In kernel space, anything that helps to achieve (even limited) wait-freeness is rather welcome. You often don't have the luxury of long lasting locks. Short las

In kernel space, anything that helps to achieve (even limited) wait-freeness is rather welcome. You often don't have the luxury of long lasting locks. Short lasting (at most microseconds) spinlocks is usually the most important concurrency primitive you have. So nice to read about techniques that can be implemented in kernel space. Inability to do a context switches often works against you there [1]. [1]: Many usermode waiting based constructs cannot be used, because they require yielding or pre-emption. You can't yield if you can't schedule (or block).