From 95ac48d605dccab3917a87f1cf07130556ac6919 Mon Sep 17 00:00:00 2001
From: Ryan Houdek <Sonicadvance1@gmail.com>
Date: Sun, 18 Jan 2015 15:26:25 -0600
Subject: [PATCH] Improve the LLVM disassembler in the debug window.

There are a couple things in this PR.
Fixes a bug where if we hit an invalid instruction we would infinite loop.
Fixes an issue where on AArch64 it would show invalid instructions for all NEON instructions.
This was due to asimd and crc being optional extensions and LLVM not enabling them by default.
So we have to specify a CPU which has the feature. LLVM 3.6 will let us select by features instead of CPUs, but we don't have a release of that quite
yet.

If we are on an architecture that has a known instruction size, we will continue onward after hitting the invalid instruction. If we don't have a
known instruction size like on x86, we will instead just dump the rest of the block.
---
 Source/Core/DolphinWX/Debugger/JitWindow.cpp | 52 ++++++++++++++++----
 1 file changed, 43 insertions(+), 9 deletions(-)

diff --git a/Source/Core/DolphinWX/Debugger/JitWindow.cpp b/Source/Core/DolphinWX/Debugger/JitWindow.cpp
index 909a86e77c..94c4334aa2 100644
--- a/Source/Core/DolphinWX/Debugger/JitWindow.cpp
+++ b/Source/Core/DolphinWX/Debugger/JitWindow.cpp
@@ -49,7 +49,7 @@
 class HostDisassemblerLLVM : public HostDisassembler
 {
 public:
-	HostDisassemblerLLVM(const std::string host_disasm);
+	HostDisassemblerLLVM(const std::string host_disasm, int inst_size = -1, const std::string cpu = "");
 	~HostDisassemblerLLVM()
 	{
 		if (m_can_disasm)
@@ -59,22 +59,24 @@ public:
 private:
 	bool m_can_disasm;
 	LLVMDisasmContextRef m_llvm_context;
+	int m_instruction_size;
 
 	std::string DisassembleHostBlock(const u8* code_start, const u32 code_size, u32* host_instructions_count) override;
 };
 
-HostDisassemblerLLVM::HostDisassemblerLLVM(const std::string host_disasm)
-	: m_can_disasm(false)
+HostDisassemblerLLVM::HostDisassemblerLLVM(const std::string host_disasm, int inst_size, const std::string cpu)
+	: m_can_disasm(false), m_instruction_size(inst_size)
 {
 	LLVMInitializeAllTargetInfos();
 	LLVMInitializeAllTargetMCs();
 	LLVMInitializeAllDisassemblers();
 
-	m_llvm_context = LLVMCreateDisasm(host_disasm.c_str(), nullptr, 0, 0, nullptr);
+	m_llvm_context = LLVMCreateDisasmCPU(host_disasm.c_str(), cpu.c_str(), nullptr, 0, 0, nullptr);
 
 	// Couldn't create llvm context
 	if (!m_llvm_context)
 		return;
+
 	LLVMSetDisasmOptions(m_llvm_context,
 		LLVMDisassembler_Option_AsmPrinterVariant |
 		LLVMDisassembler_Option_PrintLatency);
@@ -87,15 +89,47 @@ std::string HostDisassemblerLLVM::DisassembleHostBlock(const u8* code_start, con
 	if (!m_can_disasm)
 		return "(No LLVM context)";
 
-	u64 disasmPtr = (u64)code_start;
+	u8* disasmPtr = (u8*)code_start;
 	const u8 *end = code_start + code_size;
 
 	std::ostringstream x86_disasm;
 	while ((u8*)disasmPtr < end)
 	{
 		char inst_disasm[256];
-		disasmPtr += LLVMDisasmInstruction(m_llvm_context, (u8*)disasmPtr, (u64)(end - disasmPtr), (u64)disasmPtr, inst_disasm, 256);
-		x86_disasm << inst_disasm << std::endl;
+		size_t inst_size = LLVMDisasmInstruction(m_llvm_context, disasmPtr, (u64)(end - disasmPtr), (u64)disasmPtr, inst_disasm, 256);
+		if (!inst_size)
+		{
+			x86_disasm << "Invalid inst:";
+
+			if (m_instruction_size != -1)
+			{
+				// If we are on an architecture that has a fixed instruction size
+				// We can continue onward past this bad instruction.
+				std::string inst_str = "";
+				for (int i = 0; i < m_instruction_size; ++i)
+					inst_str += StringFromFormat("%02x", disasmPtr[i]);
+
+				x86_disasm << inst_str << std::endl;
+				disasmPtr += m_instruction_size;
+			}
+			else
+			{
+				// We can't continue if we are on an architecture that has flexible instruction sizes
+				// Dump the rest of the block instead
+				std::string code_block = "";
+				for (int i = 0; (disasmPtr + i) < end; ++i)
+					code_block += StringFromFormat("%02x", disasmPtr[i]);
+
+				x86_disasm << code_block << std::endl;
+				break;
+			}
+		}
+		else
+		{
+			x86_disasm << inst_disasm << std::endl;
+			disasmPtr += inst_size;
+		}
+
 		(*host_instructions_count)++;
 	}
 
@@ -199,9 +233,9 @@ CJitWindow::CJitWindow(wxWindow* parent, wxWindowID id, const wxPoint& pos,
 #elif defined(_M_X86)
 	m_disassembler.reset(new HostDisassemblerX86());
 #elif defined(_M_ARM_64) && defined(HAS_LLVM)
-	m_disassembler.reset(new HostDisassemblerLLVM("aarch64-none-unknown"));
+	m_disassembler.reset(new HostDisassemblerLLVM("aarch64-none-unknown", 4, "cortex-a57"));
 #elif defined(_M_ARM_32) && defined(HAS_LLVM)
-	m_disassembler.reset(new HostDisassemblerLLVM("armv7-none-unknown"));
+	m_disassembler.reset(new HostDisassemblerLLVM("armv7-none-unknown", 4, "cortex-a15"));
 #else
 	m_disassembler.reset(new HostDisassembler());
 #endif