CODEWITHSUNDEEP
multithreadingdeadlock
Anti Programmer
Anti Programmer

Reputation: 23

Resource pool deadlock

This pseudocode run concurrently in multiple threads causes deadlock:

Resource res1 = pool.get();
...
Resource res2 = pool.get();
...
pool.release(res2);
pool.release(res1);

Is that an expected behavior? (I guess it is)

Deadlock doesn't seem to happen if the size of pool is twice of the number of threads - is this statement correct?

Is there a "scientific" name for this kind of deadlock in computer science? Would appreciate some links.

Thanks

Upvotes: 2

Views: 121

Answers (1)

nneonneo
nneonneo

Reputation: 179687

Suppose you have two threads and two resources.

Thread A executes res1 = pool.get() and acquires one resource. Thread B executes res1 = pool.get() and acquires one resource. Now the pool is exhausted, so neither thread can finish res2 = pool.get().

The problem goes away if you have one spare resource (i.e. at least n+1 resources for n threads), since in that case one thread is guaranteed to get two resources, finish, and thus release enough resources for the other threads to use. Increasing the number of resources improves parallelism; with twice as many resources as threads, every thread can get resources without blocking and thus all threads can execute in parallel.

AFAIK this is simply known as "deadlock" since it's essentially a textbook example of the phenomena. Both tasks are now waiting on each other to release the necessary resource.

Upvotes: 1

Related Questions