Warning

This portion of the book is under construction, not ready to be read

34. Introduction to Concurrency, Synchronization and Deadlock

Earlier, we saw how computers can virtualize their CPUs through the use of threads. Virtualizing the CPU makes it seem like there are multiple CPUs running at the same time to the user. However, when a process requires threads to be ran in a certain order, we run into problems such as race conditions, deadlocks, and security. One of the key responsibilities of an operating system is that of synchronization—handling concurrent events in a reasonable way, and providing mechanisms for user applications to do so as well.

1. interprocess communication and need to share

   • system calls to enable: shared memory region, shmat, (reference back pipes)

   • short statement we have parallel hardware, lots of cores…

   • leads to threads from the perspective of we want work on same data…

2. locking

3. deadlock

4. challenge modern hardware

   • NUMA, caches…

   • sequential consistency, processor consistency… C++11 memory model (possible part 4)

   • scalability, lock granularity

5. locking in linux