A hand-rolled spinlock built on std::atomic_flag

SpinLock is hand-implemented using std::atomic_flag's test_and_set()/clear() interface. Four threads each run the loop shown, incrementing a shared counter 100,000 times under the spinlock. What is the most accurate statement about this code and about std::atomic_flag in general?

main.cpp
class SpinLock {
    std::atomic_flag flag = ATOMIC_FLAG_INIT;
public:
    void lock() {
        while (flag.test_and_set(std::memory_order_acquire)) {
            // busy-wait
        }
    }
    void unlock() {
        flag.clear(std::memory_order_release);
    }
};

SpinLock sl;
long counter = 0;

void worker() {
    for (int i = 0; i < 100000; ++i) {
        sl.lock();
        ++counter;
        sl.unlock();
    }
}
// 4 threads each run worker(), then join; main prints counter

36 of 51 in Concurrency