LLVM 19.0.0git
InstrOrderFile.cpp
Go to the documentation of this file.
1//===- InstrOrderFile.cpp ---- Late IR instrumentation for order file ----===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9//===----------------------------------------------------------------------===//
10
12#include "llvm/IR/Constants.h"
13#include "llvm/IR/Function.h"
14#include "llvm/IR/GlobalValue.h"
15#include "llvm/IR/IRBuilder.h"
17#include "llvm/IR/Module.h"
23#include <fstream>
24#include <mutex>
25#include <sstream>
26
27using namespace llvm;
28#define DEBUG_TYPE "instrorderfile"
29
31 "orderfile-write-mapping", cl::init(""),
33 "Dump functions and their MD5 hash to deobfuscate profile data"),
35
36namespace {
37
38// We need a global bitmap to tell if a function is executed. We also
39// need a global variable to save the order of functions. We can use a
40// fixed-size buffer that saves the MD5 hash of the function. We need
41// a global variable to save the index into the buffer.
42
43std::mutex MappingMutex;
44
45struct InstrOrderFile {
46private:
47 GlobalVariable *OrderFileBuffer;
48 GlobalVariable *BufferIdx;
49 GlobalVariable *BitMap;
50 ArrayType *BufferTy;
51 ArrayType *MapTy;
52
53public:
54 InstrOrderFile() = default;
55
56 void createOrderFileData(Module &M) {
57 LLVMContext &Ctx = M.getContext();
58 int NumFunctions = 0;
59 for (Function &F : M) {
60 if (!F.isDeclaration())
61 NumFunctions++;
62 }
63
64 BufferTy =
65 ArrayType::get(Type::getInt64Ty(Ctx), INSTR_ORDER_FILE_BUFFER_SIZE);
66 Type *IdxTy = Type::getInt32Ty(Ctx);
67 MapTy = ArrayType::get(Type::getInt8Ty(Ctx), NumFunctions);
68
69 // Create the global variables.
70 std::string SymbolName = INSTR_PROF_ORDERFILE_BUFFER_NAME_STR;
71 OrderFileBuffer = new GlobalVariable(M, BufferTy, false, GlobalValue::LinkOnceODRLinkage,
72 Constant::getNullValue(BufferTy), SymbolName);
73 Triple TT = Triple(M.getTargetTriple());
74 OrderFileBuffer->setSection(
75 getInstrProfSectionName(IPSK_orderfile, TT.getObjectFormat()));
76
77 std::string IndexName = INSTR_PROF_ORDERFILE_BUFFER_IDX_NAME_STR;
78 BufferIdx = new GlobalVariable(M, IdxTy, false, GlobalValue::LinkOnceODRLinkage,
79 Constant::getNullValue(IdxTy), IndexName);
80
81 std::string BitMapName = "bitmap_0";
82 BitMap = new GlobalVariable(M, MapTy, false, GlobalValue::PrivateLinkage,
83 Constant::getNullValue(MapTy), BitMapName);
84 }
85
86 // Generate the code sequence in the entry block of each function to
87 // update the buffer.
88 void generateCodeSequence(Module &M, Function &F, int FuncId) {
89 if (!ClOrderFileWriteMapping.empty()) {
90 std::lock_guard<std::mutex> LogLock(MappingMutex);
91 std::error_code EC;
94 if (EC) {
96 " to save mapping file for order file instrumentation\n");
97 } else {
98 std::stringstream stream;
99 stream << std::hex << MD5Hash(F.getName());
100 std::string singleLine = "MD5 " + stream.str() + " " +
101 std::string(F.getName()) + '\n';
102 OS << singleLine;
103 }
104 }
105
106 BasicBlock *OrigEntry = &F.getEntryBlock();
107
108 LLVMContext &Ctx = M.getContext();
110 IntegerType *Int8Ty = Type::getInt8Ty(Ctx);
111
112 // Create a new entry block for instrumentation. We will check the bitmap
113 // in this basic block.
114 BasicBlock *NewEntry =
115 BasicBlock::Create(M.getContext(), "order_file_entry", &F, OrigEntry);
116 IRBuilder<> entryB(NewEntry);
117 // Create a basic block for updating the circular buffer.
118 BasicBlock *UpdateOrderFileBB =
119 BasicBlock::Create(M.getContext(), "order_file_set", &F, OrigEntry);
120 IRBuilder<> updateB(UpdateOrderFileBB);
121
122 // Check the bitmap, if it is already 1, do nothing.
123 // Otherwise, set the bit, grab the index, update the buffer.
124 Value *IdxFlags[] = {ConstantInt::get(Int32Ty, 0),
125 ConstantInt::get(Int32Ty, FuncId)};
126 Value *MapAddr = entryB.CreateGEP(MapTy, BitMap, IdxFlags, "");
127 LoadInst *loadBitMap = entryB.CreateLoad(Int8Ty, MapAddr, "");
128 entryB.CreateStore(ConstantInt::get(Int8Ty, 1), MapAddr);
129 Value *IsNotExecuted =
130 entryB.CreateICmpEQ(loadBitMap, ConstantInt::get(Int8Ty, 0));
131 entryB.CreateCondBr(IsNotExecuted, UpdateOrderFileBB, OrigEntry);
132
133 // Fill up UpdateOrderFileBB: grab the index, update the buffer!
134 Value *IdxVal = updateB.CreateAtomicRMW(
135 AtomicRMWInst::Add, BufferIdx, ConstantInt::get(Int32Ty, 1),
136 MaybeAlign(), AtomicOrdering::SequentiallyConsistent);
137 // We need to wrap around the index to fit it inside the buffer.
138 Value *WrappedIdx = updateB.CreateAnd(
139 IdxVal, ConstantInt::get(Int32Ty, INSTR_ORDER_FILE_BUFFER_MASK));
140 Value *BufferGEPIdx[] = {ConstantInt::get(Int32Ty, 0), WrappedIdx};
141 Value *BufferAddr =
142 updateB.CreateGEP(BufferTy, OrderFileBuffer, BufferGEPIdx, "");
143 updateB.CreateStore(ConstantInt::get(Type::getInt64Ty(Ctx), MD5Hash(F.getName())),
144 BufferAddr);
145 updateB.CreateBr(OrigEntry);
146 }
147
148 bool run(Module &M) {
149 createOrderFileData(M);
150
151 int FuncId = 0;
152 for (Function &F : M) {
153 if (F.isDeclaration())
154 continue;
155 generateCodeSequence(M, F, FuncId);
156 ++FuncId;
157 }
158
159 return true;
160 }
161
162}; // End of InstrOrderFile struct
163} // End anonymous namespace
164
167 if (InstrOrderFile().run(M))
169 return PreservedAnalyses::all();
170}
This file contains the declarations for the subclasses of Constant, which represent the different fla...
static cl::opt< std::string > ClOrderFileWriteMapping("orderfile-write-mapping", cl::init(""), cl::desc("Dump functions and their MD5 hash to deobfuscate profile data"), cl::Hidden)
#define F(x, y, z)
Definition: MD5.cpp:55
Module.h This file contains the declarations for the Module class.
IntegerType * Int32Ty
Profile::FuncID FuncId
Definition: Profile.cpp:321
raw_pwrite_stream & OS
A container for analyses that lazily runs them and caches their results.
Definition: PassManager.h:348
@ Add
*p = old + v
Definition: Instructions.h:764
LLVM Basic Block Representation.
Definition: BasicBlock.h:60
static BasicBlock * Create(LLVMContext &Context, const Twine &Name="", Function *Parent=nullptr, BasicBlock *InsertBefore=nullptr)
Creates a new BasicBlock.
Definition: BasicBlock.h:198
static Constant * getNullValue(Type *Ty)
Constructor to create a '0' constant of arbitrary type.
Definition: Constants.cpp:370
void setSection(StringRef S)
Change the section for this global.
Definition: Globals.cpp:251
@ PrivateLinkage
Like Internal, but omit from symbol table.
Definition: GlobalValue.h:60
@ LinkOnceODRLinkage
Same, but only replaced by something equivalent.
Definition: GlobalValue.h:55
This provides a uniform API for creating instructions and inserting them into a basic block: either a...
Definition: IRBuilder.h:2649
PreservedAnalyses run(Module &M, ModuleAnalysisManager &AM)
Class to represent integer types.
Definition: DerivedTypes.h:40
This is an important class for using LLVM in a threaded context.
Definition: LLVMContext.h:67
An instruction for reading from memory.
Definition: Instructions.h:184
A Module instance is used to store all the information related to an LLVM module.
Definition: Module.h:65
A set of analyses that are preserved following a run of a transformation pass.
Definition: Analysis.h:109
static PreservedAnalyses none()
Convenience factory function for the empty preserved set.
Definition: Analysis.h:112
static PreservedAnalyses all()
Construct a special preserved set that preserves all passes.
Definition: Analysis.h:115
Triple - Helper class for working with autoconf configuration names.
Definition: Triple.h:44
Twine - A lightweight data structure for efficiently representing the concatenation of temporary valu...
Definition: Twine.h:81
The instances of the Type class are immutable: once they are created, they are never changed.
Definition: Type.h:45
static IntegerType * getInt8Ty(LLVMContext &C)
static IntegerType * getInt32Ty(LLVMContext &C)
static IntegerType * getInt64Ty(LLVMContext &C)
LLVM Value Representation.
Definition: Value.h:74
A raw_ostream that writes to a file descriptor.
Definition: raw_ostream.h:470
constexpr char SymbolName[]
Key for Kernel::Metadata::mSymbolName.
initializer< Ty > init(const Ty &Val)
Definition: CommandLine.h:450
PointerTypeMap run(const Module &M)
Compute the PointerTypeMap for the module M.
uint64_t MD5Hash(const FunctionId &Obj)
Definition: FunctionId.h:167
@ OF_Append
The file should be opened in append mode.
Definition: FileSystem.h:771
This is an optimization pass for GlobalISel generic memory operations.
Definition: AddressRanges.h:18
std::string getInstrProfSectionName(InstrProfSectKind IPSK, Triple::ObjectFormatType OF, bool AddSegmentInfo=true)
Return the name of the profile section corresponding to IPSK.
Definition: InstrProf.cpp:222
void report_fatal_error(Error Err, bool gen_crash_diag=true)
Report a serious error, calling any installed error handler.
Definition: Error.cpp:156
This struct is a compact representation of a valid (power of two) or undefined (0) alignment.
Definition: Alignment.h:117