Cigarette Smokers Problem

By Michael Thompson, for Mirkwood only

CMAC 4000 - Operating Systems

Overview

The Cigarette Smokers Problem was introduced by Suhas Patil (1971) as a synchronization puzzle illustrating the limits of semaphores and the difficulty of coordinating multiple consumers with overlapping resource needs.

Three smokers sit around a table. Each smoker has an infinite supply of exactly one ingredient:

A fourth process, the agent, has an infinite supply of all three. To smoke a cigarette, a smoker needs all three ingredients. The agent repeatedly places two of the three items on the table. The smoker who has the third item should take the two from the table, make a cigarette, smoke, and then signal the agent to continue.

Why the Problem Is Interesting

Unlike the Dining Philosophers problem, the Cigarette Smokers problem is provably impossible to solve using only classical Dijkstra semaphores under Patil's original constraints. These constraints forbid the agent from directly waking a specific smoker or revealing which ingredients are available.

Modern textbook versions relax these constraints, allowing the agent to signal the correct smoker. Under these relaxed rules, a clean semaphore-based solution is possible.

Standard Semaphore-Based Solution (Relaxed Rules)

The practical solution:

This avoids the race where two smokers each take one ingredient and both block forever waiting for the other.

Agent Pseudocode


semaphore table = 1
semaphore smoker_tobacco = 0
semaphore smoker_paper   = 0
semaphore smoker_match   = 0
semaphore agent_done     = 0

agent():
  while true:
    wait(table)
    choose random pair
    if (paper, match):   signal(smoker_tobacco)
    if (tobacco, match): signal(smoker_paper)
    if (tobacco, paper): signal(smoker_match)
    wait(agent_done)
    signal(table)

Smoker Pseudocode (example: smoker with tobacco)


smoker_with_tobacco():
  while true:
    wait(smoker_tobacco)
    // take paper and match
    make_and_smoke()
    signal(agent_done)

Why a Naive Semaphore Solution Fails

If smokers wait on semaphores for individual ingredients (e.g., one semaphore for paper, one for matches), two smokers may each take one ingredient and then deadlock waiting for the other. Patil showed that under his strict rules, this cannot be solved with semaphores alone.

The key issue is that semaphores cannot express the conditional logic "wake exactly the smoker who has the missing ingredient" without violating Patil's constraints.

Relation to Other Classic Problems

This makes the Cigarette Smokers problem a stronger demonstration of the limits of semaphores than Dining Philosophers.

References