Systems engineer who lives below the abstraction layer. Networking, kernel internals, performance engineering — I write about the things most people hand-wave past.https://harshith.in/en-usSun, 26 Apr 2026 07:09:53 GMTharshithsunku@gmail.com (Harshith Sunku)harshithsunku@gmail.com (Harshith Sunku)https://harshith.in/blog/gdb-multithreaded-c-debugging/https://harshith.in/blog/gdb-multithreaded-c-debugging/Multi-threaded bugs vanish under the debugger. Stack traces lie. Variables seem to change value between two prints. The GDB techniques that actually find races: thread-pinned breakpoints, non-stop mode, watchpoints, TSan-then-GDB, and rr.Sun, 26 Apr 2026 07:07:12 GMTdebuggingcthreadinglinuxhttps://harshith.in/blog/spinlocks-mutexes-futexes/https://harshith.in/blog/spinlocks-mutexes-futexes/Three primitives cover almost every locking decision: spinlocks, mutexes, and futexes. They solve the same problem with very different cost models — burning CPU vs trapping to the kernel vs the hybrid approach modern mutexes actually use.Sun, 26 Apr 2026 07:06:25 GMTthreadingkernelperformancelinuxhttps://harshith.in/blog/decoding-elf-how-linux-loads-binaries/https://harshith.in/blog/decoding-elf-how-linux-loads-binaries/You type ./a.out and a program runs. Behind that: a four-byte magic number, two parallel header tables, mmap-and-fault, and a dynamic linker that bootstraps itself. A guided tour of ELF and what execve actually does.Sun, 26 Apr 2026 07:05:42 GMTlinuxsystemscdebugginghttps://harshith.in/blog/dpdk-kernel-bypass-networking/https://harshith.in/blog/dpdk-kernel-bypass-networking/The Linux network stack tops out at ~2 Mpps per core. DPDK reaches 20 Mpps. The difference isn't speed — it's removing work: no syscalls, no allocations, no demux, no locks. Why kernel-bypass exists and when it earns its complexity.Sun, 26 Apr 2026 07:04:58 GMTnetworkingperformancesystemslinuxhttps://harshith.in/blog/ebpf-maps-explained/https://harshith.in/blog/ebpf-maps-explained/eBPF programs are stateless — maps are how they remember. A walk through the five map types that cover most real programs: hash, array, per-CPU, ring buffer, and LRU, with concrete code for each.Sun, 26 Apr 2026 07:04:13 GMTebpfkernellinuxhttps://harshith.in/blog/first-linux-kernel-module/https://harshith.in/blog/first-linux-kernel-module/Kernel modules sound intimidating, but the ceremony is small. In about a hundred lines: a working /dev/echo0 character device, with proper error unwinding, copy_to_user, and the patterns the rest of the kernel uses everywhere.Sun, 26 Apr 2026 07:03:29 GMTkernellinuxcsystemshttps://harshith.in/blog/received-packet-journey-nic-to-recv/https://harshith.in/blog/received-packet-journey-nic-to-recv/When recv() returns, dozens of layers ran behind your back — DMA, NAPI, sk_buff, protocol demux, the socket queue. A trace of the entire receive path with the cost of each hop and where the latency actually lives.Sun, 26 Apr 2026 07:02:49 GMTnetworkingkernellinuxtcp-iphttps://harshith.in/blog/memory-barriers-acquire-release/https://harshith.in/blog/memory-barriers-acquire-release/Compilers reorder. CPUs reorder. Single-threaded code never notices — multi-threaded code crashes mysteriously. A clear walk through acquire and release semantics, what the hardware actually does, and the spinlock that ties it all together.Sun, 26 Apr 2026 07:02:08 GMTsystemsthreadingcperformancehttps://harshith.in/blog/cache-lines-and-false-sharing/https://harshith.in/blog/cache-lines-and-false-sharing/Two threads, two variables, no shared state — and a 4x slowdown. False sharing is the cache coherence tax you pay when unrelated data lands on the same 64-byte line. How to spot it and fix it.Sun, 26 Apr 2026 07:01:30 GMTperformancesystemsthreadingchttps://harshith.in/blog/how-epoll-actually-works/https://harshith.in/blog/how-epoll-actually-works/select and poll were O(n). epoll is O(1). The reason isn't smarter scanning — it's that the kernel pushes ready events into a list as they happen. A walk through the data structures that make it work.Sun, 26 Apr 2026 07:00:53 GMTnetworkingkernellinuxperformancehttps://harshith.in/blog/ebpf-production-debugging/https://harshith.in/blog/ebpf-production-debugging/How eBPF changed production debugging — attaching probes to kernel functions, tracing syscalls, measuring latency histograms, and XDP packet processing, all without touching application code or rebooting.Sat, 21 Mar 2026 07:53:59 GMTebpfkernelperformancedebugginglinuxhttps://harshith.in/blog/esp32-mqtt-wake-on-lan/https://harshith.in/blog/esp32-mqtt-wake-on-lan/Using an ESP32 and ESP-IDF to build a network-aware Wake-on-LAN trigger — power on machines remotely over MQTT without an always-on server in between.Sat, 21 Mar 2026 07:53:59 GMTembeddednetworkingsystemschttps://harshith.in/blog/real-mode-bootloader-assembly/https://harshith.in/blog/real-mode-bootloader-assembly/The first 512 bytes that run when your machine powers on. A walkthrough of x86 real mode, the BIOS interrupt interface, segment:offset addressing, and what it takes to load a second stage.Sat, 21 Mar 2026 07:53:59 GMTassemblysystemskernellinuxhttps://harshith.in/blog/thread-safe-logging-c-lock-free/https://harshith.in/blog/thread-safe-logging-c-lock-free/Inside ThreadSafeCLogger: how to build a high-performance logging library that doesn't serialize your threads, using per-thread SPSC ring buffers and careful memory ordering.Sat, 21 Mar 2026 07:53:59 GMTcthreadingperformancesystemshttps://harshith.in/blog/software-tcp-ip-stack-c/https://harshith.in/blog/software-tcp-ip-stack-c/A deep dive into implementing a functional TCP/IP stack entirely in software — covering Ethernet framing, ARP, IP routing, TCP state machines, and the painful lessons learned along the way.Sat, 21 Mar 2026 07:53:59 GMTnetworkingctcp-ipsystems