Re: [PATCH] 2.4.18-pre9, trylock for read/write semaphores

[Brian J. Watson]

David Howells wrote:
> I think the following would be more elegant:
> 
> [snip]

I agree.

> I'm also not sure that the cast has any effect in the following excerpt from
> the above:
> 
>         old = (volatile signed long)sem->count;
> 

You're right. I looked at the generated assembly, and the volatile cast
makes no difference.

> What you may actually want is:
> [snip]

Although you're right that a volatile pointer is the proper way to do
it, it turns out that a volatile declaration isn't necessary at all. The
cmpxchg() function is a memory barrier that forces the count to be
refetched the next time through the loop.

> Using this inline assembly has three advantages over mixing lots of C into it:
> [snip]

I'm not much of an assembly programmer, so implementing it this way
never crossed my mind. It looks much more efficient than the code
generated from the C version. A drawback is that it is not as easy to
port to other architectures, particularly those that already have a
cmpxchg() function.

It's up to you whether you prefer C or assembly. Let me know, and I'll
test that version and regenerate the patch.

Brian