[Lldb-commits] [lldb] dc46ae6 - [lldb] Add support to load object files from thin archives
Nico Weber via lldb-commits
lldb-commits at lists.llvm.org
Tue Jul 5 01:53:03 PDT 2022
Author: Kaining Zhong
Date: 2022-07-05T10:52:26+02:00
New Revision: dc46ae6df5f769b80b56a30f6d77f962fa90833d
URL: https://github.com/llvm/llvm-project/commit/dc46ae6df5f769b80b56a30f6d77f962fa90833d
DIFF: https://github.com/llvm/llvm-project/commit/dc46ae6df5f769b80b56a30f6d77f962fa90833d.diff
LOG: [lldb] Add support to load object files from thin archives
This fixes https://github.com/llvm/llvm-project/issues/50114 where lldb/mac
can't load object files from thin archives. This patch allows lldb to identify
thin archives, and load object files contained in them.
Differential Revision: https://reviews.llvm.org/D126464
Added:
lldb/test/API/functionalities/archives/c.c
Modified:
lldb/source/Plugins/ObjectContainer/BSD-Archive/ObjectContainerBSDArchive.cpp
lldb/source/Plugins/ObjectContainer/BSD-Archive/ObjectContainerBSDArchive.h
lldb/test/API/functionalities/archives/Makefile
lldb/test/API/functionalities/archives/TestBSDArchives.py
lldb/test/API/lit.cfg.py
lldb/test/CMakeLists.txt
Removed:
################################################################################
diff --git a/lldb/source/Plugins/ObjectContainer/BSD-Archive/ObjectContainerBSDArchive.cpp b/lldb/source/Plugins/ObjectContainer/BSD-Archive/ObjectContainerBSDArchive.cpp
index a2522372f5af4..9fe222eceedca 100644
--- a/lldb/source/Plugins/ObjectContainer/BSD-Archive/ObjectContainerBSDArchive.cpp
+++ b/lldb/source/Plugins/ObjectContainer/BSD-Archive/ObjectContainerBSDArchive.cpp
@@ -40,6 +40,8 @@ typedef struct ar_hdr {
using namespace lldb;
using namespace lldb_private;
+using namespace llvm::object;
+
LLDB_PLUGIN_DEFINE(ObjectContainerBSDArchive)
ObjectContainerBSDArchive::Object::Object() : ar_name() {}
@@ -55,6 +57,74 @@ void ObjectContainerBSDArchive::Object::Clear() {
file_size = 0;
}
+lldb::offset_t ObjectContainerBSDArchive::Object::ExtractFromThin(
+ const DataExtractor &data, lldb::offset_t offset,
+ llvm::StringRef stringTable) {
+ size_t ar_name_len = 0;
+ std::string str;
+ char *err;
+
+ // File header
+ //
+ // The common format is as follows.
+ //
+ // Offset Length Name Format
+ // 0 16 File name ASCII right padded with spaces (no spaces
+ // allowed in file name)
+ // 16 12 File mod Decimal as cstring right padded with
+ // spaces
+ // 28 6 Owner ID Decimal as cstring right padded with
+ // spaces
+ // 34 6 Group ID Decimal as cstring right padded with
+ // spaces
+ // 40 8 File mode Octal as cstring right padded with
+ // spaces
+ // 48 10 File byte size Decimal as cstring right padded with
+ // spaces
+ // 58 2 File magic 0x60 0x0A
+
+ // Make sure there is enough data for the file header and bail if not
+ if (!data.ValidOffsetForDataOfSize(offset, 60))
+ return LLDB_INVALID_OFFSET;
+
+ str.assign((const char *)data.GetData(&offset, 16), 16);
+ if (!(llvm::StringRef(str).startswith("//") || stringTable.empty())) {
+ // Strip off any trailing spaces.
+ const size_t last_pos = str.find_last_not_of(' ');
+ if (last_pos != std::string::npos) {
+ if (last_pos + 1 < 16)
+ str.erase(last_pos + 1);
+ }
+ int start = strtoul(str.c_str() + 1, &err, 10);
+ int end = stringTable.find('\n', start);
+ str.assign(stringTable.data() + start, end - start - 1);
+ ar_name.SetCString(str.c_str());
+ }
+
+ str.assign((const char *)data.GetData(&offset, 12), 12);
+ modification_time = strtoul(str.c_str(), &err, 10);
+
+ str.assign((const char *)data.GetData(&offset, 6), 6);
+ uid = strtoul(str.c_str(), &err, 10);
+
+ str.assign((const char *)data.GetData(&offset, 6), 6);
+ gid = strtoul(str.c_str(), &err, 10);
+
+ str.assign((const char *)data.GetData(&offset, 8), 8);
+ mode = strtoul(str.c_str(), &err, 8);
+
+ str.assign((const char *)data.GetData(&offset, 10), 10);
+ size = strtoul(str.c_str(), &err, 10);
+
+ str.assign((const char *)data.GetData(&offset, 2), 2);
+ if (str == ARFMAG) {
+ file_offset = offset;
+ file_size = size - ar_name_len;
+ return offset;
+ }
+ return LLDB_INVALID_OFFSET;
+}
+
lldb::offset_t
ObjectContainerBSDArchive::Object::Extract(const DataExtractor &data,
lldb::offset_t offset) {
@@ -136,9 +206,10 @@ ObjectContainerBSDArchive::Object::Extract(const DataExtractor &data,
ObjectContainerBSDArchive::Archive::Archive(const lldb_private::ArchSpec &arch,
const llvm::sys::TimePoint<> &time,
lldb::offset_t file_offset,
- lldb_private::DataExtractor &data)
+ lldb_private::DataExtractor &data,
+ ArchiveType archive_type)
: m_arch(arch), m_modification_time(time), m_file_offset(file_offset),
- m_objects(), m_data(data) {}
+ m_objects(), m_data(data), m_archive_type(archive_type) {}
ObjectContainerBSDArchive::Archive::~Archive() = default;
@@ -161,6 +232,48 @@ size_t ObjectContainerBSDArchive::Archive::ParseObjects() {
obj.Clear();
} while (data.ValidOffset(offset));
+ // Now sort all of the object name pointers
+ m_object_name_to_index_map.Sort();
+ } else if (str == ThinArchiveMagic) {
+ Object obj;
+ size_t obj_idx;
+
+ // Retrieve symbol table
+ offset = obj.ExtractFromThin(data, offset, "");
+ if (offset == LLDB_INVALID_OFFSET)
+ return m_objects.size();
+ obj_idx = m_objects.size();
+ m_objects.push_back(obj);
+ // Insert all of the C strings out of order for now...
+ m_object_name_to_index_map.Append(obj.ar_name, obj_idx);
+ offset += obj.file_size;
+ obj.Clear();
+
+ // Retrieve string table
+ offset = obj.ExtractFromThin(data, offset, "");
+ if (offset == LLDB_INVALID_OFFSET)
+ return m_objects.size();
+ obj_idx = m_objects.size();
+ m_objects.push_back(obj);
+ // Insert all of the C strings out of order for now...
+ m_object_name_to_index_map.Append(obj.ar_name, obj_idx);
+ // Extract string table
+ llvm::StringRef strtab((const char *)data.GetData(&offset, obj.size),
+ obj.size);
+ obj.Clear();
+
+ // Retrieve object files
+ do {
+ offset = obj.ExtractFromThin(data, offset, strtab);
+ if (offset == LLDB_INVALID_OFFSET)
+ break;
+ obj_idx = m_objects.size();
+ m_objects.push_back(obj);
+ // Insert all of the C strings out of order for now...
+ m_object_name_to_index_map.Append(obj.ar_name, obj_idx);
+ obj.Clear();
+ } while (data.ValidOffset(offset));
+
// Now sort all of the object name pointers
m_object_name_to_index_map.Sort();
}
@@ -237,8 +350,9 @@ ObjectContainerBSDArchive::Archive::shared_ptr
ObjectContainerBSDArchive::Archive::ParseAndCacheArchiveForFile(
const FileSpec &file, const ArchSpec &arch,
const llvm::sys::TimePoint<> &time, lldb::offset_t file_offset,
- DataExtractor &data) {
- shared_ptr archive_sp(new Archive(arch, time, file_offset, data));
+ DataExtractor &data, ArchiveType archive_type) {
+ shared_ptr archive_sp(
+ new Archive(arch, time, file_offset, data, archive_type));
if (archive_sp) {
const size_t num_objects = archive_sp->ParseObjects();
if (num_objects > 0) {
@@ -288,7 +402,8 @@ ObjectContainer *ObjectContainerBSDArchive::CreateInstance(
// contents for the archive and cache it
DataExtractor data;
data.SetData(data_sp, data_offset, length);
- if (file && data_sp && ObjectContainerBSDArchive::MagicBytesMatch(data)) {
+ ArchiveType archive_type = ObjectContainerBSDArchive::MagicBytesMatch(data);
+ if (file && data_sp && archive_type != ArchiveType::Invalid) {
LLDB_SCOPED_TIMERF(
"ObjectContainerBSDArchive::CreateInstance (module = %s, file = "
"%p, file_offset = 0x%8.8" PRIx64 ", file_size = 0x%8.8" PRIx64 ")",
@@ -312,7 +427,7 @@ ObjectContainer *ObjectContainerBSDArchive::CreateInstance(
std::unique_ptr<ObjectContainerBSDArchive> container_up(
new ObjectContainerBSDArchive(module_sp, archive_data_sp,
archive_data_offset, file, file_offset,
- length));
+ length, archive_type));
if (container_up) {
if (archive_sp) {
@@ -331,7 +446,8 @@ ObjectContainer *ObjectContainerBSDArchive::CreateInstance(
if (archive_sp) {
std::unique_ptr<ObjectContainerBSDArchive> container_up(
new ObjectContainerBSDArchive(module_sp, data_sp, data_offset, file,
- file_offset, length));
+ file_offset, length,
+ archive_sp->GetArchiveType()));
if (container_up) {
// We already have this archive in our cache, use it
@@ -343,23 +459,35 @@ ObjectContainer *ObjectContainerBSDArchive::CreateInstance(
return nullptr;
}
-bool ObjectContainerBSDArchive::MagicBytesMatch(const DataExtractor &data) {
+ArchiveType
+ObjectContainerBSDArchive::MagicBytesMatch(const DataExtractor &data) {
uint32_t offset = 0;
const char *armag = (const char *)data.PeekData(offset, sizeof(ar_hdr));
- if (armag && ::strncmp(armag, ARMAG, SARMAG) == 0) {
+ if (armag == nullptr)
+ return ArchiveType::Invalid;
+ if (::strncmp(armag, ARMAG, SARMAG) == 0) {
armag += offsetof(struct ar_hdr, ar_fmag) + SARMAG;
if (strncmp(armag, ARFMAG, 2) == 0)
- return true;
+ return ArchiveType::Archive;
+ } else if (::strncmp(armag, ThinArchiveMagic, strlen(ThinArchiveMagic)) ==
+ 0) {
+ armag += offsetof(struct ar_hdr, ar_fmag) + strlen(ThinArchiveMagic);
+ if (strncmp(armag, ARFMAG, 2) == 0) {
+ return ArchiveType::ThinArchive;
+ }
}
- return false;
+ return ArchiveType::Invalid;
}
ObjectContainerBSDArchive::ObjectContainerBSDArchive(
const lldb::ModuleSP &module_sp, DataBufferSP &data_sp,
lldb::offset_t data_offset, const lldb_private::FileSpec *file,
- lldb::offset_t file_offset, lldb::offset_t size)
+ lldb::offset_t file_offset, lldb::offset_t size, ArchiveType archive_type)
: ObjectContainer(module_sp, file, file_offset, size, data_sp, data_offset),
- m_archive_sp() {}
+ m_archive_sp() {
+ m_archive_type = archive_type;
+}
+
void ObjectContainerBSDArchive::SetArchive(Archive::shared_ptr &archive_sp) {
m_archive_sp = archive_sp;
}
@@ -373,7 +501,7 @@ bool ObjectContainerBSDArchive::ParseHeader() {
if (module_sp) {
m_archive_sp = Archive::ParseAndCacheArchiveForFile(
m_file, module_sp->GetArchitecture(),
- module_sp->GetModificationTime(), m_offset, m_data);
+ module_sp->GetModificationTime(), m_offset, m_data, m_archive_type);
}
// Clear the m_data that contains the entire archive data and let our
// m_archive_sp hold onto the data.
@@ -407,6 +535,19 @@ void ObjectContainerBSDArchive::Dump(Stream *s) const {
s->EOL();
}
+FileSpec GetChildFileSpecificationsFromThin(llvm::StringRef childPath,
+ const FileSpec &parentFileSpec) {
+ llvm::SmallString<128> FullPath;
+ if (llvm::sys::path::is_absolute(childPath)) {
+ FullPath = childPath;
+ } else {
+ FullPath = parentFileSpec.GetDirectory().GetStringRef();
+ llvm::sys::path::append(FullPath, childPath);
+ }
+ FileSpec child = FileSpec(FullPath.str(), llvm::sys::path::Style::posix);
+ return child;
+}
+
ObjectFileSP ObjectContainerBSDArchive::GetObjectFile(const FileSpec *file) {
ModuleSP module_sp(GetModule());
if (module_sp) {
@@ -414,6 +555,22 @@ ObjectFileSP ObjectContainerBSDArchive::GetObjectFile(const FileSpec *file) {
Object *object = m_archive_sp->FindObject(
module_sp->GetObjectName(), module_sp->GetObjectModificationTime());
if (object) {
+ if (m_archive_type == ArchiveType::ThinArchive) {
+ // Set file to child object file
+ FileSpec child = GetChildFileSpecificationsFromThin(
+ object->ar_name.GetStringRef(), m_file);
+ lldb::offset_t file_offset = 0;
+ lldb::offset_t file_size = object->size;
+ std::shared_ptr<DataBuffer> child_data_sp =
+ FileSystem::Instance().CreateDataBuffer(child, file_size,
+ file_offset);
+ if (child_data_sp->GetByteSize() != object->file_size)
+ return ObjectFileSP();
+ lldb::offset_t data_offset = 0;
+ return ObjectFile::FindPlugin(
+ module_sp, &child, m_offset + object->file_offset,
+ object->file_size, child_data_sp, data_offset);
+ }
lldb::offset_t data_offset = object->file_offset;
return ObjectFile::FindPlugin(
module_sp, file, m_offset + object->file_offset, object->file_size,
@@ -434,7 +591,8 @@ size_t ObjectContainerBSDArchive::GetModuleSpecifications(
// contents for the archive and cache it
DataExtractor data;
data.SetData(data_sp, data_offset, data_sp->GetByteSize());
- if (!file || !data_sp || !ObjectContainerBSDArchive::MagicBytesMatch(data))
+ ArchiveType archive_type = ObjectContainerBSDArchive::MagicBytesMatch(data);
+ if (!file || !data_sp || archive_type == ArchiveType::Invalid)
return 0;
const size_t initial_count = specs.GetSize();
@@ -449,7 +607,7 @@ size_t ObjectContainerBSDArchive::GetModuleSpecifications(
if (data_sp) {
data.SetData(data_sp, 0, data_sp->GetByteSize());
archive_sp = Archive::ParseAndCacheArchiveForFile(
- file, ArchSpec(), file_mod_time, file_offset, data);
+ file, ArchSpec(), file_mod_time, file_offset, data, archive_type);
}
}
@@ -458,6 +616,24 @@ size_t ObjectContainerBSDArchive::GetModuleSpecifications(
for (size_t idx = 0; idx < num_objects; ++idx) {
const Object *object = archive_sp->GetObjectAtIndex(idx);
if (object) {
+ if (archive_sp->GetArchiveType() == ArchiveType::ThinArchive) {
+ if (object->ar_name.IsEmpty())
+ continue;
+ FileSpec child = GetChildFileSpecificationsFromThin(
+ object->ar_name.GetStringRef(), file);
+ if (ObjectFile::GetModuleSpecifications(child, 0, object->file_size,
+ specs)) {
+ ModuleSpec &spec =
+ specs.GetModuleSpecRefAtIndex(specs.GetSize() - 1);
+ llvm::sys::TimePoint<> object_mod_time(
+ std::chrono::seconds(object->modification_time));
+ spec.GetObjectName() = object->ar_name;
+ spec.SetObjectOffset(0);
+ spec.SetObjectSize(object->file_size);
+ spec.GetObjectModificationTime() = object_mod_time;
+ }
+ continue;
+ }
const lldb::offset_t object_file_offset =
file_offset + object->file_offset;
if (object->file_offset < file_size && file_size > object_file_offset) {
diff --git a/lldb/source/Plugins/ObjectContainer/BSD-Archive/ObjectContainerBSDArchive.h b/lldb/source/Plugins/ObjectContainer/BSD-Archive/ObjectContainerBSDArchive.h
index 21106d7b85907..ace072cbe1498 100644
--- a/lldb/source/Plugins/ObjectContainer/BSD-Archive/ObjectContainerBSDArchive.h
+++ b/lldb/source/Plugins/ObjectContainer/BSD-Archive/ObjectContainerBSDArchive.h
@@ -15,19 +15,24 @@
#include "lldb/Utility/ConstString.h"
#include "lldb/Utility/FileSpec.h"
+#include "llvm/Object/Archive.h"
#include "llvm/Support/Chrono.h"
+#include "llvm/Support/Path.h"
#include <map>
#include <memory>
#include <mutex>
+enum class ArchiveType { Invalid, Archive, ThinArchive };
+
class ObjectContainerBSDArchive : public lldb_private::ObjectContainer {
public:
ObjectContainerBSDArchive(const lldb::ModuleSP &module_sp,
lldb::DataBufferSP &data_sp,
lldb::offset_t data_offset,
const lldb_private::FileSpec *file,
- lldb::offset_t offset, lldb::offset_t length);
+ lldb::offset_t offset, lldb::offset_t length,
+ ArchiveType archive_type);
~ObjectContainerBSDArchive() override;
@@ -54,7 +59,7 @@ class ObjectContainerBSDArchive : public lldb_private::ObjectContainer {
lldb::offset_t length,
lldb_private::ModuleSpecList &specs);
- static bool MagicBytesMatch(const lldb_private::DataExtractor &data);
+ static ArchiveType MagicBytesMatch(const lldb_private::DataExtractor &data);
// Member Functions
bool ParseHeader() override;
@@ -78,6 +83,10 @@ class ObjectContainerBSDArchive : public lldb_private::ObjectContainer {
void Clear();
+ lldb::offset_t ExtractFromThin(const lldb_private::DataExtractor &data,
+ lldb::offset_t offset,
+ llvm::StringRef stringTable);
+
lldb::offset_t Extract(const lldb_private::DataExtractor &data,
lldb::offset_t offset);
/// Object name in the archive.
@@ -112,7 +121,7 @@ class ObjectContainerBSDArchive : public lldb_private::ObjectContainer {
Archive(const lldb_private::ArchSpec &arch,
const llvm::sys::TimePoint<> &mod_time, lldb::offset_t file_offset,
- lldb_private::DataExtractor &data);
+ lldb_private::DataExtractor &data, ArchiveType archive_type);
~Archive();
@@ -127,7 +136,7 @@ class ObjectContainerBSDArchive : public lldb_private::ObjectContainer {
static Archive::shared_ptr ParseAndCacheArchiveForFile(
const lldb_private::FileSpec &file, const lldb_private::ArchSpec &arch,
const llvm::sys::TimePoint<> &mod_time, lldb::offset_t file_offset,
- lldb_private::DataExtractor &data);
+ lldb_private::DataExtractor &data, ArchiveType archive_type);
size_t GetNumObjects() const { return m_objects.size(); }
@@ -156,6 +165,8 @@ class ObjectContainerBSDArchive : public lldb_private::ObjectContainer {
lldb_private::DataExtractor &GetData() { return m_data; }
+ ArchiveType GetArchiveType() { return m_archive_type; }
+
protected:
typedef lldb_private::UniqueCStringMap<uint32_t> ObjectNameToIndexMap;
// Member Variables
@@ -167,11 +178,14 @@ class ObjectContainerBSDArchive : public lldb_private::ObjectContainer {
lldb_private::DataExtractor m_data; ///< The data for this object container
///so we don't lose data if the .a files
///gets modified
+ ArchiveType m_archive_type;
};
void SetArchive(Archive::shared_ptr &archive_sp);
Archive::shared_ptr m_archive_sp;
+
+ ArchiveType m_archive_type;
};
#endif // LLDB_SOURCE_PLUGINS_OBJECTCONTAINER_BSD_ARCHIVE_OBJECTCONTAINERBSDARCHIVE_H
diff --git a/lldb/test/API/functionalities/archives/Makefile b/lldb/test/API/functionalities/archives/Makefile
index d5d119e440f6c..868e2d698c936 100644
--- a/lldb/test/API/functionalities/archives/Makefile
+++ b/lldb/test/API/functionalities/archives/Makefile
@@ -1,8 +1,8 @@
-C_SOURCES := main.c a.c b.c
+C_SOURCES := main.c a.c b.c c.c
EXE := # Define a.out explicitly
MAKE_DSYM := NO
-all: a.out
+all: a.out libbar.a
a.out: main.o libfoo.a
$(LD) $(LDFLAGS) $^ -o $@
@@ -11,4 +11,11 @@ libfoo.a: a.o b.o
$(AR) $(ARFLAGS) $@ $^
$(RM) $^
+# This tests whether lldb can load a thin archive
+libbar.a: c.o
+ $(eval LLVM_AR := $(LLVM_TOOLS_DIR)/llvm-ar)
+ $(eval LLVM_ARFLAGS := -rcsDT)
+ $(LLVM_AR) $(LLVM_ARFLAGS) $@ $^
+ # Note for thin archive case, we cannot remove c.o
+
include Makefile.rules
diff --git a/lldb/test/API/functionalities/archives/TestBSDArchives.py b/lldb/test/API/functionalities/archives/TestBSDArchives.py
index c79e618db58b5..f0edc62673655 100644
--- a/lldb/test/API/functionalities/archives/TestBSDArchives.py
+++ b/lldb/test/API/functionalities/archives/TestBSDArchives.py
@@ -58,3 +58,10 @@ def test(self):
self.expect("frame variable", VARIABLES_DISPLAYED_CORRECTLY,
substrs=['(int) arg = 2'])
self.expect_var_path("__b_global", type="int", value="2")
+
+ # Test loading thin archives
+ archive_path = self.getBuildArtifact("libbar.a")
+ module_specs = lldb.SBModuleSpecList.GetModuleSpecifications(archive_path)
+ num_specs = module_specs.GetSize()
+ self.assertEqual(num_specs, 1)
+ self.assertEqual(module_specs.GetSpecAtIndex(0).GetObjectName(), "c.o")
diff --git a/lldb/test/API/functionalities/archives/c.c b/lldb/test/API/functionalities/archives/c.c
new file mode 100644
index 0000000000000..08581fe4f37a8
--- /dev/null
+++ b/lldb/test/API/functionalities/archives/c.c
@@ -0,0 +1,11 @@
+static int __c_global = 3;
+
+int c(int arg) {
+ int result = arg + __c_global;
+ return result;
+}
+
+int cc(int arg1) {
+ int result3 = arg1 - __c_global;
+ return result3;
+}
diff --git a/lldb/test/API/lit.cfg.py b/lldb/test/API/lit.cfg.py
index 48bb5662dc8b5..abd396d015dff 100644
--- a/lldb/test/API/lit.cfg.py
+++ b/lldb/test/API/lit.cfg.py
@@ -162,6 +162,10 @@ def delete_module_cache(path):
if is_configured('llvm_libs_dir'):
dotest_cmd += ['--env', 'LLVM_LIBS_DIR=' + config.llvm_libs_dir]
+# This path may be needed to locate required llvm tools
+if is_configured('llvm_tools_dir'):
+ dotest_cmd += ['--env', 'LLVM_TOOLS_DIR=' + config.llvm_tools_dir]
+
# Forward ASan-specific environment variables to tests, as a test may load an
# ASan-ified dylib.
for env_var in ('ASAN_OPTIONS', 'DYLD_INSERT_LIBRARIES'):
diff --git a/lldb/test/CMakeLists.txt b/lldb/test/CMakeLists.txt
index e23ac6651a557..61106e31d7813 100644
--- a/lldb/test/CMakeLists.txt
+++ b/lldb/test/CMakeLists.txt
@@ -148,6 +148,7 @@ add_lldb_test_dependency(
llvm-objcopy
llvm-pdbutil
llvm-readobj
+ llvm-ar
)
if(TARGET lld)
More information about the lldb-commits
mailing list