Exam Guide COMPSCI 386

Transcription

1 FOUNDATIONS We discussed in broad terms the three primary responsibilities of an operating system. Describe each. What is a process? What is a thread? What parts of a process are shared by threads? What kinds of data and data structures are associated with individual threads? Identify a few special-purpose CPU registers. Discuss the purpose and use of the runtime stack? What information is stored in a stack frame? We discussed two general types of system programs. Identify them and give a few examples. What can you say about the boundary between system programs and the OS? Define kernel. Discuss privileged instructions and dual mode operation. How is dual mode operation supported in hardware? How does a user process cause privileged instructions to be performed on its behalf? What is the purpose of an interrupt? What is a trap? Give one or two examples of each. When an interrupt occurs, how is the appropriate service routine located by the OS? What is a signal in UNIX? What is meant in this context by the term upcall? Give some examples of signals caused by an executing process. Give two examples of signals created by certain key combinations entered from a terminal. Suppose a running process never terminates voluntarily. How does the OS regain control of the CPU? In what sense can a Java program make system calls? Define batch processing and multiprogramming. What problem in the dinosaur days of batch processing gave rise to multiprogramming? What is time-sharing? What is a microkernel? Discuss the relative merits of a microkernel vs. a monolithic kernel. UNIX CONCEPTS What was the design goal of UNIX? Describe in one sentence the relationship between C and UNIX. Discuss the ancestry of C. What features of C make it ideal as a language for systems development? Know how to use these shell commands: cat, cd, chmod, cp, ls, ps, pwd. Why must cd be implemented as a built-in command? Know how to combine simple commands with pipes at the command line. Know how to perform I/O redirection at the command line. Know how to use GREP for very simple types of filtering.

2 What is a hard link? Explain in terms of directory entries and inodes. What is the purpose of the GNU project? What does GNU stand for? Instead of releasing software with a copyleft license, why not simply put it in the public domain? What was the motivation for the creation of Linux? Why is it not precisely correct to refer to Linux as an operating system? What is a Linux distribution? Describe the two general strategies for implementing a shell. What are the advantages of the UNIX approach? Name two or three famous UNIX shells. In a UNIX shell, what do the following three symbolic names denote:... ~ Describe inodes and their use in the traditional UNIX-style file system. What info is contained in a directory? Where are filenames stored? What is a hard link? Explain in terms of directory entries and inodes. What is a soft link? Explain in terms of directory entries and inodes. Which kind of link becomes unusable when the target file is deleted? What can a soft link be used for that is not possible for a hard link? Which kind of link is more efficient in terms of time and memory? PROCESS MANAGEMENT The address space of a process can be partitioned the text section, data section, heap, and stack. Explain. What are the five states of a process? What information is stored inside a PCB? How is the waiting queue for a device implemented? Under what conditions would a process go from the running state to the ready state? Give brief descriptions of the fork, wait, and exec functions. What is an orphan process? What happens to an orphan in UNIX? What is a zombie process? How and why do zombies arise? What happens to them? Compare and contrast the two main mechanisms for IPC. What are file descriptors? Explain: every process begins with three open files. Be sure to understand the code we wrote in class for forking children. What can go wrong if a parent and child process each have a file descriptor that maps to the same entry in the global file table? Explain how using dup solves this problem. THREADS We discussed four benefits to multithreading. Explain. Describe the many-to-one threading model. Disadvantages?

3 Describe the one-to-one threading model. What is its primary disadvantage? Describe the many-to-many thread model. In what sense is it a compromise approach? Disadvantages? Be sure to understand the code we will study in class to create and join threads. CPU SCHEDULING What does it mean to say that the graph of CPU burst lengths for a process tends to be exponential? How does preemptive scheduling complicate kernel design? How do UNIX systems deal with this? What does the dispatcher do? What is dispatch latency? Define: CPU utilization, throughput, turnaround time, waiting time, response time. Describe FCFS scheduling. Identify one benefit and one disadvantage of this scheme. What situation leads to the convoy effect? How does SJF work? In what sense is it provably optimal? Why is it not used? Describe the recurrence relation used to generate an exponential average of CPU burst lengths. Describe a problem that arises from the use of priority scheduling. Typical solution? Describe round-robin scheduling. What kind of systems benefit especially from RR scheduling? By what scheduling criterion does RR usually perform worse than other scheduling algorithms? Describe multilevel queue scheduling. How does multilevel feedback queue scheduling differ? Describe how processor affinity and load balancing work at cross-purposes in CPU scheduling. Distinguish push migration from pull migration. MAIN MEMORY The OS keeps a list of holes that can be used for new processes. For contiguous memory allocation, a hole that is large enough for the entering process must be found. Describe 3 heuristics for this task. What is internal fragmentation? Why is it permitted to occur? What is the main problem with contiguous allocation? Describe compaction. Is it a practical solution to the problem of external fragmentation? What is paging? Explain in terms of pages, frames, and page tables. The DEC PDP-11 used a set of dedicated registers to implement page tables. It had a 16-bit address space and 8 KB page size. How many registers are needed in this case? Why is this approach to implementing page tables not feasible for 32-bit machines? Page tables can be stored in main memory using special registers to store the location and size of the page table for the current process. What problem does this introduce? How does a TLB solve this problem? Suppose the TLB has an 80% hit ratio, it takes 20 nanoseconds to search the TLB, and 100 nanoseconds to access main memory. What is the effective access time? Consider a system with 32-bit logical address space and 4KB page size. How big is the page table? We would not

4 want to store it contiguously in memory. Name three techniques used to deal with this problem. How does hierarchical paging work? Give a concrete example of a 2-level scheme for a 32-bit machine with 4 KB page size. Draw a picture to illustrate the idea. Why is hierarchical paging not appropriate for 64-bit machines? How do hashed page tables work? What is an inverted page table? What is the format of a logical address? What is the main advantage and disadvantage of this approach to structuring page tables? What is the difference between paging and segmentation? VIRTUAL MEMORY Describe the steps involved in handling a page fault. What is copy-on-write? Explain how a dirty bit can be used to reduce the effective access time. Given a reference string and the number of frames allocated to a process, compute the number of page faults using FIFO, OPT, and LRU. What is Belady's Anomaly? Illustrate how it might manifest itself when using FIFO. What is the defining property of a stack page replacement algorithm? Describe the counter implementation of LRU. Why is hardware support needed? Describe the stack implementation of LRU. How can a reference bit be used to roughly approximate LRU? Describe the second-chance algorithm and the enhanced second-chance algorithm. What problem arises from LFU page replacement? Describe an idea for dealing with this problem. Give an example in which LRU generates fewer page faults than Least Frequently Used for the same reference string. Give another example to show the opposite result. Give an example to show that LRU generates fewer page faults than Most Frequently Used for the same reference string. Give another example to show the opposite result. Give a justification for Most Frequently Used page replacement. What is thrashing? Describe a strategy for dealing with it based on locality of reference. What happens if the working set parameter Δ is chosen too small? Too large? Some systems approximate the working set by keeping track of the pages referenced within a certain number of time units. What is the potential problem with this scheme? Why do some operating systems impose a lower limit on the page fault rate of processes? Discuss local replacement and global replacement. What are their relative merits? Thrashing processes can impact non-thrashing processes even with local replacement. Explain. Give two reasons why kernel memory is often allocated from a free-memory pool different from the pool used to satisfy user-mode processes? Compare the buddy system allocator with slab allocation.

5 You can implement a solution to a programming problem using a host of different languages. How might the choice of language affect the page fault rate? DISK SCHEDULING Define: track, sector, cylinder. Define: seek time and rotational latency. Which is the bottleneck for transfer rate? Why? Understand how FCFS, SSTF, SCAN, and C-SCAN work. FCFS is often used for solid state disks. Why? The seek time to service a request is not proportional to the distance traveled by the disk arm. Explain. SYNCHRONIZATION What is a race condition? Describe a simple way to prevent race conditions in kernel code. Why is it impractical? What is meant by the term critical section? What are the three necessary properties of a solution to the critical section problem? Understand Peterson s solution to the critical section problem. What is a semaphore? What are semaphores used for? A mutex lock is a kind of semaphore. Explain. Linux provides an opaque type atomic_t. Explain. What is a synchronized method in Java? What syntax does Java provide for synchronizing blocks? Static methods can be synchronized. Explain. How are the wait and notify methods used? Where do they come from? Show how to create a lock and obtain a condition object. Show how the current thread can wait on a condition object. Show how the current thread can wake up threads that are waiting on a condition object. Understand how to use lock objects where previously the synchronized keyword would have been used.