Re: [patch 0/9] mutex subsystem, -V4

[Thomas Gleixner]

On Thu, 2005-12-22 at 05:44 -0800, Andrew Morton wrote:

> There's no need for two sets of behaviour.  What I'm saying is that we
> could add similar debugging features to semaphores (if useful).  Yes, we
> would have to tell the kernel at the source level to defeat that debugging
> if a particular lock isn't being used as a mutex.  That's rather less
> intrusive than adding a whole new type of lock.  But I'd question the value
> even of doing that, given the general historical non-bugginess of existing
> semaphore users.

Semaphores need a counter, mutexes only a binary representation of the
locked/unlocked state, which makes it possible to use instructions like
xchg, swap, test_and_set_bit depending on what atomic operations are
available on the architecture. On many architectures this is more
efficient than the counter based implementation.

Also the wakeup rules allow smaller and faster implementations for
mutexes.

	tglx


-
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/