How do you analyze deadlocks in a multithreaded application and how do you avoid them?

To analyze deadlocks in a multithreaded application, the following steps are usually performed:

1. Collect thread dumps — in Java, this can be done using jstack or monitoring tools.
2. Analyze thread states — look for threads in BLOCKED or WAITING states, and see which objects they are blocking or waiting for.
3. Identify the lock chain — determine which threads hold which locks and what they are waiting for, to find cyclic dependencies.

To avoid deadlocks:

• Always acquire multiple locks in the same order across all threads.
• Use higher-level synchronization constructs (e.g., java.util.concurrent) that reduce the risk of errors.
• Minimize the time locks are held.
• Avoid nested locking whenever possible.

Example of analysis using jstack:

jstack <pid> > threaddump.txt

In the threaddump.txt file, look for lines with BLOCKED and waiting to lock to understand which objects are causing deadlocks.