
+      It may be tied to either 0 or 1.
+
+      Behavior of the dcsr.MPRVEN bit:
+        * 'read-only-0': tied to 0
+        * 'read-only-1': tied to 1
+        * 'rw': read-write
+  DCSR_STEPIE_TYPE:
+    schema:
+      type: string
+      enum: [read-only-0, read-only-1, rw]
+    description: |
+      Implementation of dcsr.STEPIE is optional.
+      It may be tied to either 0 or 1.
+
+      Behavior of the dcsr.STEPIE bit:
+        * 'read-only-0': tied to 0
+        * 'read-only-1': tied to 1
+        * 'rw': read-write
+  DCSR_STOPCOUNT_TYPE:
+    schema:
+      type: string
+      enum: [read-only-0, read-only-1, rw]
+    description: |
+      Implementation of dcsr.STOPCOUNT is optional.
+      It may be tied to either 0 or 1.
+
+      Behavior of the dcsr.STOPCOUNT bit:
+        * 'read-only-0': tied to 0
+        * 'read-only-1': tied to 1
+        * 'rw': read-write
+  DCSR_STOPTIME_TYPE:
+    schema:
+      type: string
+      enum: [read-only-0, read-only-1, rw]
+    description: |
+      Implementation of dcsr.STOPTIME is optional.
+      It may be tied to either 0 or 1.
+
+      Behavior of the dcsr.STOPTIME bit:
+        * 'read-only-0': tied to 0
+        * 'read-only-1': tied to 1
+        * 'rw': read-write