sem_trywait changes broke sparc

[David Miller]

From: Torvald Riegel <triegel@redhat.com>
Date: Mon, 02 Feb 2015 10:29:14 +0100

> Thanks Dave.  I scanned through the patches you committed and they
> looked good to me.
 
BTW, the original 32-bit sparc semaphore implementation did something
similar, by using an in-semaphore spinlock to implement the
synchronization.

They didn't use the spinlock array, just like my adaptation of your
code does not.