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/Host/FileSystem.h"
21 #include "lldb/Interpreter/OptionValue.h"
22 #include "lldb/Interpreter/OptionValueArray.h"
23 #include "lldb/Interpreter/OptionValueDictionary.h"
24 #include "lldb/Interpreter/OptionValueRegex.h"
25 #include "lldb/Interpreter/OptionValueString.h"
26 #include "lldb/Interpreter/OptionValueUInt64.h"
27 #include "lldb/Symbol/Function.h"
28 #include "lldb/Symbol/Symbol.h" // for Symbol
29 #include "lldb/Symbol/SymbolContext.h" // for SymbolContext
30 #include "lldb/Target/ExecutionContext.h"
31 #include "lldb/Target/SectionLoadList.h"
32 #include "lldb/Target/StackFrame.h"
33 #include "lldb/Target/Target.h"
34 #include "lldb/Target/Thread.h" // for Thread
35 #include "lldb/Utility/DataBufferHeap.h"
36 #include "lldb/Utility/DataExtractor.h"
37 #include "lldb/Utility/RegularExpression.h"
38 #include "lldb/Utility/Status.h"
39 #include "lldb/Utility/Stream.h" // for Stream
40 #include "lldb/Utility/StreamString.h" // for StreamString
41 #include "lldb/Utility/Timer.h"
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 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION);
63 Timer scoped_timer(func_cat,
64 "Disassembler::FindPlugin (arch = %s, plugin_name = %s)",
65 arch.GetArchitectureName(), plugin_name);
67 DisassemblerCreateInstance create_callback = nullptr;
70 ConstString const_plugin_name(plugin_name);
71 create_callback = PluginManager::GetDisassemblerCreateCallbackForPluginName(
73 if (create_callback) {
74 DisassemblerSP disassembler_sp(create_callback(arch, flavor));
77 return disassembler_sp;
80 for (uint32_t idx = 0;
81 (create_callback = PluginManager::GetDisassemblerCreateCallbackAtIndex(
84 DisassemblerSP disassembler_sp(create_callback(arch, flavor));
87 return disassembler_sp;
90 return DisassemblerSP();
93 DisassemblerSP Disassembler::FindPluginForTarget(const TargetSP target_sp,
96 const char *plugin_name) {
97 if (target_sp && flavor == nullptr) {
98 // FIXME - we don't have the mechanism in place to do per-architecture
99 // settings. But since we know that for now we only support flavors on x86
101 if (arch.GetTriple().getArch() == llvm::Triple::x86 ||
102 arch.GetTriple().getArch() == llvm::Triple::x86_64)
103 flavor = target_sp->GetDisassemblyFlavor();
105 return FindPlugin(arch, flavor, plugin_name);
108 static void ResolveAddress(const ExecutionContext &exe_ctx, const Address &addr,
109 Address &resolved_addr) {
110 if (!addr.IsSectionOffset()) {
111 // If we weren't passed in a section offset address range, try and resolve
113 Target *target = exe_ctx.GetTargetPtr();
116 target->GetSectionLoadList().IsEmpty() ?
117 target->GetImages().ResolveFileAddress(addr.GetOffset(),
119 target->GetSectionLoadList().ResolveLoadAddress(addr.GetOffset(),
122 // We weren't able to resolve the address, just treat it as a raw address
123 if (is_resolved && resolved_addr.IsValid())
127 resolved_addr = addr;
130 size_t Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
131 const char *plugin_name, const char *flavor,
132 const ExecutionContext &exe_ctx,
133 SymbolContextList &sc_list,
134 uint32_t num_instructions,
135 bool mixed_source_and_assembly,
136 uint32_t num_mixed_context_lines,
137 uint32_t options, Stream &strm) {
138 size_t success_count = 0;
139 const size_t count = sc_list.GetSize();
142 const uint32_t scope =
143 eSymbolContextBlock | eSymbolContextFunction | eSymbolContextSymbol;
144 const bool use_inline_block_range = true;
145 for (size_t i = 0; i < count; ++i) {
146 if (!sc_list.GetContextAtIndex(i, sc))
148 for (uint32_t range_idx = 0;
149 sc.GetAddressRange(scope, range_idx, use_inline_block_range, range);
151 if (Disassemble(debugger, arch, plugin_name, flavor, exe_ctx, range,
152 num_instructions, mixed_source_and_assembly,
153 num_mixed_context_lines, options, strm)) {
159 return success_count;
162 bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
163 const char *plugin_name, const char *flavor,
164 const ExecutionContext &exe_ctx,
165 const ConstString &name, Module *module,
166 uint32_t num_instructions,
167 bool mixed_source_and_assembly,
168 uint32_t num_mixed_context_lines,
169 uint32_t options, Stream &strm) {
170 SymbolContextList sc_list;
172 const bool include_symbols = true;
173 const bool include_inlines = true;
175 module->FindFunctions(name, nullptr, eFunctionNameTypeAuto,
176 include_symbols, include_inlines, true, sc_list);
177 } else if (exe_ctx.GetTargetPtr()) {
178 exe_ctx.GetTargetPtr()->GetImages().FindFunctions(
179 name, eFunctionNameTypeAuto, include_symbols, include_inlines, false,
184 if (sc_list.GetSize()) {
185 return Disassemble(debugger, arch, plugin_name, flavor, exe_ctx, sc_list,
186 num_instructions, mixed_source_and_assembly,
187 num_mixed_context_lines, options, strm);
192 lldb::DisassemblerSP Disassembler::DisassembleRange(
193 const ArchSpec &arch, const char *plugin_name, const char *flavor,
194 const ExecutionContext &exe_ctx, const AddressRange &range,
195 bool prefer_file_cache) {
196 lldb::DisassemblerSP disasm_sp;
197 if (range.GetByteSize() > 0 && range.GetBaseAddress().IsValid()) {
198 disasm_sp = Disassembler::FindPluginForTarget(exe_ctx.GetTargetSP(), arch,
199 flavor, plugin_name);
202 size_t bytes_disassembled = disasm_sp->ParseInstructions(
203 &exe_ctx, range, nullptr, prefer_file_cache);
204 if (bytes_disassembled == 0)
212 Disassembler::DisassembleBytes(const ArchSpec &arch, const char *plugin_name,
213 const char *flavor, const Address &start,
214 const void *src, size_t src_len,
215 uint32_t num_instructions, bool data_from_file) {
216 lldb::DisassemblerSP disasm_sp;
219 disasm_sp = Disassembler::FindPlugin(arch, flavor, plugin_name);
222 DataExtractor data(src, src_len, arch.GetByteOrder(),
223 arch.GetAddressByteSize());
225 (void)disasm_sp->DecodeInstructions(start, data, 0, num_instructions,
226 false, data_from_file);
233 bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
234 const char *plugin_name, const char *flavor,
235 const ExecutionContext &exe_ctx,
236 const AddressRange &disasm_range,
237 uint32_t num_instructions,
238 bool mixed_source_and_assembly,
239 uint32_t num_mixed_context_lines,
240 uint32_t options, Stream &strm) {
241 if (disasm_range.GetByteSize()) {
242 lldb::DisassemblerSP disasm_sp(Disassembler::FindPluginForTarget(
243 exe_ctx.GetTargetSP(), arch, flavor, plugin_name));
247 ResolveAddress(exe_ctx, disasm_range.GetBaseAddress(),
248 range.GetBaseAddress());
249 range.SetByteSize(disasm_range.GetByteSize());
250 const bool prefer_file_cache = false;
251 size_t bytes_disassembled = disasm_sp->ParseInstructions(
252 &exe_ctx, range, &strm, prefer_file_cache);
253 if (bytes_disassembled == 0)
256 return PrintInstructions(disasm_sp.get(), debugger, arch, exe_ctx,
257 num_instructions, mixed_source_and_assembly,
258 num_mixed_context_lines, options, strm);
264 bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
265 const char *plugin_name, const char *flavor,
266 const ExecutionContext &exe_ctx,
267 const Address &start_address,
268 uint32_t num_instructions,
269 bool mixed_source_and_assembly,
270 uint32_t num_mixed_context_lines,
271 uint32_t options, Stream &strm) {
272 if (num_instructions > 0) {
273 lldb::DisassemblerSP disasm_sp(Disassembler::FindPluginForTarget(
274 exe_ctx.GetTargetSP(), arch, flavor, plugin_name));
277 ResolveAddress(exe_ctx, start_address, addr);
278 const bool prefer_file_cache = false;
279 size_t bytes_disassembled = disasm_sp->ParseInstructions(
280 &exe_ctx, addr, num_instructions, prefer_file_cache);
281 if (bytes_disassembled == 0)
283 return PrintInstructions(disasm_sp.get(), debugger, arch, exe_ctx,
284 num_instructions, mixed_source_and_assembly,
285 num_mixed_context_lines, options, strm);
291 Disassembler::SourceLine
292 Disassembler::GetFunctionDeclLineEntry(const SymbolContext &sc) {
293 SourceLine decl_line;
294 if (sc.function && sc.line_entry.IsValid()) {
295 LineEntry prologue_end_line = sc.line_entry;
296 FileSpec func_decl_file;
297 uint32_t func_decl_line;
298 sc.function->GetStartLineSourceInfo(func_decl_file, func_decl_line);
299 if (func_decl_file == prologue_end_line.file ||
300 func_decl_file == prologue_end_line.original_file) {
301 decl_line.file = func_decl_file;
302 decl_line.line = func_decl_line;
303 // TODO do we care about column on these entries? If so, we need to
304 // plumb that through GetStartLineSourceInfo.
305 decl_line.column = 0;
311 void Disassembler::AddLineToSourceLineTables(
313 std::map<FileSpec, std::set<uint32_t>> &source_lines_seen) {
314 if (line.IsValid()) {
315 auto source_lines_seen_pos = source_lines_seen.find(line.file);
316 if (source_lines_seen_pos == source_lines_seen.end()) {
317 std::set<uint32_t> lines;
318 lines.insert(line.line);
319 source_lines_seen.emplace(line.file, lines);
321 source_lines_seen_pos->second.insert(line.line);
326 bool Disassembler::ElideMixedSourceAndDisassemblyLine(
327 const ExecutionContext &exe_ctx, const SymbolContext &sc,
330 // TODO: should we also check target.process.thread.step-avoid-libraries ?
332 const RegularExpression *avoid_regex = nullptr;
334 // Skip any line #0 entries - they are implementation details
338 ThreadSP thread_sp = exe_ctx.GetThreadSP();
340 avoid_regex = thread_sp->GetSymbolsToAvoidRegexp();
342 TargetSP target_sp = exe_ctx.GetTargetSP();
345 OptionValueSP value_sp = target_sp->GetDebugger().GetPropertyValue(
346 &exe_ctx, "target.process.thread.step-avoid-regexp", false, error);
347 if (value_sp && value_sp->GetType() == OptionValue::eTypeRegex) {
348 OptionValueRegex *re = value_sp->GetAsRegex();
350 avoid_regex = re->GetCurrentValue();
355 if (avoid_regex && sc.symbol != nullptr) {
356 const char *function_name =
357 sc.GetFunctionName(Mangled::ePreferDemangledWithoutArguments)
360 RegularExpression::Match regex_match(1);
361 if (avoid_regex->Execute(function_name, ®ex_match)) {
362 // skip this source line
367 // don't skip this source line
371 bool Disassembler::PrintInstructions(Disassembler *disasm_ptr,
372 Debugger &debugger, const ArchSpec &arch,
373 const ExecutionContext &exe_ctx,
374 uint32_t num_instructions,
375 bool mixed_source_and_assembly,
376 uint32_t num_mixed_context_lines,
377 uint32_t options, Stream &strm) {
378 // We got some things disassembled...
379 size_t num_instructions_found = disasm_ptr->GetInstructionList().GetSize();
381 if (num_instructions > 0 && num_instructions < num_instructions_found)
382 num_instructions_found = num_instructions;
384 const uint32_t max_opcode_byte_size =
385 disasm_ptr->GetInstructionList().GetMaxOpcocdeByteSize();
387 SymbolContext prev_sc;
388 AddressRange current_source_line_range;
389 const Address *pc_addr_ptr = nullptr;
390 StackFrame *frame = exe_ctx.GetFramePtr();
392 TargetSP target_sp(exe_ctx.GetTargetSP());
393 SourceManager &source_manager =
394 target_sp ? target_sp->GetSourceManager() : debugger.GetSourceManager();
397 pc_addr_ptr = &frame->GetFrameCodeAddress();
399 const uint32_t scope =
400 eSymbolContextLineEntry | eSymbolContextFunction | eSymbolContextSymbol;
401 const bool use_inline_block_range = false;
403 const FormatEntity::Entry *disassembly_format = nullptr;
404 FormatEntity::Entry format;
405 if (exe_ctx.HasTargetScope()) {
407 exe_ctx.GetTargetRef().GetDebugger().GetDisassemblyFormat();
409 FormatEntity::Parse("${addr}: ", format);
410 disassembly_format = &format;
413 // First pass: step through the list of instructions, find how long the
414 // initial addresses strings are, insert padding in the second pass so the
415 // opcodes all line up nicely.
417 // Also build up the source line mapping if this is mixed source & assembly
418 // mode. Calculate the source line for each assembly instruction (eliding
419 // inlined functions which the user wants to skip).
421 std::map<FileSpec, std::set<uint32_t>> source_lines_seen;
422 Symbol *previous_symbol = nullptr;
424 size_t address_text_size = 0;
425 for (size_t i = 0; i < num_instructions_found; ++i) {
427 disasm_ptr->GetInstructionList().GetInstructionAtIndex(i).get();
429 const Address &addr = inst->GetAddress();
430 ModuleSP module_sp(addr.GetModule());
432 const uint32_t resolve_mask = eSymbolContextFunction |
433 eSymbolContextSymbol |
434 eSymbolContextLineEntry;
435 uint32_t resolved_mask =
436 module_sp->ResolveSymbolContextForAddress(addr, resolve_mask, sc);
438 StreamString strmstr;
439 Debugger::FormatDisassemblerAddress(disassembly_format, &sc, nullptr,
440 &exe_ctx, &addr, strmstr);
441 size_t cur_line = strmstr.GetSizeOfLastLine();
442 if (cur_line > address_text_size)
443 address_text_size = cur_line;
445 // Add entries to our "source_lines_seen" map+set which list which
446 // sources lines occur in this disassembly session. We will print
447 // lines of context around a source line, but we don't want to print
448 // a source line that has a line table entry of its own - we'll leave
449 // that source line to be printed when it actually occurs in the
452 if (mixed_source_and_assembly && sc.line_entry.IsValid()) {
453 if (sc.symbol != previous_symbol) {
454 SourceLine decl_line = GetFunctionDeclLineEntry(sc);
455 if (ElideMixedSourceAndDisassemblyLine(exe_ctx, sc, decl_line) ==
457 AddLineToSourceLineTables(decl_line, source_lines_seen);
459 if (sc.line_entry.IsValid()) {
460 SourceLine this_line;
461 this_line.file = sc.line_entry.file;
462 this_line.line = sc.line_entry.line;
463 this_line.column = sc.line_entry.column;
464 if (ElideMixedSourceAndDisassemblyLine(exe_ctx, sc, this_line) ==
466 AddLineToSourceLineTables(this_line, source_lines_seen);
475 previous_symbol = nullptr;
476 SourceLine previous_line;
477 for (size_t i = 0; i < num_instructions_found; ++i) {
479 disasm_ptr->GetInstructionList().GetInstructionAtIndex(i).get();
482 const Address &addr = inst->GetAddress();
483 const bool inst_is_at_pc = pc_addr_ptr && addr == *pc_addr_ptr;
484 SourceLinesToDisplay source_lines_to_display;
488 ModuleSP module_sp(addr.GetModule());
490 uint32_t resolved_mask = module_sp->ResolveSymbolContextForAddress(
491 addr, eSymbolContextEverything, sc);
493 if (mixed_source_and_assembly) {
495 // If we've started a new function (non-inlined), print all of the
496 // source lines from the function declaration until the first line
497 // table entry - typically the opening curly brace of the function.
498 if (previous_symbol != sc.symbol) {
499 // The default disassembly format puts an extra blank line
500 // between functions - so when we're displaying the source
501 // context for a function, we don't want to add a blank line
502 // after the source context or we'll end up with two of them.
503 if (previous_symbol != nullptr)
504 source_lines_to_display.print_source_context_end_eol = false;
506 previous_symbol = sc.symbol;
507 if (sc.function && sc.line_entry.IsValid()) {
508 LineEntry prologue_end_line = sc.line_entry;
509 if (ElideMixedSourceAndDisassemblyLine(
510 exe_ctx, sc, prologue_end_line) == false) {
511 FileSpec func_decl_file;
512 uint32_t func_decl_line;
513 sc.function->GetStartLineSourceInfo(func_decl_file,
515 if (func_decl_file == prologue_end_line.file ||
516 func_decl_file == prologue_end_line.original_file) {
517 // Add all the lines between the function declaration and
518 // the first non-prologue source line to the list of lines
520 for (uint32_t lineno = func_decl_line;
521 lineno <= prologue_end_line.line; lineno++) {
522 SourceLine this_line;
523 this_line.file = func_decl_file;
524 this_line.line = lineno;
525 source_lines_to_display.lines.push_back(this_line);
527 // Mark the last line as the "current" one. Usually this
528 // is the open curly brace.
529 if (source_lines_to_display.lines.size() > 0)
530 source_lines_to_display.current_source_line =
531 source_lines_to_display.lines.size() - 1;
535 sc.GetAddressRange(scope, 0, use_inline_block_range,
536 current_source_line_range);
539 // If we've left a previous source line's address range, print a
541 if (!current_source_line_range.ContainsFileAddress(addr)) {
542 sc.GetAddressRange(scope, 0, use_inline_block_range,
543 current_source_line_range);
545 if (sc != prev_sc && sc.comp_unit && sc.line_entry.IsValid()) {
546 SourceLine this_line;
547 this_line.file = sc.line_entry.file;
548 this_line.line = sc.line_entry.line;
550 if (ElideMixedSourceAndDisassemblyLine(exe_ctx, sc,
551 this_line) == false) {
552 // Only print this source line if it is different from the
553 // last source line we printed. There may have been inlined
554 // functions between these lines that we elided, resulting in
555 // the same line being printed twice in a row for a
556 // contiguous block of assembly instructions.
557 if (this_line != previous_line) {
559 std::vector<uint32_t> previous_lines;
561 i < num_mixed_context_lines &&
562 (this_line.line - num_mixed_context_lines) > 0;
565 this_line.line - num_mixed_context_lines + i;
566 auto pos = source_lines_seen.find(this_line.file);
567 if (pos != source_lines_seen.end()) {
568 if (pos->second.count(line) == 1) {
569 previous_lines.clear();
571 previous_lines.push_back(line);
575 for (size_t i = 0; i < previous_lines.size(); i++) {
576 SourceLine previous_line;
577 previous_line.file = this_line.file;
578 previous_line.line = previous_lines[i];
579 auto pos = source_lines_seen.find(previous_line.file);
580 if (pos != source_lines_seen.end()) {
581 pos->second.insert(previous_line.line);
583 source_lines_to_display.lines.push_back(previous_line);
586 source_lines_to_display.lines.push_back(this_line);
587 source_lines_to_display.current_source_line =
588 source_lines_to_display.lines.size() - 1;
590 for (uint32_t i = 0; i < num_mixed_context_lines; i++) {
591 SourceLine next_line;
592 next_line.file = this_line.file;
593 next_line.line = this_line.line + i + 1;
594 auto pos = source_lines_seen.find(next_line.file);
595 if (pos != source_lines_seen.end()) {
596 if (pos->second.count(next_line.line) == 1)
598 pos->second.insert(next_line.line);
600 source_lines_to_display.lines.push_back(next_line);
603 previous_line = this_line;
613 if (source_lines_to_display.lines.size() > 0) {
615 for (size_t idx = 0; idx < source_lines_to_display.lines.size();
617 SourceLine ln = source_lines_to_display.lines[idx];
618 const char *line_highlight = "";
619 if (inst_is_at_pc && (options & eOptionMarkPCSourceLine)) {
620 line_highlight = "->";
621 } else if (idx == source_lines_to_display.current_source_line) {
622 line_highlight = "**";
624 source_manager.DisplaySourceLinesWithLineNumbers(
625 ln.file, ln.line, ln.column, 0, 0, line_highlight, &strm);
627 if (source_lines_to_display.print_source_context_end_eol)
631 const bool show_bytes = (options & eOptionShowBytes) != 0;
632 inst->Dump(&strm, max_opcode_byte_size, true, show_bytes, &exe_ctx, &sc,
633 &prev_sc, nullptr, address_text_size);
643 bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
644 const char *plugin_name, const char *flavor,
645 const ExecutionContext &exe_ctx,
646 uint32_t num_instructions,
647 bool mixed_source_and_assembly,
648 uint32_t num_mixed_context_lines,
649 uint32_t options, Stream &strm) {
651 StackFrame *frame = exe_ctx.GetFramePtr();
654 frame->GetSymbolContext(eSymbolContextFunction | eSymbolContextSymbol));
656 range = sc.function->GetAddressRange();
657 } else if (sc.symbol && sc.symbol->ValueIsAddress()) {
658 range.GetBaseAddress() = sc.symbol->GetAddressRef();
659 range.SetByteSize(sc.symbol->GetByteSize());
661 range.GetBaseAddress() = frame->GetFrameCodeAddress();
664 if (range.GetBaseAddress().IsValid() && range.GetByteSize() == 0)
665 range.SetByteSize(DEFAULT_DISASM_BYTE_SIZE);
668 return Disassemble(debugger, arch, plugin_name, flavor, exe_ctx, range,
669 num_instructions, mixed_source_and_assembly,
670 num_mixed_context_lines, options, strm);
673 Instruction::Instruction(const Address &address, AddressClass addr_class)
674 : m_address(address), m_address_class(addr_class), m_opcode(),
675 m_calculated_strings(false) {}
677 Instruction::~Instruction() = default;
679 AddressClass Instruction::GetAddressClass() {
680 if (m_address_class == AddressClass::eInvalid)
681 m_address_class = m_address.GetAddressClass();
682 return m_address_class;
685 void Instruction::Dump(lldb_private::Stream *s, uint32_t max_opcode_byte_size,
686 bool show_address, bool show_bytes,
687 const ExecutionContext *exe_ctx,
688 const SymbolContext *sym_ctx,
689 const SymbolContext *prev_sym_ctx,
690 const FormatEntity::Entry *disassembly_addr_format,
691 size_t max_address_text_size) {
692 size_t opcode_column_width = 7;
693 const size_t operand_column_width = 25;
695 CalculateMnemonicOperandsAndCommentIfNeeded(exe_ctx);
700 Debugger::FormatDisassemblerAddress(disassembly_addr_format, sym_ctx,
701 prev_sym_ctx, exe_ctx, &m_address, ss);
702 ss.FillLastLineToColumn(max_address_text_size, ' ');
706 if (m_opcode.GetType() == Opcode::eTypeBytes) {
707 // x86_64 and i386 are the only ones that use bytes right now so pad out
708 // the byte dump to be able to always show 15 bytes (3 chars each) plus a
710 if (max_opcode_byte_size > 0)
711 m_opcode.Dump(&ss, max_opcode_byte_size * 3 + 1);
713 m_opcode.Dump(&ss, 15 * 3 + 1);
715 // Else, we have ARM or MIPS which can show up to a uint32_t 0x00000000
716 // (10 spaces) plus two for padding...
717 if (max_opcode_byte_size > 0)
718 m_opcode.Dump(&ss, max_opcode_byte_size * 3 + 1);
720 m_opcode.Dump(&ss, 12);
724 const size_t opcode_pos = ss.GetSizeOfLastLine();
726 // The default opcode size of 7 characters is plenty for most architectures
727 // but some like arm can pull out the occasional vqrshrun.s16. We won't get
728 // consistent column spacing in these cases, unfortunately.
729 if (m_opcode_name.length() >= opcode_column_width) {
730 opcode_column_width = m_opcode_name.length() + 1;
733 ss.PutCString(m_opcode_name);
734 ss.FillLastLineToColumn(opcode_pos + opcode_column_width, ' ');
735 ss.PutCString(m_mnemonics);
737 if (!m_comment.empty()) {
738 ss.FillLastLineToColumn(
739 opcode_pos + opcode_column_width + operand_column_width, ' ');
740 ss.PutCString(" ; ");
741 ss.PutCString(m_comment);
743 s->PutCString(ss.GetString());
746 bool Instruction::DumpEmulation(const ArchSpec &arch) {
747 std::unique_ptr<EmulateInstruction> insn_emulator_ap(
748 EmulateInstruction::FindPlugin(arch, eInstructionTypeAny, nullptr));
749 if (insn_emulator_ap) {
750 insn_emulator_ap->SetInstruction(GetOpcode(), GetAddress(), nullptr);
751 return insn_emulator_ap->EvaluateInstruction(0);
757 bool Instruction::CanSetBreakpoint () {
758 return !HasDelaySlot();
761 bool Instruction::HasDelaySlot() {
766 OptionValueSP Instruction::ReadArray(FILE *in_file, Stream *out_stream,
767 OptionValue::Type data_type) {
771 auto option_value_sp = std::make_shared<OptionValueArray>(1u << data_type);
775 if (!fgets(buffer, 1023, in_file)) {
777 "Instruction::ReadArray: Error reading file (fgets).\n");
778 option_value_sp.reset();
779 return option_value_sp;
782 std::string line(buffer);
784 size_t len = line.size();
785 if (line[len - 1] == '\n') {
786 line[len - 1] = '\0';
787 line.resize(len - 1);
790 if ((line.size() == 1) && line[0] == ']') {
797 static RegularExpression g_reg_exp(
798 llvm::StringRef("^[ \t]*([^ \t]+)[ \t]*$"));
799 RegularExpression::Match regex_match(1);
800 bool reg_exp_success = g_reg_exp.Execute(line, ®ex_match);
802 regex_match.GetMatchAtIndex(line.c_str(), 1, value);
806 OptionValueSP data_value_sp;
808 case OptionValue::eTypeUInt64:
809 data_value_sp = std::make_shared<OptionValueUInt64>(0, 0);
810 data_value_sp->SetValueFromString(value);
812 // Other types can be added later as needed.
814 data_value_sp = std::make_shared<OptionValueString>(value.c_str(), "");
818 option_value_sp->GetAsArray()->InsertValue(idx, data_value_sp);
823 return option_value_sp;
826 OptionValueSP Instruction::ReadDictionary(FILE *in_file, Stream *out_stream) {
830 auto option_value_sp = std::make_shared<OptionValueDictionary>();
831 static ConstString encoding_key("data_encoding");
832 OptionValue::Type data_type = OptionValue::eTypeInvalid;
835 // Read the next line in the file
836 if (!fgets(buffer, 1023, in_file)) {
838 "Instruction::ReadDictionary: Error reading file (fgets).\n");
839 option_value_sp.reset();
840 return option_value_sp;
843 // Check to see if the line contains the end-of-dictionary marker ("}")
844 std::string line(buffer);
846 size_t len = line.size();
847 if (line[len - 1] == '\n') {
848 line[len - 1] = '\0';
849 line.resize(len - 1);
852 if ((line.size() == 1) && (line[0] == '}')) {
857 // Try to find a key-value pair in the current line and add it to the
860 static RegularExpression g_reg_exp(llvm::StringRef(
861 "^[ \t]*([a-zA-Z_][a-zA-Z0-9_]*)[ \t]*=[ \t]*(.*)[ \t]*$"));
862 RegularExpression::Match regex_match(2);
864 bool reg_exp_success = g_reg_exp.Execute(line, ®ex_match);
867 if (reg_exp_success) {
868 regex_match.GetMatchAtIndex(line.c_str(), 1, key);
869 regex_match.GetMatchAtIndex(line.c_str(), 2, value);
871 out_stream->Printf("Instruction::ReadDictionary: Failure executing "
872 "regular expression.\n");
873 option_value_sp.reset();
874 return option_value_sp;
877 ConstString const_key(key.c_str());
878 // Check value to see if it's the start of an array or dictionary.
880 lldb::OptionValueSP value_sp;
881 assert(value.empty() == false);
882 assert(key.empty() == false);
884 if (value[0] == '{') {
885 assert(value.size() == 1);
886 // value is a dictionary
887 value_sp = ReadDictionary(in_file, out_stream);
889 option_value_sp.reset();
890 return option_value_sp;
892 } else if (value[0] == '[') {
893 assert(value.size() == 1);
895 value_sp = ReadArray(in_file, out_stream, data_type);
897 option_value_sp.reset();
898 return option_value_sp;
900 // We've used the data_type to read an array; re-set the type to
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. Go
1106 // backwards until we find an instruction marked end-of-packet, or until we
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 bits 15:14 will be
1127 // 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 "armv..."
1256 if (arch.IsAlwaysThumbInstructions()) {
1257 std::string thumb_arch_name(arch.GetTriple().getArchName().str());
1258 // Replace "arm" with "thumb" so we get all thumb variants correct
1259 if (thumb_arch_name.size() > 3) {
1260 thumb_arch_name.erase(0, 3);
1261 thumb_arch_name.insert(0, "thumb");
1263 m_arch.SetTriple(thumb_arch_name.c_str());
1267 Disassembler::~Disassembler() = default;
1269 InstructionList &Disassembler::GetInstructionList() {
1270 return m_instruction_list;
1273 const InstructionList &Disassembler::GetInstructionList() const {
1274 return m_instruction_list;
1277 //----------------------------------------------------------------------
1278 // Class PseudoInstruction
1279 //----------------------------------------------------------------------
1281 PseudoInstruction::PseudoInstruction()
1282 : Instruction(Address(), AddressClass::eUnknown), m_description() {}
1284 PseudoInstruction::~PseudoInstruction() = default;
1286 bool PseudoInstruction::DoesBranch() {
1287 // This is NOT a valid question for a pseudo instruction.
1291 bool PseudoInstruction::HasDelaySlot() {
1292 // This is NOT a valid question for a pseudo instruction.
1296 size_t PseudoInstruction::Decode(const lldb_private::Disassembler &disassembler,
1297 const lldb_private::DataExtractor &data,
1298 lldb::offset_t data_offset) {
1299 return m_opcode.GetByteSize();
1302 void PseudoInstruction::SetOpcode(size_t opcode_size, void *opcode_data) {
1306 switch (opcode_size) {
1308 uint8_t value8 = *((uint8_t *)opcode_data);
1309 m_opcode.SetOpcode8(value8, eByteOrderInvalid);
1313 uint16_t value16 = *((uint16_t *)opcode_data);
1314 m_opcode.SetOpcode16(value16, eByteOrderInvalid);
1318 uint32_t value32 = *((uint32_t *)opcode_data);
1319 m_opcode.SetOpcode32(value32, eByteOrderInvalid);
1323 uint64_t value64 = *((uint64_t *)opcode_data);
1324 m_opcode.SetOpcode64(value64, eByteOrderInvalid);
1332 void PseudoInstruction::SetDescription(llvm::StringRef description) {
1333 m_description = description;
1336 Instruction::Operand Instruction::Operand::BuildRegister(ConstString &r) {
1338 ret.m_type = Type::Register;
1343 Instruction::Operand Instruction::Operand::BuildImmediate(lldb::addr_t imm,
1346 ret.m_type = Type::Immediate;
1347 ret.m_immediate = imm;
1348 ret.m_negative = neg;
1352 Instruction::Operand Instruction::Operand::BuildImmediate(int64_t imm) {
1354 ret.m_type = Type::Immediate;
1356 ret.m_immediate = -imm;
1357 ret.m_negative = true;
1359 ret.m_immediate = imm;
1360 ret.m_negative = false;
1365 Instruction::Operand
1366 Instruction::Operand::BuildDereference(const Operand &ref) {
1368 ret.m_type = Type::Dereference;
1369 ret.m_children = {ref};
1373 Instruction::Operand Instruction::Operand::BuildSum(const Operand &lhs,
1374 const Operand &rhs) {
1376 ret.m_type = Type::Sum;
1377 ret.m_children = {lhs, rhs};
1381 Instruction::Operand Instruction::Operand::BuildProduct(const Operand &lhs,
1382 const Operand &rhs) {
1384 ret.m_type = Type::Product;
1385 ret.m_children = {lhs, rhs};
1389 std::function<bool(const Instruction::Operand &)>
1390 lldb_private::OperandMatchers::MatchBinaryOp(
1391 std::function<bool(const Instruction::Operand &)> base,
1392 std::function<bool(const Instruction::Operand &)> left,
1393 std::function<bool(const Instruction::Operand &)> right) {
1394 return [base, left, right](const Instruction::Operand &op) -> bool {
1395 return (base(op) && op.m_children.size() == 2 &&
1396 ((left(op.m_children[0]) && right(op.m_children[1])) ||
1397 (left(op.m_children[1]) && right(op.m_children[0]))));
1401 std::function<bool(const Instruction::Operand &)>
1402 lldb_private::OperandMatchers::MatchUnaryOp(
1403 std::function<bool(const Instruction::Operand &)> base,
1404 std::function<bool(const Instruction::Operand &)> child) {
1405 return [base, child](const Instruction::Operand &op) -> bool {
1406 return (base(op) && op.m_children.size() == 1 && child(op.m_children[0]));
1410 std::function<bool(const Instruction::Operand &)>
1411 lldb_private::OperandMatchers::MatchRegOp(const RegisterInfo &info) {
1412 return [&info](const Instruction::Operand &op) {
1413 return (op.m_type == Instruction::Operand::Type::Register &&
1414 (op.m_register == ConstString(info.name) ||
1415 op.m_register == ConstString(info.alt_name)));
1419 std::function<bool(const Instruction::Operand &)>
1420 lldb_private::OperandMatchers::FetchRegOp(ConstString ®) {
1421 return [®](const Instruction::Operand &op) {
1422 if (op.m_type != Instruction::Operand::Type::Register) {
1425 reg = op.m_register;
1430 std::function<bool(const Instruction::Operand &)>
1431 lldb_private::OperandMatchers::MatchImmOp(int64_t imm) {
1432 return [imm](const Instruction::Operand &op) {
1433 return (op.m_type == Instruction::Operand::Type::Immediate &&
1434 ((op.m_negative && op.m_immediate == (uint64_t)-imm) ||
1435 (!op.m_negative && op.m_immediate == (uint64_t)imm)));
1439 std::function<bool(const Instruction::Operand &)>
1440 lldb_private::OperandMatchers::FetchImmOp(int64_t &imm) {
1441 return [&imm](const Instruction::Operand &op) {
1442 if (op.m_type != Instruction::Operand::Type::Immediate) {
1445 if (op.m_negative) {
1446 imm = -((int64_t)op.m_immediate);
1448 imm = ((int64_t)op.m_immediate);
1454 std::function<bool(const Instruction::Operand &)>
1455 lldb_private::OperandMatchers::MatchOpType(Instruction::Operand::Type type) {
1456 return [type](const Instruction::Operand &op) { return op.m_type == type; };