Re: Aw: Re: [PATCH] mlx4: give precise rx/tx bytes/packets counters

[Eric Dumazet <eric.dumazet@gmail.com>]

On Wed, 2016-11-30 at 15:28 +0000, David Laight wrote:

> Are you sure??

Yes I am

> Last I looked gcc seemed to convert 'foo++' to 'foo = foo + 1' before
> generating any code.

Your gcc might need a refresh then.

> It might then optimise that back to a memory increment, but that would
> also happen if you'd coded the latter form.
> 
> > Which is kind of unfortunate, given it is the fast path.
> > 
> > Better add a comment, like :
> > 
> > /* We should use WRITE_ONCE() to pair with the READ_ONCE() found in xxxx()
> >  * But gcc would generate non optimal code.
> >  */
> 
> Actually while READ_ONCE() is generally useful - to get a snapshot of a changing value.
> 
> WRITE_ONCE() isn't a pairing - the compiler is highly unlikely to write a
> location twice.

WOW. I can not believe what you just said.

We had numerous bugs because compiler was writing on the location
temporary computations. Just take a look at git history to find some
gems.

> You might want an annotation to ensure is doesn't assume it can read the value
> back (write through volatile pointer). But that has nothing to do with how readers behave.

Completely wrong.