Tags: #◇Computer Science #◇CSOPESY #◇Operating Systems #◇Quiz #◇Reviewer
CSOPESY Practice Quiz 1: OS Foundations (Weeks 1-5)
◇Overview
This quiz covers the conceptual foundations of CSOPESY from Weeks 1 through 5: OS introduction and roles, OS architectural structures, I/O subsystems, process management, and multithreading. All questions are multiple choice. Aim for a passing score of 70% (14/20) before the midterm.
Coverage:
- Module 1: OS Introduction and OS Structure (Notes 01, 02)
- Module 2: I/O and Display, Processes, Threads (Notes 03, 04, 05)
◇References & Navigation
ACTIVE RECALL VERIFICATION
Select an answer to verify conceptual comprehension and reveal the technical rationale.
A large factory automation system runs with no keyboard, mouse, or display. Its operating system must run silently, managing sensors and actuators without human interaction. From which OS perspective does this best exemplify?
A hardware device has finished an operation and needs to notify the CPU. It asserts an electrical signal on the system bus. The CPU immediately suspends its current task, saves its context, and jumps to a handler routine. What is the table of function pointers that maps signal numbers to these handler routines called?
A system has a cache access time of ns and a main memory access time of ns. Under the sequential access model (cache is checked first, memory accessed only on a miss), what is the Effective Access Time (EAT) when the cache hit ratio ?
Which of the following represents a Non-Maskable Interrupt (NMI)?
A developer wants to call a function that requires full hardware privileges — for example, modifying the CPU's memory management registers. Under Dual-Mode Operation, what must happen for this to execute safely?
Which OS architectural design runs all kernel services — including device drivers, file systems, and memory management — in a single large binary sharing one address space?
A device driver crashes on a microkernel-based operating system. What is the most likely outcome, and why?
A programmer writes printf("Hello") in C. This never directly invokes a system call. Instead, it calls a function in the C Standard Library, which eventually triggers a kernel trap. What is the name for the layer between the programmer's API call and the raw kernel system call?
A disk controller needs to transfer 512 KB of data into RAM. Using Interrupt-Driven I/O with one interrupt per byte and a context-switch cost of 50 CPU cycles per interrupt, how many total CPU cycles are consumed in interrupt overhead?
An application calls read() on a network socket. The call returns immediately with the error EWOULDBLOCK because no data has arrived yet. The application proceeds to render a UI frame and then polls the socket again. Which I/O model is this?
Which I/O subsystem service is responsible for managing an exclusive device (e.g., a printer) that cannot handle interleaved data streams, by queuing jobs on disk and feeding them sequentially?
A USB mouse is connected to the system. The OS classifies it as what type of device, and why?
In a standard process memory layout, which two regions grow towards each other and can cause an error if they collide?
A process has called exit(). Its resources have been partially cleaned up, but its parent process has not yet called wait() to collect the exit status. What state is this process in?
The OS needs to transfer CPU control from Process A (currently running) to Process B (in the Ready state). What sequence of operations occurs?
A web server process uses a strategy where each incoming request is handled by a separate worker — but instead of creating a new process with fork() each time (which is expensive), each worker shares the same code and data sections while only getting its own stack and registers. What construct is being described?
Which type of scheduler operates most frequently (potentially hundreds of times per second) and is responsible for selecting which process in the Ready Queue gets the CPU next?
An application is 40% sequential and 60% parallelizable. Using Amdahl's Law, what is the maximum theoretical speedup achievable with an infinite number of processors? (Use , )
In the Many-to-One threading model, all user-level threads map to a single kernel thread. What is the critical failure mode of this model?
A developer uses #pragma omp parallel for in a C program, and the compiler automatically splits a for-loop across all available CPU cores without the developer manually creating threads. What threading paradigm is this an example of?