Re: compiler bug gcc4.6/4.7 with ACCESS_ONCE and workarounds

[Linus Torvalds]

On Thu, Nov 20, 2014 at 3:39 AM, Christian Borntraeger
<borntraeger@xxxxxxxxxx> wrote:
>
> So It looks like we could make a change to ACCESS_ONCE. Would something like
>
> CONFIG_ARCH_SCALAR_ACCESS_ONCE be a good start?

No, if it's just a handful of places to be fixed, let's not add config
options for broken cases.

> This would boil down to
> Patch1: Provide stricter ACCESS_ONCE if CONFIG_ARCH_SCALAR_ACCESS_ONCE is set + docu update + comments
> Patch2: Change mm/* to barriers
> Patch3: Change x86 locks
> Patch4: Change x86 gup
> Patch4: Enable CONFIG_ARCH_SCALAR_ACCESS_ONCE for s390x and x86

Just do patches 2-4 first, and then patch 1 unconditionally.

Obviously you'd need to spread the word on linux-arch to see how bad
it is for other cases, but if other architectures are at all like x86
and s390, and just require a few trivial patches, let's not make this
some config option.

                   Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/