Fastrace v1.0.0 - Major Performance Release

This is the first stable major release of fastrace, featuring significant performance improvements and new capabilities.

Performance Improvements

• 20x faster than standard traceroute in typical scenarios
• Batch I/O with recvmmsg on Linux (up to 60% less syscall overhead)
• Adaptive concurrency window based on network latency
• Optimized timeouts (500ms hop, 1ms poll) for faster trace completion
• Lock-free DNS queue with atomic operations
• CLOCK_MONOTONIC_RAW for precise RTT measurements

New Features

Quick Start

# Build
make

# Standard trace
sudo ./fastrace google.com

# IPv6 with ICMP
sudo ./fastrace -6 -I ipv6.google.com

# JSON output with metrics
sudo ./fastrace --json --metrics cloudflare.com

Test Suite

make test        # Basic tests (no root)
sudo make test-full   # Full test suite
sudo make benchmark   # Performance benchmark

See CHANGELOG.md for full details.

Performance

• Nanosleep Precision: Replaced usleep() with nanosleep() for more accurate inter-probe delays and POSIX compliance.
• DNS Pool Allocator: Added pre-allocated pool for DNS queue items, reducing malloc overhead in high-throughput scenarios.
• Faster Early Exit: Optimized main loop to check finish condition more frequently, reducing latency when destination is reached.

Added

• Trace Statistics: Displays hop count and response rate on completion for better observability.

Fixed

• Hop Display with Concurrent Probing: Fixed a bug where intermediate hops were not displayed when the destination was reached by a concurrent probe. The trace now correctly shows all hops up to the destination.

Code Quality

• Improved code organization with dedicated pool allocator functions.
• Added statistics counters for probes sent and responses received.

Performance

• TTL Batching: Sets socket TTL once per hop and sends probes in bursts, reducing per-packet syscalls and improving send-side throughput under high concurrency.
• Adaptive Waiting: Poll wait now aligns to the earliest hop deadline, cutting idle spins while still draining the ICMP socket aggressively.

Added

• Probe Delay Flag: New -d <microseconds> option exposes inter-probe pacing (default 250µs) to tune burstiness versus jitter.

Changed

• Clearer Hop Completion Logic: Hop readiness now factors both full receipt and TTL deadlines, yielding more deterministic timeout behavior in noisy networks.

Fixed

• Critical RTT Bug: Fixed incorrect RTT measurements showing values like 1764330397890.64 ms. The bug was caused by mixing SO_TIMESTAMP (wall-clock time from kernel) with CLOCK_MONOTONIC for sent timestamps. Now consistently uses monotonic clock for all RTT calculations.

Performance

• Asynchronous DNS Resolution: Moved reverse DNS lookups to a dedicated background thread. This eliminates the "stop-and-wait" behavior during printing, ensuring the packet probing loop never stalls while waiting for a hostname to resolve.
• Kernel Timestamps: Implemented SO_TIMESTAMP support to retrieve packet arrival times directly from the kernel. This bypasses user-space scheduling jitter, providing significantly more accurate RTT measurements.

Changed

• Thread-Safe Architecture: Refactored internal data structures (host cache, print queue) to be thread-safe using mutexes.
• Build System: Added -pthread flag to compiler options and introduced a make test target.

Added

• Test Suite: Added a basic regression test script (tests/test_basic.sh) to verify core functionality and argument parsing.

Fixed

• Removed artificial RTT clamping that was hiding legitimate network measurements
  • Previously clamped RTT between 0.05ms and 800ms
  • Now reports actual values including sub-50µs localhost responses and >800ms satellite/long-distance links
  • Added sanity check for negative RTT to detect clock issues without corrupting data

This fix restores full diagnostic capability for detecting network anomalies like bufferbloat and measuring true round-trip times across all network types.

Full Changelog: v0.2.0...v0.2.1

A major performance and feature release delivering state-of-the-art traceroute capabilities with blazing speed and minimal resource usage.

Highlights

• 65-75% faster trace completion with non-blocking event loop
• 75-80% less memory usage through dynamic allocation
• 70-80% lower CPU utilization with poll-driven I/O
• Runtime tuning via comprehensive CLI options
• DNS caching to eliminate redundant lookups
• Monotonic timing for nanosecond-precision RTT tracking

Performance Improvements

• Non-blocking event loop: Replaced select()-based architecture with poll() to minimize wakeups and efficiently drain ICMP response bursts
• Monotonic timing: Switched from wall-clock to CLOCK_MONOTONIC for nanosecond-precision RTT tracking and tighter hop completion deadlines
• Adaptive concurrency: Configurable probe counts, TTL windows, and wait intervals enable faster route convergence across diverse network conditions
• Optimized packet emission: Pre-generated probe payloads and tuned UDP transmission cadence reduce per-packet overhead

New Features

Runtime Configuration

New command-line options for fine-grained control:

• -n — Disable reverse DNS lookups for faster output
• -m <hops> — Set maximum hop count (1-128, default 30)
• -q <probes> — Configure probes per hop (1-10, default 3)
• -c <count> — Adjust concurrent TTL window (default 6)
• -W <ms> — Set poll wait timeout (default 2ms)
• -t <ms> — Configure hop completion timeout (default 700ms)
• -P <port> — Specify base UDP port (default 33434)
• -V — Display version information

DNS Caching

Lightweight hostname cache prevents redundant reverse lookups when encountering load-balanced hops, significantly accelerating output generation.

Enhanced Output

Dynamic configuration summary displays active settings (protocol, hop limits, probe count) in the banner for better observability.

🛡️ Reliability Enhancements

• Modern name resolution: Migrated to getaddrinfo()/getnameinfo() for thread-safe, standards-compliant DNS operations
• Hardened ICMP parsing: Comprehensive bounds checking across all packet layers prevents buffer overruns
• Resource management: Expanded socket buffers and safer cleanup routines improve stability under heavy traffic

📊 Benchmark Comparison

Security & Reliability

• Added proper file descriptor flags with FD_CLOEXEC for socket security
• Implemented comprehensive bounds checking to prevent buffer overflows
• Enhanced memory safety with null pointer checks and proper initialization
• Improved error handling throughout the codebase

Performance

• Further optimized socket configuration with larger buffer sizes
• Reduced probe delay to 1ms for faster tracing operations
• Improved RTT measurement accuracy with better timing algorithm
• Added constants for configurable response processing iterations

Code Quality

• Refactored socket management with proper cleanup in error paths
• Added additional debug information for troubleshooting
• Improved source code organization with clearer separation of functions
• Enhanced input validation for more robust operation
• Fixed potential issues with type casting and numeric comparisons

Fixed

• Significantly improved RTT measurement accuracy to align with standard traceroute tools
• Added proper timersub macro for correct time difference calculations with microsecond precision
• Fixed timestamp handling to ensure accurate RTT measurements
• Optimized timeout values for more responsive trace results

Changed

• Reduced probe delay from 50ms to 5ms for faster tracing
• Improved response processing with more frequent checks and shorter timeouts
• Added sanity checks and reasonable caps for RTT values (maximum 1000ms)
• Reduced socket receive timeout from 5s to 1s
• Reduced TTL timeout from 5000ms to 1000ms for better responsiveness
• Improved socket resource cleanup on program exit

[0.1.0] - 2025-03-13

Added

• Initial release of FastRace
• High-performance traceroute implementation in pure C with zero dependencies
• Dual socket architecture with UDP for probes and raw ICMP for responses
• Concurrent TTL probing for faster route discovery (up to 5x faster than standard traceroute)
• Visual tree-like representation of network paths showing load-balanced routes
• Efficient memory management with minimal memory footprint
• Optimized for speed with parallel probing and efficient packet handling
• Smart timeout management with adaptive timeouts based on hop distance
• Comprehensive route visualization with RTT measurements and hostname resolution

Technical Details

• Implemented using standard C libraries and system calls
• UDP probes with incremental TTL values to discover network path
• Raw socket implementation for capturing ICMP responses
• Port-based probe identification for accurate response matching
• Non-blocking I/O with select() for efficient packet handling
• Probe batching system with multiple probes per TTL
• Flexible build system with optimization options

Code Improvements

• Added proper random number generator seeding
• Streamlined command line argument handling
• Improved error messages and socket handling
• Removed unnecessary privilege checking in favor of runtime capability detection
• Eliminated redundant cleanup code for better performance
• Improved code structure and variable declaration

Known Issues

• IPv6 support not yet implemented
• Limited support for non-POSIX systems
• May require elevated privileges depending on system configuration