Re: Litmus test for question from Al Viro

From: Alan Stern
Date: Mon Oct 05 2020 - 11:16:42 EST


On Mon, Oct 05, 2020 at 04:13:13PM +0100, Will Deacon wrote:
> > The failure to recognize the dependency in P0 should be considered a
> > combined limitation of the memory model and herd7. It's not a simple
> > mistake that can be fixed by a small rewrite of herd7; rather it's a
> > deliberate choice we made based on herd7's inherent design. We
> > explicitly said that control dependencies extend only to the code in the
> > branches of an "if" statement; anything beyond the end of the statement
> > is not considered to be dependent.
>
> Interesting. How does this interact with loops that are conditionally broken
> out of, e.g. a relaxed cmpxchg() loop or an smp_cond_load_relaxed() call
> prior to a WRITE_ONCE()?

Heh -- We finesse this issue by not supporting loops at all! :-)

Alan