Merge pull request #377 from PCSX2/tlb_goemon_v2

goemon gamefix improvement

pcsx2/DebugTools/DisR5900asm.cpp

@@ -25,6 +25,10 @@
 #include "DisASM.h"
 #include "R5900OpcodeTables.h"
 
+// Allow to print register content when you print dissassembler info
+// Note only a subset of the opcodes are supported. It is intended as a cheap debugger
+//#define PRINT_REG_CONTENT
+
 unsigned long opcode_addr;
 u32 disasmOpcode;
 bool disSimplify;
@@ -710,16 +714,46 @@ const char* signedImmediate(s32 imm, int len = 0)
 
 const char* disDestSource(int dest, int source)
 {
-	static char buffer[32];
+	static char buffer[64];
-
+#ifdef PRINT_REG_CONTENT
+	sprintf(buffer,"%s,%s(0x%8.8x)",GPR_REG[dest],GPR_REG[source], cpuRegs.GPR.r[source].UL[0]);
+#else
 	if (disSimplify && dest == source)
 		sprintf(buffer,"%s",GPR_REG[dest]);
 	else
 		sprintf(buffer,"%s,%s",GPR_REG[dest],GPR_REG[source]);
 
+#endif
+
 	return buffer;
 }
 
+void disBranch(std::string& output, const char* op)
+{
+	ssappendf(output, "%s\t", op);
+	offset_decode(output);
+}
+
+void disBranch(std::string& output, const char* op, int rs)
+{
+#ifdef PRINT_REG_CONTENT
+	ssappendf(output, "%s\t%s(0x%8.8x), ", op, GPR_REG[rs], cpuRegs.GPR.r[rs].UL[0]);
+#else
+	ssappendf(output, "%s\t%s, ", op, GPR_REG[rs]);
+#endif
+	offset_decode(output);
+}
+
+void disBranch(std::string& output, const char* op, int rs, int rt)
+{
+#ifdef PRINT_REG_CONTENT
+	ssappendf(output, "%s\t%s(0x%8.8x), %s(0x%8.8x), ", op, GPR_REG[rs], cpuRegs.GPR.r[rs].UL[0], GPR_REG[rt], cpuRegs.GPR.r[rt].UL[0]);
+#else
+	ssappendf(output, "%s\t%s, %s, ", op, GPR_REG[rs], GPR_REG[rt]);
+#endif
+	offset_decode(output);
+}
+
 //********************* Standard Opcodes***********************
 void J( std::string& output )      { output += "j\t";        jump_decode(output);}
 void JAL( std::string& output )    { output += "jal\t";      jump_decode(output);}
@@ -730,15 +764,13 @@ void BEQ( std::string& output )
 	int rt = DECODE_RT;
 
 	if (disSimplify && rs == rt)
-		ssappendf(output, "b\t");
+		disBranch(output, "b");
 	else if (disSimplify && rs == 0 && rt != 0)
-		ssappendf(output, "beqz\t%s, ",GPR_REG[rt]);
+		disBranch(output, "beqz", rt);
 	else if (disSimplify && rs != 0 && rt == 0)
-		ssappendf(output, "beqz\t%s, ",GPR_REG[rs]);
+		disBranch(output, "beqz", rs);
 	else
-		ssappendf(output, "beq\t%s, %s, ",GPR_REG[rs], GPR_REG[rt]);
+		disBranch(output, "beq", rs, rt);
-
-	offset_decode(output);
 }
 
 void BNE( std::string& output )
@@ -747,17 +779,15 @@ void BNE( std::string& output )
 	int rt = DECODE_RT;
 
 	if (disSimplify && rs == 0 && rt != 0)
-		ssappendf(output, "bnez\t%s, ",GPR_REG[rt]);
+		disBranch(output, "bnez", rt);
 	else if (disSimplify && rs != 0 && rt == 0)
-		ssappendf(output, "bnez\t%s, ",GPR_REG[rs]);
+		disBranch(output, "bnez", rs);
 	else
-		ssappendf(output, "bne\t%s, %s, ",GPR_REG[rs], GPR_REG[rt]);
+		disBranch(output, "bne", rt, rs);
-
-	offset_decode(output);
 }
 
-void BLEZ( std::string& output )   { _sap("blez\t%s, ")             GPR_REG[DECODE_RS]); offset_decode(output); }
+void BLEZ( std::string& output )   { disBranch(output, "blez", DECODE_RS); }
-void BGTZ( std::string& output )   { _sap("bgtz\t%s, ")             GPR_REG[DECODE_RS]); offset_decode(output); }
+void BGTZ( std::string& output )   { disBranch(output, "bgtz", DECODE_RS); }
 void ADDI( std::string& output )   { _sap("addi\t%s, %s, 0x%04X")   GPR_REG[DECODE_RT], GPR_REG[DECODE_RS], DECODE_IMMED);}
 
 void ADDIU( std::string& output )
@@ -798,15 +828,13 @@ void BEQL( std::string& output )
 	int rt = DECODE_RT;
 
 	if (disSimplify && rs == rt)
-		ssappendf(output, "b\t");
+		disBranch(output, "bl");
 	else if (disSimplify && rs == 0 && rt != 0)
-		ssappendf(output, "beqzl\t%s, ",GPR_REG[rt]);
+		disBranch(output, "beqzl", rt);
 	else if (disSimplify && rs != 0 && rt == 0)
-		ssappendf(output, "beqzl\t%s, ",GPR_REG[rs]);
+		disBranch(output, "beqzl", rs);
 	else
-		ssappendf(output, "beql\t%s, %s, ",GPR_REG[rs], GPR_REG[rt]);
+		disBranch(output, "beql", rs, rt);
-
-	offset_decode(output);
 }
 
 void BNEL( std::string& output )
@@ -815,17 +843,15 @@ void BNEL( std::string& output )
 	int rt = DECODE_RT;
 
 	if (disSimplify && rs == 0 && rt != 0)
-		ssappendf(output, "bnezl\t%s, ",GPR_REG[rt]);
+		disBranch(output, "bnezl", rt);
 	else if (disSimplify && rs != 0 && rt == 0)
-		ssappendf(output, "bnezl\t%s, ",GPR_REG[rs]);
+		disBranch(output, "bnezl", rs);
 	else
-		ssappendf(output, "bnel\t%s, %s, ",GPR_REG[rs], GPR_REG[rt]);
+		disBranch(output, "bnel", rt, rs);
-
-	offset_decode(output);
 }
 
-void BLEZL( std::string& output )  { _sap("blezl\t%s, ")            GPR_REG[DECODE_RS]); offset_decode(output); }
+void BLEZL( std::string& output )  { disBranch(output, "blezl", DECODE_RS); }
-void BGTZL( std::string& output )  { _sap("bgtzl\t%s, ")            GPR_REG[DECODE_RS]); offset_decode(output); }
+void BGTZL( std::string& output )  { disBranch(output, "bgtzl", DECODE_RS); }
 void DADDI( std::string& output )  { _sap("daddi\t%s, 0x%04X")      disDestSource(DECODE_RT, DECODE_RS), DECODE_IMMED); }
 void DADDIU( std::string& output ) { _sap("daddiu\t%s, 0x%04X")     disDestSource(DECODE_RT, DECODE_RS), DECODE_IMMED); }
 
@@ -994,20 +1020,20 @@ void SYSCALL( std::string& output ) { output +="syscall ---";/*_sap("syscall\t0x
 void BREAK( std::string& output )   { output += "break   ---";/*_sap("break\t0x%05X")     DECODE_BREAK); */}
 void CACHE( std::string& output )   { output += "cache   ---";/*_sap("cache\t%s, 0x%04X(%s)")  GPR_REG[DECODE_RT], DECODE_IMMED, GPR_REG[DECODE_RS]); */}
 //************************REGIMM OPCODES***************************
-void BLTZ( std::string& output )    { _sap("bltz\t%s, ")       GPR_REG[DECODE_RS]); offset_decode(output); }
+void BLTZ( std::string& output )    { disBranch(output, "bltz", DECODE_RS); }
-void BGEZ( std::string& output )    { _sap("bgez\t%s, ")       GPR_REG[DECODE_RS]); offset_decode(output); }
+void BGEZ( std::string& output )    { disBranch(output, "bgez", DECODE_RS); }
-void BLTZL( std::string& output )   { _sap("bltzl\t%s, ")      GPR_REG[DECODE_RS]); offset_decode(output); }
+void BLTZL( std::string& output )   { disBranch(output, "bltzl", DECODE_RS); }
-void BGEZL( std::string& output )   { _sap("bgezl\t%s, ")      GPR_REG[DECODE_RS]); offset_decode(output); }
+void BGEZL( std::string& output )   { disBranch(output, "bgezl", DECODE_RS); }
 void TGEI( std::string& output )    { _sap("tgei\t%s, 0x%04X") GPR_REG[DECODE_RS], DECODE_IMMED); }
 void TGEIU( std::string& output )   { _sap("tgeiu\t%s,0x%04X") GPR_REG[DECODE_RS], DECODE_IMMED); }
 void TLTI( std::string& output )    { _sap("tlti\t%s, 0x%04X") GPR_REG[DECODE_RS], DECODE_IMMED); }
 void TLTIU( std::string& output )   { _sap("tltiu\t%s,0x%04X") GPR_REG[DECODE_RS], DECODE_IMMED); }
 void TEQI( std::string& output )    { _sap("teqi\t%s, 0x%04X") GPR_REG[DECODE_RS], DECODE_IMMED); }
 void TNEI( std::string& output )    { _sap("tnei\t%s, 0x%04X") GPR_REG[DECODE_RS], DECODE_IMMED); }
-void BLTZAL( std::string& output )  { _sap("bltzal\t%s, ")     GPR_REG[DECODE_RS]); offset_decode(output); }
+void BLTZAL( std::string& output )  { disBranch(output, "bltzal", DECODE_RS); }
-void BGEZAL( std::string& output )  { _sap("bgezal\t%s, ")     GPR_REG[DECODE_RS]); offset_decode(output); }
+void BGEZAL( std::string& output )  { disBranch(output, "bgezal", DECODE_RS); }
-void BLTZALL( std::string& output ) { _sap("bltzall\t%s, ")    GPR_REG[DECODE_RS]); offset_decode(output); }
+void BLTZALL( std::string& output ) { disBranch(output, "bltzall", DECODE_RS); }
-void BGEZALL( std::string& output ) { _sap("bgezall\t%s, ")    GPR_REG[DECODE_RS]); offset_decode(output); }
+void BGEZALL( std::string& output ) { disBranch(output, "bgezall", DECODE_RS); }
 void MTSAB( std::string& output )   { _sap("mtsab\t%s, 0x%04X") GPR_REG[DECODE_RS], DECODE_IMMED);}
 void MTSAH( std::string& output )   { _sap("mtsah\t%s, 0x%04X") GPR_REG[DECODE_RS], DECODE_IMMED);}
 

pcsx2/Interpreter.cpp

@@ -170,7 +170,7 @@ static void execI()
 	// Based on cycle
 	// if( cpuRegs.cycle > 0x4f24d714 )
 	// Or dump from a particular PC (useful to debug handler/syscall)
-	if (cpuRegs.pc == 0x80000000) {
+	if (pc == 0x80000000) {
 		print_me = 2000;
 	}
 	if (print_me) {
@@ -254,6 +254,11 @@ void J()
 
 void JAL()
 {
+	// 0x3563b8 is the start address of the function that invalidate entry in TLB cache
+	if (EmuConfig.Gamefixes.GoemonTlbHack) {
+		if (_JumpTarget_ == 0x3563b8)
+			GoemonUnloadTlb(cpuRegs.GPR.n.a0.UL[0]);
+	}
 	_SetLink(31);
 	doBranch(_JumpTarget_);
 }
@@ -455,9 +460,11 @@ void BGEZALL()   // Branch if Rs >= 0 and link
 *********************************************************/
 void JR()
 {
-	// 0x33ad48 is the return address of the function that populate the TLB cache
+	// 0x33ad48 and 0x35060c are the return address of the function (0x356250) that populate the TLB cache
-	if (cpuRegs.GPR.r[_Rs_].UL[0] == 0x33ad48 && EmuConfig.Gamefixes.GoemonTlbHack) {
+	if (EmuConfig.Gamefixes.GoemonTlbHack) {
-		GoemonPreloadTlb();
+		u32 add = cpuRegs.GPR.r[_Rs_].UL[0];
+		if (add == 0x33ad48 || add == 0x35060c)
+			GoemonPreloadTlb();
 	}
 	doBranch(cpuRegs.GPR.r[_Rs_].UL[0]);
 }

pcsx2/R5900.h

@@ -430,6 +430,7 @@ extern void cpuTlbMissW(u32 addr, u32 bd);
 extern void cpuTestHwInts();
 extern void cpuClearInt(uint n);
 extern void __fastcall GoemonPreloadTlb();
+extern void __fastcall GoemonUnloadTlb(u32 key);
 
 extern void cpuSetNextEvent( u32 startCycle, s32 delta );
 extern void cpuSetNextEventDelta( s32 delta );

pcsx2/vtlb.cpp

@@ -339,7 +339,9 @@ static void GoemonTlbMissDebug()
 
 	for (u32 i = 0; i < 150; i++) {
 		if (tlb[i].valid == 0x1 && tlb[i].low_add != tlb[i].high_add)
-			DevCon.WriteLn("Entry %d is valid. From V:0x%8.8x to V:0x%8.8x (P:0x%8.8x)", i, tlb[i].low_add, tlb[i].high_add, tlb[i].physical_add);
+			DevCon.WriteLn("GoemonTlbMissDebug: Entry %d is valid. Key %x. From V:0x%8.8x to V:0x%8.8x (P:0x%8.8x)", i, tlb[i].key, tlb[i].low_add, tlb[i].high_add, tlb[i].physical_add);
+		else if (tlb[i].low_add != tlb[i].high_add)
+			DevCon.WriteLn("GoemonTlbMissDebug: Entry %d is invalid. Key %x. From V:0x%8.8x to V:0x%8.8x (P:0x%8.8x)", i, tlb[i].key, tlb[i].low_add, tlb[i].high_add, tlb[i].physical_add);
 	}
 }
 
@@ -356,7 +358,7 @@ void __fastcall GoemonPreloadTlb()
 			u32 paddr = tlb[i].physical_add;
 
 			if ((uptr)vtlbdata.vmap[vaddr>>VTLB_PAGE_BITS] == POINTER_SIGN_BIT) {
-				DevCon.WriteLn("Preload TLB[%d]: From V:0x%8.8x to P:0x%8.8x (%d pages)", i, vaddr, paddr, size >> VTLB_PAGE_BITS);
+				DevCon.WriteLn("GoemonPreloadTlb: Entry %d. Key %x. From V:0x%8.8x to P:0x%8.8x (%d pages)", i, tlb[i].key, vaddr, paddr, size >> VTLB_PAGE_BITS);
 				vtlb_VMap(           vaddr , paddr, size);
 				vtlb_VMap(0x20000000|vaddr , paddr, size);
 			}
@@ -364,6 +366,33 @@ void __fastcall GoemonPreloadTlb()
 	}
 }
 
+void __fastcall GoemonUnloadTlb(u32 key)
+{
+	// 0x3d5580 is the address of the TLB cache table
+	GoemonTlb* tlb = (GoemonTlb*)&eeMem->Main[0x3d5580];
+	for (u32 i = 0; i < 150; i++) {
+		if (tlb[i].key == key) {
+			if (tlb[i].valid == 0x1) {
+				u32 size  = tlb[i].high_add - tlb[i].low_add;
+				u32 vaddr = tlb[i].low_add;
+				DevCon.WriteLn("GoemonUnloadTlb: Entry %d. Key %x. From V:0x%8.8x to V:0x%8.8x (%d pages)", i, tlb[i].key, vaddr, vaddr+size, size >> VTLB_PAGE_BITS);
+
+				vtlb_VMapUnmap(           vaddr , size);
+				vtlb_VMapUnmap(0x20000000|vaddr , size);
+
+				// Unmap the tlb in game cache table
+				// Note: Game copy FEFEFEFE for others data
+				tlb[i].valid    = 0;
+				tlb[i].key      = 0xFEFEFEFE;
+				tlb[i].low_add  = 0xFEFEFEFE;
+				tlb[i].high_add = 0xFEFEFEFE;
+			} else {
+				DevCon.Error("GoemonUnloadTlb: Entry %d is not valid. Key %x", i, tlb[i].key);
+			}
+		}
+	}
+}
+
 // Generates a tlbMiss Exception
 static __ri void vtlb_Miss(u32 addr,u32 mode)
 {

pcsx2/vtlb.h

@@ -225,6 +225,6 @@ struct GoemonTlb {
 	u32 physical_add;
 	u32 unk3; // likely the size
 	u32 high_add;
-	u32 unk4;
+	u32 key; // uniq number attached to an allocation
 	u32 unk5;
 };

pcsx2/x86/ix86-32/iR5900-32.cpp

@@ -1739,9 +1739,17 @@ static void __fastcall recRecompile( const u32 startpc )
 		xCALL(PreBlockCheck);
 	}
 
-	// 0x33ad48 is the return address of the function that populate the TLB cache
+	if (EmuConfig.Gamefixes.GoemonTlbHack) {
-	if (pc == 0x33ad48 && EmuConfig.Gamefixes.GoemonTlbHack) {
+		if (pc == 0x33ad48 || pc == 0x35060c) {
-		xCALL(GoemonPreloadTlb);
+			// 0x33ad48 and 0x35060c are the return address of the function (0x356250) that populate the TLB cache
+			xCALL(GoemonPreloadTlb);
+		} else if (pc == 0x3563b8) {
+			// Game will unmap some virtual addresses. If a constant address were hardcoded in the block, we would be in a bad situation.
+			AtomicExchange( eeRecNeedsReset, true );
+			// 0x3563b8 is the start address of the function that invalidate entry in TLB cache
+			MOV32MtoR(ECX, (int)&cpuRegs.GPR.n.a0.UL[ 0 ] );
+			xCALL(GoemonUnloadTlb);
+		}
 	}
 
 	// go until the next branch

plugins/GSdx/GSClut.cpp

@@ -109,7 +109,7 @@ bool GSClut::WriteTest(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT)
 	case 4: if(m_CBP[0] == TEX0.CBP) return false; m_CBP[0] = TEX0.CBP; break;
 	case 5: if(m_CBP[1] == TEX0.CBP) return false; m_CBP[1] = TEX0.CBP; break;
 	case 6: ASSERT(0); return false; // ffx2 menu
-	case 7: ASSERT(0); return false; // ford mustang racing
+	case 7: ASSERT(0); return false; // ford mustang racing // Bouken Jidai Katsugeki Goemon
 	default: __assume(0);
 	}
 

plugins/GSdx/GSState.cpp

@@ -851,7 +851,7 @@ template<int i> void GSState::GIFRegHandlerTEX0(const GIFReg* RESTRICT r)
 
 	if((TEX0.TBW & 1) && (TEX0.PSM == PSM_PSMT8 || TEX0.PSM == PSM_PSMT4))
 	{
-		ASSERT(TEX0.TBW == 1); // TODO
+		ASSERT(TEX0.TBW == 1); // TODO // Bouken Jidai Katsugeki Goemon
 
 		TEX0.TBW &= ~1; // GS User 2.6
 	}