Re: [PATCH v2 00/33] locking/atomic: convert all architectures to ARCH_ATOMIC

[Mark Rutland]

On Fri, Jun 04, 2021 at 10:56:16PM -0700, Randy Dunlap wrote:
> On 5/25/21 7:01 AM, Mark Rutland wrote:
> > This series (based on v5.13-rc2) converts all architectures to
> > ARCH_ATOMIC. This will allow the use of instrumented atomics on all
> > architectures (e.g. for KASAN and similar), and simplifies the core
> > atomic code (which should allow for easier rework of the fallbacks and
> > other bits in future).

[...]

> Hi Mark,
> Sorry for the late reply. 

Hi Randy,

Likewise, apologies in the delay in getting to this!

> I was just trying to update a patch
> to arch/sh/include/asm/cmpxchg.h, in its xchg() macro:
> 
> https://lore.kernel.org/lkml/20210602231443.4670-2-rdunlap@xxxxxxxxxxxxx/
> 
> The patch simply converts xchg() to a GCC statement expression to
> eliminate a build warning.
> 
> Arnd has done this for m68k and I have done it for sparc in the past.
> 
> Is there any (good) reason that all versions of arch_xchg() are not
> statement expressions?  In this patch series, they seem to be quite
> mixed (as they were before this patch series). I count 11 arches
> that use a statement expression and 4 that do not (including arch/sh/).

Largely I tried to make the minimal change from what was there before,
and I didn't have any specific reason to either use or avoid statement
expressions.

This series has been queued in the tip tree's locking/core branch for a
while now, but we could spin a patch atop. Do you want to spin a patch
to convert the remaining 4 architectures in one go?

Thanks,
Mark.