Learning Resources

This module is populated from the local chunked books in library/raw/semester-05-os-networking/books. Use this page as a source map, not as an instruction to read everything.

Source Stack

Book	Role	How to use it in this module
Operating Systems: Three Easy Pieces (OSTEP), persistence chapters	Primary teaching source	Default escalation for every FS concept; the clearest operational explanations of inodes, layout, journaling, LFS, and disks
Operating System Concepts (Silberschatz)	Selective support	Use for the I/O subsystem (drivers, DMA), VFS framing, and consistency-checking perspective
Unix Network Programming (Stevens)	Selective support	Use for select, poll, and advanced polling; also the canonical I/O models framing
kernel.org documentation	Current Linux truth	Use for ext4, block layer, io_uring, and anything the textbooks predate
LWN.net	Evolution and nuance	Use for fsync failure modes, COW vs fsync interactions, and io_uring history

Resource Map by Cluster

Cluster 1: The File Abstraction

Need	Best local chunk	Why
what a file and directory are	OSTEP: Files and directories	Operational introduction to the abstraction
reading and writing files	OSTEP: Reading and writing files	How syscalls touch the inode
non-sequential I/O	OSTEP: Reading and writing but not sequentially	lseek and offsets
directory structure	OSTEP: Directory organization	Directory as a special kind of file
making directories	OSTEP: Making directories	mkdir trace
hard links	OSTEP: Hard links	Link count semantics
symbolic links	OSTEP: Symbolic links	Path-based indirection
the inode data structure	OSTEP: Aside - the inode data structure	Layout and pointer scheme
access paths	OSTEP: Access paths - reading and writing	End-to-end trace through kernel tables
VFS framing	OS Concepts: Virtual file systems	How Linux unifies many FS types

Cluster 2: On-Disk Layout and Structure

Need	Best local chunk	Why
disk interface	OSTEP: The interface	The block device contract
disk model	OSTEP: A simple disk drive	Platters, sectors, LBAs
I/O timing math	OSTEP: I/O time - doing the math	Quantitative model
disk scheduling	OSTEP: Disk scheduling	SSTF, C-SCAN, and the elevator
FS design philosophy	OSTEP: The way to think	Framing for layout decisions
overall layout	OSTEP: Overall organization	Canonical block arrangement
FFS - why block groups	OSTEP: The problem - poor performance	Motivation for locality
cylinder groups	OSTEP: Organizing structure - the cylinder group	Grouping related data
allocation policy	OSTEP: Policies - how to allocate files and directories	Where new data goes
locality measurement	OSTEP: Measuring file locality	Real measurements
large files	OSTEP: The large file exception	Why large files break naive locality
LFS design	OSTEP: Writing to disk sequentially	The sequential-only idea
LFS buffering	OSTEP: Writing sequentially and effectively	Segment buffering
LFS indirection	OSTEP: Solution through indirection - the inode map	Finding latest inodes
LFS garbage collection	OSTEP: A new problem - garbage collection	Cost of log-structured
ext4 format reference	ext4 documentation (kernel.org)	Authoritative on-disk layout

Cluster 3: Crash Consistency

Need	Best local chunk	Why
worked crash example	OSTEP: A detailed example	The canonical four-case analysis
fsck	OSTEP: Solution 1 - the file system checker	Pre-journal recovery
data journaling	OSTEP: Data journaling	Commit ordering and transaction model
forcing writes	OSTEP: Aside - forcing writes to disk	Barriers and FLUSH
log optimization	OSTEP: Aside - optimizing log writes	Reducing journal cost
log optimization (part 2)	OSTEP: Optimizing log writes - part 2	Continued
other approaches	OSTEP: Solution 3 - other approaches	Soft updates, optimistic crash consistency
COW framing	OSTEP: Copy-on-write mappings	COW pattern in general
ZFS	OSTEP: ZFS	COW FS in production
journaling FS appendix	OSTEP: Journaling file system	Compact summary
consistency checking	OS Concepts: Consistency checking	FSCK-style perspective
disk failure modes	OSTEP: Disk failure modes	Beyond power loss
checksums	OSTEP: Detecting corruption - the checksum	Silent corruption
misdirected writes	OSTEP: Misdirected writes	Why just a checksum isn't enough

Cluster 4: Caching and Performance

Need	Best local chunk	Why
caching and buffering	OSTEP: Caching and buffering	Page and buffer cache
cache management	OSTEP: Cache management (paging)	Replacement policies that carry over
forcing writes	OSTEP: Forcing writes to disk	fsync barrier discussion
sequentiality	OSTEP: I/O time - doing the math	Quantitative gap
measuring locality	OSTEP: Measuring file locality	Real workload data
Linux performance	Brendan Gregg: Linux performance tools	Current practical reference
fsync failures	LWN: PostgreSQL's fsync() surprise	Real-world gotcha
fsync error handling	LWN: Improved fsync error handling	Evolution of the semantics

Cluster 5: I/O Models and the Syscall Path

Need	Best local chunk	Why
I/O models framing	UNP: I/O models	Canonical five-model taxonomy
select	UNP: select function (part 1)	Classical multiplexing
select continued	UNP: select (part 2)	Usage details
select worked	UNP: select (part 3)	Echo server example
pselect	UNP: pselect function	Signal-safe variant
advanced polling	UNP: Advanced polling (part 1)	poll and extensions
advanced polling 2	UNP: Advanced polling (part 2)	Continued
device architecture	OSTEP: System architecture	How devices attach
device protocol	OSTEP: The canonical protocol	Command/status/interrupt
DMA	OSTEP: More efficient data movement with DMA	Why CPU is not the bottleneck
drivers	OSTEP: Fitting into the OS - the device driver	Where drivers sit
IDE case study	OSTEP: Case study - a simple IDE disk driver	Worked simple driver
epoll man page	epoll(7) man page	Authoritative semantics
io_uring primer	Lord of the io_uring (unixism.net)	Best hands-on intro
io_uring paper	Jens Axboe: Efficient IO with io_uring (PDF)	Canonical design document
io_uring evolution	LWN: rapid growth of io_uring	Historical arc

External Resources (Read-If-Curious)

Only use these if you want a second exposition or a supplementary problem source. The module is completable from the local chunks alone.

• kernel.org: Filesystems documentation - ext4, xfs, btrfs, f2fs specs.
• OpenZFS documentation - practical and design-level ZFS reference.
• Btrfs documentation - Btrfs on-disk format and operations.
• Brendan Gregg: Linux performance tools - diagnostic tool catalog for I/O and storage.
• Brendan Gregg: Disk / storage latency - latency anatomy and tracing.
• Dan Kegel: The C10K problem - historical context that motivated epoll.
• LWN: The page cache - deep dive into the Linux page cache.
• fio documentation - the benchmarking tool used in Practice 3.
• man7.org syscall index - definitive syscall reference.

Use Rules

• If you are stuck on any FS layout or crash scenario, go to OSTEP first. It is the best operational reference for persistence.
• If you need the I/O models taxonomy or socket-style readiness patterns, go to UNP.
• If you need current Linux specifics (ext4 format, io_uring, fsync corner cases), go to kernel.org and LWN.
• Open one chunk for one concept gap; do not wander through a whole appendix sequence.
• If rereading does not fix the problem, stop and draw the on-disk structures in your own hand before reading more.