Everywhere: Core dump => Coredump

We all know what a coredump is, and it feels more natural to refer to
it as a coredump (most code already does), so let's be consistent.

Userland/Libraries/LibCoredump/Backtrace.cpp

@@ -0,0 +1,140 @@
+/*
+ * Copyright (c) 2020, Linus Groh <linusg@serenityos.org>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#include <AK/LexicalPath.h>
+#include <AK/MappedFile.h>
+#include <AK/Platform.h>
+#include <AK/StringBuilder.h>
+#include <AK/Types.h>
+#include <LibCore/File.h>
+#include <LibCoredump/Backtrace.h>
+#include <LibCoredump/Reader.h>
+#include <LibELF/Core.h>
+#include <LibELF/Image.h>
+
+namespace Coredump {
+
+ELFObjectInfo const* Backtrace::object_info_for_region(ELF::Core::MemoryRegionInfo const& region)
+{
+    auto path = region.object_name();
+    if (!path.starts_with('/') && path.ends_with(".so"sv))
+        path = LexicalPath::join("/usr/lib", path).string();
+
+    auto maybe_ptr = m_debug_info_cache.get(path);
+    if (maybe_ptr.has_value())
+        return *maybe_ptr;
+
+    if (!Core::File::exists(path))
+        return nullptr;
+
+    auto file_or_error = MappedFile::map(path);
+    if (file_or_error.is_error())
+        return nullptr;
+
+    auto image = make<ELF::Image>(file_or_error.value()->bytes());
+    auto& image_reference = *image;
+    auto info = make<ELFObjectInfo>(file_or_error.release_value(), make<Debug::DebugInfo>(image_reference), move(image));
+    auto* info_ptr = info.ptr();
+    m_debug_info_cache.set(path, move(info));
+    return info_ptr;
+}
+
+Backtrace::Backtrace(const Reader& coredump, const ELF::Core::ThreadInfo& thread_info)
+    : m_thread_info(move(thread_info))
+{
+    FlatPtr* bp;
+    FlatPtr* ip;
+#if ARCH(I386)
+    bp = (FlatPtr*)m_thread_info.regs.ebp;
+    ip = (FlatPtr*)m_thread_info.regs.eip;
+#else
+    bp = (FlatPtr*)m_thread_info.regs.rbp;
+    ip = (FlatPtr*)m_thread_info.regs.rip;
+#endif
+
+    bool first_frame = true;
+    while (bp && ip) {
+        // We use eip - 1 because the return address from a function frame
+        // is the instruction that comes after the 'call' instruction.
+        // However, because the first frame represents the faulting
+        // instruction rather than the return address we don't subtract
+        // 1 there.
+        VERIFY((FlatPtr)ip > 0);
+        add_entry(coredump, (FlatPtr)ip - (first_frame ? 0 : 1));
+        first_frame = false;
+        auto next_ip = coredump.peek_memory((FlatPtr)(bp + 1));
+        auto next_bp = coredump.peek_memory((FlatPtr)(bp));
+        if (!next_ip.has_value() || !next_bp.has_value())
+            break;
+        ip = (FlatPtr*)next_ip.value();
+        bp = (FlatPtr*)next_bp.value();
+    }
+}
+
+Backtrace::~Backtrace()
+{
+}
+
+void Backtrace::add_entry(const Reader& coredump, FlatPtr ip)
+{
+    auto* ip_region = coredump.region_containing((FlatPtr)ip);
+    if (!ip_region) {
+        m_entries.append({ ip, {}, {}, {} });
+        return;
+    }
+    auto object_name = ip_region->object_name();
+    if (object_name == "Loader.so")
+        return;
+    // We need to find the first region for the object, just in case
+    // the PT_LOAD header for the .text segment isn't the first one
+    // in the object file.
+    auto region = coredump.first_region_for_object(object_name);
+    auto* object_info = object_info_for_region(*region);
+    if (!object_info)
+        return;
+
+    auto function_name = object_info->debug_info->elf().symbolicate(ip - region->region_start);
+    auto source_position = object_info->debug_info->get_source_position_with_inlines(ip - region->region_start);
+    m_entries.append({ ip, object_name, function_name, source_position });
+}
+
+String Backtrace::Entry::to_string(bool color) const
+{
+    StringBuilder builder;
+    builder.appendff("{:p}: ", eip);
+    if (object_name.is_empty()) {
+        builder.append("???");
+        return builder.build();
+    }
+    builder.appendff("[{}] {}", object_name, function_name.is_empty() ? "???" : function_name);
+    builder.append(" (");
+
+    Vector<Debug::DebugInfo::SourcePosition> source_positions;
+
+    for (auto& position : source_position_with_inlines.inline_chain) {
+        if (!source_positions.contains_slow(position))
+            source_positions.append(position);
+    }
+
+    if (source_position_with_inlines.source_position.has_value() && !source_positions.contains_slow(source_position_with_inlines.source_position.value())) {
+        source_positions.insert(0, source_position_with_inlines.source_position.value());
+    }
+
+    for (size_t i = 0; i < source_positions.size(); ++i) {
+        auto& position = source_positions[i];
+        auto fmt = color ? "\033[34;1m{}\033[0m:{}" : "{}:{}";
+        builder.appendff(fmt, LexicalPath::basename(position.file_path), position.line_number);
+        if (i != source_positions.size() - 1) {
+            builder.append(" => ");
+        }
+    }
+
+    builder.append(")");
+
+    return builder.build();
+}
+
+}

Userland/Libraries/LibCoredump/Backtrace.h

@@ -0,0 +1,55 @@
+/*
+ * Copyright (c) 2020, Linus Groh <linusg@serenityos.org>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+#include <AK/Types.h>
+#include <LibCoredump/Reader.h>
+#include <LibDebug/DebugInfo.h>
+#include <LibELF/Core.h>
+
+namespace Coredump {
+
+struct ELFObjectInfo {
+    ELFObjectInfo(NonnullRefPtr<MappedFile> file, NonnullOwnPtr<Debug::DebugInfo>&& debug_info, NonnullOwnPtr<ELF::Image> image)
+        : file(move(file))
+        , debug_info(move(debug_info))
+        , image(move(image))
+    {
+    }
+
+    NonnullRefPtr<MappedFile> file;
+    NonnullOwnPtr<Debug::DebugInfo> debug_info;
+    NonnullOwnPtr<ELF::Image> image;
+};
+
+class Backtrace {
+public:
+    struct Entry {
+        FlatPtr eip;
+        String object_name;
+        String function_name;
+        Debug::DebugInfo::SourcePositionWithInlines source_position_with_inlines;
+
+        String to_string(bool color = false) const;
+    };
+
+    Backtrace(const Reader&, const ELF::Core::ThreadInfo&);
+    ~Backtrace();
+
+    const ELF::Core::ThreadInfo thread_info() const { return m_thread_info; }
+    const Vector<Entry> entries() const { return m_entries; }
+
+private:
+    void add_entry(const Reader&, FlatPtr ip);
+    ELFObjectInfo const* object_info_for_region(ELF::Core::MemoryRegionInfo const&);
+
+    ELF::Core::ThreadInfo m_thread_info;
+    Vector<Entry> m_entries;
+    HashMap<String, NonnullOwnPtr<ELFObjectInfo>> m_debug_info_cache;
+};
+
+}

Userland/Libraries/LibCoredump/CMakeLists.txt

@@ -0,0 +1,7 @@
+set(SOURCES
+    Backtrace.cpp
+    Reader.cpp
+)
+
+serenity_lib(LibCoredump Coredump)
+target_link_libraries(LibCoredump LibC LibCompress LibCore LibDebug)

Userland/Libraries/LibCoredump/Forward.h

@@ -0,0 +1,14 @@
+/*
+ * Copyright (c) 2020, Linus Groh <linusg@serenityos.org>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+namespace Coredump {
+
+class Backtrace;
+class Reader;
+
+}

Userland/Libraries/LibCoredump/Reader.cpp

@@ -0,0 +1,275 @@
+/*
+ * Copyright (c) 2020, Itamar S. <itamar8910@gmail.com>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#include <AK/JsonObject.h>
+#include <AK/JsonValue.h>
+#include <LibCompress/Gzip.h>
+#include <LibCoredump/Reader.h>
+#include <signal_numbers.h>
+#include <string.h>
+
+namespace Coredump {
+
+OwnPtr<Reader> Reader::create(const String& path)
+{
+    auto file_or_error = MappedFile::map(path);
+    if (file_or_error.is_error())
+        return {};
+    return adopt_own(*new Reader(file_or_error.value()->bytes()));
+}
+
+Reader::Reader(ReadonlyBytes coredump_bytes)
+    : m_coredump_buffer(decompress_coredump(coredump_bytes))
+    , m_coredump_image(m_coredump_buffer.bytes())
+{
+    size_t index = 0;
+    m_coredump_image.for_each_program_header([this, &index](auto pheader) {
+        if (pheader.type() == PT_NOTE) {
+            m_notes_segment_index = index;
+            return IterationDecision::Break;
+        }
+        ++index;
+        return IterationDecision::Continue;
+    });
+    VERIFY(m_notes_segment_index != -1);
+}
+
+ByteBuffer Reader::decompress_coredump(const ReadonlyBytes& raw_coredump)
+{
+    if (!Compress::GzipDecompressor::is_likely_compressed(raw_coredump))
+        return ByteBuffer::copy(raw_coredump); // handle old format coredumps (uncompressed)
+    auto decompressed_coredump = Compress::GzipDecompressor::decompress_all(raw_coredump);
+    if (!decompressed_coredump.has_value())
+        return ByteBuffer::copy(raw_coredump); // if we didn't manage to decompress it, try and parse it as decompressed coredump
+    return decompressed_coredump.value();
+}
+
+Reader::~Reader()
+{
+}
+
+Reader::NotesEntryIterator::NotesEntryIterator(const u8* notes_data)
+    : m_current((const ELF::Core::NotesEntry*)notes_data)
+    , start(notes_data)
+{
+}
+
+ELF::Core::NotesEntryHeader::Type Reader::NotesEntryIterator::type() const
+{
+    VERIFY(m_current->header.type == ELF::Core::NotesEntryHeader::Type::ProcessInfo
+        || m_current->header.type == ELF::Core::NotesEntryHeader::Type::MemoryRegionInfo
+        || m_current->header.type == ELF::Core::NotesEntryHeader::Type::ThreadInfo
+        || m_current->header.type == ELF::Core::NotesEntryHeader::Type::Metadata
+        || m_current->header.type == ELF::Core::NotesEntryHeader::Type::Null);
+    return m_current->header.type;
+}
+
+const ELF::Core::NotesEntry* Reader::NotesEntryIterator::current() const
+{
+    return m_current;
+}
+
+void Reader::NotesEntryIterator::next()
+{
+    VERIFY(!at_end());
+    switch (type()) {
+    case ELF::Core::NotesEntryHeader::Type::ProcessInfo: {
+        const auto* current = reinterpret_cast<const ELF::Core::ProcessInfo*>(m_current);
+        m_current = reinterpret_cast<const ELF::Core::NotesEntry*>(current->json_data + strlen(current->json_data) + 1);
+        break;
+    }
+    case ELF::Core::NotesEntryHeader::Type::ThreadInfo: {
+        const auto* current = reinterpret_cast<const ELF::Core::ThreadInfo*>(m_current);
+        m_current = reinterpret_cast<const ELF::Core::NotesEntry*>(current + 1);
+        break;
+    }
+    case ELF::Core::NotesEntryHeader::Type::MemoryRegionInfo: {
+        const auto* current = reinterpret_cast<const ELF::Core::MemoryRegionInfo*>(m_current);
+        m_current = reinterpret_cast<const ELF::Core::NotesEntry*>(current->region_name + strlen(current->region_name) + 1);
+        break;
+    }
+    case ELF::Core::NotesEntryHeader::Type::Metadata: {
+        const auto* current = reinterpret_cast<const ELF::Core::Metadata*>(m_current);
+        m_current = reinterpret_cast<const ELF::Core::NotesEntry*>(current->json_data + strlen(current->json_data) + 1);
+        break;
+    }
+    default:
+        VERIFY_NOT_REACHED();
+    }
+}
+
+bool Reader::NotesEntryIterator::at_end() const
+{
+    return type() == ELF::Core::NotesEntryHeader::Type::Null;
+}
+
+Optional<FlatPtr> Reader::peek_memory(FlatPtr address) const
+{
+    const auto* region = region_containing(address);
+    if (!region)
+        return {};
+
+    FlatPtr offset_in_region = address - region->region_start;
+    const char* region_data = image().program_header(region->program_header_index).raw_data();
+    return *(const FlatPtr*)(&region_data[offset_in_region]);
+}
+
+const JsonObject Reader::process_info() const
+{
+    const ELF::Core::ProcessInfo* process_info_notes_entry = nullptr;
+    for (NotesEntryIterator it((const u8*)m_coredump_image.program_header(m_notes_segment_index).raw_data()); !it.at_end(); it.next()) {
+        if (it.type() != ELF::Core::NotesEntryHeader::Type::ProcessInfo)
+            continue;
+        process_info_notes_entry = reinterpret_cast<const ELF::Core::ProcessInfo*>(it.current());
+        break;
+    }
+    if (!process_info_notes_entry)
+        return {};
+    auto process_info_json_value = JsonValue::from_string(process_info_notes_entry->json_data);
+    if (!process_info_json_value.has_value())
+        return {};
+    if (!process_info_json_value.value().is_object())
+        return {};
+    return process_info_json_value.value().as_object();
+    // FIXME: Maybe just cache this on the Reader instance after first access.
+}
+
+ELF::Core::MemoryRegionInfo const* Reader::first_region_for_object(StringView object_name) const
+{
+    ELF::Core::MemoryRegionInfo const* ret = nullptr;
+    for_each_memory_region_info([&ret, &object_name](auto& region_info) {
+        if (region_info.object_name() == object_name) {
+            ret = &region_info;
+            return IterationDecision::Break;
+        }
+        return IterationDecision::Continue;
+    });
+    return ret;
+}
+
+const ELF::Core::MemoryRegionInfo* Reader::region_containing(FlatPtr address) const
+{
+    const ELF::Core::MemoryRegionInfo* ret = nullptr;
+    for_each_memory_region_info([&ret, address](const ELF::Core::MemoryRegionInfo& region_info) {
+        if (region_info.region_start <= address && region_info.region_end >= address) {
+            ret = &region_info;
+            return IterationDecision::Break;
+        }
+        return IterationDecision::Continue;
+    });
+    return ret;
+}
+
+int Reader::process_pid() const
+{
+    auto process_info = this->process_info();
+    auto pid = process_info.get("pid");
+    return pid.to_number<int>();
+}
+
+u8 Reader::process_termination_signal() const
+{
+    auto process_info = this->process_info();
+    auto termination_signal = process_info.get("termination_signal");
+    auto signal_number = termination_signal.to_number<int>();
+    if (signal_number <= SIGINVAL || signal_number >= NSIG)
+        return SIGINVAL;
+    return (u8)signal_number;
+}
+
+String Reader::process_executable_path() const
+{
+    auto process_info = this->process_info();
+    auto executable_path = process_info.get("executable_path");
+    return executable_path.as_string_or({});
+}
+
+Vector<String> Reader::process_arguments() const
+{
+    auto process_info = this->process_info();
+    auto arguments = process_info.get("arguments");
+    if (!arguments.is_array())
+        return {};
+    Vector<String> vector;
+    arguments.as_array().for_each([&](auto& value) {
+        if (value.is_string())
+            vector.append(value.as_string());
+    });
+    return vector;
+}
+
+Vector<String> Reader::process_environment() const
+{
+    auto process_info = this->process_info();
+    auto environment = process_info.get("environment");
+    if (!environment.is_array())
+        return {};
+    Vector<String> vector;
+    environment.as_array().for_each([&](auto& value) {
+        if (value.is_string())
+            vector.append(value.as_string());
+    });
+    return vector;
+}
+
+HashMap<String, String> Reader::metadata() const
+{
+    const ELF::Core::Metadata* metadata_notes_entry = nullptr;
+    for (NotesEntryIterator it((const u8*)m_coredump_image.program_header(m_notes_segment_index).raw_data()); !it.at_end(); it.next()) {
+        if (it.type() != ELF::Core::NotesEntryHeader::Type::Metadata)
+            continue;
+        metadata_notes_entry = reinterpret_cast<const ELF::Core::Metadata*>(it.current());
+        break;
+    }
+    if (!metadata_notes_entry)
+        return {};
+    auto metadata_json_value = JsonValue::from_string(metadata_notes_entry->json_data);
+    if (!metadata_json_value.has_value())
+        return {};
+    if (!metadata_json_value.value().is_object())
+        return {};
+    HashMap<String, String> metadata;
+    metadata_json_value.value().as_object().for_each_member([&](auto& key, auto& value) {
+        metadata.set(key, value.as_string_or({}));
+    });
+    return metadata;
+}
+
+struct LibraryData {
+    String name;
+    OwnPtr<MappedFile> file;
+    ELF::Image lib_elf;
+};
+
+const Reader::LibraryData* Reader::library_containing(FlatPtr address) const
+{
+    static HashMap<String, OwnPtr<LibraryData>> cached_libs;
+    auto* region = region_containing(address);
+    if (!region)
+        return {};
+
+    auto name = region->object_name();
+
+    String path;
+    if (name.contains(".so"))
+        path = String::formatted("/usr/lib/{}", name);
+    else {
+        path = name;
+    }
+
+    if (!cached_libs.contains(path)) {
+        auto file_or_error = MappedFile::map(path);
+        if (file_or_error.is_error())
+            return {};
+        auto image = ELF::Image(file_or_error.value()->bytes());
+        cached_libs.set(path, make<LibraryData>(name, (FlatPtr)region->region_start, file_or_error.release_value(), move(image)));
+    }
+
+    auto lib_data = cached_libs.get(path).value();
+    return lib_data;
+}
+
+}

Userland/Libraries/LibCoredump/Reader.h

@@ -0,0 +1,109 @@
+/*
+ * Copyright (c) 2020, Itamar S. <itamar8910@gmail.com>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+#include <AK/HashMap.h>
+#include <AK/MappedFile.h>
+#include <AK/Noncopyable.h>
+#include <AK/OwnPtr.h>
+#include <LibELF/Core.h>
+#include <LibELF/Image.h>
+
+namespace Coredump {
+
+class Reader {
+    AK_MAKE_NONCOPYABLE(Reader);
+    AK_MAKE_NONMOVABLE(Reader);
+
+public:
+    static OwnPtr<Reader> create(const String&);
+    ~Reader();
+
+    template<typename Func>
+    void for_each_memory_region_info(Func func) const;
+
+    template<typename Func>
+    void for_each_thread_info(Func func) const;
+
+    const ELF::Image& image() const { return m_coredump_image; }
+
+    Optional<FlatPtr> peek_memory(FlatPtr address) const;
+    ELF::Core::MemoryRegionInfo const* first_region_for_object(StringView object_name) const;
+    const ELF::Core::MemoryRegionInfo* region_containing(FlatPtr address) const;
+
+    struct LibraryData {
+        String name;
+        FlatPtr base_address { 0 };
+        NonnullRefPtr<MappedFile> file;
+        ELF::Image lib_elf;
+    };
+    const LibraryData* library_containing(FlatPtr address) const;
+
+    int process_pid() const;
+    u8 process_termination_signal() const;
+    String process_executable_path() const;
+    Vector<String> process_arguments() const;
+    Vector<String> process_environment() const;
+    HashMap<String, String> metadata() const;
+
+private:
+    Reader(ReadonlyBytes);
+
+    static ByteBuffer decompress_coredump(const ReadonlyBytes&);
+
+    class NotesEntryIterator {
+    public:
+        NotesEntryIterator(const u8* notes_data);
+
+        ELF::Core::NotesEntryHeader::Type type() const;
+        const ELF::Core::NotesEntry* current() const;
+
+        void next();
+        bool at_end() const;
+
+    private:
+        const ELF::Core::NotesEntry* m_current { nullptr };
+        const u8* start { nullptr };
+    };
+
+    // Private as we don't need anyone poking around in this JsonObject
+    // manually - we know very well what should be included and expose that
+    // as getters with the appropriate (non-JsonValue) types.
+    const JsonObject process_info() const;
+
+    ByteBuffer m_coredump_buffer;
+    ELF::Image m_coredump_image;
+    ssize_t m_notes_segment_index { -1 };
+};
+
+template<typename Func>
+void Reader::for_each_memory_region_info(Func func) const
+{
+    for (NotesEntryIterator it((const u8*)m_coredump_image.program_header(m_notes_segment_index).raw_data()); !it.at_end(); it.next()) {
+        if (it.type() != ELF::Core::NotesEntryHeader::Type::MemoryRegionInfo)
+            continue;
+        auto& memory_region_info = reinterpret_cast<const ELF::Core::MemoryRegionInfo&>(*it.current());
+        IterationDecision decision = func(memory_region_info);
+        if (decision == IterationDecision::Break)
+            return;
+    }
+}
+
+template<typename Func>
+void Reader::for_each_thread_info(Func func) const
+{
+    for (NotesEntryIterator it((const u8*)m_coredump_image.program_header(m_notes_segment_index).raw_data()); !it.at_end(); it.next()) {
+        if (it.type() != ELF::Core::NotesEntryHeader::Type::ThreadInfo)
+            continue;
+        auto& thread_info = reinterpret_cast<const ELF::Core::ThreadInfo&>(*it.current());
+        IterationDecision decision = func(thread_info);
+        if (decision == IterationDecision::Break)
+            return;
+    }
+}
+
+}