File I/O Buffering in Linux: Complete Developer Guide

File I/O Buffering

How Linux buffers your data — from write() to the actual disk — and how to control it.

5

Core Topics

30+

Code Examples

40+

Interview Q&A

Free

No Cost

What You Will Learn

When you call write() in C, does data go to the disk immediately? No — and understanding why is critical for writing correct, fast Linux programs. This chapter explains the two layers of buffering that sit between your code and the physical disk: the stdio library buffer (in your process memory) and the kernel buffer cache (in kernel memory). You will learn to control both layers, force disk writes when needed, and avoid common bugs caused by mixing stdio and system calls.

🔧 Prerequisites

You should know: basic C programming, what open() / read() / write() / close() do, and have a rough idea that Linux has a kernel and a user space. That’s all.

🔑 Key Terms in This Chapter

Buffer Cache Page Cache stdio Buffering setvbuf() fflush() fsync() fdatasync() sync() O_SYNC O_DIRECT posix_fadvise() fileno() fdopen() Direct I/O Synchronized I/O _IONBF _IOLBF _IOFBF

The Big Picture: Two Layers of Buffering

Your C Code
printf(), fwrite(), fputs()

stdio Buffer (Layer 1)
Lives in your process memory · Controlled by setvbuf(), fflush()

System Call Boundary
write() / read()

Kernel Buffer Cache (Layer 2)
Lives in kernel memory · Controlled by fsync(), O_SYNC

Disk Hardware
Physical disk / SSD

Data travels through two buffers before reaching the disk. Each layer can be controlled independently.

Topics in This Chapter

🗄️

Part 1 Kernel Buffer Cache

Learn how Linux caches all disk reads and writes in RAM. Understand why a larger write buffer means fewer system calls and massively better performance. Includes real benchmark data from the book.

Buffer Cache Internals Page Cache Buffer Size vs Performance read() / write() mechanics

📦

Part 2 stdio Library Buffering

The C standard library buffers data before calling write(). Learn the three buffering modes, how to change them with setvbuf(), and how to force a flush using fflush(). Critical for embedded and real-time systems.

_IONBF · _IOLBF · _IOFBF setvbuf() setbuf() fflush()

💾

Part 3 Controlling Kernel Buffering

Sometimes you must guarantee data is physically on disk — think databases, journaling, crash recovery. Learn fsync(), fdatasync(), sync(), and the O_SYNC flag, plus their real performance cost.

fsync() fdatasync() sync() O_SYNC flag O_DSYNC · O_RSYNC Synchronized I/O

🚀

Part 4 posix_fadvise() and Direct I/O

Tell the kernel how you plan to access a file so it can optimize caching. And for databases and specialized apps, learn how to bypass the buffer cache entirely using O_DIRECT with strict alignment rules.

posix_fadvise() O_DIRECT Alignment Rules memalign()

🔀

Part 5 Mixing stdio and System Calls

Mixing printf() with write() on the same file? This can cause output to appear out of order. Learn fileno() and fdopen(), and understand how to safely combine both approaches with sockets and pipes.

fileno() fdopen() Output ordering bugs Sockets + stdio

Start Learning Now — It’s Free

All content on EmbeddedPathashala is free for students. No login needed.

Start with Part 1 →

embeddedpathashala.com