refactor(walltime): extract symbolication module from perf/

Move ELF/DWARF artifact extraction (elf_helper, module_symbols,
unwind_data, debug_info, jit_dump) and the LoadedModule struct out
of perf/ into a shared symbolication/ module. These helpers are
agnostic to which sampling tool produced the module list and will
be reused by future Linux profilers.

Author: not-matthias

src/executor/wall_time/mod.rs

@@ -3,3 +3,4 @@ pub mod helpers;
 pub mod isolation;
 pub mod perf;
 pub mod profiler;
+pub mod symbolication;

src/executor/wall_time/perf/mod.rs

@@ -28,18 +28,13 @@ use std::path::Path;
 use std::path::PathBuf;
 use std::{cell::OnceCell, process::ExitStatus};
 
-mod jit_dump;
 mod naming;
 mod parse_perf_file;
 mod save_artifacts;
 pub(crate) mod setup;
 
-pub mod debug_info;
-pub mod elf_helper;
 pub mod fifo;
-pub mod module_symbols;
 pub mod perf_executable;
-pub mod unwind_data;
 
 const PERF_METADATA_CURRENT_VERSION: u64 = 1;
 const PERF_PIPEDATA_FILE_NAME: &str = "perf.pipedata";
@@ -318,7 +313,7 @@ impl BenchmarkData {
                 BenchmarkDataSaveError::FailedToHarvestPerfMaps
             })?;
         let jit_unwind_data_by_pid =
-            jit_dump::save_symbols_and_harvest_unwind_data_for_pids(path_ref, &tracked_pids)
+            crate::executor::wall_time::symbolication::jit_dump::save_symbols_and_harvest_unwind_data_for_pids(path_ref, &tracked_pids)
                 .await
                 .map_err(|e| {
                     error!("Failed to harvest jit dumps: {e}");

src/executor/wall_time/perf/parse_perf_file.rs

@@ -1,43 +1,17 @@
-use super::module_symbols::ModuleSymbols;
-use super::unwind_data::unwind_data_from_elf;
+use crate::executor::wall_time::symbolication::loaded_module::LoadedModule;
+use crate::executor::wall_time::symbolication::module_symbols::ModuleSymbols;
+use crate::executor::wall_time::symbolication::unwind_data::unwind_data_from_elf;
 use crate::prelude::*;
 use libc::pid_t;
 use linux_perf_data::PerfFileReader;
 use linux_perf_data::PerfFileRecord;
 use linux_perf_data::linux_perf_event_reader::EventRecord;
 use linux_perf_data::linux_perf_event_reader::RecordType;
-use runner_shared::unwind_data::ProcessUnwindData;
-use runner_shared::unwind_data::UnwindData;
 use std::collections::HashMap;
 use std::collections::HashSet;
 use std::path::Path;
 use std::path::PathBuf;
 
-#[derive(Default)]
-pub struct LoadedModule {
-    /// Symbols extracted from the mapped ELF file
-    pub module_symbols: Option<ModuleSymbols>,
-    /// Unwind data extracted from the mapped ELF file
-    pub unwind_data: Option<UnwindData>,
-    /// Per-process mounting information
-    pub process_loaded_modules: HashMap<pid_t, ProcessLoadedModule>,
-}
-
-#[derive(Default)]
-pub struct ProcessLoadedModule {
-    /// Load bias used to adjust declared elf addresses to their actual runtime addresses
-    /// The bias is the difference between where the segment *actually* is in memory versus where the ELF file *preferred* it to be
-    pub symbols_load_bias: Option<u64>,
-    /// Unwind data specific to the process mounting, derived from both load bias and the actual unwind data
-    pub process_unwind_data: Option<ProcessUnwindData>,
-}
-
-impl LoadedModule {
-    pub fn pids(&self) -> impl Iterator<Item = pid_t> {
-        self.process_loaded_modules.keys().copied()
-    }
-}
-
 pub struct MemmapRecordsOutput {
     /// Module symbols and the computed load bias for each pid that maps the ELF path.
     pub loaded_modules_by_path: HashMap<PathBuf, LoadedModule>,

src/executor/wall_time/perf/save_artifacts.rs

@@ -1,7 +1,7 @@
 use crate::executor::valgrind::helpers::ignored_objects_path::get_objects_path_to_ignore;
-use crate::executor::wall_time::perf::debug_info::debug_info_by_path;
 use crate::executor::wall_time::perf::naming;
-use crate::executor::wall_time::perf::parse_perf_file::LoadedModule;
+use crate::executor::wall_time::symbolication::debug_info::debug_info_by_path;
+use crate::executor::wall_time::symbolication::loaded_module::LoadedModule;
 use crate::prelude::*;
 use libc::pid_t;
 use rayon::prelude::*;

src/executor/wall_time/perf/snapshots/codspeed_runner__executor__wall_time__perf__debug_info__tests__ruff_debug_info.snap

@@ -1,6 +1,6 @@
 use super::elf_helper::find_debug_file;
-use super::parse_perf_file::LoadedModule;
-use crate::executor::wall_time::perf::module_symbols::ModuleSymbols;
+use super::loaded_module::LoadedModule;
+use super::module_symbols::ModuleSymbols;
 use crate::prelude::*;
 use addr2line::{fallible_iterator::FallibleIterator, gimli};
 use object::{Object, ObjectSection};

src/executor/wall_time/perf/snapshots/codspeed_runner__executor__wall_time__perf__module_symbols__tests__ruff_symbols.snap

@@ -1,5 +1,5 @@
 use crate::{
-    executor::wall_time::perf::module_symbols::{ModuleSymbols, Symbol},
+    executor::wall_time::symbolication::module_symbols::{ModuleSymbols, Symbol},
     prelude::*,
 };
 use linux_perf_data::jitdump::{JitDumpReader, JitDumpRecord};

src/executor/wall_time/perf/debug_info.rs …or/wall_time/symbolication/debug_info.rssrc/executor/wall_time/perf/debug_info.rs renamed to src/executor/wall_time/symbolication/debug_info.rs src/executor/wall_time/perf/debug_info.rs renamed to src/executor/wall_time/symbolication/debug_info.rs

@@ -0,0 +1,34 @@
+use crate::executor::wall_time::symbolication::module_symbols::ModuleSymbols;
+use libc::pid_t;
+use runner_shared::unwind_data::{ProcessUnwindData, UnwindData};
+use std::collections::HashMap;
+
+/// A loaded ELF module discovered while parsing a profiler's sample stream.
+///
+/// Holds the symbol/unwind data extracted from the file plus the per-process
+/// mounting metadata (load bias and rebased unwind data) for every pid that
+/// mapped this module.
+#[derive(Default)]
+pub struct LoadedModule {
+    /// Symbols extracted from the mapped ELF file
+    pub module_symbols: Option<ModuleSymbols>,
+    /// Unwind data extracted from the mapped ELF file
+    pub unwind_data: Option<UnwindData>,
+    /// Per-process mounting information
+    pub process_loaded_modules: HashMap<pid_t, ProcessLoadedModule>,
+}
+
+#[derive(Default)]
+pub struct ProcessLoadedModule {
+    /// Load bias used to adjust declared elf addresses to their actual runtime addresses
+    /// The bias is the difference between where the segment *actually* is in memory versus where the ELF file *preferred* it to be
+    pub symbols_load_bias: Option<u64>,
+    /// Unwind data specific to the process mounting, derived from both load bias and the actual unwind data
+    pub process_unwind_data: Option<ProcessUnwindData>,
+}
+
+impl LoadedModule {
+    pub fn pids(&self) -> impl Iterator<Item = pid_t> {
+        self.process_loaded_modules.keys().copied()
+    }
+}