Parse __eh_frame DWARF info on macOS for per-PC frame descriptions by jbachorik · Pull Request #430 · DataDog/java-profiler

jbachorik · 2026-03-22T23:04:01Z

What

Parse raw __eh_frame DWARF CFI data on macOS to provide per-PC frame descriptions for GCC-built libraries, improving stack unwinding accuracy for functions with non-standard frame layouts (leaf functions, -fomit-frame-pointer).

Changes

`ddprof-lib/src/main/cpp/dwarf.h` / `dwarf.cpp`

Add new DwarfParser constructor accepting a raw (eh_frame, size) pair (no binary-search index).
Implement parseEhFrame(): linearly iterates CIE/FDE records, extracts code/data alignment from each CIE, decodes per-PC frame descriptions from FDEs, then qsorts the result.
Add _has_z_augmentation field: set when the CIE augmentation string starts with 'z'. The FDE augmentation-data-length LEB field is only present in that case; guarding the skip prevents corrupting the instruction stream for CIEs with empty or non-z augmentation. The parser assumes a single CIE per module — the DWARF spec allows multiple CIEs with different augmentation strings, but macOS binaries compiled by clang always emit one CIE per module and this path is only exercised for macOS __eh_frame sections.
Pass detectedDefaultFrame() to setDwarfTable(), matching the Linux path and correctly selecting the clang frame layout on aarch64.
Add bounds check in CIE parsing before getLeb()/getSLeb() calls to guard against malformed or truncated __eh_frame data.

`ddprof-lib/src/main/cpp/symbols_macos.cpp`

Detect __eh_frame section in the __TEXT segment during Mach-O parsing.
When DWARF_SUPPORTED and the section is present, construct a DwarfParser and call cc->setDwarfTable() with the detected default frame.
Add missing Symbols::clearParsingCaches() definition (fixes linker error on macOS).

`ddprof-lib/src/test/cpp/dwarf_ut.cpp` (new)

Four gtest unit tests covering parseEhFrame(): empty section, terminator-only, CIE-only, and CIE+FDE.

Why

On macOS, GCC-built shared libraries (libgcc, Homebrew packages) emit __eh_frame with standard DWARF CFI, but no __unwind_info. Previously the profiler fell back to a single library-wide heuristic frame layout for these libraries. Parsing the DWARF FDE table yields exact CFA/FP/PC rules per instruction address, eliminating unwinding failures at function prologues and epilogues.

JIRA: PROF-14088

dd-octo-sts · 2026-03-22T23:19:03Z

CI Test Results

Run: #24088148738 | Commit: fc2c27a | Duration: 22m 32s (longest job)

✅ All 32 test jobs passed

Status Overview

JDK	glibc-aarch64/debug	glibc-amd64/debug	musl-aarch64/debug	musl-amd64/debug
8	-	✅	-	-
8-ibm	-	✅	-	-
8-j9	✅	✅	-	-
8-librca	-	-	✅	✅
8-orcl	-	✅	-	-
11	-	✅	-	-
11-j9	✅	✅	-	-
11-librca	-	-	✅	✅
17	✅	✅	-	-
17-graal	✅	✅	-	-
17-j9	✅	✅	-	-
17-librca	-	-	✅	✅
21	✅	✅	-	-
21-graal	✅	✅	-	-
21-librca	-	-	✅	✅
25	✅	✅	-	-
25-graal	✅	✅	-	-
25-librca	-	-	✅	✅

Legend: ✅ passed | ❌ failed | ⚪ skipped | 🚫 cancelled

Summary: Total: 32 | Passed: 32 | Failed: 0

Updated: 2026-04-07 15:23:06 UTC

Copilot

Pull request overview

Adds macOS support for parsing raw __eh_frame DWARF CFI to produce per-PC frame descriptions, improving unwinding for binaries that don’t provide __unwind_info (e.g., GCC-built libs).

Changes:

Introduces a DwarfParser constructor + parseEhFrame() path to parse linear .eh_frame / __eh_frame without an .eh_frame_hdr index.
Extends macOS Mach-O parsing to locate __eh_frame and populate CodeCache’s DWARF table; also adds Symbols::clearParsingCaches() for macOS.
Adds new gtest unit tests for parseEhFrame().

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

File	Description
`ddprof-lib/src/main/cpp/dwarf.h`	Declares the new raw-eh-frame constructor and `parseEhFrame()` API.
`ddprof-lib/src/main/cpp/dwarf.cpp`	Implements linear `__eh_frame` parsing and augmentation handling.
`ddprof-lib/src/main/cpp/symbols_macos.cpp`	Finds `__eh_frame` in Mach-O `__TEXT` and wires parsed DWARF into `CodeCache`.
`ddprof-lib/src/test/cpp/dwarf_ut.cpp`	Adds unit tests exercising the new linear `.eh_frame` parsing path.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

ddprof-lib/src/main/cpp/dwarf.cpp

ddprof-lib/src/main/cpp/symbols_macos.cpp

ddprof-lib/src/test/cpp/dwarf_ut.cpp

dd-octo-sts · 2026-04-03T14:21:37Z

Scan-Build Report

User:	runner@runnervmrg6be
Working Directory:	/home/runner/work/java-profiler/java-profiler/ddprof-lib/src/test/make
Command Line:	make -j4 all
Clang Version:	Ubuntu clang version 18.1.3 (1ubuntu1)
Date:	Fri Apr 3 14:40:28 2026

Bug Summary

Logic error
Bug Type	Quantity	Display?
All Bugs	1
Stack address stored into global variable	1

Reports

Bug Group	Bug Type ▾	File	Function/Method	Line	Path Length
Logic error	Stack address stored into global variable	stackWalker.cpp	walkVM	848	37

Revert static thread_local jmp_buf in walkVM — thread_local access via __tls_get_addr is not async-signal-safe in a dlopen'd agent. Suppress core.StackAddressEscape in scan-build instead: the jmp_buf address is always restored to saved_exception before walkVM returns. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 4 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-07T10:44:05Z

ddprof-lib/src/main/cpp/dwarf.cpp

+    const char *record_end = record_start + 4 + length;
+    if (record_end > section_end) {
+      break;
+    }


record_end is computed via pointer arithmetic (record_start + 4 + length) without validating that length fits within the remaining section size first. With malformed/truncated data, this can trigger pointer overflow/UB before the record_end > section_end guard. Consider checking length > (size_t)(section_end - record_start - 4) (or similar) before forming record_end and bailing out early.

Suggested change

const char *record_end = record_start + 4 + length;

if (record_end > section_end) {

break;

}

size_t remaining = (size_t)(section_end - record_start);

if ((size_t)length > remaining - 4) {

break;

}

const char *record_end = record_start + 4 + length;

Fixed. Added if (length > (size_t)(section_end - record_start) - 4) { break; } before forming record_end.

Copilot · 2026-04-07T10:44:06Z

ddprof-lib/src/main/cpp/dwarf.cpp

+      // binaries are not produced by the toolchains we target here.
+      _ptr++;  // skip version


CIE parsing advances and dereferences _ptr (_ptr++ / *_ptr == 'z') without first ensuring there’s at least 1 byte available in the record. With a short/empty CIE body this can read past record_end. Add explicit if (_ptr >= record_end) { _ptr = record_end; continue; } (or similar) before skipping the version / reading the augmentation.

Suggested change

// binaries are not produced by the toolchains we target here.

_ptr++; // skip version

// binaries are not produced by the toolchains we target here.

if (_ptr >= record_end) {

_ptr = record_end;

continue;

}

_ptr++; // skip version

if (_ptr >= record_end) {

_ptr = record_end;

continue;

}

Fixed. Added if (_ptr >= record_end) { _ptr = record_end; continue; } guards before the version skip and before the augmentation string read.

Copilot · 2026-04-07T10:44:06Z

ddprof-lib/src/main/cpp/dwarf.cpp

+      if (_ptr + 2 > record_end) {
+        _ptr = record_end;
+        continue;
+      }
+      _code_align = getLeb();
+      _data_align = getSLeb();


The new “bounds check” (if (_ptr + 2 > record_end)) is not sufficient to make getLeb() / getSLeb() safe: LEB128 values can span more than 2 bytes, and getLeb() will keep reading until it finds a byte without the 0x80 continuation bit. With truncated data this can overrun record_end and read arbitrary memory. Consider adding end-bounded LEB decode helpers (e.g., getLeb(end) / getSLeb(end) returning a status) and using them here (and for the FDE augmentation-length LEB).

Fixed. Added bounded getLeb(end) and getSLeb(end) overloads that stop reading at end, and switched the CIE code_align/data_align reads to use them with record_end as the bound.

Copilot · 2026-04-07T10:44:06Z

ddprof-lib/src/main/cpp/dwarf.cpp

+      if (_has_z_augmentation) {
+        _ptr += getLeb();  // skip augmentation data length + data
+        if (_ptr > record_end) {
+          break;
+        }


In the FDE path, _ptr += getLeb() assumes the augmentation-data-length LEB is fully readable. If the record ends mid-LEB, getLeb() can read past record_end (same issue as in CIE parsing). Use a bounded LEB decoder (or at least check _ptr < record_end per-byte) and fail the record if the LEB doesn’t complete within record_end.

Fixed. The FDE augmentation-data-length read now uses getLeb(record_end) (the bounded overload).

- Add overflow guard before record_end in parseEhFrame - Guard _ptr accesses in short CIE body path - Add bounded getLeb/getSLeb overloads; use in parseEhFrame - Extract DwarfParser::init() to remove constructor duplication - Demote _eh_frame/_eh_frame_size to locals in MachOParser::parse() - Add bounds-guard tests: TruncatedRecord, ShortCieBody, FdeAugDataOverrun - Document table() ownership, CIE/FDE layout assumptions, EOF handling Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

jbachorik changed the title ~~PROF-14088~~ Parse __eh_frame DWARF info on macOS for per-PC frame descriptions Mar 22, 2026

jbachorik requested a review from Copilot March 23, 2026 08:37

Copilot started reviewing on behalf of jbachorik March 23, 2026 08:37 View session

Copilot AI reviewed Mar 23, 2026

View reviewed changes

ddprof-lib/src/main/cpp/dwarf.cpp Show resolved Hide resolved

ddprof-lib/src/main/cpp/dwarf.cpp Show resolved Hide resolved

ddprof-lib/src/main/cpp/symbols_macos.cpp Outdated Show resolved Hide resolved

ddprof-lib/src/test/cpp/dwarf_ut.cpp Show resolved Hide resolved

jbachorik force-pushed the jb/prof-14088 branch from 2f622b0 to bea854f Compare April 3, 2026 14:18

jbachorik force-pushed the jb/prof-14088 branch 3 times, most recently from dfe7b83 to 4466cfb Compare April 3, 2026 14:38

jbachorik marked this pull request as ready for review April 3, 2026 14:51

jbachorik requested a review from a team as a code owner April 3, 2026 14:51

PROF-14088

2a6e628

jbachorik force-pushed the jb/prof-14088 branch from 4466cfb to 2a6e628 Compare April 3, 2026 15:01

jbachorik marked this pull request as draft April 3, 2026 15:43

jbachorik and others added 2 commits April 3, 2026 18:14

fix: single-dash for scan-build -disable-checker flag

9aba288

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

jbachorik marked this pull request as ready for review April 3, 2026 16:49

jbachorik requested a review from Copilot April 7, 2026 10:37

Copilot started reviewing on behalf of jbachorik April 7, 2026 10:38 View session

Copilot AI reviewed Apr 7, 2026

View reviewed changes

		// binaries are not produced by the toolchains we target here.
		_ptr++; // skip version

Conversation

jbachorik commented Mar 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Changes

ddprof-lib/src/main/cpp/dwarf.h / dwarf.cpp

ddprof-lib/src/main/cpp/symbols_macos.cpp

ddprof-lib/src/test/cpp/dwarf_ut.cpp (new)

Why

Uh oh!

dd-octo-sts bot commented Mar 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CI Test Results

Status Overview

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

dd-octo-sts bot commented Apr 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Scan-Build Report

Bug Summary

Reports

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

jbachorik Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

jbachorik Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

jbachorik Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

jbachorik Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

jbachorik commented Mar 22, 2026 •

edited

Loading

`ddprof-lib/src/main/cpp/dwarf.h` / `dwarf.cpp`

`ddprof-lib/src/main/cpp/symbols_macos.cpp`

`ddprof-lib/src/test/cpp/dwarf_ut.cpp` (new)

dd-octo-sts bot commented Mar 22, 2026 •

edited

Loading

dd-octo-sts bot commented Apr 3, 2026 •

edited

Loading