1277323Sdim//===-- MCExternalSymbolizer.cpp - External symbolizer --------------------===// 2277323Sdim// 3353358Sdim// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4353358Sdim// See https://llvm.org/LICENSE.txt for license information. 5353358Sdim// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6277323Sdim// 7277323Sdim//===----------------------------------------------------------------------===// 8277323Sdim 9309124Sdim#include "llvm/MC/MCDisassembler/MCExternalSymbolizer.h" 10277323Sdim#include "llvm/MC/MCContext.h" 11277323Sdim#include "llvm/MC/MCExpr.h" 12277323Sdim#include "llvm/MC/MCInst.h" 13277323Sdim#include "llvm/Support/raw_ostream.h" 14277323Sdim#include <cstring> 15277323Sdim 16277323Sdimusing namespace llvm; 17277323Sdim 18288943Sdimnamespace llvm { 19288943Sdimclass Triple; 20288943Sdim} 21288943Sdim 22277323Sdim// This function tries to add a symbolic operand in place of the immediate 23277323Sdim// Value in the MCInst. The immediate Value has had any PC adjustment made by 24277323Sdim// the caller. If the instruction is a branch instruction then IsBranch is true, 25277323Sdim// else false. If the getOpInfo() function was set as part of the 26277323Sdim// setupForSymbolicDisassembly() call then that function is called to get any 27277323Sdim// symbolic information at the Address for this instruction. If that returns 28277323Sdim// non-zero then the symbolic information it returns is used to create an MCExpr 29277323Sdim// and that is added as an operand to the MCInst. If getOpInfo() returns zero 30277323Sdim// and IsBranch is true then a symbol look up for Value is done and if a symbol 31277323Sdim// is found an MCExpr is created with that, else an MCExpr with Value is 32277323Sdim// created. This function returns true if it adds an operand to the MCInst and 33277323Sdim// false otherwise. 34277323Sdimbool MCExternalSymbolizer::tryAddingSymbolicOperand(MCInst &MI, 35277323Sdim raw_ostream &cStream, 36277323Sdim int64_t Value, 37277323Sdim uint64_t Address, 38277323Sdim bool IsBranch, 39277323Sdim uint64_t Offset, 40277323Sdim uint64_t InstSize) { 41277323Sdim struct LLVMOpInfo1 SymbolicOp; 42277323Sdim std::memset(&SymbolicOp, '\0', sizeof(struct LLVMOpInfo1)); 43277323Sdim SymbolicOp.Value = Value; 44277323Sdim 45277323Sdim if (!GetOpInfo || 46277323Sdim !GetOpInfo(DisInfo, Address, Offset, InstSize, 1, &SymbolicOp)) { 47277323Sdim // Clear SymbolicOp.Value from above and also all other fields. 48277323Sdim std::memset(&SymbolicOp, '\0', sizeof(struct LLVMOpInfo1)); 49277323Sdim 50277323Sdim // At this point, GetOpInfo() did not find any relocation information about 51277323Sdim // this operand and we are left to use the SymbolLookUp() call back to guess 52277323Sdim // if the Value is the address of a symbol. In the case this is a branch 53277323Sdim // that always makes sense to guess. But in the case of an immediate it is 54277323Sdim // a bit more questionable if it is an address of a symbol or some other 55277323Sdim // reference. So if the immediate Value comes from a width of 1 byte, 56277323Sdim // InstSize, we will not guess it is an address of a symbol. Because in 57277323Sdim // object files assembled starting at address 0 this usually leads to 58277323Sdim // incorrect symbolication. 59277323Sdim if (!SymbolLookUp || (InstSize == 1 && !IsBranch)) 60277323Sdim return false; 61277323Sdim 62277323Sdim uint64_t ReferenceType; 63277323Sdim if (IsBranch) 64277323Sdim ReferenceType = LLVMDisassembler_ReferenceType_In_Branch; 65277323Sdim else 66277323Sdim ReferenceType = LLVMDisassembler_ReferenceType_InOut_None; 67277323Sdim const char *ReferenceName; 68277323Sdim const char *Name = SymbolLookUp(DisInfo, Value, &ReferenceType, Address, 69277323Sdim &ReferenceName); 70277323Sdim if (Name) { 71277323Sdim SymbolicOp.AddSymbol.Name = Name; 72277323Sdim SymbolicOp.AddSymbol.Present = true; 73277323Sdim // If Name is a C++ symbol name put the human readable name in a comment. 74277323Sdim if(ReferenceType == LLVMDisassembler_ReferenceType_DeMangled_Name) 75277323Sdim cStream << ReferenceName; 76277323Sdim } 77277323Sdim // For branches always create an MCExpr so it gets printed as hex address. 78277323Sdim else if (IsBranch) { 79277323Sdim SymbolicOp.Value = Value; 80277323Sdim } 81277323Sdim if(ReferenceType == LLVMDisassembler_ReferenceType_Out_SymbolStub) 82277323Sdim cStream << "symbol stub for: " << ReferenceName; 83277323Sdim else if(ReferenceType == LLVMDisassembler_ReferenceType_Out_Objc_Message) 84277323Sdim cStream << "Objc message: " << ReferenceName; 85277323Sdim if (!Name && !IsBranch) 86277323Sdim return false; 87277323Sdim } 88277323Sdim 89277323Sdim const MCExpr *Add = nullptr; 90277323Sdim if (SymbolicOp.AddSymbol.Present) { 91277323Sdim if (SymbolicOp.AddSymbol.Name) { 92277323Sdim StringRef Name(SymbolicOp.AddSymbol.Name); 93288943Sdim MCSymbol *Sym = Ctx.getOrCreateSymbol(Name); 94288943Sdim Add = MCSymbolRefExpr::create(Sym, Ctx); 95277323Sdim } else { 96288943Sdim Add = MCConstantExpr::create((int)SymbolicOp.AddSymbol.Value, Ctx); 97277323Sdim } 98277323Sdim } 99277323Sdim 100277323Sdim const MCExpr *Sub = nullptr; 101277323Sdim if (SymbolicOp.SubtractSymbol.Present) { 102277323Sdim if (SymbolicOp.SubtractSymbol.Name) { 103277323Sdim StringRef Name(SymbolicOp.SubtractSymbol.Name); 104288943Sdim MCSymbol *Sym = Ctx.getOrCreateSymbol(Name); 105288943Sdim Sub = MCSymbolRefExpr::create(Sym, Ctx); 106277323Sdim } else { 107288943Sdim Sub = MCConstantExpr::create((int)SymbolicOp.SubtractSymbol.Value, Ctx); 108277323Sdim } 109277323Sdim } 110277323Sdim 111277323Sdim const MCExpr *Off = nullptr; 112277323Sdim if (SymbolicOp.Value != 0) 113288943Sdim Off = MCConstantExpr::create(SymbolicOp.Value, Ctx); 114277323Sdim 115277323Sdim const MCExpr *Expr; 116277323Sdim if (Sub) { 117277323Sdim const MCExpr *LHS; 118277323Sdim if (Add) 119288943Sdim LHS = MCBinaryExpr::createSub(Add, Sub, Ctx); 120277323Sdim else 121288943Sdim LHS = MCUnaryExpr::createMinus(Sub, Ctx); 122277323Sdim if (Off) 123288943Sdim Expr = MCBinaryExpr::createAdd(LHS, Off, Ctx); 124277323Sdim else 125277323Sdim Expr = LHS; 126277323Sdim } else if (Add) { 127277323Sdim if (Off) 128288943Sdim Expr = MCBinaryExpr::createAdd(Add, Off, Ctx); 129277323Sdim else 130277323Sdim Expr = Add; 131277323Sdim } else { 132277323Sdim if (Off) 133277323Sdim Expr = Off; 134277323Sdim else 135288943Sdim Expr = MCConstantExpr::create(0, Ctx); 136277323Sdim } 137277323Sdim 138277323Sdim Expr = RelInfo->createExprForCAPIVariantKind(Expr, SymbolicOp.VariantKind); 139277323Sdim if (!Expr) 140277323Sdim return false; 141277323Sdim 142288943Sdim MI.addOperand(MCOperand::createExpr(Expr)); 143277323Sdim return true; 144277323Sdim} 145277323Sdim 146277323Sdim// This function tries to add a comment as to what is being referenced by a load 147277323Sdim// instruction with the base register that is the Pc. These can often be values 148277323Sdim// in a literal pool near the Address of the instruction. The Address of the 149277323Sdim// instruction and its immediate Value are used as a possible literal pool entry. 150277323Sdim// The SymbolLookUp call back will return the name of a symbol referenced by the 151277323Sdim// literal pool's entry if the referenced address is that of a symbol. Or it 152277323Sdim// will return a pointer to a literal 'C' string if the referenced address of 153277323Sdim// the literal pool's entry is an address into a section with C string literals. 154277323Sdim// Or if the reference is to an Objective-C data structure it will return a 155277323Sdim// specific reference type for it and a string. 156277323Sdimvoid MCExternalSymbolizer::tryAddingPcLoadReferenceComment(raw_ostream &cStream, 157277323Sdim int64_t Value, 158277323Sdim uint64_t Address) { 159277323Sdim if (SymbolLookUp) { 160277323Sdim uint64_t ReferenceType = LLVMDisassembler_ReferenceType_In_PCrel_Load; 161277323Sdim const char *ReferenceName; 162277323Sdim (void)SymbolLookUp(DisInfo, Value, &ReferenceType, Address, &ReferenceName); 163277323Sdim if(ReferenceType == LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr) 164277323Sdim cStream << "literal pool symbol address: " << ReferenceName; 165277323Sdim else if(ReferenceType == 166277323Sdim LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr) { 167277323Sdim cStream << "literal pool for: \""; 168277323Sdim cStream.write_escaped(ReferenceName); 169277323Sdim cStream << "\""; 170277323Sdim } 171277323Sdim else if(ReferenceType == 172277323Sdim LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref) 173277323Sdim cStream << "Objc cfstring ref: @\"" << ReferenceName << "\""; 174277323Sdim else if(ReferenceType == 175277323Sdim LLVMDisassembler_ReferenceType_Out_Objc_Message) 176277323Sdim cStream << "Objc message: " << ReferenceName; 177277323Sdim else if(ReferenceType == 178277323Sdim LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref) 179277323Sdim cStream << "Objc message ref: " << ReferenceName; 180277323Sdim else if(ReferenceType == 181277323Sdim LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref) 182277323Sdim cStream << "Objc selector ref: " << ReferenceName; 183277323Sdim else if(ReferenceType == 184277323Sdim LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref) 185277323Sdim cStream << "Objc class ref: " << ReferenceName; 186277323Sdim } 187277323Sdim} 188277323Sdim 189277323Sdimnamespace llvm { 190288943SdimMCSymbolizer *createMCSymbolizer(const Triple &TT, LLVMOpInfoCallback GetOpInfo, 191277323Sdim LLVMSymbolLookupCallback SymbolLookUp, 192288943Sdim void *DisInfo, MCContext *Ctx, 193288943Sdim std::unique_ptr<MCRelocationInfo> &&RelInfo) { 194277323Sdim assert(Ctx && "No MCContext given for symbolic disassembly"); 195277323Sdim 196288943Sdim return new MCExternalSymbolizer(*Ctx, std::move(RelInfo), GetOpInfo, 197288943Sdim SymbolLookUp, DisInfo); 198277323Sdim} 199277323Sdim} 200