What is the wrong with this algorithm for ensuring the critical section isn't interupted?

Question

From here

The global variable turn is used to indicate the next process to enter the critical section. The initial value of turn can be 0 or 1.

int turn = 1;

T0:

while (true) {
  while (turn != 0) { ; } (1)
  critical section (2)
  turn = 1; (3)
  non-critical section (4)
}

T1:

while (true) {
  while (turn != 1) { ; } (1)
  critical section (2)
  turn = 0; (3)
  non-critical section (4)
}

I don't understand what the problem is. Why would T0 repeat while (turn != 1) forever? If the context switches to T1 then it will enter it's critical section then set turn=0.

EDIT: I see now why T0 will wait forever. Is there a name to the "rule" that's being violated? For example in the context of threads there's "mutual exclusion", "progress", "bounded waiting" and "No assumptions about relative speed of threads/processes" so is one of these not being met?

Answer 1

You are missing the second assumption in the problem description: a thread may terminate in the non-critical section. The description you copied here specifies "T1 terminates in non-critical section", so that T1 will never set turn=0 anymore.