Basic virtual memory implementation #163

nemecad · 2025-08-01T15:58:42Z

This PR implements basic virtual memory support:

Set-associative TLB (machine::TLB): Implements a set-associative Translation Lookaside Buffer (TLB) frontend over physical memory, handling virtual to physical translation, flush, and replacement policy.
Pluggable Replacement Policies (machine::TLBPolicy): Abstract TLB replacement policy interface & implementations (RAND, LRU, LFU, PLRU) for set-associative tables.
SV32 Page-Table Walker (machine::PageTableWalker): Performs multi-level page-table walks (SV32) in memory to resolve a virtual address to a physical one.
Sv32Pte Bitfield Helpers (sv32.h): SV32-specific definitions: page-table entry (PTE) bitfields, shifts/masks, and PTE to physical address helpers.
VirtualAddress (virtual_address.h): Lightweight VirtualAddress wrapper offering raw access, alignment checks, arithmetic, and comparisons.
Page Fault Handling (machine::PageFaultHandler): Catches page-fault exceptions, invokes page allocation, and recovers execution.
Qt “BigSlider” Widget (BigSlider): Qt widget subclass exposing 64-bit slider range/value designed for selecting a base virtual memory address from 0x0000_0000 up to 0xFFFF_F000.

UI Components:

Configuration Dialog:

Memory window: “Show virtual” checkbox to toggle between physical and virtual address spaces.

…age‑table walker)

jdupak

Overall this is pretty good work! Thank you

Most of my comments are style nits, so don't worry to much :) Btw, please make sure the the code is properly formated by clangformat that with config that is in the root of the project.

There are some cases I think we should move the components around to make this better encapslated.

There is one major decision that we need to make regarding the use of virtual memory. I think the current compomise is not great.

jdupak · 2025-08-16T11:21:39Z

CMakeLists.txt.user

Please delete the accidentally committed files.

jdupak · 2025-08-16T11:26:20Z

.gitignore

+/cmake-build-debug-wsl/
+/cmake-build-release-wsl/


The way I do this on my machine is that I configure the files to be build/Debug and build/Release. You can also ignore thise files localy. It should not be here.

jdupak · 2025-08-16T11:31:15Z

src/assembler/simpleasm.h

@@ -79,6 +80,7 @@ class SimpleAsm : public QObject {
 private:
    QStringList include_stack;
    machine::FrontendMemory *mem {};
+    machine::MachineConfig config {};
    machine::RelocExpressionList reloc;
 };


I think this should be a pointer - it feels weirt to have multiple copies of machine config in the program.

jdupak · 2025-08-16T11:32:14Z

src/assembler/simpleasm.cpp

@@ -286,7 +288,12 @@ bool SimpleAsm::process_line(
            if (error_ptr != nullptr) { *error_ptr = error; }
            return false;
        }
-        address = machine::Address(value);
+        qint64 offset = qint64(value);
+        qint64 base = 0;


Nit: please prefer standard cpp types over qt ones whenever possible.

jdupak · 2025-08-16T11:34:57Z

src/assembler/simpleasm.cpp

+        if (config.get_vm_enabled() && config.get_vm_mode() == machine::MachineConfig::VM_SV32) {
+            base = config.get_kernel_virt_base();
+        }
+        address = machine::Address(base + offset);


Now looking at this, wouldn't it be simpler to just pass the base to the assembler constructor and keep this logic outside?

jdupak · 2025-08-16T14:36:59Z

src/machine/machine.h

    aclint::AclintSswi *aclint_sswi = nullptr;
+    std::optional<TLB> tlb_program;
+    std::optional<TLB> tlb_data;
+    FrontendMemory *instr_if_;


This will require a deeper design decision @ppisa

I see 2 ways to go here:

We go all in on virtual addresses. We say that the CPU always operates with virtual memory. For CPU without MMU we make the mapping 1:1. This means that the memory interface will always be an instance of TLB (possibly trivial).

We give up on virtuall addresses, we just make TLB just another instance of FrontendMemory. We use the same type for both virtual and physical address.

This code is somewhere in the middle which IMO does not bring any benefit of special type but brings all the cons. This one is on me, I should have explained it better.

jdupak · 2025-08-16T14:46:21Z

src/machine/memory/virtual/page_fault_handler.cpp

+    return false;
+}
+
+void PageFaultHandler::perform_page_allocation(const VirtualAddress &va, uint32_t raw_satp) {


@ppisa I am wondering if this should rather be in osemu.

jdupak · 2025-08-16T14:49:56Z

src/gui/windows/program/programmodel.cpp

    if (machine == nullptr) { return nullptr; }
-    if (machine->memory_data_bus() != nullptr) { return machine->memory_data_bus(); }
+    return machine->instr_frontend();


This is wrong, thos never used to point to cache.

jdupak · 2025-08-16T14:51:22Z

src/gui/windows/program/programmodel.cpp

    this->machine = machine;
+    if (machine && machine->config().get_vm_enabled()
+        && machine->config().get_vm_mode() == machine::MachineConfig::VM_SV32) {
+        index0_offset = machine::Address(machine->config().get_va_base_addr());


I think this place is a really good ilustration why separate virtual memory type will increase clarity a lot.

jdupak · 2025-08-16T14:53:19Z

src/gui/windows/memory/memorymodel.cpp

-        if (mem == nullptr) { return QString(""); }
-        if ((access_through_cache > 0) && (machine->cache_data() != nullptr)) {


We might need this for proper startup/reset,

jdupak · 2025-08-16T15:01:38Z

Also, please consider that the next person reading this code will not be an expert on MMU like you, so try to add comments and use less cryptic names whenever possible (don't be afraid of long names too much).

ppisa · 2025-08-17T21:35:43Z

I did send direct e-mail to @nemecad and @jdupak at 2025-08-08. It has been mainly about generated content in the commits and some remarks for clarification etc. I hope that you have received it. I plan ti provide review there when the technical/format problems are resolved and updated series is posted.

jdupak · 2025-08-18T10:28:56Z

I did send direct e-mail to @nemecad and @jdupak at 2025-08-08. It has been mainly about generated content in the commits and some remarks for clarification etc. I hope that you have received it. I plan ti provide review there when the technical/format problems are resolved and updated series is posted.

I did get it, but the generated files did not cause too much trouble for me to review this.

petr-stepan · 2025-08-20T15:31:52Z

Nice work, I have problem to cmake this project. I had to change CMakeList.txt in machine directory (remove memory and some others tests). Then in cli main.cpp, there is TODO at line 500, than was necessary to comment.
Finally, I run qtrvsim_gui and started virtual memory. In dialog number of sets should be only power of 2. The virtual memory base address input is not working. If I tried to insert concrete number, then there was error.

nemecad added 4 commits July 21, 2025 14:01

feat(vm): implement basic SV32 virtual memory infrastructure (TLB + p…

ef495a6

…age‑table walker)

feat(vm): add basic VM GUI

dd735fd

feat(vm): improve TLB and virtual memory infrastructure

b29b13f

Ignore build outputs and untrack large binaries

8c23cd8

jdupak reviewed Aug 16, 2025

View reviewed changes

		if (mem == nullptr) { return QString(""); }
		if ((access_through_cache > 0) && (machine->cache_data() != nullptr)) {

		/cmake-build-debug-wsl/
		/cmake-build-release-wsl/

Basic virtual memory implementation #163

Are you sure you want to change the base?

Basic virtual memory implementation #163

Uh oh!

Conversation

nemecad commented Aug 1, 2025

Uh oh!

jdupak left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jdupak commented Aug 16, 2025

Uh oh!

ppisa commented Aug 17, 2025

Uh oh!

jdupak commented Aug 18, 2025

Uh oh!

petr-stepan commented Aug 20, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants