Add support for Breakpad format

.github/workflows/test.yml

@@ -33,6 +33,10 @@ jobs:
             rust: stable
             profile: dev
             args: "--lib --no-default-features --features=apk"
+          - runs-on: ubuntu-latest
+            rust: stable
+            profile: dev
+            args: "--lib --no-default-features --features=breakpad"
           - runs-on: ubuntu-latest
             rust: stable
             profile: dev
@@ -153,8 +157,11 @@ jobs:
           cat Cargo.toml
     - name: cargo test -Zsanitizer=${{ matrix.sanitizer }}
       env:
-        CFLAGS: "-fsanitize=${{ matrix.sanitizer }}"
-        CXXFLAGS: "-fsanitize=${{ matrix.sanitizer }}"
+        # TODO: Ideally we'd set CFLAGS and CXXFLAGS as well, but some
+        #       of our transitive dependencies don't play nice with
+        #       ASAN.
+        #CFLAGS: "-fsanitize=${{ matrix.sanitizer }}"
+        #CXXFLAGS: "-fsanitize=${{ matrix.sanitizer }}"
         RUSTFLAGS: "-Zsanitizer=${{ matrix.sanitizer }}"
         ASAN_OPTIONS: "detect_odr_violation=0:detect_leaks=0"
         LSAN_OPTIONS: ""

.gitignore

@@ -1,6 +1,7 @@
 /data/*.bin
 /data/*.gsym
 /data/*.so
+/data/*.sym
 /data/*.zip
 /data/kallsyms
 /data/vmlinux-5.17.12-100.fc34.x86_64*

CHANGELOG.md

@@ -1,3 +1,9 @@
+Unreleased
+----------
+- Added support for Breakpad format behind `breakpad` feature (disabled
+  by default)
+
+
 0.2.0-alpha.10
 --------------
 - Introduced `symbolize::Reason` enum to provide best guess at why symbolization

Cargo.toml

@@ -47,6 +47,8 @@ apk = []
 # Note that by default backtraces will not be collected unless opted in with
 # environment variables.
 backtrace = []
+# Enable this feature to enable Breakpad support.
+breakpad = ["circular", "nom"]
 # Enable this feature to get transparent symbol demangling.
 demangle = ["cpp_demangle", "rustc-demangle"]
 # Enable this feature to enable DWARF support.
@@ -59,10 +61,10 @@ gsym = []
 
 # Enable this feature to opt in to the generation of unit test files.
 # Having these test files created is necessary for running tests.
-generate-unit-test-files = ["xz2", "zip"]
+generate-unit-test-files = ["dump_syms", "xz2", "zip"]
 # Enable this feature to opt in to the generation of large benchmark
 # files (also used for regression testing).
-generate-large-test-files = ["reqwest", "xz2"]
+generate-large-test-files = ["dump_syms", "reqwest", "xz2"]
 # Disable generation of test files. This feature takes preference over
 # `generate-unit-test-files`.
 dont-generate-unit-test-files = []
@@ -82,9 +84,11 @@ lto = true
 codegen-units = 1
 
 [dependencies]
+circular = {version = "0.3", optional = true}
 cpp_demangle = {version = "0.4", optional = true}
 gimli = {version = "0.28", optional = true}
 libc = "0.2.137"
+nom = {version = "7", optional = true}
 rustc-demangle = {version = "0.1.4", optional = true}
 tracing = {version = "0.1.27", default-features = false, features = ["attributes"], optional = true}
 
@@ -93,7 +97,7 @@ tracing = {version = "0.1.27", default-features = false, features = ["attributes
 # APIs.
 addr2line = "=0.21.0"
 anyhow = "1.0.71"
-blazesym = {path = ".", features = ["generate-unit-test-files", "apk", "gsym", "tracing"]}
+blazesym = {path = ".", features = ["generate-unit-test-files", "apk", "breakpad", "gsym", "tracing"]}
 criterion = {version = "0.5.1", default-features = false, features = ["rayon", "cargo_bench_support"]}
 env_logger = "0.10"
 scopeguard = "1.2"
@@ -102,13 +106,14 @@ test-log = {version = "0.2", default-features = false, features = ["trace"]}
 tracing-subscriber = {version = "0.3", default-features = false, features = ["env-filter", "fmt"]}
 
 [build-dependencies]
+dump_syms = {version = "2.2", optional = true, default-features = false}
 libc = "0.2.137"
 reqwest = {version = "0.11.18", optional = true, features = ["blocking"]}
 xz2 = {version = "0.1.7", optional = true}
 zip = {version = "0.6.4", optional = true, default-features = false}
 
 # https://docs.rs/about/metadata
 [package.metadata.docs.rs]
-features = ["apk", "backtrace", "demangle", "dwarf", "gsym"]
+features = ["apk", "backtrace", "breakpad", "demangle", "dwarf", "gsym"]
 # defines the configuration attribute `docsrs`
 rustdoc-args = ["--cfg", "docsrs"]

README.md

@@ -43,6 +43,9 @@ format supports and whether **blazesym** can currently use this feature:
 
 | Format        | Feature                          | Supported by format?     | Supported by blazesym?   |
 | ------------- | -------------------------------- | ------------------------ | ------------------------ |
+| Breakpad      | symbol size                      | :heavy_check_mark:       | :heavy_check_mark:       |
+|               | source code location information | :heavy_check_mark:       | :heavy_check_mark:       |
+|               | inlined function information     | :heavy_check_mark:       | :heavy_check_mark:       |
 | ELF           | symbol size                      | :heavy_check_mark:       | :heavy_check_mark:       |
 |               | source code location information | :heavy_multiplication_x: | :heavy_multiplication_x: |
 |               | inlined function information     | :heavy_multiplication_x: | :heavy_multiplication_x: |

benches/symbolize.rs

@@ -3,6 +3,7 @@
 use std::hint::black_box;
 use std::path::Path;
 
+use blazesym::symbolize::Breakpad;
 use blazesym::symbolize::Elf;
 use blazesym::symbolize::GsymFile;
 use blazesym::symbolize::Input;
@@ -35,6 +36,25 @@ fn symbolize_process() {
     assert_eq!(results.len(), addrs.len());
 }
 
+/// Symbolize an address in a Breakpad (*.sym) file, end-to-end, i.e.,
+/// including all necessary setup.
+fn symbolize_breakpad() {
+    let sym_vmlinux = Path::new(&env!("CARGO_MANIFEST_DIR"))
+        .join("data")
+        .join("vmlinux-5.17.12-100.fc34.x86_64.sym");
+    let src = Source::from(Breakpad::new(sym_vmlinux));
+    let symbolizer = Symbolizer::new();
+
+    let result = symbolizer
+        .symbolize_single(black_box(&src), black_box(Input::FileOffset(0x10ecb0)))
+        .unwrap()
+        .into_sym()
+        .unwrap();
+
+    assert_eq!(result.name, "abort_creds");
+    assert_eq!(result.code_info.as_ref().unwrap().line, Some(534));
+}
+
 /// Symbolize an address in an ELF file, end-to-end, i.e., including all
 /// necessary setup.
 fn symbolize_elf() {
@@ -161,6 +181,7 @@ where
     M: Measurement,
 {
     bench_fn!(group, symbolize_process);
+    bench_fn!(group, symbolize_breakpad);
     bench_fn!(group, symbolize_elf);
     bench_fn!(group, symbolize_dwarf_no_lines);
     bench_fn!(group, symbolize_dwarf);

build.rs

@@ -205,6 +205,41 @@ fn dwarf(src: &Path, dst: impl AsRef<OsStr>) {
     strip(src, dst, &["--keep-section=.debug_*"])
 }
 
+/// Generate a Breakpad .sym file for the given source.
+#[cfg(feature = "dump_syms")]
+fn syms(src: &Path, dst: impl AsRef<OsStr>) {
+    use std::env::consts::ARCH;
+
+    use dump_syms::dumper;
+    use dump_syms::dumper::Config;
+    use dump_syms::dumper::FileOutput;
+    use dump_syms::dumper::Output;
+
+    let dst = src.with_file_name(dst);
+
+    let config = Config {
+        output: Output::File(FileOutput::Path(dst)),
+        symbol_server: None,
+        debug_id: None,
+        code_id: None,
+        arch: ARCH,
+        num_jobs: 1,
+        check_cfi: false,
+        emit_inlines: true,
+        mapping_var: None,
+        mapping_src: None,
+        mapping_dest: None,
+        mapping_file: None,
+    };
+    let path = src.to_str().unwrap();
+    let () = dumper::single_file(&config, path).unwrap();
+}
+
+#[cfg(not(feature = "dump_syms"))]
+fn syms(_src: &Path, _dst: impl AsRef<OsStr>) {
+    unimplemented!()
+}
+
 /// Unpack an xz compressed file.
 #[cfg(feature = "xz2")]
 fn unpack_xz(src: &Path, dst: &Path) {
@@ -382,6 +417,9 @@ fn prepare_test_files(crate_root: &Path) {
     gsym(&src, "test-stable-addresses.gsym");
     dwarf(&src, "test-stable-addresses-dwarf-only.bin");
     strip(&src, "test-stable-addresses-stripped.bin", &[]);
+    if cfg!(feature = "dump_syms") {
+        syms(&src, "test-stable-addresses.sym");
+    }
 
     let src = crate_root.join("data").join("kallsyms.xz");
     let mut dst = src.clone();
@@ -474,10 +512,15 @@ fn prepare_bench_files(crate_root: &Path) {
     let dst = dst.file_name().unwrap();
     gsym(&vmlinux, dst);
 
-    let mut dst = vmlinux_xz;
+    let mut dst = vmlinux_xz.clone();
     assert!(dst.set_extension("dwarf"));
     let dst = dst.file_name().unwrap();
     dwarf(&vmlinux, dst);
+
+    let mut dst = vmlinux_xz;
+    assert!(dst.set_extension("sym"));
+    let dst = dst.file_name().unwrap();
+    syms(&vmlinux, dst);
 }
 
 fn main() {

cli/CHANGELOG.md

@@ -1,3 +1,8 @@
+Unreleased
+----------
+- Added support for symbolization using Breakpad (`*.sym`) files
+
+
 0.1.2
 -----
 - Bumped `blazesym` dependency to `0.2.0-alpha.10`

cli/Cargo.toml

@@ -21,7 +21,7 @@ grev = "0.1.3"
 
 [dependencies]
 anyhow = "1.0.68"
-blazesym = {version = "=0.2.0-alpha.10", path = "../", features = ["apk", "demangle", "dwarf", "gsym", "tracing"]}
+blazesym = {version = "=0.2.0-alpha.10", path = "../", features = ["apk", "breakpad", "demangle", "dwarf", "gsym", "tracing"]}
 clap = {version = "4.1.7", features = ["derive"]}
 clap_complete = {version = "4.1.1", optional = true}
 tracing = "0.1"

cli/src/args.rs

@@ -78,11 +78,25 @@ pub struct User {
 /// A type representing the `symbolize` command.
 #[derive(Debug, Subcommand)]
 pub enum Symbolize {
+    Breakpad(Breakpad),
     Elf(Elf),
     Gsym(Gsym),
     Process(Process),
 }
 
+#[derive(Debug, Arguments)]
+pub struct Breakpad {
+    /// The path to the Breakpad (*.sym) file.
+    #[clap(short, long)]
+    pub path: PathBuf,
+    /// The addresses to symbolize.
+    ///
+    /// Addresses are assumed to be file offsets as they would be used on the
+    /// original (ELF/DWARF/...) source file.
+    #[arg(value_parser = parse_addr)]
+    pub addrs: Vec<Addr>,
+}
+
 #[derive(Debug, Arguments)]
 pub struct Elf {
     /// The path to the ELF file.

cli/src/main.rs

@@ -111,6 +111,12 @@ fn print_frame(
 fn symbolize(symbolize: args::Symbolize) -> Result<()> {
     let symbolizer = Symbolizer::new();
     let (src, input, addrs) = match symbolize {
+        args::Symbolize::Breakpad(args::Breakpad { path, ref addrs }) => {
+            let src = symbolize::Source::from(symbolize::Breakpad::new(path));
+            let addrs = addrs.as_slice();
+            let input = symbolize::Input::FileOffset(addrs);
+            (src, input, addrs)
+        }
         args::Symbolize::Elf(args::Elf { path, ref addrs }) => {
             let src = symbolize::Source::from(symbolize::Elf::new(path));
             let addrs = addrs.as_slice();