| 1 | //===-- Disassembler.cpp --------------------------------------------------===// |
| 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 | #include "lldb/Core/Disassembler.h" |
| 10 | |
| 11 | #include "lldb/Core/AddressRange.h" |
| 12 | #include "lldb/Core/Debugger.h" |
| 13 | #include "lldb/Core/EmulateInstruction.h" |
| 14 | #include "lldb/Core/Mangled.h" |
| 15 | #include "lldb/Core/Module.h" |
| 16 | #include "lldb/Core/ModuleList.h" |
| 17 | #include "lldb/Core/PluginManager.h" |
| 18 | #include "lldb/Core/SourceManager.h" |
| 19 | #include "lldb/Host/FileSystem.h" |
| 20 | #include "lldb/Interpreter/OptionValue.h" |
| 21 | #include "lldb/Interpreter/OptionValueArray.h" |
| 22 | #include "lldb/Interpreter/OptionValueDictionary.h" |
| 23 | #include "lldb/Interpreter/OptionValueRegex.h" |
| 24 | #include "lldb/Interpreter/OptionValueString.h" |
| 25 | #include "lldb/Interpreter/OptionValueUInt64.h" |
| 26 | #include "lldb/Symbol/Function.h" |
| 27 | #include "lldb/Symbol/Symbol.h" |
| 28 | #include "lldb/Symbol/SymbolContext.h" |
| 29 | #include "lldb/Target/ExecutionContext.h" |
| 30 | #include "lldb/Target/SectionLoadList.h" |
| 31 | #include "lldb/Target/StackFrame.h" |
| 32 | #include "lldb/Target/Target.h" |
| 33 | #include "lldb/Target/Thread.h" |
| 34 | #include "lldb/Utility/DataBufferHeap.h" |
| 35 | #include "lldb/Utility/DataExtractor.h" |
| 36 | #include "lldb/Utility/RegularExpression.h" |
| 37 | #include "lldb/Utility/Status.h" |
| 38 | #include "lldb/Utility/Stream.h" |
| 39 | #include "lldb/Utility/StreamString.h" |
| 40 | #include "lldb/Utility/Timer.h" |
| 41 | #include "lldb/lldb-private-enumerations.h" |
| 42 | #include "lldb/lldb-private-interfaces.h" |
| 43 | #include "lldb/lldb-private-types.h" |
| 44 | #include "llvm/Support/Compiler.h" |
| 45 | #include "llvm/TargetParser/Triple.h" |
| 46 | |
| 47 | #include <cstdint> |
| 48 | #include <cstring> |
| 49 | #include <utility> |
| 50 | |
| 51 | #include <cassert> |
| 52 | |
| 53 | #define DEFAULT_DISASM_BYTE_SIZE 32 |
| 54 | |
| 55 | using namespace lldb; |
| 56 | using namespace lldb_private; |
| 57 | |
| 58 | DisassemblerSP Disassembler::FindPlugin(const ArchSpec &arch, |
| 59 | const char *flavor, const char *cpu, |
| 60 | const char *features, |
| 61 | const char *plugin_name) { |
| 62 | LLDB_SCOPED_TIMERF("Disassembler::FindPlugin (arch = %s, plugin_name = %s)" , |
| 63 | arch.GetArchitectureName(), plugin_name); |
| 64 | |
| 65 | DisassemblerCreateInstance create_callback = nullptr; |
| 66 | |
| 67 | if (plugin_name) { |
| 68 | create_callback = |
| 69 | PluginManager::GetDisassemblerCreateCallbackForPluginName(name: plugin_name); |
| 70 | if (create_callback) { |
| 71 | if (auto disasm_sp = create_callback(arch, flavor, cpu, features)) |
| 72 | return disasm_sp; |
| 73 | } |
| 74 | } else { |
| 75 | for (uint32_t idx = 0; |
| 76 | (create_callback = PluginManager::GetDisassemblerCreateCallbackAtIndex( |
| 77 | idx)) != nullptr; |
| 78 | ++idx) { |
| 79 | if (auto disasm_sp = create_callback(arch, flavor, cpu, features)) |
| 80 | return disasm_sp; |
| 81 | } |
| 82 | } |
| 83 | return DisassemblerSP(); |
| 84 | } |
| 85 | |
| 86 | DisassemblerSP Disassembler::FindPluginForTarget( |
| 87 | const Target &target, const ArchSpec &arch, const char *flavor, |
| 88 | const char *cpu, const char *features, const char *plugin_name) { |
| 89 | if (!flavor) { |
| 90 | // FIXME - we don't have the mechanism in place to do per-architecture |
| 91 | // settings. But since we know that for now we only support flavors on x86 |
| 92 | // & x86_64, |
| 93 | if (arch.GetTriple().getArch() == llvm::Triple::x86 || |
| 94 | arch.GetTriple().getArch() == llvm::Triple::x86_64) |
| 95 | flavor = target.GetDisassemblyFlavor(); |
| 96 | } |
| 97 | if (!cpu) |
| 98 | cpu = target.GetDisassemblyCPU(); |
| 99 | if (!features) |
| 100 | features = target.GetDisassemblyFeatures(); |
| 101 | |
| 102 | return FindPlugin(arch, flavor, cpu, features, plugin_name); |
| 103 | } |
| 104 | |
| 105 | static Address ResolveAddress(Target &target, const Address &addr) { |
| 106 | if (!addr.IsSectionOffset()) { |
| 107 | Address resolved_addr; |
| 108 | // If we weren't passed in a section offset address range, try and resolve |
| 109 | // it to something |
| 110 | bool is_resolved = |
| 111 | target.HasLoadedSections() |
| 112 | ? target.ResolveLoadAddress(load_addr: addr.GetOffset(), so_addr&: resolved_addr) |
| 113 | : target.GetImages().ResolveFileAddress(vm_addr: addr.GetOffset(), |
| 114 | so_addr&: resolved_addr); |
| 115 | |
| 116 | // We weren't able to resolve the address, just treat it as a raw address |
| 117 | if (is_resolved && resolved_addr.IsValid()) |
| 118 | return resolved_addr; |
| 119 | } |
| 120 | return addr; |
| 121 | } |
| 122 | |
| 123 | lldb::DisassemblerSP Disassembler::DisassembleRange( |
| 124 | const ArchSpec &arch, const char *plugin_name, const char *flavor, |
| 125 | const char *cpu, const char *features, Target &target, |
| 126 | llvm::ArrayRef<AddressRange> disasm_ranges, bool force_live_memory) { |
| 127 | lldb::DisassemblerSP disasm_sp = Disassembler::FindPluginForTarget( |
| 128 | target, arch, flavor, cpu, features, plugin_name); |
| 129 | |
| 130 | if (!disasm_sp) |
| 131 | return {}; |
| 132 | |
| 133 | size_t bytes_disassembled = 0; |
| 134 | for (const AddressRange &range : disasm_ranges) { |
| 135 | bytes_disassembled += disasm_sp->AppendInstructions( |
| 136 | target, address: range.GetBaseAddress(), limit: {.kind: Limit::Bytes, .value: range.GetByteSize()}, |
| 137 | error_strm_ptr: nullptr, force_live_memory); |
| 138 | } |
| 139 | if (bytes_disassembled == 0) |
| 140 | return {}; |
| 141 | |
| 142 | return disasm_sp; |
| 143 | } |
| 144 | |
| 145 | lldb::DisassemblerSP |
| 146 | Disassembler::DisassembleBytes(const ArchSpec &arch, const char *plugin_name, |
| 147 | const char *flavor, const char *cpu, |
| 148 | const char *features, const Address &start, |
| 149 | const void *src, size_t src_len, |
| 150 | uint32_t num_instructions, bool data_from_file) { |
| 151 | if (!src) |
| 152 | return {}; |
| 153 | |
| 154 | lldb::DisassemblerSP disasm_sp = |
| 155 | Disassembler::FindPlugin(arch, flavor, cpu, features, plugin_name); |
| 156 | |
| 157 | if (!disasm_sp) |
| 158 | return {}; |
| 159 | |
| 160 | DataExtractor data(src, src_len, arch.GetByteOrder(), |
| 161 | arch.GetAddressByteSize()); |
| 162 | |
| 163 | (void)disasm_sp->DecodeInstructions(base_addr: start, data, data_offset: 0, num_instructions, append: false, |
| 164 | data_from_file); |
| 165 | return disasm_sp; |
| 166 | } |
| 167 | |
| 168 | bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch, |
| 169 | const char *plugin_name, const char *flavor, |
| 170 | const char *cpu, const char *features, |
| 171 | const ExecutionContext &exe_ctx, |
| 172 | const Address &address, Limit limit, |
| 173 | bool mixed_source_and_assembly, |
| 174 | uint32_t num_mixed_context_lines, |
| 175 | uint32_t options, Stream &strm) { |
| 176 | if (!exe_ctx.GetTargetPtr()) |
| 177 | return false; |
| 178 | |
| 179 | lldb::DisassemblerSP disasm_sp(Disassembler::FindPluginForTarget( |
| 180 | target: exe_ctx.GetTargetRef(), arch, flavor, cpu, features, plugin_name)); |
| 181 | if (!disasm_sp) |
| 182 | return false; |
| 183 | |
| 184 | const bool force_live_memory = true; |
| 185 | size_t bytes_disassembled = disasm_sp->ParseInstructions( |
| 186 | target&: exe_ctx.GetTargetRef(), address, limit, error_strm_ptr: &strm, force_live_memory); |
| 187 | if (bytes_disassembled == 0) |
| 188 | return false; |
| 189 | |
| 190 | disasm_sp->PrintInstructions(debugger, arch, exe_ctx, |
| 191 | mixed_source_and_assembly, |
| 192 | num_mixed_context_lines, options, strm); |
| 193 | return true; |
| 194 | } |
| 195 | |
| 196 | Disassembler::SourceLine |
| 197 | Disassembler::GetFunctionDeclLineEntry(const SymbolContext &sc) { |
| 198 | if (!sc.function) |
| 199 | return {}; |
| 200 | |
| 201 | if (!sc.line_entry.IsValid()) |
| 202 | return {}; |
| 203 | |
| 204 | LineEntry prologue_end_line = sc.line_entry; |
| 205 | SupportFileSP func_decl_file_sp; |
| 206 | uint32_t func_decl_line; |
| 207 | sc.function->GetStartLineSourceInfo(source_file_sp&: func_decl_file_sp, line_no&: func_decl_line); |
| 208 | |
| 209 | if (!func_decl_file_sp) |
| 210 | return {}; |
| 211 | if (!func_decl_file_sp->Equal(other: *prologue_end_line.file_sp, |
| 212 | equality: SupportFile::eEqualFileSpecAndChecksumIfSet) && |
| 213 | !func_decl_file_sp->Equal(other: *prologue_end_line.original_file_sp, |
| 214 | equality: SupportFile::eEqualFileSpecAndChecksumIfSet)) |
| 215 | return {}; |
| 216 | |
| 217 | SourceLine decl_line; |
| 218 | decl_line.file = func_decl_file_sp->GetSpecOnly(); |
| 219 | decl_line.line = func_decl_line; |
| 220 | // TODO: Do we care about column on these entries? If so, we need to plumb |
| 221 | // that through GetStartLineSourceInfo. |
| 222 | decl_line.column = 0; |
| 223 | return decl_line; |
| 224 | } |
| 225 | |
| 226 | void Disassembler::AddLineToSourceLineTables( |
| 227 | SourceLine &line, |
| 228 | std::map<FileSpec, std::set<uint32_t>> &source_lines_seen) { |
| 229 | if (line.IsValid()) { |
| 230 | auto source_lines_seen_pos = source_lines_seen.find(x: line.file); |
| 231 | if (source_lines_seen_pos == source_lines_seen.end()) { |
| 232 | std::set<uint32_t> lines; |
| 233 | lines.insert(x: line.line); |
| 234 | source_lines_seen.emplace(args&: line.file, args&: lines); |
| 235 | } else { |
| 236 | source_lines_seen_pos->second.insert(x: line.line); |
| 237 | } |
| 238 | } |
| 239 | } |
| 240 | |
| 241 | bool Disassembler::ElideMixedSourceAndDisassemblyLine( |
| 242 | const ExecutionContext &exe_ctx, const SymbolContext &sc, |
| 243 | SourceLine &line) { |
| 244 | |
| 245 | // TODO: should we also check target.process.thread.step-avoid-libraries ? |
| 246 | |
| 247 | const RegularExpression *avoid_regex = nullptr; |
| 248 | |
| 249 | // Skip any line #0 entries - they are implementation details |
| 250 | if (line.line == 0) |
| 251 | return true; |
| 252 | |
| 253 | ThreadSP thread_sp = exe_ctx.GetThreadSP(); |
| 254 | if (thread_sp) { |
| 255 | avoid_regex = thread_sp->GetSymbolsToAvoidRegexp(); |
| 256 | } else { |
| 257 | TargetSP target_sp = exe_ctx.GetTargetSP(); |
| 258 | if (target_sp) { |
| 259 | Status error; |
| 260 | OptionValueSP value_sp = target_sp->GetDebugger().GetPropertyValue( |
| 261 | exe_ctx: &exe_ctx, property_path: "target.process.thread.step-avoid-regexp" , error); |
| 262 | if (value_sp && value_sp->GetType() == OptionValue::eTypeRegex) { |
| 263 | OptionValueRegex *re = value_sp->GetAsRegex(); |
| 264 | if (re) { |
| 265 | avoid_regex = re->GetCurrentValue(); |
| 266 | } |
| 267 | } |
| 268 | } |
| 269 | } |
| 270 | if (avoid_regex && sc.symbol != nullptr) { |
| 271 | const char *function_name = |
| 272 | sc.GetFunctionName(preference: Mangled::ePreferDemangledWithoutArguments) |
| 273 | .GetCString(); |
| 274 | if (function_name && avoid_regex->Execute(string: function_name)) { |
| 275 | // skip this source line |
| 276 | return true; |
| 277 | } |
| 278 | } |
| 279 | // don't skip this source line |
| 280 | return false; |
| 281 | } |
| 282 | |
| 283 | void Disassembler::PrintInstructions(Debugger &debugger, const ArchSpec &arch, |
| 284 | const ExecutionContext &exe_ctx, |
| 285 | bool mixed_source_and_assembly, |
| 286 | uint32_t num_mixed_context_lines, |
| 287 | uint32_t options, Stream &strm) { |
| 288 | // We got some things disassembled... |
| 289 | size_t num_instructions_found = GetInstructionList().GetSize(); |
| 290 | |
| 291 | const uint32_t max_opcode_byte_size = |
| 292 | GetInstructionList().GetMaxOpcocdeByteSize(); |
| 293 | SymbolContext sc; |
| 294 | SymbolContext prev_sc; |
| 295 | AddressRange current_source_line_range; |
| 296 | const Address *pc_addr_ptr = nullptr; |
| 297 | StackFrame *frame = exe_ctx.GetFramePtr(); |
| 298 | |
| 299 | TargetSP target_sp(exe_ctx.GetTargetSP()); |
| 300 | SourceManager &source_manager = |
| 301 | target_sp ? target_sp->GetSourceManager() : debugger.GetSourceManager(); |
| 302 | |
| 303 | if (frame) { |
| 304 | pc_addr_ptr = &frame->GetFrameCodeAddress(); |
| 305 | } |
| 306 | const uint32_t scope = |
| 307 | eSymbolContextLineEntry | eSymbolContextFunction | eSymbolContextSymbol; |
| 308 | const bool use_inline_block_range = false; |
| 309 | |
| 310 | const FormatEntity::Entry *disassembly_format = nullptr; |
| 311 | FormatEntity::Entry format; |
| 312 | if (exe_ctx.HasTargetScope()) { |
| 313 | format = exe_ctx.GetTargetRef().GetDebugger().GetDisassemblyFormat(); |
| 314 | disassembly_format = &format; |
| 315 | } else { |
| 316 | FormatEntity::Parse(format: "${addr}: " , entry&: format); |
| 317 | disassembly_format = &format; |
| 318 | } |
| 319 | |
| 320 | // First pass: step through the list of instructions, find how long the |
| 321 | // initial addresses strings are, insert padding in the second pass so the |
| 322 | // opcodes all line up nicely. |
| 323 | |
| 324 | // Also build up the source line mapping if this is mixed source & assembly |
| 325 | // mode. Calculate the source line for each assembly instruction (eliding |
| 326 | // inlined functions which the user wants to skip). |
| 327 | |
| 328 | std::map<FileSpec, std::set<uint32_t>> source_lines_seen; |
| 329 | Symbol *previous_symbol = nullptr; |
| 330 | |
| 331 | size_t address_text_size = 0; |
| 332 | for (size_t i = 0; i < num_instructions_found; ++i) { |
| 333 | Instruction *inst = GetInstructionList().GetInstructionAtIndex(idx: i).get(); |
| 334 | if (inst) { |
| 335 | const Address &addr = inst->GetAddress(); |
| 336 | ModuleSP module_sp(addr.GetModule()); |
| 337 | if (module_sp) { |
| 338 | const SymbolContextItem resolve_mask = eSymbolContextFunction | |
| 339 | eSymbolContextSymbol | |
| 340 | eSymbolContextLineEntry; |
| 341 | uint32_t resolved_mask = |
| 342 | module_sp->ResolveSymbolContextForAddress(so_addr: addr, resolve_scope: resolve_mask, sc); |
| 343 | if (resolved_mask) { |
| 344 | StreamString strmstr; |
| 345 | Debugger::FormatDisassemblerAddress(format: disassembly_format, sc: &sc, prev_sc: nullptr, |
| 346 | exe_ctx: &exe_ctx, addr: &addr, s&: strmstr); |
| 347 | size_t cur_line = strmstr.GetSizeOfLastLine(); |
| 348 | if (cur_line > address_text_size) |
| 349 | address_text_size = cur_line; |
| 350 | |
| 351 | // Add entries to our "source_lines_seen" map+set which list which |
| 352 | // sources lines occur in this disassembly session. We will print |
| 353 | // lines of context around a source line, but we don't want to print |
| 354 | // a source line that has a line table entry of its own - we'll leave |
| 355 | // that source line to be printed when it actually occurs in the |
| 356 | // disassembly. |
| 357 | |
| 358 | if (mixed_source_and_assembly && sc.line_entry.IsValid()) { |
| 359 | if (sc.symbol != previous_symbol) { |
| 360 | SourceLine decl_line = GetFunctionDeclLineEntry(sc); |
| 361 | if (!ElideMixedSourceAndDisassemblyLine(exe_ctx, sc, line&: decl_line)) |
| 362 | AddLineToSourceLineTables(line&: decl_line, source_lines_seen); |
| 363 | } |
| 364 | if (sc.line_entry.IsValid()) { |
| 365 | SourceLine this_line; |
| 366 | this_line.file = sc.line_entry.GetFile(); |
| 367 | this_line.line = sc.line_entry.line; |
| 368 | this_line.column = sc.line_entry.column; |
| 369 | if (!ElideMixedSourceAndDisassemblyLine(exe_ctx, sc, line&: this_line)) |
| 370 | AddLineToSourceLineTables(line&: this_line, source_lines_seen); |
| 371 | } |
| 372 | } |
| 373 | } |
| 374 | sc.Clear(clear_target: false); |
| 375 | } |
| 376 | } |
| 377 | } |
| 378 | |
| 379 | previous_symbol = nullptr; |
| 380 | SourceLine previous_line; |
| 381 | for (size_t i = 0; i < num_instructions_found; ++i) { |
| 382 | Instruction *inst = GetInstructionList().GetInstructionAtIndex(idx: i).get(); |
| 383 | |
| 384 | if (inst) { |
| 385 | const Address &addr = inst->GetAddress(); |
| 386 | const bool inst_is_at_pc = pc_addr_ptr && addr == *pc_addr_ptr; |
| 387 | SourceLinesToDisplay source_lines_to_display; |
| 388 | |
| 389 | prev_sc = sc; |
| 390 | |
| 391 | ModuleSP module_sp(addr.GetModule()); |
| 392 | if (module_sp) { |
| 393 | uint32_t resolved_mask = module_sp->ResolveSymbolContextForAddress( |
| 394 | so_addr: addr, resolve_scope: eSymbolContextEverything, sc); |
| 395 | if (resolved_mask) { |
| 396 | if (mixed_source_and_assembly) { |
| 397 | |
| 398 | // If we've started a new function (non-inlined), print all of the |
| 399 | // source lines from the function declaration until the first line |
| 400 | // table entry - typically the opening curly brace of the function. |
| 401 | if (previous_symbol != sc.symbol) { |
| 402 | // The default disassembly format puts an extra blank line |
| 403 | // between functions - so when we're displaying the source |
| 404 | // context for a function, we don't want to add a blank line |
| 405 | // after the source context or we'll end up with two of them. |
| 406 | if (previous_symbol != nullptr) |
| 407 | source_lines_to_display.print_source_context_end_eol = false; |
| 408 | |
| 409 | previous_symbol = sc.symbol; |
| 410 | if (sc.function && sc.line_entry.IsValid()) { |
| 411 | LineEntry prologue_end_line = sc.line_entry; |
| 412 | if (!ElideMixedSourceAndDisassemblyLine(exe_ctx, sc, |
| 413 | line&: prologue_end_line)) { |
| 414 | SupportFileSP func_decl_file_sp; |
| 415 | uint32_t func_decl_line; |
| 416 | sc.function->GetStartLineSourceInfo(source_file_sp&: func_decl_file_sp, |
| 417 | line_no&: func_decl_line); |
| 418 | if (func_decl_file_sp && |
| 419 | (func_decl_file_sp->Equal( |
| 420 | other: *prologue_end_line.file_sp, |
| 421 | equality: SupportFile::eEqualFileSpecAndChecksumIfSet) || |
| 422 | func_decl_file_sp->Equal( |
| 423 | other: *prologue_end_line.original_file_sp, |
| 424 | equality: SupportFile::eEqualFileSpecAndChecksumIfSet))) { |
| 425 | // Add all the lines between the function declaration and |
| 426 | // the first non-prologue source line to the list of lines |
| 427 | // to print. |
| 428 | for (uint32_t lineno = func_decl_line; |
| 429 | lineno <= prologue_end_line.line; lineno++) { |
| 430 | SourceLine this_line; |
| 431 | this_line.file = func_decl_file_sp->GetSpecOnly(); |
| 432 | this_line.line = lineno; |
| 433 | source_lines_to_display.lines.push_back(x: this_line); |
| 434 | } |
| 435 | // Mark the last line as the "current" one. Usually this |
| 436 | // is the open curly brace. |
| 437 | if (source_lines_to_display.lines.size() > 0) |
| 438 | source_lines_to_display.current_source_line = |
| 439 | source_lines_to_display.lines.size() - 1; |
| 440 | } |
| 441 | } |
| 442 | } |
| 443 | sc.GetAddressRange(scope, range_idx: 0, use_inline_block_range, |
| 444 | range&: current_source_line_range); |
| 445 | } |
| 446 | |
| 447 | // If we've left a previous source line's address range, print a |
| 448 | // new source line |
| 449 | if (!current_source_line_range.ContainsFileAddress(so_addr: addr)) { |
| 450 | sc.GetAddressRange(scope, range_idx: 0, use_inline_block_range, |
| 451 | range&: current_source_line_range); |
| 452 | |
| 453 | if (sc != prev_sc && sc.comp_unit && sc.line_entry.IsValid()) { |
| 454 | SourceLine this_line; |
| 455 | this_line.file = sc.line_entry.GetFile(); |
| 456 | this_line.line = sc.line_entry.line; |
| 457 | |
| 458 | if (!ElideMixedSourceAndDisassemblyLine(exe_ctx, sc, |
| 459 | line&: this_line)) { |
| 460 | // Only print this source line if it is different from the |
| 461 | // last source line we printed. There may have been inlined |
| 462 | // functions between these lines that we elided, resulting in |
| 463 | // the same line being printed twice in a row for a |
| 464 | // contiguous block of assembly instructions. |
| 465 | if (this_line != previous_line) { |
| 466 | |
| 467 | std::vector<uint32_t> previous_lines; |
| 468 | for (uint32_t i = 0; |
| 469 | i < num_mixed_context_lines && |
| 470 | (this_line.line - num_mixed_context_lines) > 0; |
| 471 | i++) { |
| 472 | uint32_t line = |
| 473 | this_line.line - num_mixed_context_lines + i; |
| 474 | auto pos = source_lines_seen.find(x: this_line.file); |
| 475 | if (pos != source_lines_seen.end()) { |
| 476 | if (pos->second.count(x: line) == 1) { |
| 477 | previous_lines.clear(); |
| 478 | } else { |
| 479 | previous_lines.push_back(x: line); |
| 480 | } |
| 481 | } |
| 482 | } |
| 483 | for (size_t i = 0; i < previous_lines.size(); i++) { |
| 484 | SourceLine previous_line; |
| 485 | previous_line.file = this_line.file; |
| 486 | previous_line.line = previous_lines[i]; |
| 487 | auto pos = source_lines_seen.find(x: previous_line.file); |
| 488 | if (pos != source_lines_seen.end()) { |
| 489 | pos->second.insert(x: previous_line.line); |
| 490 | } |
| 491 | source_lines_to_display.lines.push_back(x: previous_line); |
| 492 | } |
| 493 | |
| 494 | source_lines_to_display.lines.push_back(x: this_line); |
| 495 | source_lines_to_display.current_source_line = |
| 496 | source_lines_to_display.lines.size() - 1; |
| 497 | |
| 498 | for (uint32_t i = 0; i < num_mixed_context_lines; i++) { |
| 499 | SourceLine next_line; |
| 500 | next_line.file = this_line.file; |
| 501 | next_line.line = this_line.line + i + 1; |
| 502 | auto pos = source_lines_seen.find(x: next_line.file); |
| 503 | if (pos != source_lines_seen.end()) { |
| 504 | if (pos->second.count(x: next_line.line) == 1) |
| 505 | break; |
| 506 | pos->second.insert(x: next_line.line); |
| 507 | } |
| 508 | source_lines_to_display.lines.push_back(x: next_line); |
| 509 | } |
| 510 | } |
| 511 | previous_line = this_line; |
| 512 | } |
| 513 | } |
| 514 | } |
| 515 | } |
| 516 | } else { |
| 517 | sc.Clear(clear_target: true); |
| 518 | } |
| 519 | } |
| 520 | |
| 521 | if (source_lines_to_display.lines.size() > 0) { |
| 522 | strm.EOL(); |
| 523 | for (size_t idx = 0; idx < source_lines_to_display.lines.size(); |
| 524 | idx++) { |
| 525 | SourceLine ln = source_lines_to_display.lines[idx]; |
| 526 | const char *line_highlight = "" ; |
| 527 | if (inst_is_at_pc && (options & eOptionMarkPCSourceLine)) { |
| 528 | line_highlight = "->" ; |
| 529 | } else if (idx == source_lines_to_display.current_source_line) { |
| 530 | line_highlight = "**" ; |
| 531 | } |
| 532 | source_manager.DisplaySourceLinesWithLineNumbers( |
| 533 | support_file_sp: std::make_shared<SupportFile>(args&: ln.file), line: ln.line, column: ln.column, context_before: 0, context_after: 0, |
| 534 | current_line_cstr: line_highlight, s: &strm); |
| 535 | } |
| 536 | if (source_lines_to_display.print_source_context_end_eol) |
| 537 | strm.EOL(); |
| 538 | } |
| 539 | |
| 540 | const bool show_bytes = (options & eOptionShowBytes) != 0; |
| 541 | const bool show_control_flow_kind = |
| 542 | (options & eOptionShowControlFlowKind) != 0; |
| 543 | inst->Dump(s: &strm, max_opcode_byte_size, show_address: true, show_bytes, |
| 544 | show_control_flow_kind, exe_ctx: &exe_ctx, sym_ctx: &sc, prev_sym_ctx: &prev_sc, disassembly_addr_format: nullptr, |
| 545 | max_address_text_size: address_text_size); |
| 546 | strm.EOL(); |
| 547 | } else { |
| 548 | break; |
| 549 | } |
| 550 | } |
| 551 | } |
| 552 | |
| 553 | bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch, |
| 554 | StackFrame &frame, Stream &strm) { |
| 555 | constexpr const char *plugin_name = nullptr; |
| 556 | constexpr const char *flavor = nullptr; |
| 557 | constexpr const char *cpu = nullptr; |
| 558 | constexpr const char *features = nullptr; |
| 559 | constexpr bool mixed_source_and_assembly = false; |
| 560 | constexpr uint32_t num_mixed_context_lines = 0; |
| 561 | constexpr uint32_t options = 0; |
| 562 | |
| 563 | SymbolContext sc( |
| 564 | frame.GetSymbolContext(resolve_scope: eSymbolContextFunction | eSymbolContextSymbol)); |
| 565 | if (sc.function) { |
| 566 | if (DisassemblerSP disasm_sp = DisassembleRange( |
| 567 | arch, plugin_name, flavor, cpu, features, target&: *frame.CalculateTarget(), |
| 568 | disasm_ranges: sc.function->GetAddressRanges())) { |
| 569 | disasm_sp->PrintInstructions(debugger, arch, exe_ctx: frame, |
| 570 | mixed_source_and_assembly, |
| 571 | num_mixed_context_lines, options, strm); |
| 572 | return true; |
| 573 | } |
| 574 | return false; |
| 575 | } |
| 576 | |
| 577 | AddressRange range; |
| 578 | if (sc.symbol && sc.symbol->ValueIsAddress()) { |
| 579 | range.GetBaseAddress() = sc.symbol->GetAddressRef(); |
| 580 | range.SetByteSize(sc.symbol->GetByteSize()); |
| 581 | } else { |
| 582 | range.GetBaseAddress() = frame.GetFrameCodeAddress(); |
| 583 | } |
| 584 | |
| 585 | if (range.GetBaseAddress().IsValid() && range.GetByteSize() == 0) |
| 586 | range.SetByteSize(DEFAULT_DISASM_BYTE_SIZE); |
| 587 | |
| 588 | Disassembler::Limit limit = {.kind: Disassembler::Limit::Bytes, .value: range.GetByteSize()}; |
| 589 | if (limit.value == 0) |
| 590 | limit.value = DEFAULT_DISASM_BYTE_SIZE; |
| 591 | |
| 592 | return Disassemble(debugger, arch, plugin_name, flavor, cpu, features, exe_ctx: frame, |
| 593 | address: range.GetBaseAddress(), limit, mixed_source_and_assembly, |
| 594 | num_mixed_context_lines, options, strm); |
| 595 | } |
| 596 | |
| 597 | Instruction::Instruction(const Address &address, AddressClass addr_class) |
| 598 | : m_address(address), m_address_class(addr_class), m_opcode(), |
| 599 | m_calculated_strings(false) {} |
| 600 | |
| 601 | Instruction::~Instruction() = default; |
| 602 | |
| 603 | AddressClass Instruction::GetAddressClass() { |
| 604 | if (m_address_class == AddressClass::eInvalid) |
| 605 | m_address_class = m_address.GetAddressClass(); |
| 606 | return m_address_class; |
| 607 | } |
| 608 | |
| 609 | const char *Instruction::GetNameForInstructionControlFlowKind( |
| 610 | lldb::InstructionControlFlowKind instruction_control_flow_kind) { |
| 611 | switch (instruction_control_flow_kind) { |
| 612 | case eInstructionControlFlowKindUnknown: |
| 613 | return "unknown" ; |
| 614 | case eInstructionControlFlowKindOther: |
| 615 | return "other" ; |
| 616 | case eInstructionControlFlowKindCall: |
| 617 | return "call" ; |
| 618 | case eInstructionControlFlowKindReturn: |
| 619 | return "return" ; |
| 620 | case eInstructionControlFlowKindJump: |
| 621 | return "jump" ; |
| 622 | case eInstructionControlFlowKindCondJump: |
| 623 | return "cond jump" ; |
| 624 | case eInstructionControlFlowKindFarCall: |
| 625 | return "far call" ; |
| 626 | case eInstructionControlFlowKindFarReturn: |
| 627 | return "far return" ; |
| 628 | case eInstructionControlFlowKindFarJump: |
| 629 | return "far jump" ; |
| 630 | } |
| 631 | llvm_unreachable("Fully covered switch above!" ); |
| 632 | } |
| 633 | |
| 634 | void Instruction::Dump(lldb_private::Stream *s, uint32_t max_opcode_byte_size, |
| 635 | bool show_address, bool show_bytes, |
| 636 | bool show_control_flow_kind, |
| 637 | const ExecutionContext *exe_ctx, |
| 638 | const SymbolContext *sym_ctx, |
| 639 | const SymbolContext *prev_sym_ctx, |
| 640 | const FormatEntity::Entry *disassembly_addr_format, |
| 641 | size_t max_address_text_size) { |
| 642 | size_t opcode_column_width = 7; |
| 643 | const size_t operand_column_width = 25; |
| 644 | |
| 645 | CalculateMnemonicOperandsAndCommentIfNeeded(exe_ctx); |
| 646 | |
| 647 | StreamString ss; |
| 648 | |
| 649 | if (show_address) { |
| 650 | Debugger::FormatDisassemblerAddress(format: disassembly_addr_format, sc: sym_ctx, |
| 651 | prev_sc: prev_sym_ctx, exe_ctx, addr: &m_address, s&: ss); |
| 652 | ss.FillLastLineToColumn(column: max_address_text_size, fill_char: ' '); |
| 653 | } |
| 654 | |
| 655 | if (show_bytes) { |
| 656 | if (m_opcode.GetType() == Opcode::eTypeBytes) { |
| 657 | // x86_64 and i386 are the only ones that use bytes right now so pad out |
| 658 | // the byte dump to be able to always show 15 bytes (3 chars each) plus a |
| 659 | // space |
| 660 | if (max_opcode_byte_size > 0) |
| 661 | m_opcode.Dump(s: &ss, min_byte_width: max_opcode_byte_size * 3 + 1); |
| 662 | else |
| 663 | m_opcode.Dump(s: &ss, min_byte_width: 15 * 3 + 1); |
| 664 | } else { |
| 665 | // Else, we have ARM or MIPS which can show up to a uint32_t 0x00000000 |
| 666 | // (10 spaces) plus two for padding... |
| 667 | if (max_opcode_byte_size > 0) |
| 668 | m_opcode.Dump(s: &ss, min_byte_width: max_opcode_byte_size * 3 + 1); |
| 669 | else |
| 670 | m_opcode.Dump(s: &ss, min_byte_width: 12); |
| 671 | } |
| 672 | } |
| 673 | |
| 674 | if (show_control_flow_kind) { |
| 675 | lldb::InstructionControlFlowKind instruction_control_flow_kind = |
| 676 | GetControlFlowKind(exe_ctx); |
| 677 | ss.Printf(format: "%-12s" , GetNameForInstructionControlFlowKind( |
| 678 | instruction_control_flow_kind)); |
| 679 | } |
| 680 | |
| 681 | bool show_color = false; |
| 682 | if (exe_ctx) { |
| 683 | if (TargetSP target_sp = exe_ctx->GetTargetSP()) { |
| 684 | show_color = target_sp->GetDebugger().GetUseColor(); |
| 685 | } |
| 686 | } |
| 687 | const size_t opcode_pos = ss.GetSizeOfLastLine(); |
| 688 | const std::string &opcode_name = |
| 689 | show_color ? m_markup_opcode_name : m_opcode_name; |
| 690 | const std::string &mnemonics = show_color ? m_markup_mnemonics : m_mnemonics; |
| 691 | |
| 692 | // The default opcode size of 7 characters is plenty for most architectures |
| 693 | // but some like arm can pull out the occasional vqrshrun.s16. We won't get |
| 694 | // consistent column spacing in these cases, unfortunately. Also note that we |
| 695 | // need to directly use m_opcode_name here (instead of opcode_name) so we |
| 696 | // don't include color codes as characters. |
| 697 | if (m_opcode_name.length() >= opcode_column_width) { |
| 698 | opcode_column_width = m_opcode_name.length() + 1; |
| 699 | } |
| 700 | |
| 701 | ss.PutCString(cstr: opcode_name); |
| 702 | ss.FillLastLineToColumn(column: opcode_pos + opcode_column_width, fill_char: ' '); |
| 703 | ss.PutCString(cstr: mnemonics); |
| 704 | |
| 705 | if (!m_comment.empty()) { |
| 706 | ss.FillLastLineToColumn( |
| 707 | column: opcode_pos + opcode_column_width + operand_column_width, fill_char: ' '); |
| 708 | ss.PutCString(cstr: " ; " ); |
| 709 | ss.PutCString(cstr: m_comment); |
| 710 | } |
| 711 | s->PutCString(cstr: ss.GetString()); |
| 712 | } |
| 713 | |
| 714 | bool Instruction::DumpEmulation(const ArchSpec &arch) { |
| 715 | std::unique_ptr<EmulateInstruction> insn_emulator_up( |
| 716 | EmulateInstruction::FindPlugin(arch, supported_inst_type: eInstructionTypeAny, plugin_name: nullptr)); |
| 717 | if (insn_emulator_up) { |
| 718 | insn_emulator_up->SetInstruction(insn_opcode: GetOpcode(), inst_addr: GetAddress(), target: nullptr); |
| 719 | return insn_emulator_up->EvaluateInstruction(evaluate_options: 0); |
| 720 | } |
| 721 | |
| 722 | return false; |
| 723 | } |
| 724 | |
| 725 | bool Instruction::CanSetBreakpoint () { |
| 726 | return !HasDelaySlot(); |
| 727 | } |
| 728 | |
| 729 | bool Instruction::HasDelaySlot() { |
| 730 | // Default is false. |
| 731 | return false; |
| 732 | } |
| 733 | |
| 734 | OptionValueSP Instruction::ReadArray(FILE *in_file, Stream &out_stream, |
| 735 | OptionValue::Type data_type) { |
| 736 | bool done = false; |
| 737 | char buffer[1024]; |
| 738 | |
| 739 | auto option_value_sp = std::make_shared<OptionValueArray>(args: 1u << data_type); |
| 740 | |
| 741 | int idx = 0; |
| 742 | while (!done) { |
| 743 | if (!fgets(s: buffer, n: 1023, stream: in_file)) { |
| 744 | out_stream.Printf( |
| 745 | format: "Instruction::ReadArray: Error reading file (fgets).\n" ); |
| 746 | option_value_sp.reset(); |
| 747 | return option_value_sp; |
| 748 | } |
| 749 | |
| 750 | std::string line(buffer); |
| 751 | |
| 752 | size_t len = line.size(); |
| 753 | if (line[len - 1] == '\n') { |
| 754 | line[len - 1] = '\0'; |
| 755 | line.resize(n: len - 1); |
| 756 | } |
| 757 | |
| 758 | if ((line.size() == 1) && line[0] == ']') { |
| 759 | done = true; |
| 760 | line.clear(); |
| 761 | } |
| 762 | |
| 763 | if (!line.empty()) { |
| 764 | std::string value; |
| 765 | static RegularExpression g_reg_exp( |
| 766 | llvm::StringRef("^[ \t]*([^ \t]+)[ \t]*$" )); |
| 767 | llvm::SmallVector<llvm::StringRef, 2> matches; |
| 768 | if (g_reg_exp.Execute(string: line, matches: &matches)) |
| 769 | value = matches[1].str(); |
| 770 | else |
| 771 | value = line; |
| 772 | |
| 773 | OptionValueSP data_value_sp; |
| 774 | switch (data_type) { |
| 775 | case OptionValue::eTypeUInt64: |
| 776 | data_value_sp = std::make_shared<OptionValueUInt64>(args: 0, args: 0); |
| 777 | data_value_sp->SetValueFromString(value); |
| 778 | break; |
| 779 | // Other types can be added later as needed. |
| 780 | default: |
| 781 | data_value_sp = std::make_shared<OptionValueString>(args: value.c_str(), args: "" ); |
| 782 | break; |
| 783 | } |
| 784 | |
| 785 | option_value_sp->GetAsArray()->InsertValue(idx, value_sp: data_value_sp); |
| 786 | ++idx; |
| 787 | } |
| 788 | } |
| 789 | |
| 790 | return option_value_sp; |
| 791 | } |
| 792 | |
| 793 | OptionValueSP Instruction::ReadDictionary(FILE *in_file, Stream &out_stream) { |
| 794 | bool done = false; |
| 795 | char buffer[1024]; |
| 796 | |
| 797 | auto option_value_sp = std::make_shared<OptionValueDictionary>(); |
| 798 | static constexpr llvm::StringLiteral encoding_key("data_encoding" ); |
| 799 | OptionValue::Type data_type = OptionValue::eTypeInvalid; |
| 800 | |
| 801 | while (!done) { |
| 802 | // Read the next line in the file |
| 803 | if (!fgets(s: buffer, n: 1023, stream: in_file)) { |
| 804 | out_stream.Printf( |
| 805 | format: "Instruction::ReadDictionary: Error reading file (fgets).\n" ); |
| 806 | option_value_sp.reset(); |
| 807 | return option_value_sp; |
| 808 | } |
| 809 | |
| 810 | // Check to see if the line contains the end-of-dictionary marker ("}") |
| 811 | std::string line(buffer); |
| 812 | |
| 813 | size_t len = line.size(); |
| 814 | if (line[len - 1] == '\n') { |
| 815 | line[len - 1] = '\0'; |
| 816 | line.resize(n: len - 1); |
| 817 | } |
| 818 | |
| 819 | if ((line.size() == 1) && (line[0] == '}')) { |
| 820 | done = true; |
| 821 | line.clear(); |
| 822 | } |
| 823 | |
| 824 | // Try to find a key-value pair in the current line and add it to the |
| 825 | // dictionary. |
| 826 | if (!line.empty()) { |
| 827 | static RegularExpression g_reg_exp(llvm::StringRef( |
| 828 | "^[ \t]*([a-zA-Z_][a-zA-Z0-9_]*)[ \t]*=[ \t]*(.*)[ \t]*$" )); |
| 829 | |
| 830 | llvm::SmallVector<llvm::StringRef, 3> matches; |
| 831 | |
| 832 | bool reg_exp_success = g_reg_exp.Execute(string: line, matches: &matches); |
| 833 | std::string key; |
| 834 | std::string value; |
| 835 | if (reg_exp_success) { |
| 836 | key = matches[1].str(); |
| 837 | value = matches[2].str(); |
| 838 | } else { |
| 839 | out_stream.Printf(format: "Instruction::ReadDictionary: Failure executing " |
| 840 | "regular expression.\n" ); |
| 841 | option_value_sp.reset(); |
| 842 | return option_value_sp; |
| 843 | } |
| 844 | |
| 845 | // Check value to see if it's the start of an array or dictionary. |
| 846 | |
| 847 | lldb::OptionValueSP value_sp; |
| 848 | assert(value.empty() == false); |
| 849 | assert(key.empty() == false); |
| 850 | |
| 851 | if (value[0] == '{') { |
| 852 | assert(value.size() == 1); |
| 853 | // value is a dictionary |
| 854 | value_sp = ReadDictionary(in_file, out_stream); |
| 855 | if (!value_sp) { |
| 856 | option_value_sp.reset(); |
| 857 | return option_value_sp; |
| 858 | } |
| 859 | } else if (value[0] == '[') { |
| 860 | assert(value.size() == 1); |
| 861 | // value is an array |
| 862 | value_sp = ReadArray(in_file, out_stream, data_type); |
| 863 | if (!value_sp) { |
| 864 | option_value_sp.reset(); |
| 865 | return option_value_sp; |
| 866 | } |
| 867 | // We've used the data_type to read an array; re-set the type to |
| 868 | // Invalid |
| 869 | data_type = OptionValue::eTypeInvalid; |
| 870 | } else if ((value[0] == '0') && (value[1] == 'x')) { |
| 871 | value_sp = std::make_shared<OptionValueUInt64>(args: 0, args: 0); |
| 872 | value_sp->SetValueFromString(value); |
| 873 | } else { |
| 874 | size_t len = value.size(); |
| 875 | if ((value[0] == '"') && (value[len - 1] == '"')) |
| 876 | value = value.substr(pos: 1, n: len - 2); |
| 877 | value_sp = std::make_shared<OptionValueString>(args: value.c_str(), args: "" ); |
| 878 | } |
| 879 | |
| 880 | if (key == encoding_key) { |
| 881 | // A 'data_encoding=..." is NOT a normal key-value pair; it is meta-data |
| 882 | // indicating the data type of an upcoming array (usually the next bit |
| 883 | // of data to be read in). |
| 884 | if (llvm::StringRef(value) == "uint32_t" ) |
| 885 | data_type = OptionValue::eTypeUInt64; |
| 886 | } else |
| 887 | option_value_sp->GetAsDictionary()->SetValueForKey(key, value_sp, |
| 888 | can_replace: false); |
| 889 | } |
| 890 | } |
| 891 | |
| 892 | return option_value_sp; |
| 893 | } |
| 894 | |
| 895 | bool Instruction::TestEmulation(Stream &out_stream, const char *file_name) { |
| 896 | if (!file_name) { |
| 897 | out_stream.Printf(format: "Instruction::TestEmulation: Missing file_name." ); |
| 898 | return false; |
| 899 | } |
| 900 | FILE *test_file = FileSystem::Instance().Fopen(path: file_name, mode: "r" ); |
| 901 | if (!test_file) { |
| 902 | out_stream.Printf( |
| 903 | format: "Instruction::TestEmulation: Attempt to open test file failed." ); |
| 904 | return false; |
| 905 | } |
| 906 | |
| 907 | char buffer[256]; |
| 908 | if (!fgets(s: buffer, n: 255, stream: test_file)) { |
| 909 | out_stream.Printf( |
| 910 | format: "Instruction::TestEmulation: Error reading first line of test file.\n" ); |
| 911 | fclose(stream: test_file); |
| 912 | return false; |
| 913 | } |
| 914 | |
| 915 | if (strncmp(s1: buffer, s2: "InstructionEmulationState={" , n: 27) != 0) { |
| 916 | out_stream.Printf(format: "Instructin::TestEmulation: Test file does not contain " |
| 917 | "emulation state dictionary\n" ); |
| 918 | fclose(stream: test_file); |
| 919 | return false; |
| 920 | } |
| 921 | |
| 922 | // Read all the test information from the test file into an |
| 923 | // OptionValueDictionary. |
| 924 | |
| 925 | OptionValueSP data_dictionary_sp(ReadDictionary(in_file: test_file, out_stream)); |
| 926 | if (!data_dictionary_sp) { |
| 927 | out_stream.Printf( |
| 928 | format: "Instruction::TestEmulation: Error reading Dictionary Object.\n" ); |
| 929 | fclose(stream: test_file); |
| 930 | return false; |
| 931 | } |
| 932 | |
| 933 | fclose(stream: test_file); |
| 934 | |
| 935 | OptionValueDictionary *data_dictionary = |
| 936 | data_dictionary_sp->GetAsDictionary(); |
| 937 | static constexpr llvm::StringLiteral description_key("assembly_string" ); |
| 938 | static constexpr llvm::StringLiteral triple_key("triple" ); |
| 939 | |
| 940 | OptionValueSP value_sp = data_dictionary->GetValueForKey(key: description_key); |
| 941 | |
| 942 | if (!value_sp) { |
| 943 | out_stream.Printf(format: "Instruction::TestEmulation: Test file does not " |
| 944 | "contain description string.\n" ); |
| 945 | return false; |
| 946 | } |
| 947 | |
| 948 | SetDescription(value_sp->GetValueAs<llvm::StringRef>().value_or(u: "" )); |
| 949 | |
| 950 | value_sp = data_dictionary->GetValueForKey(key: triple_key); |
| 951 | if (!value_sp) { |
| 952 | out_stream.Printf( |
| 953 | format: "Instruction::TestEmulation: Test file does not contain triple.\n" ); |
| 954 | return false; |
| 955 | } |
| 956 | |
| 957 | ArchSpec arch; |
| 958 | arch.SetTriple( |
| 959 | llvm::Triple(value_sp->GetValueAs<llvm::StringRef>().value_or(u: "" ))); |
| 960 | |
| 961 | bool success = false; |
| 962 | std::unique_ptr<EmulateInstruction> insn_emulator_up( |
| 963 | EmulateInstruction::FindPlugin(arch, supported_inst_type: eInstructionTypeAny, plugin_name: nullptr)); |
| 964 | if (insn_emulator_up) |
| 965 | success = |
| 966 | insn_emulator_up->TestEmulation(out_stream, arch, test_data: data_dictionary); |
| 967 | |
| 968 | if (success) |
| 969 | out_stream.Printf(format: "Emulation test succeeded." ); |
| 970 | else |
| 971 | out_stream.Printf(format: "Emulation test failed." ); |
| 972 | |
| 973 | return success; |
| 974 | } |
| 975 | |
| 976 | bool Instruction::Emulate( |
| 977 | const ArchSpec &arch, uint32_t evaluate_options, void *baton, |
| 978 | EmulateInstruction::ReadMemoryCallback read_mem_callback, |
| 979 | EmulateInstruction::WriteMemoryCallback write_mem_callback, |
| 980 | EmulateInstruction::ReadRegisterCallback read_reg_callback, |
| 981 | EmulateInstruction::WriteRegisterCallback write_reg_callback) { |
| 982 | std::unique_ptr<EmulateInstruction> insn_emulator_up( |
| 983 | EmulateInstruction::FindPlugin(arch, supported_inst_type: eInstructionTypeAny, plugin_name: nullptr)); |
| 984 | if (insn_emulator_up) { |
| 985 | insn_emulator_up->SetBaton(baton); |
| 986 | insn_emulator_up->SetCallbacks(read_mem_callback, write_mem_callback, |
| 987 | read_reg_callback, write_reg_callback); |
| 988 | insn_emulator_up->SetInstruction(insn_opcode: GetOpcode(), inst_addr: GetAddress(), target: nullptr); |
| 989 | return insn_emulator_up->EvaluateInstruction(evaluate_options); |
| 990 | } |
| 991 | |
| 992 | return false; |
| 993 | } |
| 994 | |
| 995 | uint32_t Instruction::(DataExtractor &data) { |
| 996 | return m_opcode.GetData(data); |
| 997 | } |
| 998 | |
| 999 | InstructionList::InstructionList() : m_instructions() {} |
| 1000 | |
| 1001 | InstructionList::~InstructionList() = default; |
| 1002 | |
| 1003 | size_t InstructionList::GetSize() const { return m_instructions.size(); } |
| 1004 | |
| 1005 | uint32_t InstructionList::GetMaxOpcocdeByteSize() const { |
| 1006 | uint32_t max_inst_size = 0; |
| 1007 | collection::const_iterator pos, end; |
| 1008 | for (pos = m_instructions.begin(), end = m_instructions.end(); pos != end; |
| 1009 | ++pos) { |
| 1010 | uint32_t inst_size = (*pos)->GetOpcode().GetByteSize(); |
| 1011 | if (max_inst_size < inst_size) |
| 1012 | max_inst_size = inst_size; |
| 1013 | } |
| 1014 | return max_inst_size; |
| 1015 | } |
| 1016 | |
| 1017 | InstructionSP InstructionList::GetInstructionAtIndex(size_t idx) const { |
| 1018 | InstructionSP inst_sp; |
| 1019 | if (idx < m_instructions.size()) |
| 1020 | inst_sp = m_instructions[idx]; |
| 1021 | return inst_sp; |
| 1022 | } |
| 1023 | |
| 1024 | InstructionSP InstructionList::GetInstructionAtAddress(const Address &address) { |
| 1025 | uint32_t index = GetIndexOfInstructionAtAddress(addr: address); |
| 1026 | if (index != UINT32_MAX) |
| 1027 | return GetInstructionAtIndex(idx: index); |
| 1028 | return nullptr; |
| 1029 | } |
| 1030 | |
| 1031 | void InstructionList::Dump(Stream *s, bool show_address, bool show_bytes, |
| 1032 | bool show_control_flow_kind, |
| 1033 | const ExecutionContext *exe_ctx) { |
| 1034 | const uint32_t max_opcode_byte_size = GetMaxOpcocdeByteSize(); |
| 1035 | collection::const_iterator pos, begin, end; |
| 1036 | |
| 1037 | const FormatEntity::Entry *disassembly_format = nullptr; |
| 1038 | FormatEntity::Entry format; |
| 1039 | if (exe_ctx && exe_ctx->HasTargetScope()) { |
| 1040 | format = exe_ctx->GetTargetRef().GetDebugger().GetDisassemblyFormat(); |
| 1041 | disassembly_format = &format; |
| 1042 | } else { |
| 1043 | FormatEntity::Parse(format: "${addr}: " , entry&: format); |
| 1044 | disassembly_format = &format; |
| 1045 | } |
| 1046 | |
| 1047 | for (begin = m_instructions.begin(), end = m_instructions.end(), pos = begin; |
| 1048 | pos != end; ++pos) { |
| 1049 | if (pos != begin) |
| 1050 | s->EOL(); |
| 1051 | (*pos)->Dump(s, max_opcode_byte_size, show_address, show_bytes, |
| 1052 | show_control_flow_kind, exe_ctx, sym_ctx: nullptr, prev_sym_ctx: nullptr, |
| 1053 | disassembly_addr_format: disassembly_format, max_address_text_size: 0); |
| 1054 | } |
| 1055 | } |
| 1056 | |
| 1057 | void InstructionList::Clear() { m_instructions.clear(); } |
| 1058 | |
| 1059 | void InstructionList::Append(lldb::InstructionSP &inst_sp) { |
| 1060 | if (inst_sp) |
| 1061 | m_instructions.push_back(x: inst_sp); |
| 1062 | } |
| 1063 | |
| 1064 | uint32_t |
| 1065 | InstructionList::GetIndexOfNextBranchInstruction(uint32_t start, |
| 1066 | bool ignore_calls, |
| 1067 | bool *found_calls) const { |
| 1068 | size_t num_instructions = m_instructions.size(); |
| 1069 | |
| 1070 | uint32_t next_branch = UINT32_MAX; |
| 1071 | |
| 1072 | if (found_calls) |
| 1073 | *found_calls = false; |
| 1074 | for (size_t i = start; i < num_instructions; i++) { |
| 1075 | if (m_instructions[i]->DoesBranch()) { |
| 1076 | if (ignore_calls && m_instructions[i]->IsCall()) { |
| 1077 | if (found_calls) |
| 1078 | *found_calls = true; |
| 1079 | continue; |
| 1080 | } |
| 1081 | next_branch = i; |
| 1082 | break; |
| 1083 | } |
| 1084 | } |
| 1085 | |
| 1086 | return next_branch; |
| 1087 | } |
| 1088 | |
| 1089 | uint32_t |
| 1090 | InstructionList::GetIndexOfInstructionAtAddress(const Address &address) { |
| 1091 | size_t num_instructions = m_instructions.size(); |
| 1092 | uint32_t index = UINT32_MAX; |
| 1093 | for (size_t i = 0; i < num_instructions; i++) { |
| 1094 | if (m_instructions[i]->GetAddress() == address) { |
| 1095 | index = i; |
| 1096 | break; |
| 1097 | } |
| 1098 | } |
| 1099 | return index; |
| 1100 | } |
| 1101 | |
| 1102 | uint32_t |
| 1103 | InstructionList::GetIndexOfInstructionAtLoadAddress(lldb::addr_t load_addr, |
| 1104 | Target &target) { |
| 1105 | Address address; |
| 1106 | address.SetLoadAddress(load_addr, target: &target); |
| 1107 | return GetIndexOfInstructionAtAddress(address); |
| 1108 | } |
| 1109 | |
| 1110 | size_t Disassembler::AppendInstructions(Target &target, Address start, |
| 1111 | Limit limit, Stream *error_strm_ptr, |
| 1112 | bool force_live_memory) { |
| 1113 | if (!start.IsValid()) |
| 1114 | return 0; |
| 1115 | |
| 1116 | start = ResolveAddress(target, addr: start); |
| 1117 | |
| 1118 | addr_t byte_size = limit.value; |
| 1119 | if (limit.kind == Limit::Instructions) |
| 1120 | byte_size *= m_arch.GetMaximumOpcodeByteSize(); |
| 1121 | auto data_sp = std::make_shared<DataBufferHeap>(args&: byte_size, args: '\0'); |
| 1122 | |
| 1123 | Status error; |
| 1124 | lldb::addr_t load_addr = LLDB_INVALID_ADDRESS; |
| 1125 | const size_t bytes_read = |
| 1126 | target.ReadMemory(addr: start, dst: data_sp->GetBytes(), dst_len: data_sp->GetByteSize(), |
| 1127 | error, force_live_memory, load_addr_ptr: &load_addr); |
| 1128 | const bool data_from_file = load_addr == LLDB_INVALID_ADDRESS; |
| 1129 | |
| 1130 | if (bytes_read == 0) { |
| 1131 | if (error_strm_ptr) { |
| 1132 | if (const char *error_cstr = error.AsCString()) |
| 1133 | error_strm_ptr->Printf(format: "error: %s\n" , error_cstr); |
| 1134 | } |
| 1135 | return 0; |
| 1136 | } |
| 1137 | |
| 1138 | if (bytes_read != data_sp->GetByteSize()) |
| 1139 | data_sp->SetByteSize(bytes_read); |
| 1140 | DataExtractor data(data_sp, m_arch.GetByteOrder(), |
| 1141 | m_arch.GetAddressByteSize()); |
| 1142 | return DecodeInstructions(base_addr: start, data, data_offset: 0, |
| 1143 | num_instructions: limit.kind == Limit::Instructions ? limit.value |
| 1144 | : UINT32_MAX, |
| 1145 | /*append=*/true, data_from_file); |
| 1146 | } |
| 1147 | |
| 1148 | // Disassembler copy constructor |
| 1149 | Disassembler::Disassembler(const ArchSpec &arch, const char *flavor) |
| 1150 | : m_arch(arch), m_instruction_list(), m_flavor() { |
| 1151 | if (flavor == nullptr) |
| 1152 | m_flavor.assign(s: "default" ); |
| 1153 | else |
| 1154 | m_flavor.assign(s: flavor); |
| 1155 | |
| 1156 | // If this is an arm variant that can only include thumb (T16, T32) |
| 1157 | // instructions, force the arch triple to be "thumbv.." instead of "armv..." |
| 1158 | if (arch.IsAlwaysThumbInstructions()) { |
| 1159 | std::string thumb_arch_name(arch.GetTriple().getArchName().str()); |
| 1160 | // Replace "arm" with "thumb" so we get all thumb variants correct |
| 1161 | if (thumb_arch_name.size() > 3) { |
| 1162 | thumb_arch_name.erase(pos: 0, n: 3); |
| 1163 | thumb_arch_name.insert(pos: 0, s: "thumb" ); |
| 1164 | } |
| 1165 | m_arch.SetTriple(thumb_arch_name.c_str()); |
| 1166 | } |
| 1167 | } |
| 1168 | |
| 1169 | Disassembler::~Disassembler() = default; |
| 1170 | |
| 1171 | InstructionList &Disassembler::GetInstructionList() { |
| 1172 | return m_instruction_list; |
| 1173 | } |
| 1174 | |
| 1175 | const InstructionList &Disassembler::GetInstructionList() const { |
| 1176 | return m_instruction_list; |
| 1177 | } |
| 1178 | |
| 1179 | // Class PseudoInstruction |
| 1180 | |
| 1181 | PseudoInstruction::PseudoInstruction() |
| 1182 | : Instruction(Address(), AddressClass::eUnknown), m_description() {} |
| 1183 | |
| 1184 | PseudoInstruction::~PseudoInstruction() = default; |
| 1185 | |
| 1186 | bool PseudoInstruction::DoesBranch() { |
| 1187 | // This is NOT a valid question for a pseudo instruction. |
| 1188 | return false; |
| 1189 | } |
| 1190 | |
| 1191 | bool PseudoInstruction::HasDelaySlot() { |
| 1192 | // This is NOT a valid question for a pseudo instruction. |
| 1193 | return false; |
| 1194 | } |
| 1195 | |
| 1196 | bool PseudoInstruction::IsLoad() { return false; } |
| 1197 | |
| 1198 | bool PseudoInstruction::IsAuthenticated() { return false; } |
| 1199 | |
| 1200 | size_t PseudoInstruction::(const lldb_private::Disassembler &disassembler, |
| 1201 | const lldb_private::DataExtractor &data, |
| 1202 | lldb::offset_t data_offset) { |
| 1203 | return m_opcode.GetByteSize(); |
| 1204 | } |
| 1205 | |
| 1206 | void PseudoInstruction::SetOpcode(size_t opcode_size, void *opcode_data) { |
| 1207 | if (!opcode_data) |
| 1208 | return; |
| 1209 | |
| 1210 | switch (opcode_size) { |
| 1211 | case 8: { |
| 1212 | uint8_t value8 = *((uint8_t *)opcode_data); |
| 1213 | m_opcode.SetOpcode8(inst: value8, order: eByteOrderInvalid); |
| 1214 | break; |
| 1215 | } |
| 1216 | case 16: { |
| 1217 | uint16_t value16 = *((uint16_t *)opcode_data); |
| 1218 | m_opcode.SetOpcode16(inst: value16, order: eByteOrderInvalid); |
| 1219 | break; |
| 1220 | } |
| 1221 | case 32: { |
| 1222 | uint32_t value32 = *((uint32_t *)opcode_data); |
| 1223 | m_opcode.SetOpcode32(inst: value32, order: eByteOrderInvalid); |
| 1224 | break; |
| 1225 | } |
| 1226 | case 64: { |
| 1227 | uint64_t value64 = *((uint64_t *)opcode_data); |
| 1228 | m_opcode.SetOpcode64(inst: value64, order: eByteOrderInvalid); |
| 1229 | break; |
| 1230 | } |
| 1231 | default: |
| 1232 | break; |
| 1233 | } |
| 1234 | } |
| 1235 | |
| 1236 | void PseudoInstruction::SetDescription(llvm::StringRef description) { |
| 1237 | m_description = std::string(description); |
| 1238 | } |
| 1239 | |
| 1240 | Instruction::Operand Instruction::Operand::BuildRegister(ConstString &r) { |
| 1241 | Operand ret; |
| 1242 | ret.m_type = Type::Register; |
| 1243 | ret.m_register = r; |
| 1244 | return ret; |
| 1245 | } |
| 1246 | |
| 1247 | Instruction::Operand Instruction::Operand::BuildImmediate(lldb::addr_t imm, |
| 1248 | bool neg) { |
| 1249 | Operand ret; |
| 1250 | ret.m_type = Type::Immediate; |
| 1251 | ret.m_immediate = imm; |
| 1252 | ret.m_negative = neg; |
| 1253 | return ret; |
| 1254 | } |
| 1255 | |
| 1256 | Instruction::Operand Instruction::Operand::BuildImmediate(int64_t imm) { |
| 1257 | Operand ret; |
| 1258 | ret.m_type = Type::Immediate; |
| 1259 | if (imm < 0) { |
| 1260 | ret.m_immediate = -imm; |
| 1261 | ret.m_negative = true; |
| 1262 | } else { |
| 1263 | ret.m_immediate = imm; |
| 1264 | ret.m_negative = false; |
| 1265 | } |
| 1266 | return ret; |
| 1267 | } |
| 1268 | |
| 1269 | Instruction::Operand |
| 1270 | Instruction::Operand::BuildDereference(const Operand &ref) { |
| 1271 | Operand ret; |
| 1272 | ret.m_type = Type::Dereference; |
| 1273 | ret.m_children = {ref}; |
| 1274 | return ret; |
| 1275 | } |
| 1276 | |
| 1277 | Instruction::Operand Instruction::Operand::BuildSum(const Operand &lhs, |
| 1278 | const Operand &rhs) { |
| 1279 | Operand ret; |
| 1280 | ret.m_type = Type::Sum; |
| 1281 | ret.m_children = {lhs, rhs}; |
| 1282 | return ret; |
| 1283 | } |
| 1284 | |
| 1285 | Instruction::Operand Instruction::Operand::BuildProduct(const Operand &lhs, |
| 1286 | const Operand &rhs) { |
| 1287 | Operand ret; |
| 1288 | ret.m_type = Type::Product; |
| 1289 | ret.m_children = {lhs, rhs}; |
| 1290 | return ret; |
| 1291 | } |
| 1292 | |
| 1293 | std::function<bool(const Instruction::Operand &)> |
| 1294 | lldb_private::OperandMatchers::MatchBinaryOp( |
| 1295 | std::function<bool(const Instruction::Operand &)> base, |
| 1296 | std::function<bool(const Instruction::Operand &)> left, |
| 1297 | std::function<bool(const Instruction::Operand &)> right) { |
| 1298 | return [base, left, right](const Instruction::Operand &op) -> bool { |
| 1299 | return (base(op) && op.m_children.size() == 2 && |
| 1300 | ((left(op.m_children[0]) && right(op.m_children[1])) || |
| 1301 | (left(op.m_children[1]) && right(op.m_children[0])))); |
| 1302 | }; |
| 1303 | } |
| 1304 | |
| 1305 | std::function<bool(const Instruction::Operand &)> |
| 1306 | lldb_private::OperandMatchers::MatchUnaryOp( |
| 1307 | std::function<bool(const Instruction::Operand &)> base, |
| 1308 | std::function<bool(const Instruction::Operand &)> child) { |
| 1309 | return [base, child](const Instruction::Operand &op) -> bool { |
| 1310 | return (base(op) && op.m_children.size() == 1 && child(op.m_children[0])); |
| 1311 | }; |
| 1312 | } |
| 1313 | |
| 1314 | std::function<bool(const Instruction::Operand &)> |
| 1315 | lldb_private::OperandMatchers::MatchRegOp(const RegisterInfo &info) { |
| 1316 | return [&info](const Instruction::Operand &op) { |
| 1317 | return (op.m_type == Instruction::Operand::Type::Register && |
| 1318 | (op.m_register == ConstString(info.name) || |
| 1319 | op.m_register == ConstString(info.alt_name))); |
| 1320 | }; |
| 1321 | } |
| 1322 | |
| 1323 | std::function<bool(const Instruction::Operand &)> |
| 1324 | lldb_private::OperandMatchers::FetchRegOp(ConstString ®) { |
| 1325 | return [®](const Instruction::Operand &op) { |
| 1326 | if (op.m_type != Instruction::Operand::Type::Register) { |
| 1327 | return false; |
| 1328 | } |
| 1329 | reg = op.m_register; |
| 1330 | return true; |
| 1331 | }; |
| 1332 | } |
| 1333 | |
| 1334 | std::function<bool(const Instruction::Operand &)> |
| 1335 | lldb_private::OperandMatchers::MatchImmOp(int64_t imm) { |
| 1336 | return [imm](const Instruction::Operand &op) { |
| 1337 | return (op.m_type == Instruction::Operand::Type::Immediate && |
| 1338 | ((op.m_negative && op.m_immediate == (uint64_t)-imm) || |
| 1339 | (!op.m_negative && op.m_immediate == (uint64_t)imm))); |
| 1340 | }; |
| 1341 | } |
| 1342 | |
| 1343 | std::function<bool(const Instruction::Operand &)> |
| 1344 | lldb_private::OperandMatchers::FetchImmOp(int64_t &imm) { |
| 1345 | return [&imm](const Instruction::Operand &op) { |
| 1346 | if (op.m_type != Instruction::Operand::Type::Immediate) { |
| 1347 | return false; |
| 1348 | } |
| 1349 | if (op.m_negative) { |
| 1350 | imm = -((int64_t)op.m_immediate); |
| 1351 | } else { |
| 1352 | imm = ((int64_t)op.m_immediate); |
| 1353 | } |
| 1354 | return true; |
| 1355 | }; |
| 1356 | } |
| 1357 | |
| 1358 | std::function<bool(const Instruction::Operand &)> |
| 1359 | lldb_private::OperandMatchers::MatchOpType(Instruction::Operand::Type type) { |
| 1360 | return [type](const Instruction::Operand &op) { return op.m_type == type; }; |
| 1361 | } |
| 1362 | |