1. Semester Review Brian Kocoloski
CSE 422S - Operating Systems Organization
Washington University in St. Louis
St. Louis, MO 63130

2. CSE 422S –Operating Systems Organization
Final Exam
8-10 questions; types of questions same as midterm
True/False
Multiple Choice
Open answer / code snippets
Not cumulative -- covers only material since the midterm
How you should study (in order from most to least important)
Go back over the LKD text, consulting the slides as you go to reinforce the material
Study the slides on paging and program execution/linking/layout, which are not covered in the LKD text
Review the studios/labs to further reinforce concepts
8-10 am on Thursday, Dec 13th, Urbauer 218
Exam length similar to midterm, should be possible to complete the exam in 80 minutes though you have the full 120 should you need it
CSE 422S –Operating Systems Organization

3. Core Concepts (Synchronization)
LKD Ch 9
Critical Regions and Race Conditions (pp. 162)
Causes of Concurrency (pp. 167)
Note the difference between “true” concurrency - i.e., parallelism – and concurrency
Deadlock (pp. 169)
LKD Ch 10
Atomic hardware operations
How to build a spinlock with atomic instructions
When to use spinlocks vs. semaphores/mutexes (pp. 197)
CSE 422S –Operating Systems Organization

4. Core Concepts (Virtual Memory)
Slides from Mon, Oct. 29th
What does the kernel need to do to handle memory allocation (mmap or brk) system calls?
What is the difference between prepaging and demand paging
Why does Linux use the latter?
Understand how VA->PA translations work via the page tables
CSE 422S –Operating Systems Organization

5. Core Concepts (Kernel Memory)
LKD Ch 12
Difference between physically contiguous and virtually contiguous memory allocators
kmalloc() vs vmalloc()
When should vmalloc be chosen over kmalloc?
What is the purpose of the slab allocator? (pp. 245)
CSE 422S –Operating Systems Organization

6. Core Concepts (Address Spaces+ Shared Memory)
LKD Ch 15
Understand information presented in /proc/<pid>/maps (pp. 314)
Short discussion on paging and page tables (pp. 320)
Review Studio 13
How to set up shared memory regions via shm functions and mmap()
CSE 422S –Operating Systems Organization

7. Core Concepts (Program Execution, Linking, and Layout)
Slides from Wed, Nov 7th
Performance differences between static and dynamic linking
Usage of PLT and GOT for performing dynamic linking/loading
CSE 422S –Operating Systems Organization

8. CSE 422S –Operating Systems Organization
Core Concepts (I/O)
LKD Ch 13
Understand at a high level the information managed in the 4 VFS objects:
Superblock, inode, dentry, and file (pp. 265)
LKD Ch 14
I/O schedulers (pp. 297)
Prioritization of reads over writes (why?)
LKD Ch 16
High level idea behind page cache
Temporal and spatial locality
Understand the two-list LRU approach (pp. 325)
CSE 422S –Operating Systems Organization

9. Core Concepts (Post-Midterm)
Kernel synchronization
Virtual Memory
Atomic Variables
- Process address spaces
Mutexes
- Page tables
Spinlocks
- Copy-on-write via page tables
Race Conditions
- What page tables of different threads of the same process look like
Kernel memory
Shared Memory
Page level management
Only use kernel to map initially and unmap after done
Kernel virtual to physical address translation
Use same physical memory
Map to (different) virtual addresses in the processes
CSE 422S –Operating Systems Organization

10. Core Concepts (Post-Midterm)
VFS layer
Superblocks, filesystems, paths, namespaces, inodes, directory entries
Page cache and writeback
Directory entry hierarchy
Mapping files into memory
Block I/O layer
Read/write caching
Eviction strategies
Character vs. block devices
Flusher threads, writebacks
I/O scheduling algorithms
CSE 422S –Operating Systems Organization

11. CSE 422S –Operating Systems Organization
Example Question #1
Why are disk I/O schedulers sometimes called “elevators?” (4 points) Why do most I/O schedulers merge requests that target adjacent data blocks? (3 points) Why does Linux cache write requests in memory, rather than always writing them directly back to disk? (3 points)
CSE 422S –Operating Systems Organization

12. CSE 422S –Operating Systems Organization
Example Question #2
User-space programs may be linked either statically or dynamically
Describe the differences between static and dynamic linking. In your answer, consider the following metrics: size of the resulting binaries, speed of linking, speed of loading into memory
Was this process statically or dynamically linked, and how can you tell?
CSE 422S –Operating Systems Organization

13. CSE 422S –Operating Systems Organization
Example Question #3
Please circle the letter next to each of the following descriptions that illustrates the idea of “lock data, not code” (4 points).
Put locks around a sequence of reads and writes to same memory location
Make critical sections as long as possible
Lock a data structure when accessing or removing elements
Use a mutex around a function implementing a complex sequence of operations on multiple pieces of shared data
CSE 422S –Operating Systems Organization