LLVM 17.0.0git
WebAssemblyDisassembler.cpp
Go to the documentation of this file.
1//==- WebAssemblyDisassembler.cpp - Disassembler for WebAssembly -*- C++ -*-==//
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/// \file
10/// This file is part of the WebAssembly Disassembler.
11///
12/// It contains code to translate the data produced by the decoder into
13/// MCInsts.
14///
15//===----------------------------------------------------------------------===//
16
19#include "llvm/MC/MCContext.h"
22#include "llvm/MC/MCInst.h"
23#include "llvm/MC/MCInstrInfo.h"
25#include "llvm/MC/MCSymbol.h"
28#include "llvm/Support/Endian.h"
29#include "llvm/Support/LEB128.h"
30
31using namespace llvm;
32
33#define DEBUG_TYPE "wasm-disassembler"
34
36
37#include "WebAssemblyGenDisassemblerTables.inc"
38
39namespace {
40static constexpr int WebAssemblyInstructionTableSize = 256;
41
42class WebAssemblyDisassembler final : public MCDisassembler {
43 std::unique_ptr<const MCInstrInfo> MCII;
44
46 ArrayRef<uint8_t> Bytes, uint64_t Address,
47 raw_ostream &CStream) const override;
48 std::optional<DecodeStatus>
50 uint64_t Address, raw_ostream &CStream) const override;
51
52public:
53 WebAssemblyDisassembler(const MCSubtargetInfo &STI, MCContext &Ctx,
54 std::unique_ptr<const MCInstrInfo> MCII)
55 : MCDisassembler(STI, Ctx), MCII(std::move(MCII)) {}
56};
57} // end anonymous namespace
58
60 const MCSubtargetInfo &STI,
61 MCContext &Ctx) {
62 std::unique_ptr<const MCInstrInfo> MCII(T.createMCInstrInfo());
63 return new WebAssemblyDisassembler(STI, Ctx, std::move(MCII));
64}
65
66extern "C" LLVM_EXTERNAL_VISIBILITY void
68 // Register the disassembler for each target.
73}
74
76 if (Size >= Bytes.size())
77 return -1;
78 auto V = Bytes[Size];
79 Size++;
80 return V;
81}
82
83static bool nextLEB(int64_t &Val, ArrayRef<uint8_t> Bytes, uint64_t &Size,
84 bool Signed) {
85 unsigned N = 0;
86 const char *Error = nullptr;
87 Val = Signed ? decodeSLEB128(Bytes.data() + Size, &N,
88 Bytes.data() + Bytes.size(), &Error)
89 : static_cast<int64_t>(decodeULEB128(Bytes.data() + Size, &N,
90 Bytes.data() + Bytes.size(),
91 &Error));
92 if (Error)
93 return false;
94 Size += N;
95 return true;
96}
97
99 ArrayRef<uint8_t> Bytes, bool Signed) {
100 int64_t Val;
101 if (!nextLEB(Val, Bytes, Size, Signed))
102 return false;
103 MI.addOperand(MCOperand::createImm(Val));
104 return true;
105}
106
107template <typename T>
109 if (Size + sizeof(T) > Bytes.size())
110 return false;
111 T Val = support::endian::read<T, support::endianness::little, 1>(
112 Bytes.data() + Size);
113 Size += sizeof(T);
114 if (std::is_floating_point<T>::value) {
115 MI.addOperand(
116 MCOperand::createDFPImm(bit_cast<uint64_t>(static_cast<double>(Val))));
117 } else {
118 MI.addOperand(MCOperand::createImm(static_cast<int64_t>(Val)));
119 }
120 return true;
121}
122
123std::optional<MCDisassembler::DecodeStatus>
124WebAssemblyDisassembler::onSymbolStart(SymbolInfoTy &Symbol, uint64_t &Size,
125 ArrayRef<uint8_t> Bytes,
126 uint64_t Address,
127 raw_ostream &CStream) const {
128 Size = 0;
129 if (Address == 0) {
130 // Start of a code section: we're parsing only the function count.
131 int64_t FunctionCount;
132 if (!nextLEB(FunctionCount, Bytes, Size, false))
133 return std::nullopt;
134 outs() << " # " << FunctionCount << " functions in section.";
135 } else {
136 // Parse the start of a single function.
137 int64_t BodySize, LocalEntryCount;
138 if (!nextLEB(BodySize, Bytes, Size, false) ||
139 !nextLEB(LocalEntryCount, Bytes, Size, false))
140 return std::nullopt;
141 if (LocalEntryCount) {
142 outs() << " .local ";
143 for (int64_t I = 0; I < LocalEntryCount; I++) {
144 int64_t Count, Type;
145 if (!nextLEB(Count, Bytes, Size, false) ||
146 !nextLEB(Type, Bytes, Size, false))
147 return std::nullopt;
148 for (int64_t J = 0; J < Count; J++) {
149 if (I || J)
150 outs() << ", ";
152 }
153 }
154 }
155 }
156 outs() << "\n";
158}
159
160MCDisassembler::DecodeStatus WebAssemblyDisassembler::getInstruction(
161 MCInst &MI, uint64_t &Size, ArrayRef<uint8_t> Bytes, uint64_t /*Address*/,
162 raw_ostream &CS) const {
163 CommentStream = &CS;
164 Size = 0;
165 int Opc = nextByte(Bytes, Size);
166 if (Opc < 0)
168 const auto *WasmInst = &InstructionTable0[Opc];
169 // If this is a prefix byte, indirect to another table.
170 if (WasmInst->ET == ET_Prefix) {
171 WasmInst = nullptr;
172 // Linear search, so far only 2 entries.
173 for (auto PT = PrefixTable; PT->Table; PT++) {
174 if (PT->Prefix == Opc) {
175 WasmInst = PT->Table;
176 break;
177 }
178 }
179 if (!WasmInst)
181 int64_t PrefixedOpc;
182 if (!nextLEB(PrefixedOpc, Bytes, Size, false))
184 if (PrefixedOpc < 0 || PrefixedOpc >= WebAssemblyInstructionTableSize)
186 WasmInst += PrefixedOpc;
187 }
188 if (WasmInst->ET == ET_Unused)
190 // At this point we must have a valid instruction to decode.
191 assert(WasmInst->ET == ET_Instruction);
192 MI.setOpcode(WasmInst->Opcode);
193 // Parse any operands.
194 for (uint8_t OPI = 0; OPI < WasmInst->NumOperands; OPI++) {
195 auto OT = OperandTable[WasmInst->OperandStart + OPI];
196 switch (OT) {
197 // ULEB operands:
209 if (!parseLEBImmediate(MI, Size, Bytes, false))
211 break;
212 }
213 // SLEB operands:
216 if (!parseLEBImmediate(MI, Size, Bytes, true))
218 break;
219 }
220 // block_type operands:
222 int64_t Val;
223 uint64_t PrevSize = Size;
224 if (!nextLEB(Val, Bytes, Size, true))
226 if (Val < 0) {
227 // Negative values are single septet value types or empty types
228 if (Size != PrevSize + 1) {
229 MI.addOperand(
230 MCOperand::createImm(int64_t(WebAssembly::BlockType::Invalid)));
231 } else {
232 MI.addOperand(MCOperand::createImm(Val & 0x7f));
233 }
234 } else {
235 // We don't have access to the signature, so create a symbol without one
236 MCSymbol *Sym = getContext().createTempSymbol("typeindex", true);
237 auto *WasmSym = cast<MCSymbolWasm>(Sym);
238 WasmSym->setType(wasm::WASM_SYMBOL_TYPE_FUNCTION);
239 const MCExpr *Expr = MCSymbolRefExpr::create(
240 WasmSym, MCSymbolRefExpr::VK_WASM_TYPEINDEX, getContext());
241 MI.addOperand(MCOperand::createExpr(Expr));
242 }
243 break;
244 }
245 // FP operands.
247 if (!parseImmediate<float>(MI, Size, Bytes))
249 break;
250 }
252 if (!parseImmediate<double>(MI, Size, Bytes))
254 break;
255 }
256 // Vector lane operands (not LEB encoded).
258 if (!parseImmediate<uint8_t>(MI, Size, Bytes))
260 break;
261 }
263 if (!parseImmediate<uint16_t>(MI, Size, Bytes))
265 break;
266 }
268 if (!parseImmediate<uint32_t>(MI, Size, Bytes))
270 break;
271 }
273 if (!parseImmediate<uint64_t>(MI, Size, Bytes))
275 break;
276 }
278 int64_t TargetTableLen;
279 if (!nextLEB(TargetTableLen, Bytes, Size, false))
281 for (int64_t I = 0; I < TargetTableLen; I++) {
282 if (!parseLEBImmediate(MI, Size, Bytes, false))
284 }
285 // Default case.
286 if (!parseLEBImmediate(MI, Size, Bytes, false))
288 break;
289 }
291 // The tablegen header currently does not have any register operands since
292 // we use only the stack (_S) instructions.
293 // If you hit this that probably means a bad instruction definition in
294 // tablegen.
295 llvm_unreachable("Register operand in WebAssemblyDisassembler");
296 default:
297 llvm_unreachable("Unknown operand type in WebAssemblyDisassembler");
298 }
299 }
301}
static constexpr const ArrayRef< StringLiteral > PrefixTable(PrefixTable_init, std::size(PrefixTable_init) - 1)
#define LLVM_EXTERNAL_VISIBILITY
Definition: Compiler.h:127
uint64_t Size
IRTranslator LLVM IR MI
#define I(x, y, z)
Definition: MD5.cpp:58
assert(ImpDefSCC.getReg()==AMDGPU::SCC &&ImpDefSCC.isDef())
LLVM_EXTERNAL_VISIBILITY void LLVMInitializeWebAssemblyDisassembler()
static int nextByte(ArrayRef< uint8_t > Bytes, uint64_t &Size)
static bool nextLEB(int64_t &Val, ArrayRef< uint8_t > Bytes, uint64_t &Size, bool Signed)
static MCDisassembler * createWebAssemblyDisassembler(const Target &T, const MCSubtargetInfo &STI, MCContext &Ctx)
bool parseImmediate(MCInst &MI, uint64_t &Size, ArrayRef< uint8_t > Bytes)
static bool parseLEBImmediate(MCInst &MI, uint64_t &Size, ArrayRef< uint8_t > Bytes, bool Signed)
This file contains the declaration of the WebAssembly-specific type parsing utility functions.
This file registers the WebAssembly target.
ArrayRef - Represent a constant reference to an array (0 or more elements consecutively in memory),...
Definition: ArrayRef.h:41
size_t size() const
size - Get the array size.
Definition: ArrayRef.h:163
const T * data() const
Definition: ArrayRef.h:160
Lightweight error class with error context and mandatory checking.
Definition: Error.h:156
Context object for machine code objects.
Definition: MCContext.h:76
Superclass for all disassemblers.
virtual std::optional< DecodeStatus > onSymbolStart(SymbolInfoTy &Symbol, uint64_t &Size, ArrayRef< uint8_t > Bytes, uint64_t Address, raw_ostream &CStream) const
Used to perform separate target specific disassembly for a particular symbol.
DecodeStatus
Ternary decode status.
virtual DecodeStatus getInstruction(MCInst &Instr, uint64_t &Size, ArrayRef< uint8_t > Bytes, uint64_t Address, raw_ostream &CStream) const =0
Returns the disassembly of a single instruction.
Base class for the full range of assembler expressions which are needed for parsing.
Definition: MCExpr.h:35
Instances of this class represent a single low-level machine instruction.
Definition: MCInst.h:184
static MCOperand createExpr(const MCExpr *Val)
Definition: MCInst.h:162
static MCOperand createImm(int64_t Val)
Definition: MCInst.h:141
static MCOperand createDFPImm(uint64_t Val)
Definition: MCInst.h:155
Generic base class for all target subtargets.
static const MCSymbolRefExpr * create(const MCSymbol *Symbol, MCContext &Ctx)
Definition: MCExpr.h:386
MCSymbol - Instances of this class represent a symbol name in the MC file, and MCSymbols are created ...
Definition: MCSymbol.h:41
Target - Wrapper for Target specific information.
The instances of the Type class are immutable: once they are created, they are never changed.
Definition: Type.h:45
This class implements an extremely fast bulk output stream that can only output to a stream.
Definition: raw_ostream.h:52
#define llvm_unreachable(msg)
Marks that the current location is not supposed to be reachable.
@ OPERAND_REGISTER
Definition: MCInstrDesc.h:61
@ OPERAND_IMMEDIATE
Definition: MCInstrDesc.h:60
@ OPERAND_GLOBAL
Global index.
@ OPERAND_OFFSET64
64-bit unsigned memory offsets.
@ OPERAND_I32IMM
32-bit integer immediates.
@ OPERAND_P2ALIGN
p2align immediate for load and store address alignment.
@ OPERAND_TABLE
32-bit unsigned table number.
@ OPERAND_VEC_I64IMM
64-bit vector lane immediate
@ OPERAND_VEC_I16IMM
16-bit vector lane immediate
@ OPERAND_TYPEINDEX
type signature immediate for call_indirect.
@ OPERAND_FUNCTION32
32-bit unsigned function indices.
@ OPERAND_F32IMM
32-bit floating-point immediates.
@ OPERAND_BASIC_BLOCK
Basic block label in a branch construct.
@ OPERAND_VEC_I32IMM
32-bit vector lane immediate
@ OPERAND_BRLIST
A list of branch targets for br_list.
@ OPERAND_F64IMM
64-bit floating-point immediates.
@ OPERAND_VEC_I8IMM
8-bit vector lane immediate
@ OPERAND_SIGNATURE
signature immediate for block/loop.
@ OPERAND_I64IMM
64-bit integer immediates.
@ OPERAND_OFFSET32
32-bit unsigned memory offsets.
const char * anyTypeToString(unsigned Type)
@ WASM_SYMBOL_TYPE_FUNCTION
Definition: Wasm.h:383
This is an optimization pass for GlobalISel generic memory operations.
Definition: AddressRanges.h:18
raw_fd_ostream & outs()
This returns a reference to a raw_fd_ostream for standard output.
uint64_t decodeULEB128(const uint8_t *p, unsigned *n=nullptr, const uint8_t *end=nullptr, const char **error=nullptr)
Utility function to decode a ULEB128 value.
Definition: LEB128.h:128
int64_t decodeSLEB128(const uint8_t *p, unsigned *n=nullptr, const uint8_t *end=nullptr, const char **error=nullptr)
Utility function to decode a SLEB128 value.
Definition: LEB128.h:161
Target & getTheWebAssemblyTarget32()
Target & getTheWebAssemblyTarget64()
#define N
static void RegisterMCDisassembler(Target &T, Target::MCDisassemblerCtorTy Fn)
RegisterMCDisassembler - Register a MCDisassembler implementation for the given target.