Re: spinaphore conceptual draft (was discussion of RT patch)

[Steven Rostedt]

One thing you are forgetting is that we are not just talking about the
latencies of contention.  We are talking about the latency of a high
priority process when it wakes up to the time it runs.  Most of the time
a spin lock stops preemption, either with (CONFIG_PREEMPT)
preempt_disable or simple turning off interrupts.  With Ingo's mutexes,
the places with spin_locks are now preemptable.  So there is probably
lots of times that it would be better to just spin on contention, but
that's not what Ingo's spin_locks are saving us.  It's to keep most of
the kernel preemptable.

The priority inheritance of spin_locks is simply there to protect from
priority inversion.

-- Steve


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/