[LLDB][SBSaveCore] Add selectable memory regions to SBSaveCore (#105442)

This patch adds the option to specify specific memory ranges to be
included in a given core file. The current implementation lets user
specified ranges either be in addition to a certain save style, or
independent of them via the newly added custom enum.

To achieve being inclusive of save style, I've moved from a std::vector
of ranges to a RangeDataVector, and to join overlapping ranges to
prevent duplication of memory ranges in the core file.

As a non function bonus, when SBSavecore was initially created, the
header was included in the lldb-private interfaces, and I've fixed that
and moved it the forward declare as an oversight. CC @bulbazord in case
we need to include that into swift.

Author: Jlalond

lldb/include/lldb/API/SBMemoryRegionInfo.h

@@ -120,7 +120,7 @@ class LLDB_API SBMemoryRegionInfo {
 private:
   friend class SBProcess;
   friend class SBMemoryRegionInfoList;
-
+  friend class SBSaveCoreOptions;
   friend class lldb_private::ScriptInterpreter;
 
   lldb_private::MemoryRegionInfo &ref();

lldb/include/lldb/API/SBSaveCoreOptions.h

@@ -80,6 +80,17 @@ class LLDB_API SBSaveCoreOptions {
   /// \return True if the thread was removed, false if it was not in the list.
   bool RemoveThread(lldb::SBThread thread);
 
+  /// Add a memory region to save in the core file.
+  ///
+  /// \param region The memory region to save.
+  /// \returns An empty SBError upon success, or an error if the region is
+  /// invalid.
+  /// \note Ranges that overlapped will be unioned into a single region, this
+  /// also supercedes stack minification. Specifying full regions and a
+  /// non-custom core style will include the specified regions and union them
+  /// with all style specific regions.
+  SBError AddMemoryRegionToSave(const SBMemoryRegionInfo &region);
+
   /// Reset all options.
   void Clear();
 

lldb/include/lldb/Symbol/SaveCoreOptions.h

@@ -10,13 +10,15 @@
 #define LLDB_SOURCE_PLUGINS_OBJECTFILE_SaveCoreOPTIONS_H
 
 #include "lldb/Utility/FileSpec.h"
-#include "lldb/lldb-forward.h"
-#include "lldb/lldb-types.h"
+#include "lldb/Utility/RangeMap.h"
 
 #include <optional>
+#include <set>
 #include <string>
 #include <unordered_set>
 
+using MemoryRanges = lldb_private::RangeVector<lldb::addr_t, lldb::addr_t>;
+
 namespace lldb_private {
 
 class SaveCoreOptions {
@@ -38,8 +40,12 @@ class SaveCoreOptions {
   Status AddThread(lldb::ThreadSP thread_sp);
   bool RemoveThread(lldb::ThreadSP thread_sp);
   bool ShouldThreadBeSaved(lldb::tid_t tid) const;
+  bool HasSpecifiedThreads() const;
 
   Status EnsureValidConfiguration(lldb::ProcessSP process_sp) const;
+  const MemoryRanges &GetCoreFileMemoryRanges() const;
+
+  void AddMemoryRegionToSave(const lldb_private::MemoryRegionInfo &region);
 
   void Clear();
 
@@ -51,6 +57,7 @@ class SaveCoreOptions {
   std::optional<lldb::SaveCoreStyle> m_style;
   lldb::ProcessSP m_process_sp;
   std::unordered_set<lldb::tid_t> m_threads_to_save;
+  MemoryRanges m_regions_to_save;
 };
 } // namespace lldb_private
 

lldb/include/lldb/Target/Process.h

@@ -35,6 +35,7 @@
 #include "lldb/Host/ProcessLaunchInfo.h"
 #include "lldb/Host/ProcessRunLock.h"
 #include "lldb/Symbol/ObjectFile.h"
+#include "lldb/Symbol/SaveCoreOptions.h"
 #include "lldb/Target/ExecutionContextScope.h"
 #include "lldb/Target/InstrumentationRuntime.h"
 #include "lldb/Target/Memory.h"
@@ -733,7 +734,9 @@ class Process : public std::enable_shared_from_this<Process>,
     }
   };
 
-  using CoreFileMemoryRanges = std::vector<CoreFileMemoryRange>;
+  using CoreFileMemoryRanges =
+      lldb_private::RangeDataVector<lldb::addr_t, lldb::addr_t,
+                                    CoreFileMemoryRange>;
 
   /// Helper function for Process::SaveCore(...) that calculates the address
   /// ranges that should be saved. This allows all core file plug-ins to save

lldb/include/lldb/Utility/RangeMap.h

@@ -450,6 +450,8 @@ class RangeDataVector {
 
   void Append(const Entry &entry) { m_entries.emplace_back(entry); }
 
+  void Append(B &&b, S &&s, T &&t) { m_entries.emplace_back(Entry(b, s, t)); }
+
   bool Erase(uint32_t start, uint32_t end) {
     if (start >= end || end > m_entries.size())
       return false;

lldb/include/lldb/lldb-enumerations.h

@@ -1222,6 +1222,7 @@ enum SaveCoreStyle {
   eSaveCoreFull = 1,
   eSaveCoreDirtyOnly = 2,
   eSaveCoreStackOnly = 3,
+  eSaveCoreCustomOnly = 4,
 };
 
 /// Events that might happen during a trace session.

lldb/include/lldb/lldb-forward.h

@@ -207,6 +207,7 @@ class StackFrameRecognizer;
 class StackFrameRecognizerManager;
 class StackID;
 class Status;
+class SaveCoreOptions;
 class StopInfo;
 class Stoppoint;
 class StoppointCallbackContext;

lldb/include/lldb/lldb-private-interfaces.h

@@ -9,7 +9,6 @@
 #ifndef LLDB_LLDB_PRIVATE_INTERFACES_H
 #define LLDB_LLDB_PRIVATE_INTERFACES_H
 
-#include "lldb/Symbol/SaveCoreOptions.h"
 #include "lldb/lldb-enumerations.h"
 #include "lldb/lldb-forward.h"
 #include "lldb/lldb-private-enumerations.h"

lldb/source/API/SBSaveCoreOptions.cpp

@@ -7,6 +7,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "lldb/API/SBSaveCoreOptions.h"
+#include "lldb/API/SBMemoryRegionInfo.h"
 #include "lldb/Host/FileSystem.h"
 #include "lldb/Symbol/SaveCoreOptions.h"
 #include "lldb/Utility/Instrumentation.h"
@@ -90,6 +91,16 @@ bool SBSaveCoreOptions::RemoveThread(lldb::SBThread thread) {
   return m_opaque_up->RemoveThread(thread.GetSP());
 }
 
+lldb::SBError
+SBSaveCoreOptions::AddMemoryRegionToSave(const SBMemoryRegionInfo &region) {
+  LLDB_INSTRUMENT_VA(this, region);
+  // Currently add memory region can't fail, so we always return a success
+  // SBerror, but because these API's live forever, this is the most future
+  // proof thing to do.
+  m_opaque_up->AddMemoryRegionToSave(region.ref());
+  return SBError();
+}
+
 void SBSaveCoreOptions::Clear() {
   LLDB_INSTRUMENT_VA(this);
   m_opaque_up->Clear();

lldb/source/Commands/CommandObjectProcess.cpp

@@ -25,6 +25,7 @@
 #include "lldb/Interpreter/OptionArgParser.h"
 #include "lldb/Interpreter/OptionGroupPythonClassWithDict.h"
 #include "lldb/Interpreter/Options.h"
+#include "lldb/Symbol/SaveCoreOptions.h"
 #include "lldb/Target/Platform.h"
 #include "lldb/Target/Process.h"
 #include "lldb/Target/StopInfo.h"

lldb/source/Plugins/ObjectFile/Mach-O/ObjectFileMachO.cpp

@@ -6568,7 +6568,9 @@ bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp,
         const uint32_t addr_byte_size = target_arch.GetAddressByteSize();
         const ByteOrder byte_order = target_arch.GetByteOrder();
         std::vector<llvm::MachO::segment_command_64> segment_load_commands;
-        for (const auto &core_range : core_ranges) {
+        for (const auto &core_range_info : core_ranges) {
+          // TODO: Refactor RangeDataVector to have a data iterator.
+          const auto &core_range = core_range_info.data;
           uint32_t cmd_type = LC_SEGMENT_64;
           uint32_t segment_size = sizeof(llvm::MachO::segment_command_64);
           if (addr_byte_size == 4) {

lldb/source/Plugins/ObjectFile/Mach-O/ObjectFileMachO.h

@@ -12,6 +12,7 @@
 #include "lldb/Core/Address.h"
 #include "lldb/Host/SafeMachO.h"
 #include "lldb/Symbol/ObjectFile.h"
+#include "lldb/Symbol/SaveCoreOptions.h"
 #include "lldb/Utility/FileSpec.h"
 #include "lldb/Utility/FileSpecList.h"
 #include "lldb/Utility/RangeMap.h"

lldb/source/Plugins/ObjectFile/Minidump/MinidumpFileBuilder.cpp

@@ -823,25 +823,32 @@ Status MinidumpFileBuilder::AddMemoryList() {
   // bytes of the core file. Thread structures in minidump files can only use
   // 32 bit memory descriptiors, so we emit them first to ensure the memory is
   // in accessible with a 32 bit offset.
-  Process::CoreFileMemoryRanges ranges_32;
-  Process::CoreFileMemoryRanges ranges_64;
+  std::vector<Process::CoreFileMemoryRange> ranges_32;
+  std::vector<Process::CoreFileMemoryRange> ranges_64;
   Process::CoreFileMemoryRanges all_core_memory_ranges;
   error = m_process_sp->CalculateCoreFileSaveRanges(m_save_core_options,
                                                     all_core_memory_ranges);
+
+  std::vector<Process::CoreFileMemoryRange> all_core_memory_vec;
+  // Extract all the data into just a vector of data. So we can mutate this in
+  // place.
+  for (const auto &core_range : all_core_memory_ranges)
+    all_core_memory_vec.push_back(core_range.data);
+
   if (error.Fail())
     return error;
 
   // Start by saving all of the stacks and ensuring they fit under the 32b
   // limit.
   uint64_t total_size = GetCurrentDataEndOffset();
-  auto iterator = all_core_memory_ranges.begin();
-  while (iterator != all_core_memory_ranges.end()) {
+  auto iterator = all_core_memory_vec.begin();
+  while (iterator != all_core_memory_vec.end()) {
     if (m_saved_stack_ranges.count(iterator->range.start()) > 0) {
       // We don't save stacks twice.
       ranges_32.push_back(*iterator);
       total_size +=
           iterator->range.size() + sizeof(llvm::minidump::MemoryDescriptor);
-      iterator = all_core_memory_ranges.erase(iterator);
+      iterator = all_core_memory_vec.erase(iterator);
     } else {
       iterator++;
     }
@@ -860,11 +867,11 @@ Status MinidumpFileBuilder::AddMemoryList() {
   // Then anything overflow extends into 64b addressable space.
   // All core memeroy ranges will either container nothing on stacks only
   // or all the memory ranges including stacks
-  if (!all_core_memory_ranges.empty())
-    total_size += 256 + (all_core_memory_ranges.size() *
+  if (!all_core_memory_vec.empty())
+    total_size += 256 + (all_core_memory_vec.size() *
                          sizeof(llvm::minidump::MemoryDescriptor_64));
 
-  for (const auto &core_range : all_core_memory_ranges) {
+  for (const auto &core_range : all_core_memory_vec) {
     const addr_t range_size = core_range.range.size();
     // We don't need to check for stacks here because we already removed them
     // from all_core_memory_ranges.
@@ -949,15 +956,15 @@ Status MinidumpFileBuilder::DumpDirectories() const {
 }
 
 static uint64_t
-GetLargestRangeSize(const Process::CoreFileMemoryRanges &ranges) {
+GetLargestRangeSize(const std::vector<Process::CoreFileMemoryRange> &ranges) {
   uint64_t max_size = 0;
   for (const auto &core_range : ranges)
     max_size = std::max(max_size, core_range.range.size());
   return max_size;
 }
 
-Status
-MinidumpFileBuilder::AddMemoryList_32(Process::CoreFileMemoryRanges &ranges) {
+Status MinidumpFileBuilder::AddMemoryList_32(
+    std::vector<Process::CoreFileMemoryRange> &ranges) {
   std::vector<MemoryDescriptor> descriptors;
   Status error;
   if (ranges.size() == 0)
@@ -1033,8 +1040,8 @@ MinidumpFileBuilder::AddMemoryList_32(Process::CoreFileMemoryRanges &ranges) {
   return error;
 }
 
-Status
-MinidumpFileBuilder::AddMemoryList_64(Process::CoreFileMemoryRanges &ranges) {
+Status MinidumpFileBuilder::AddMemoryList_64(
+    std::vector<Process::CoreFileMemoryRange> &ranges) {
   Status error;
   if (ranges.empty())
     return error;

lldb/source/Plugins/ObjectFile/Minidump/MinidumpFileBuilder.h

@@ -23,6 +23,7 @@
 #include <utility>
 #include <variant>
 
+#include "lldb/Symbol/SaveCoreOptions.h"
 #include "lldb/Target/Process.h"
 #include "lldb/Target/Target.h"
 #include "lldb/Utility/DataBufferHeap.h"
@@ -119,10 +120,10 @@ class MinidumpFileBuilder {
   // trigger a flush.
   lldb_private::Status AddData(const void *data, uint64_t size);
   // Add MemoryList stream, containing dumps of important memory segments
-  lldb_private::Status
-  AddMemoryList_64(lldb_private::Process::CoreFileMemoryRanges &ranges);
-  lldb_private::Status
-  AddMemoryList_32(lldb_private::Process::CoreFileMemoryRanges &ranges);
+  lldb_private::Status AddMemoryList_64(
+      std::vector<lldb_private::Process::CoreFileMemoryRange> &ranges);
+  lldb_private::Status AddMemoryList_32(
+      std::vector<lldb_private::Process::CoreFileMemoryRange> &ranges);
   // Update the thread list on disk with the newly emitted stack RVAs.
   lldb_private::Status FixThreadStacks();
   lldb_private::Status FlushBufferToDisk();

lldb/source/Plugins/ObjectFile/Minidump/ObjectFileMinidump.h

@@ -21,6 +21,7 @@
 #define LLDB_SOURCE_PLUGINS_OBJECTFILE_MINIDUMP_OBJECTFILEMINIDUMP_H
 
 #include "lldb/Symbol/ObjectFile.h"
+#include "lldb/Symbol/SaveCoreOptions.h"
 #include "lldb/Utility/ArchSpec.h"
 
 class ObjectFileMinidump : public lldb_private::PluginInterface {

lldb/source/Plugins/ObjectFile/PECOFF/ObjectFilePECOFF.cpp

@@ -17,6 +17,7 @@
 #include "lldb/Interpreter/OptionValueDictionary.h"
 #include "lldb/Interpreter/OptionValueProperties.h"
 #include "lldb/Symbol/ObjectFile.h"
+#include "lldb/Symbol/SaveCoreOptions.h"
 #include "lldb/Target/Process.h"
 #include "lldb/Target/SectionLoadList.h"
 #include "lldb/Target/Target.h"

lldb/source/Plugins/ObjectFile/PECOFF/ObjectFilePECOFF.h

@@ -13,6 +13,7 @@
 #include <vector>
 
 #include "lldb/Symbol/ObjectFile.h"
+#include "lldb/Symbol/SaveCoreOptions.h"
 #include "llvm/Object/COFF.h"
 
 class ObjectFilePECOFF : public lldb_private::ObjectFile {

lldb/source/Symbol/SaveCoreOptions.cpp

@@ -101,6 +101,19 @@ bool SaveCoreOptions::ShouldThreadBeSaved(lldb::tid_t tid) const {
   return m_threads_to_save.count(tid) > 0;
 }
 
+bool SaveCoreOptions::HasSpecifiedThreads() const {
+  return !m_threads_to_save.empty();
+}
+
+void SaveCoreOptions::AddMemoryRegionToSave(
+    const lldb_private::MemoryRegionInfo &region) {
+  m_regions_to_save.Insert(region.GetRange(), /*combine=*/true);
+}
+
+const MemoryRanges &SaveCoreOptions::GetCoreFileMemoryRanges() const {
+  return m_regions_to_save;
+}
+
 Status SaveCoreOptions::EnsureValidConfiguration(
     lldb::ProcessSP process_sp) const {
   Status error;
@@ -130,4 +143,5 @@ void SaveCoreOptions::Clear() {
   m_style = std::nullopt;
   m_threads_to_save.clear();
   m_process_sp.reset();
+  m_regions_to_save.Clear();
 }