1 //===-- Disassembler.cpp ----------------------------------------*- C++ -*-===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 #include "lldb/Core/Disassembler.h"
12 #include "lldb/Core/AddressRange.h" // for AddressRange
13 #include "lldb/Core/Debugger.h"
14 #include "lldb/Core/EmulateInstruction.h"
15 #include "lldb/Core/Mangled.h" // for Mangled, Mangled...
16 #include "lldb/Core/Module.h"
17 #include "lldb/Core/ModuleList.h" // for ModuleList
18 #include "lldb/Core/PluginManager.h"
19 #include "lldb/Core/SourceManager.h" // for SourceManager
20 #include "lldb/Core/Timer.h"
21 #include "lldb/Host/FileSystem.h"
22 #include "lldb/Interpreter/OptionValue.h"
23 #include "lldb/Interpreter/OptionValueArray.h"
24 #include "lldb/Interpreter/OptionValueDictionary.h"
25 #include "lldb/Interpreter/OptionValueRegex.h"
26 #include "lldb/Interpreter/OptionValueString.h"
27 #include "lldb/Interpreter/OptionValueUInt64.h"
28 #include "lldb/Symbol/Function.h"
29 #include "lldb/Symbol/Symbol.h" // for Symbol
30 #include "lldb/Symbol/SymbolContext.h" // for SymbolContext
31 #include "lldb/Target/ExecutionContext.h"
32 #include "lldb/Target/SectionLoadList.h"
33 #include "lldb/Target/StackFrame.h"
34 #include "lldb/Target/Target.h"
35 #include "lldb/Target/Thread.h" // for Thread
36 #include "lldb/Utility/DataBufferHeap.h"
37 #include "lldb/Utility/DataExtractor.h"
38 #include "lldb/Utility/Error.h"
39 #include "lldb/Utility/RegularExpression.h"
40 #include "lldb/Utility/Stream.h" // for Stream
41 #include "lldb/Utility/StreamString.h" // for StreamString
42 #include "lldb/lldb-private-enumerations.h" // for InstructionType:...
43 #include "lldb/lldb-private-interfaces.h" // for DisassemblerCrea...
44 #include "lldb/lldb-private-types.h" // for RegisterInfo
45 #include "llvm/ADT/Triple.h" // for Triple, Triple::...
46 #include "llvm/Support/Compiler.h" // for LLVM_PRETTY_FUNC...
48 #include <cstdint> // for uint32_t, UINT32...
50 #include <utility> // for pair
52 #include <assert.h> // for assert
54 #define DEFAULT_DISASM_BYTE_SIZE 32
57 using namespace lldb_private;
59 DisassemblerSP Disassembler::FindPlugin(const ArchSpec &arch,
61 const char *plugin_name) {
62 Timer scoped_timer(LLVM_PRETTY_FUNCTION,
63 "Disassembler::FindPlugin (arch = %s, plugin_name = %s)",
64 arch.GetArchitectureName(), plugin_name);
66 DisassemblerCreateInstance create_callback = nullptr;
69 ConstString const_plugin_name(plugin_name);
70 create_callback = PluginManager::GetDisassemblerCreateCallbackForPluginName(
72 if (create_callback) {
73 DisassemblerSP disassembler_sp(create_callback(arch, flavor));
76 return disassembler_sp;
79 for (uint32_t idx = 0;
80 (create_callback = PluginManager::GetDisassemblerCreateCallbackAtIndex(
83 DisassemblerSP disassembler_sp(create_callback(arch, flavor));
86 return disassembler_sp;
89 return DisassemblerSP();
92 DisassemblerSP Disassembler::FindPluginForTarget(const TargetSP target_sp,
95 const char *plugin_name) {
96 if (target_sp && flavor == nullptr) {
97 // FIXME - we don't have the mechanism in place to do per-architecture
98 // settings. But since we know that for now
99 // we only support flavors on x86 & x86_64,
100 if (arch.GetTriple().getArch() == llvm::Triple::x86 ||
101 arch.GetTriple().getArch() == llvm::Triple::x86_64)
102 flavor = target_sp->GetDisassemblyFlavor();
104 return FindPlugin(arch, flavor, plugin_name);
107 static void ResolveAddress(const ExecutionContext &exe_ctx, const Address &addr,
108 Address &resolved_addr) {
109 if (!addr.IsSectionOffset()) {
110 // If we weren't passed in a section offset address range,
111 // try and resolve it to something
112 Target *target = exe_ctx.GetTargetPtr();
114 if (target->GetSectionLoadList().IsEmpty()) {
115 target->GetImages().ResolveFileAddress(addr.GetOffset(), resolved_addr);
117 target->GetSectionLoadList().ResolveLoadAddress(addr.GetOffset(),
120 // We weren't able to resolve the address, just treat it as a
122 if (resolved_addr.IsValid())
126 resolved_addr = addr;
129 size_t Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
130 const char *plugin_name, const char *flavor,
131 const ExecutionContext &exe_ctx,
132 SymbolContextList &sc_list,
133 uint32_t num_instructions,
134 bool mixed_source_and_assembly,
135 uint32_t num_mixed_context_lines,
136 uint32_t options, Stream &strm) {
137 size_t success_count = 0;
138 const size_t count = sc_list.GetSize();
141 const uint32_t scope =
142 eSymbolContextBlock | eSymbolContextFunction | eSymbolContextSymbol;
143 const bool use_inline_block_range = true;
144 for (size_t i = 0; i < count; ++i) {
145 if (!sc_list.GetContextAtIndex(i, sc))
147 for (uint32_t range_idx = 0;
148 sc.GetAddressRange(scope, range_idx, use_inline_block_range, range);
150 if (Disassemble(debugger, arch, plugin_name, flavor, exe_ctx, range,
151 num_instructions, mixed_source_and_assembly,
152 num_mixed_context_lines, options, strm)) {
158 return success_count;
161 bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
162 const char *plugin_name, const char *flavor,
163 const ExecutionContext &exe_ctx,
164 const ConstString &name, Module *module,
165 uint32_t num_instructions,
166 bool mixed_source_and_assembly,
167 uint32_t num_mixed_context_lines,
168 uint32_t options, Stream &strm) {
169 SymbolContextList sc_list;
171 const bool include_symbols = true;
172 const bool include_inlines = true;
174 module->FindFunctions(name, nullptr, eFunctionNameTypeAuto,
175 include_symbols, include_inlines, true, sc_list);
176 } else if (exe_ctx.GetTargetPtr()) {
177 exe_ctx.GetTargetPtr()->GetImages().FindFunctions(
178 name, eFunctionNameTypeAuto, include_symbols, include_inlines, false,
183 if (sc_list.GetSize()) {
184 return Disassemble(debugger, arch, plugin_name, flavor, exe_ctx, sc_list,
185 num_instructions, mixed_source_and_assembly,
186 num_mixed_context_lines, options, strm);
191 lldb::DisassemblerSP Disassembler::DisassembleRange(
192 const ArchSpec &arch, const char *plugin_name, const char *flavor,
193 const ExecutionContext &exe_ctx, const AddressRange &range,
194 bool prefer_file_cache) {
195 lldb::DisassemblerSP disasm_sp;
196 if (range.GetByteSize() > 0 && range.GetBaseAddress().IsValid()) {
197 disasm_sp = Disassembler::FindPluginForTarget(exe_ctx.GetTargetSP(), arch,
198 flavor, plugin_name);
201 size_t bytes_disassembled = disasm_sp->ParseInstructions(
202 &exe_ctx, range, nullptr, prefer_file_cache);
203 if (bytes_disassembled == 0)
211 Disassembler::DisassembleBytes(const ArchSpec &arch, const char *plugin_name,
212 const char *flavor, const Address &start,
213 const void *src, size_t src_len,
214 uint32_t num_instructions, bool data_from_file) {
215 lldb::DisassemblerSP disasm_sp;
218 disasm_sp = Disassembler::FindPlugin(arch, flavor, plugin_name);
221 DataExtractor data(src, src_len, arch.GetByteOrder(),
222 arch.GetAddressByteSize());
224 (void)disasm_sp->DecodeInstructions(start, data, 0, num_instructions,
225 false, data_from_file);
232 bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
233 const char *plugin_name, const char *flavor,
234 const ExecutionContext &exe_ctx,
235 const AddressRange &disasm_range,
236 uint32_t num_instructions,
237 bool mixed_source_and_assembly,
238 uint32_t num_mixed_context_lines,
239 uint32_t options, Stream &strm) {
240 if (disasm_range.GetByteSize()) {
241 lldb::DisassemblerSP disasm_sp(Disassembler::FindPluginForTarget(
242 exe_ctx.GetTargetSP(), arch, flavor, plugin_name));
246 ResolveAddress(exe_ctx, disasm_range.GetBaseAddress(),
247 range.GetBaseAddress());
248 range.SetByteSize(disasm_range.GetByteSize());
249 const bool prefer_file_cache = false;
250 size_t bytes_disassembled = disasm_sp->ParseInstructions(
251 &exe_ctx, range, &strm, prefer_file_cache);
252 if (bytes_disassembled == 0)
255 return PrintInstructions(disasm_sp.get(), debugger, arch, exe_ctx,
256 num_instructions, mixed_source_and_assembly,
257 num_mixed_context_lines, options, strm);
263 bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
264 const char *plugin_name, const char *flavor,
265 const ExecutionContext &exe_ctx,
266 const Address &start_address,
267 uint32_t num_instructions,
268 bool mixed_source_and_assembly,
269 uint32_t num_mixed_context_lines,
270 uint32_t options, Stream &strm) {
271 if (num_instructions > 0) {
272 lldb::DisassemblerSP disasm_sp(Disassembler::FindPluginForTarget(
273 exe_ctx.GetTargetSP(), arch, flavor, plugin_name));
276 ResolveAddress(exe_ctx, start_address, addr);
277 const bool prefer_file_cache = false;
278 size_t bytes_disassembled = disasm_sp->ParseInstructions(
279 &exe_ctx, addr, num_instructions, prefer_file_cache);
280 if (bytes_disassembled == 0)
282 return PrintInstructions(disasm_sp.get(), debugger, arch, exe_ctx,
283 num_instructions, mixed_source_and_assembly,
284 num_mixed_context_lines, options, strm);
290 Disassembler::SourceLine
291 Disassembler::GetFunctionDeclLineEntry(const SymbolContext &sc) {
292 SourceLine decl_line;
293 if (sc.function && sc.line_entry.IsValid()) {
294 LineEntry prologue_end_line = sc.line_entry;
295 FileSpec func_decl_file;
296 uint32_t func_decl_line;
297 sc.function->GetStartLineSourceInfo(func_decl_file, func_decl_line);
298 if (func_decl_file == prologue_end_line.file ||
299 func_decl_file == prologue_end_line.original_file) {
300 decl_line.file = func_decl_file;
301 decl_line.line = func_decl_line;
302 // TODO do we care about column on these entries? If so, we need to
303 // plumb that through GetStartLineSourceInfo.
304 decl_line.column = 0;
310 void Disassembler::AddLineToSourceLineTables(
312 std::map<FileSpec, std::set<uint32_t>> &source_lines_seen) {
313 if (line.IsValid()) {
314 auto source_lines_seen_pos = source_lines_seen.find(line.file);
315 if (source_lines_seen_pos == source_lines_seen.end()) {
316 std::set<uint32_t> lines;
317 lines.insert(line.line);
318 source_lines_seen.emplace(line.file, lines);
320 source_lines_seen_pos->second.insert(line.line);
325 bool Disassembler::ElideMixedSourceAndDisassemblyLine(
326 const ExecutionContext &exe_ctx, const SymbolContext &sc,
329 // TODO: should we also check target.process.thread.step-avoid-libraries ?
331 const RegularExpression *avoid_regex = nullptr;
333 // Skip any line #0 entries - they are implementation details
337 ThreadSP thread_sp = exe_ctx.GetThreadSP();
339 avoid_regex = thread_sp->GetSymbolsToAvoidRegexp();
341 TargetSP target_sp = exe_ctx.GetTargetSP();
344 OptionValueSP value_sp = target_sp->GetDebugger().GetPropertyValue(
345 &exe_ctx, "target.process.thread.step-avoid-regexp", false, error);
346 if (value_sp && value_sp->GetType() == OptionValue::eTypeRegex) {
347 OptionValueRegex *re = value_sp->GetAsRegex();
349 avoid_regex = re->GetCurrentValue();
354 if (avoid_regex && sc.symbol != nullptr) {
355 const char *function_name =
356 sc.GetFunctionName(Mangled::ePreferDemangledWithoutArguments)
359 RegularExpression::Match regex_match(1);
360 if (avoid_regex->Execute(function_name, ®ex_match)) {
361 // skip this source line
366 // don't skip this source line
370 bool Disassembler::PrintInstructions(Disassembler *disasm_ptr,
371 Debugger &debugger, const ArchSpec &arch,
372 const ExecutionContext &exe_ctx,
373 uint32_t num_instructions,
374 bool mixed_source_and_assembly,
375 uint32_t num_mixed_context_lines,
376 uint32_t options, Stream &strm) {
377 // We got some things disassembled...
378 size_t num_instructions_found = disasm_ptr->GetInstructionList().GetSize();
380 if (num_instructions > 0 && num_instructions < num_instructions_found)
381 num_instructions_found = num_instructions;
383 const uint32_t max_opcode_byte_size =
384 disasm_ptr->GetInstructionList().GetMaxOpcocdeByteSize();
386 SymbolContext prev_sc;
387 AddressRange current_source_line_range;
388 const Address *pc_addr_ptr = nullptr;
389 StackFrame *frame = exe_ctx.GetFramePtr();
391 TargetSP target_sp(exe_ctx.GetTargetSP());
392 SourceManager &source_manager =
393 target_sp ? target_sp->GetSourceManager() : debugger.GetSourceManager();
396 pc_addr_ptr = &frame->GetFrameCodeAddress();
398 const uint32_t scope =
399 eSymbolContextLineEntry | eSymbolContextFunction | eSymbolContextSymbol;
400 const bool use_inline_block_range = false;
402 const FormatEntity::Entry *disassembly_format = nullptr;
403 FormatEntity::Entry format;
404 if (exe_ctx.HasTargetScope()) {
406 exe_ctx.GetTargetRef().GetDebugger().GetDisassemblyFormat();
408 FormatEntity::Parse("${addr}: ", format);
409 disassembly_format = &format;
412 // First pass: step through the list of instructions,
413 // find how long the initial addresses strings are, insert padding
414 // in the second pass so the opcodes all line up nicely.
416 // Also build up the source line mapping if this is mixed source & assembly
418 // Calculate the source line for each assembly instruction (eliding inlined
420 // which the user wants to skip).
422 std::map<FileSpec, std::set<uint32_t>> source_lines_seen;
423 Symbol *previous_symbol = nullptr;
425 size_t address_text_size = 0;
426 for (size_t i = 0; i < num_instructions_found; ++i) {
428 disasm_ptr->GetInstructionList().GetInstructionAtIndex(i).get();
430 const Address &addr = inst->GetAddress();
431 ModuleSP module_sp(addr.GetModule());
433 const uint32_t resolve_mask = eSymbolContextFunction |
434 eSymbolContextSymbol |
435 eSymbolContextLineEntry;
436 uint32_t resolved_mask =
437 module_sp->ResolveSymbolContextForAddress(addr, resolve_mask, sc);
439 StreamString strmstr;
440 Debugger::FormatDisassemblerAddress(disassembly_format, &sc, nullptr,
441 &exe_ctx, &addr, strmstr);
442 size_t cur_line = strmstr.GetSizeOfLastLine();
443 if (cur_line > address_text_size)
444 address_text_size = cur_line;
446 // Add entries to our "source_lines_seen" map+set which list which
447 // sources lines occur in this disassembly session. We will print
448 // lines of context around a source line, but we don't want to print
449 // a source line that has a line table entry of its own - we'll leave
450 // that source line to be printed when it actually occurs in the
453 if (mixed_source_and_assembly && sc.line_entry.IsValid()) {
454 if (sc.symbol != previous_symbol) {
455 SourceLine decl_line = GetFunctionDeclLineEntry(sc);
456 if (ElideMixedSourceAndDisassemblyLine(exe_ctx, sc, decl_line) ==
458 AddLineToSourceLineTables(decl_line, source_lines_seen);
460 if (sc.line_entry.IsValid()) {
461 SourceLine this_line;
462 this_line.file = sc.line_entry.file;
463 this_line.line = sc.line_entry.line;
464 this_line.column = sc.line_entry.column;
465 if (ElideMixedSourceAndDisassemblyLine(exe_ctx, sc, this_line) ==
467 AddLineToSourceLineTables(this_line, source_lines_seen);
476 previous_symbol = nullptr;
477 SourceLine previous_line;
478 for (size_t i = 0; i < num_instructions_found; ++i) {
480 disasm_ptr->GetInstructionList().GetInstructionAtIndex(i).get();
483 const Address &addr = inst->GetAddress();
484 const bool inst_is_at_pc = pc_addr_ptr && addr == *pc_addr_ptr;
485 SourceLinesToDisplay source_lines_to_display;
489 ModuleSP module_sp(addr.GetModule());
491 uint32_t resolved_mask = module_sp->ResolveSymbolContextForAddress(
492 addr, eSymbolContextEverything, sc);
494 if (mixed_source_and_assembly) {
496 // If we've started a new function (non-inlined), print all of the
497 // source lines from the
498 // function declaration until the first line table entry - typically
499 // the opening curly brace of
501 if (previous_symbol != sc.symbol) {
502 // The default disassembly format puts an extra blank line between
504 // when we're displaying the source context for a function, we
506 // a blank line after the source context or we'll end up with two
508 if (previous_symbol != nullptr)
509 source_lines_to_display.print_source_context_end_eol = false;
511 previous_symbol = sc.symbol;
512 if (sc.function && sc.line_entry.IsValid()) {
513 LineEntry prologue_end_line = sc.line_entry;
514 if (ElideMixedSourceAndDisassemblyLine(
515 exe_ctx, sc, prologue_end_line) == false) {
516 FileSpec func_decl_file;
517 uint32_t func_decl_line;
518 sc.function->GetStartLineSourceInfo(func_decl_file,
520 if (func_decl_file == prologue_end_line.file ||
521 func_decl_file == prologue_end_line.original_file) {
522 // Add all the lines between the function declaration
523 // and the first non-prologue source line to the list
524 // of lines to print.
525 for (uint32_t lineno = func_decl_line;
526 lineno <= prologue_end_line.line; lineno++) {
527 SourceLine this_line;
528 this_line.file = func_decl_file;
529 this_line.line = lineno;
530 source_lines_to_display.lines.push_back(this_line);
532 // Mark the last line as the "current" one. Usually
533 // this is the open curly brace.
534 if (source_lines_to_display.lines.size() > 0)
535 source_lines_to_display.current_source_line =
536 source_lines_to_display.lines.size() - 1;
540 sc.GetAddressRange(scope, 0, use_inline_block_range,
541 current_source_line_range);
544 // If we've left a previous source line's address range, print a new
546 if (!current_source_line_range.ContainsFileAddress(addr)) {
547 sc.GetAddressRange(scope, 0, use_inline_block_range,
548 current_source_line_range);
550 if (sc != prev_sc && sc.comp_unit && sc.line_entry.IsValid()) {
551 SourceLine this_line;
552 this_line.file = sc.line_entry.file;
553 this_line.line = sc.line_entry.line;
555 if (ElideMixedSourceAndDisassemblyLine(exe_ctx, sc,
556 this_line) == false) {
557 // Only print this source line if it is different from the
558 // last source line we printed. There may have been inlined
559 // functions between these lines that we elided, resulting in
560 // the same line being printed twice in a row for a contiguous
561 // block of assembly instructions.
562 if (this_line != previous_line) {
564 std::vector<uint32_t> previous_lines;
566 i < num_mixed_context_lines &&
567 (this_line.line - num_mixed_context_lines) > 0;
570 this_line.line - num_mixed_context_lines + i;
571 auto pos = source_lines_seen.find(this_line.file);
572 if (pos != source_lines_seen.end()) {
573 if (pos->second.count(line) == 1) {
574 previous_lines.clear();
576 previous_lines.push_back(line);
580 for (size_t i = 0; i < previous_lines.size(); i++) {
581 SourceLine previous_line;
582 previous_line.file = this_line.file;
583 previous_line.line = previous_lines[i];
584 auto pos = source_lines_seen.find(previous_line.file);
585 if (pos != source_lines_seen.end()) {
586 pos->second.insert(previous_line.line);
588 source_lines_to_display.lines.push_back(previous_line);
591 source_lines_to_display.lines.push_back(this_line);
592 source_lines_to_display.current_source_line =
593 source_lines_to_display.lines.size() - 1;
595 for (uint32_t i = 0; i < num_mixed_context_lines; i++) {
596 SourceLine next_line;
597 next_line.file = this_line.file;
598 next_line.line = this_line.line + i + 1;
599 auto pos = source_lines_seen.find(next_line.file);
600 if (pos != source_lines_seen.end()) {
601 if (pos->second.count(next_line.line) == 1)
603 pos->second.insert(next_line.line);
605 source_lines_to_display.lines.push_back(next_line);
608 previous_line = this_line;
618 if (source_lines_to_display.lines.size() > 0) {
620 for (size_t idx = 0; idx < source_lines_to_display.lines.size();
622 SourceLine ln = source_lines_to_display.lines[idx];
623 const char *line_highlight = "";
624 if (inst_is_at_pc && (options & eOptionMarkPCSourceLine)) {
625 line_highlight = "->";
626 } else if (idx == source_lines_to_display.current_source_line) {
627 line_highlight = "**";
629 source_manager.DisplaySourceLinesWithLineNumbers(
630 ln.file, ln.line, ln.column, 0, 0, line_highlight, &strm);
632 if (source_lines_to_display.print_source_context_end_eol)
636 const bool show_bytes = (options & eOptionShowBytes) != 0;
637 inst->Dump(&strm, max_opcode_byte_size, true, show_bytes, &exe_ctx, &sc,
638 &prev_sc, nullptr, address_text_size);
648 bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
649 const char *plugin_name, const char *flavor,
650 const ExecutionContext &exe_ctx,
651 uint32_t num_instructions,
652 bool mixed_source_and_assembly,
653 uint32_t num_mixed_context_lines,
654 uint32_t options, Stream &strm) {
656 StackFrame *frame = exe_ctx.GetFramePtr();
659 frame->GetSymbolContext(eSymbolContextFunction | eSymbolContextSymbol));
661 range = sc.function->GetAddressRange();
662 } else if (sc.symbol && sc.symbol->ValueIsAddress()) {
663 range.GetBaseAddress() = sc.symbol->GetAddressRef();
664 range.SetByteSize(sc.symbol->GetByteSize());
666 range.GetBaseAddress() = frame->GetFrameCodeAddress();
669 if (range.GetBaseAddress().IsValid() && range.GetByteSize() == 0)
670 range.SetByteSize(DEFAULT_DISASM_BYTE_SIZE);
673 return Disassemble(debugger, arch, plugin_name, flavor, exe_ctx, range,
674 num_instructions, mixed_source_and_assembly,
675 num_mixed_context_lines, options, strm);
678 Instruction::Instruction(const Address &address, AddressClass addr_class)
679 : m_address(address), m_address_class(addr_class), m_opcode(),
680 m_calculated_strings(false) {}
682 Instruction::~Instruction() = default;
684 AddressClass Instruction::GetAddressClass() {
685 if (m_address_class == eAddressClassInvalid)
686 m_address_class = m_address.GetAddressClass();
687 return m_address_class;
690 void Instruction::Dump(lldb_private::Stream *s, uint32_t max_opcode_byte_size,
691 bool show_address, bool show_bytes,
692 const ExecutionContext *exe_ctx,
693 const SymbolContext *sym_ctx,
694 const SymbolContext *prev_sym_ctx,
695 const FormatEntity::Entry *disassembly_addr_format,
696 size_t max_address_text_size) {
697 size_t opcode_column_width = 7;
698 const size_t operand_column_width = 25;
700 CalculateMnemonicOperandsAndCommentIfNeeded(exe_ctx);
705 Debugger::FormatDisassemblerAddress(disassembly_addr_format, sym_ctx,
706 prev_sym_ctx, exe_ctx, &m_address, ss);
707 ss.FillLastLineToColumn(max_address_text_size, ' ');
711 if (m_opcode.GetType() == Opcode::eTypeBytes) {
712 // x86_64 and i386 are the only ones that use bytes right now so
713 // pad out the byte dump to be able to always show 15 bytes (3 chars each)
715 if (max_opcode_byte_size > 0)
716 m_opcode.Dump(&ss, max_opcode_byte_size * 3 + 1);
718 m_opcode.Dump(&ss, 15 * 3 + 1);
720 // Else, we have ARM or MIPS which can show up to a uint32_t
721 // 0x00000000 (10 spaces) plus two for padding...
722 if (max_opcode_byte_size > 0)
723 m_opcode.Dump(&ss, max_opcode_byte_size * 3 + 1);
725 m_opcode.Dump(&ss, 12);
729 const size_t opcode_pos = ss.GetSizeOfLastLine();
731 // The default opcode size of 7 characters is plenty for most architectures
732 // but some like arm can pull out the occasional vqrshrun.s16. We won't get
733 // consistent column spacing in these cases, unfortunately.
734 if (m_opcode_name.length() >= opcode_column_width) {
735 opcode_column_width = m_opcode_name.length() + 1;
738 ss.PutCString(m_opcode_name);
739 ss.FillLastLineToColumn(opcode_pos + opcode_column_width, ' ');
740 ss.PutCString(m_mnemonics);
742 if (!m_comment.empty()) {
743 ss.FillLastLineToColumn(
744 opcode_pos + opcode_column_width + operand_column_width, ' ');
745 ss.PutCString(" ; ");
746 ss.PutCString(m_comment);
748 s->PutCString(ss.GetString());
751 bool Instruction::DumpEmulation(const ArchSpec &arch) {
752 std::unique_ptr<EmulateInstruction> insn_emulator_ap(
753 EmulateInstruction::FindPlugin(arch, eInstructionTypeAny, nullptr));
754 if (insn_emulator_ap) {
755 insn_emulator_ap->SetInstruction(GetOpcode(), GetAddress(), nullptr);
756 return insn_emulator_ap->EvaluateInstruction(0);
762 bool Instruction::HasDelaySlot() {
767 OptionValueSP Instruction::ReadArray(FILE *in_file, Stream *out_stream,
768 OptionValue::Type data_type) {
772 auto option_value_sp = std::make_shared<OptionValueArray>(1u << data_type);
776 if (!fgets(buffer, 1023, in_file)) {
778 "Instruction::ReadArray: Error reading file (fgets).\n");
779 option_value_sp.reset();
780 return option_value_sp;
783 std::string line(buffer);
785 size_t len = line.size();
786 if (line[len - 1] == '\n') {
787 line[len - 1] = '\0';
788 line.resize(len - 1);
791 if ((line.size() == 1) && line[0] == ']') {
798 static RegularExpression g_reg_exp(
799 llvm::StringRef("^[ \t]*([^ \t]+)[ \t]*$"));
800 RegularExpression::Match regex_match(1);
801 bool reg_exp_success = g_reg_exp.Execute(line, ®ex_match);
803 regex_match.GetMatchAtIndex(line.c_str(), 1, value);
807 OptionValueSP data_value_sp;
809 case OptionValue::eTypeUInt64:
810 data_value_sp = std::make_shared<OptionValueUInt64>(0, 0);
811 data_value_sp->SetValueFromString(value);
813 // Other types can be added later as needed.
815 data_value_sp = std::make_shared<OptionValueString>(value.c_str(), "");
819 option_value_sp->GetAsArray()->InsertValue(idx, data_value_sp);
824 return option_value_sp;
827 OptionValueSP Instruction::ReadDictionary(FILE *in_file, Stream *out_stream) {
831 auto option_value_sp = std::make_shared<OptionValueDictionary>();
832 static ConstString encoding_key("data_encoding");
833 OptionValue::Type data_type = OptionValue::eTypeInvalid;
836 // Read the next line in the file
837 if (!fgets(buffer, 1023, in_file)) {
839 "Instruction::ReadDictionary: Error reading file (fgets).\n");
840 option_value_sp.reset();
841 return option_value_sp;
844 // Check to see if the line contains the end-of-dictionary marker ("}")
845 std::string line(buffer);
847 size_t len = line.size();
848 if (line[len - 1] == '\n') {
849 line[len - 1] = '\0';
850 line.resize(len - 1);
853 if ((line.size() == 1) && (line[0] == '}')) {
858 // Try to find a key-value pair in the current line and add it to the
861 static RegularExpression g_reg_exp(llvm::StringRef(
862 "^[ \t]*([a-zA-Z_][a-zA-Z0-9_]*)[ \t]*=[ \t]*(.*)[ \t]*$"));
863 RegularExpression::Match regex_match(2);
865 bool reg_exp_success = g_reg_exp.Execute(line, ®ex_match);
868 if (reg_exp_success) {
869 regex_match.GetMatchAtIndex(line.c_str(), 1, key);
870 regex_match.GetMatchAtIndex(line.c_str(), 2, value);
872 out_stream->Printf("Instruction::ReadDictionary: Failure executing "
873 "regular expression.\n");
874 option_value_sp.reset();
875 return option_value_sp;
878 ConstString const_key(key.c_str());
879 // Check value to see if it's the start of an array or dictionary.
881 lldb::OptionValueSP value_sp;
882 assert(value.empty() == false);
883 assert(key.empty() == false);
885 if (value[0] == '{') {
886 assert(value.size() == 1);
887 // value is a dictionary
888 value_sp = ReadDictionary(in_file, out_stream);
890 option_value_sp.reset();
891 return option_value_sp;
893 } else if (value[0] == '[') {
894 assert(value.size() == 1);
896 value_sp = ReadArray(in_file, out_stream, data_type);
898 option_value_sp.reset();
899 return option_value_sp;
901 // We've used the data_type to read an array; re-set the type to Invalid
902 data_type = OptionValue::eTypeInvalid;
903 } else if ((value[0] == '0') && (value[1] == 'x')) {
904 value_sp = std::make_shared<OptionValueUInt64>(0, 0);
905 value_sp->SetValueFromString(value);
907 size_t len = value.size();
908 if ((value[0] == '"') && (value[len - 1] == '"'))
909 value = value.substr(1, len - 2);
910 value_sp = std::make_shared<OptionValueString>(value.c_str(), "");
913 if (const_key == encoding_key) {
914 // A 'data_encoding=..." is NOT a normal key-value pair; it is meta-data
916 // data type of an upcoming array (usually the next bit of data to be
918 if (strcmp(value.c_str(), "uint32_t") == 0)
919 data_type = OptionValue::eTypeUInt64;
921 option_value_sp->GetAsDictionary()->SetValueForKey(const_key, value_sp,
926 return option_value_sp;
929 bool Instruction::TestEmulation(Stream *out_stream, const char *file_name) {
934 out_stream->Printf("Instruction::TestEmulation: Missing file_name.");
937 FILE *test_file = FileSystem::Fopen(file_name, "r");
940 "Instruction::TestEmulation: Attempt to open test file failed.");
945 if (!fgets(buffer, 255, test_file)) {
947 "Instruction::TestEmulation: Error reading first line of test file.\n");
952 if (strncmp(buffer, "InstructionEmulationState={", 27) != 0) {
953 out_stream->Printf("Instructin::TestEmulation: Test file does not contain "
954 "emulation state dictionary\n");
959 // Read all the test information from the test file into an
960 // OptionValueDictionary.
962 OptionValueSP data_dictionary_sp(ReadDictionary(test_file, out_stream));
963 if (!data_dictionary_sp) {
965 "Instruction::TestEmulation: Error reading Dictionary Object.\n");
972 OptionValueDictionary *data_dictionary =
973 data_dictionary_sp->GetAsDictionary();
974 static ConstString description_key("assembly_string");
975 static ConstString triple_key("triple");
977 OptionValueSP value_sp = data_dictionary->GetValueForKey(description_key);
980 out_stream->Printf("Instruction::TestEmulation: Test file does not "
981 "contain description string.\n");
985 SetDescription(value_sp->GetStringValue());
987 value_sp = data_dictionary->GetValueForKey(triple_key);
990 "Instruction::TestEmulation: Test file does not contain triple.\n");
995 arch.SetTriple(llvm::Triple(value_sp->GetStringValue()));
997 bool success = false;
998 std::unique_ptr<EmulateInstruction> insn_emulator_ap(
999 EmulateInstruction::FindPlugin(arch, eInstructionTypeAny, nullptr));
1000 if (insn_emulator_ap)
1002 insn_emulator_ap->TestEmulation(out_stream, arch, data_dictionary);
1005 out_stream->Printf("Emulation test succeeded.");
1007 out_stream->Printf("Emulation test failed.");
1012 bool Instruction::Emulate(
1013 const ArchSpec &arch, uint32_t evaluate_options, void *baton,
1014 EmulateInstruction::ReadMemoryCallback read_mem_callback,
1015 EmulateInstruction::WriteMemoryCallback write_mem_callback,
1016 EmulateInstruction::ReadRegisterCallback read_reg_callback,
1017 EmulateInstruction::WriteRegisterCallback write_reg_callback) {
1018 std::unique_ptr<EmulateInstruction> insn_emulator_ap(
1019 EmulateInstruction::FindPlugin(arch, eInstructionTypeAny, nullptr));
1020 if (insn_emulator_ap) {
1021 insn_emulator_ap->SetBaton(baton);
1022 insn_emulator_ap->SetCallbacks(read_mem_callback, write_mem_callback,
1023 read_reg_callback, write_reg_callback);
1024 insn_emulator_ap->SetInstruction(GetOpcode(), GetAddress(), nullptr);
1025 return insn_emulator_ap->EvaluateInstruction(evaluate_options);
1031 uint32_t Instruction::GetData(DataExtractor &data) {
1032 return m_opcode.GetData(data);
1035 InstructionList::InstructionList() : m_instructions() {}
1037 InstructionList::~InstructionList() = default;
1039 size_t InstructionList::GetSize() const { return m_instructions.size(); }
1041 uint32_t InstructionList::GetMaxOpcocdeByteSize() const {
1042 uint32_t max_inst_size = 0;
1043 collection::const_iterator pos, end;
1044 for (pos = m_instructions.begin(), end = m_instructions.end(); pos != end;
1046 uint32_t inst_size = (*pos)->GetOpcode().GetByteSize();
1047 if (max_inst_size < inst_size)
1048 max_inst_size = inst_size;
1050 return max_inst_size;
1053 InstructionSP InstructionList::GetInstructionAtIndex(size_t idx) const {
1054 InstructionSP inst_sp;
1055 if (idx < m_instructions.size())
1056 inst_sp = m_instructions[idx];
1060 void InstructionList::Dump(Stream *s, bool show_address, bool show_bytes,
1061 const ExecutionContext *exe_ctx) {
1062 const uint32_t max_opcode_byte_size = GetMaxOpcocdeByteSize();
1063 collection::const_iterator pos, begin, end;
1065 const FormatEntity::Entry *disassembly_format = nullptr;
1066 FormatEntity::Entry format;
1067 if (exe_ctx && exe_ctx->HasTargetScope()) {
1068 disassembly_format =
1069 exe_ctx->GetTargetRef().GetDebugger().GetDisassemblyFormat();
1071 FormatEntity::Parse("${addr}: ", format);
1072 disassembly_format = &format;
1075 for (begin = m_instructions.begin(), end = m_instructions.end(), pos = begin;
1076 pos != end; ++pos) {
1079 (*pos)->Dump(s, max_opcode_byte_size, show_address, show_bytes, exe_ctx,
1080 nullptr, nullptr, disassembly_format, 0);
1084 void InstructionList::Clear() { m_instructions.clear(); }
1086 void InstructionList::Append(lldb::InstructionSP &inst_sp) {
1088 m_instructions.push_back(inst_sp);
1092 InstructionList::GetIndexOfNextBranchInstruction(uint32_t start,
1093 Target &target) const {
1094 size_t num_instructions = m_instructions.size();
1096 uint32_t next_branch = UINT32_MAX;
1098 for (i = start; i < num_instructions; i++) {
1099 if (m_instructions[i]->DoesBranch()) {
1105 // Hexagon needs the first instruction of the packet with the branch.
1106 // Go backwards until we find an instruction marked end-of-packet, or
1107 // until we hit start.
1108 if (target.GetArchitecture().GetTriple().getArch() == llvm::Triple::hexagon) {
1109 // If we didn't find a branch, find the last packet start.
1110 if (next_branch == UINT32_MAX) {
1111 i = num_instructions - 1;
1118 uint32_t inst_bytes;
1119 bool prefer_file_cache = false; // Read from process if process is running
1120 lldb::addr_t load_addr = LLDB_INVALID_ADDRESS;
1121 target.ReadMemory(m_instructions[i]->GetAddress(), prefer_file_cache,
1122 &inst_bytes, sizeof(inst_bytes), error, &load_addr);
1123 // If we have an error reading memory, return start
1124 if (!error.Success())
1126 // check if this is the last instruction in a packet
1127 // bits 15:14 will be 11b or 00b for a duplex
1128 if (((inst_bytes & 0xC000) == 0xC000) ||
1129 ((inst_bytes & 0xC000) == 0x0000)) {
1130 // instruction after this should be the start of next packet
1131 next_branch = i + 1;
1136 if (next_branch == UINT32_MAX) {
1137 // We couldn't find the previous packet, so return start
1138 next_branch = start;
1145 InstructionList::GetIndexOfInstructionAtAddress(const Address &address) {
1146 size_t num_instructions = m_instructions.size();
1147 uint32_t index = UINT32_MAX;
1148 for (size_t i = 0; i < num_instructions; i++) {
1149 if (m_instructions[i]->GetAddress() == address) {
1158 InstructionList::GetIndexOfInstructionAtLoadAddress(lldb::addr_t load_addr,
1161 address.SetLoadAddress(load_addr, &target);
1162 return GetIndexOfInstructionAtAddress(address);
1165 size_t Disassembler::ParseInstructions(const ExecutionContext *exe_ctx,
1166 const AddressRange &range,
1167 Stream *error_strm_ptr,
1168 bool prefer_file_cache) {
1170 Target *target = exe_ctx->GetTargetPtr();
1171 const addr_t byte_size = range.GetByteSize();
1172 if (target == nullptr || byte_size == 0 ||
1173 !range.GetBaseAddress().IsValid())
1176 auto data_sp = std::make_shared<DataBufferHeap>(byte_size, '\0');
1179 lldb::addr_t load_addr = LLDB_INVALID_ADDRESS;
1180 const size_t bytes_read = target->ReadMemory(
1181 range.GetBaseAddress(), prefer_file_cache, data_sp->GetBytes(),
1182 data_sp->GetByteSize(), error, &load_addr);
1184 if (bytes_read > 0) {
1185 if (bytes_read != data_sp->GetByteSize())
1186 data_sp->SetByteSize(bytes_read);
1187 DataExtractor data(data_sp, m_arch.GetByteOrder(),
1188 m_arch.GetAddressByteSize());
1189 const bool data_from_file = load_addr == LLDB_INVALID_ADDRESS;
1190 return DecodeInstructions(range.GetBaseAddress(), data, 0, UINT32_MAX,
1191 false, data_from_file);
1192 } else if (error_strm_ptr) {
1193 const char *error_cstr = error.AsCString();
1195 error_strm_ptr->Printf("error: %s\n", error_cstr);
1198 } else if (error_strm_ptr) {
1199 error_strm_ptr->PutCString("error: invalid execution context\n");
1204 size_t Disassembler::ParseInstructions(const ExecutionContext *exe_ctx,
1205 const Address &start,
1206 uint32_t num_instructions,
1207 bool prefer_file_cache) {
1208 m_instruction_list.Clear();
1210 if (exe_ctx == nullptr || num_instructions == 0 || !start.IsValid())
1213 Target *target = exe_ctx->GetTargetPtr();
1214 // Calculate the max buffer size we will need in order to disassemble
1215 const addr_t byte_size = num_instructions * m_arch.GetMaximumOpcodeByteSize();
1217 if (target == nullptr || byte_size == 0)
1220 DataBufferHeap *heap_buffer = new DataBufferHeap(byte_size, '\0');
1221 DataBufferSP data_sp(heap_buffer);
1224 lldb::addr_t load_addr = LLDB_INVALID_ADDRESS;
1225 const size_t bytes_read =
1226 target->ReadMemory(start, prefer_file_cache, heap_buffer->GetBytes(),
1227 byte_size, error, &load_addr);
1229 const bool data_from_file = load_addr == LLDB_INVALID_ADDRESS;
1231 if (bytes_read == 0)
1233 DataExtractor data(data_sp, m_arch.GetByteOrder(),
1234 m_arch.GetAddressByteSize());
1236 const bool append_instructions = true;
1237 DecodeInstructions(start, data, 0, num_instructions, append_instructions,
1240 return m_instruction_list.GetSize();
1243 //----------------------------------------------------------------------
1244 // Disassembler copy constructor
1245 //----------------------------------------------------------------------
1246 Disassembler::Disassembler(const ArchSpec &arch, const char *flavor)
1247 : m_arch(arch), m_instruction_list(), m_base_addr(LLDB_INVALID_ADDRESS),
1249 if (flavor == nullptr)
1250 m_flavor.assign("default");
1252 m_flavor.assign(flavor);
1254 // If this is an arm variant that can only include thumb (T16, T32)
1255 // instructions, force the arch triple to be "thumbv.." instead of
1257 if (arch.IsAlwaysThumbInstructions()) {
1258 std::string thumb_arch_name(arch.GetTriple().getArchName().str());
1259 // Replace "arm" with "thumb" so we get all thumb variants correct
1260 if (thumb_arch_name.size() > 3) {
1261 thumb_arch_name.erase(0, 3);
1262 thumb_arch_name.insert(0, "thumb");
1264 m_arch.SetTriple(thumb_arch_name.c_str());
1268 Disassembler::~Disassembler() = default;
1270 InstructionList &Disassembler::GetInstructionList() {
1271 return m_instruction_list;
1274 const InstructionList &Disassembler::GetInstructionList() const {
1275 return m_instruction_list;
1278 //----------------------------------------------------------------------
1279 // Class PseudoInstruction
1280 //----------------------------------------------------------------------
1282 PseudoInstruction::PseudoInstruction()
1283 : Instruction(Address(), eAddressClassUnknown), m_description() {}
1285 PseudoInstruction::~PseudoInstruction() = default;
1287 bool PseudoInstruction::DoesBranch() {
1288 // This is NOT a valid question for a pseudo instruction.
1292 bool PseudoInstruction::HasDelaySlot() {
1293 // This is NOT a valid question for a pseudo instruction.
1297 size_t PseudoInstruction::Decode(const lldb_private::Disassembler &disassembler,
1298 const lldb_private::DataExtractor &data,
1299 lldb::offset_t data_offset) {
1300 return m_opcode.GetByteSize();
1303 void PseudoInstruction::SetOpcode(size_t opcode_size, void *opcode_data) {
1307 switch (opcode_size) {
1309 uint8_t value8 = *((uint8_t *)opcode_data);
1310 m_opcode.SetOpcode8(value8, eByteOrderInvalid);
1314 uint16_t value16 = *((uint16_t *)opcode_data);
1315 m_opcode.SetOpcode16(value16, eByteOrderInvalid);
1319 uint32_t value32 = *((uint32_t *)opcode_data);
1320 m_opcode.SetOpcode32(value32, eByteOrderInvalid);
1324 uint64_t value64 = *((uint64_t *)opcode_data);
1325 m_opcode.SetOpcode64(value64, eByteOrderInvalid);
1333 void PseudoInstruction::SetDescription(llvm::StringRef description) {
1334 m_description = description;
1337 Instruction::Operand Instruction::Operand::BuildRegister(ConstString &r) {
1339 ret.m_type = Type::Register;
1344 Instruction::Operand Instruction::Operand::BuildImmediate(lldb::addr_t imm,
1347 ret.m_type = Type::Immediate;
1348 ret.m_immediate = imm;
1349 ret.m_negative = neg;
1353 Instruction::Operand Instruction::Operand::BuildImmediate(int64_t imm) {
1355 ret.m_type = Type::Immediate;
1357 ret.m_immediate = -imm;
1358 ret.m_negative = true;
1360 ret.m_immediate = imm;
1361 ret.m_negative = false;
1366 Instruction::Operand
1367 Instruction::Operand::BuildDereference(const Operand &ref) {
1369 ret.m_type = Type::Dereference;
1370 ret.m_children = {ref};
1374 Instruction::Operand Instruction::Operand::BuildSum(const Operand &lhs,
1375 const Operand &rhs) {
1377 ret.m_type = Type::Sum;
1378 ret.m_children = {lhs, rhs};
1382 Instruction::Operand Instruction::Operand::BuildProduct(const Operand &lhs,
1383 const Operand &rhs) {
1385 ret.m_type = Type::Product;
1386 ret.m_children = {lhs, rhs};
1390 std::function<bool(const Instruction::Operand &)>
1391 lldb_private::OperandMatchers::MatchBinaryOp(
1392 std::function<bool(const Instruction::Operand &)> base,
1393 std::function<bool(const Instruction::Operand &)> left,
1394 std::function<bool(const Instruction::Operand &)> right) {
1395 return [base, left, right](const Instruction::Operand &op) -> bool {
1396 return (base(op) && op.m_children.size() == 2 &&
1397 ((left(op.m_children[0]) && right(op.m_children[1])) ||
1398 (left(op.m_children[1]) && right(op.m_children[0]))));
1402 std::function<bool(const Instruction::Operand &)>
1403 lldb_private::OperandMatchers::MatchUnaryOp(
1404 std::function<bool(const Instruction::Operand &)> base,
1405 std::function<bool(const Instruction::Operand &)> child) {
1406 return [base, child](const Instruction::Operand &op) -> bool {
1407 return (base(op) && op.m_children.size() == 1 && child(op.m_children[0]));
1411 std::function<bool(const Instruction::Operand &)>
1412 lldb_private::OperandMatchers::MatchRegOp(const RegisterInfo &info) {
1413 return [&info](const Instruction::Operand &op) {
1414 return (op.m_type == Instruction::Operand::Type::Register &&
1415 (op.m_register == ConstString(info.name) ||
1416 op.m_register == ConstString(info.alt_name)));
1420 std::function<bool(const Instruction::Operand &)>
1421 lldb_private::OperandMatchers::FetchRegOp(ConstString ®) {
1422 return [®](const Instruction::Operand &op) {
1423 if (op.m_type != Instruction::Operand::Type::Register) {
1426 reg = op.m_register;
1431 std::function<bool(const Instruction::Operand &)>
1432 lldb_private::OperandMatchers::MatchImmOp(int64_t imm) {
1433 return [imm](const Instruction::Operand &op) {
1434 return (op.m_type == Instruction::Operand::Type::Immediate &&
1435 ((op.m_negative && op.m_immediate == (uint64_t)-imm) ||
1436 (!op.m_negative && op.m_immediate == (uint64_t)imm)));
1440 std::function<bool(const Instruction::Operand &)>
1441 lldb_private::OperandMatchers::FetchImmOp(int64_t &imm) {
1442 return [&imm](const Instruction::Operand &op) {
1443 if (op.m_type != Instruction::Operand::Type::Immediate) {
1446 if (op.m_negative) {
1447 imm = -((int64_t)op.m_immediate);
1449 imm = ((int64_t)op.m_immediate);
1455 std::function<bool(const Instruction::Operand &)>
1456 lldb_private::OperandMatchers::MatchOpType(Instruction::Operand::Type type) {
1457 return [type](const Instruction::Operand &op) { return op.m_type == type; };