Linux – development of low-level services and libraries (LNXPR1)

Course content:

Hide details

• Operating system interface
1. System calls (syscalls)
2. Standard library and POSIX.1-2024
3. Tools for tracing kernel communication

• Errors in system calls
1. Handling fatal and recoverable errors
2. Handling and propagating OS errors
3. System calls interrupted by signals

• Communication, input and output
1. Standard streams (stdin, stdout, stderr)
2. File descriptors
3. Buffered streams in C, C++ and Rust

• Filesystem and directory structure
1. Filesystem operations
2. Relative paths and current working directory
3. File permissions and attributes

• Working with files
1. Opening for reading or writing
2. Hybrid approaches to file access
3. Workarounds for lack of transactions

• Clocks and time measurement
1. Current time and time intervals
2. Timer signals, events and timerfd
3. Waiting and suspending processes (wait)

• Process and thread management
1. fork, exec and spawn functions
2. Corresponding system calls
3. Standard threads and pthreads

• Resource sharing
1. When using fork or clone
2. Closing or preserving resources across exec
3. Program return value

• Environment and the shell
1. Process groups and background processes
2. System services and the service manager
3. Combining processes started from the shell

• Operating system tasks
1. Difference between process and thread (LWP)
2. Choosing between multiprocessing and multithreading
3. Threads and processes in C, C++, Rust and Python

• Signals and their handling
1. Asynchronous signal handling
2. Synchronous handling via timerfd
3. Communication using real-time signals

• Blocking and non-blocking communication
1. When we don't want syscalls to sleep the process
2. File descriptor readiness
3. EWOULDBLOCK returned instead of blocking

• I/O multiplexing
1. Waiting for a set of possible events
2. Events come from file descriptors
3. External event-management libraries

• Basic interprocess communication
1. Pipes between processes (pipe)
2. Named pipes in the filesystem (named pipe)
3. Handling broken pipes and SIGPIPE

• Message queues
1. POSIX and System V variants
2. Multiple senders and receivers
3. In some ways more flexible than sockets

• Local (Unix) sockets
1. Forking processes and socketpair
2. Local client-server architecture
3. Socket types and their limitations

• Network sockets
1. Client-server architecture
2. The Internet and TCP/IP protocol
3. Workaround with SO_REUSEADDR

• Connection handling strategies
1. Forking processes or starting threads
2. Using a pool of processes or threads
3. Non-blocking sockets and an event queue

• Name services
1. Domain names and getaddrinfo
2. glibc and /etc/nsswitch.conf
3. Limitations, errors and alternatives

• Memory management and mapping
1. Dynamic memory allocation (heap, alloca)
2. Mapping files from the filesystem
3. Boundaries between kernel and user space

• Shared memory
1. Automatic for threads
2. Explicit POSIX or System V (shm)
3. The only communication that does not go through the kernel

• Passing file descriptors
1. The kernel can transfer file descriptors
2. Can use a pipe or a local socket, for example
3. Any file descriptor can then be sent

• Synchronization primitives
1. pthreads tools (mutex, condition variable)
2. POSIX and System V semaphores and eventfd
3. Atomic variables and lock-free operations

• Debugging options
1. ptrace system call
2. Interactive debuggers like GDB
3. Tracers like strace

• Shared dynamic libraries
1. Linking libraries at program start
2. Explicit loading via dlopen
3. Library dependencies

• Bonus: Optimized data paths
1. Optimized polling via epoll
2. Zero-copy techniques in the kernel
3. The io_uring mechanism