On Tue, Sep 10, 2019 at 08:46:07AM +0100, Will Deacon wrote:
> On Mon, Sep 09, 2019 at 10:21:35PM +0200, Arnd Bergmann wrote:
> > On arm64 build with clang, sometimes the __cmpxchg_mb is not inlined
> > when CONFIG_OPTIMIZE_INLINING is set.
>
> Hmm. Given that CONFIG_OPTIMIZE_INLINING has also been shown to break
> assignment of local 'register' variables on GCC, perhaps we should just
> disable that option for arm64 (at least) since we don't have any toolchains
> that seem to like it very much! I'd certainly prefer that over playing
> whack-a-mole with __always_inline.
I assume we're referring to stuff such as the following?
https://www.spinics.net/lists/arm-kernel/msg730329.html
Are these breakages limited to the out-of-line hacks made for LL/SC
atomics, or were there other breakages elsewhere?
Now that the out-of-line hacks have gone, I wonder if this is actually
still a problem anymore. In any case isn't the right thing to do there
to add the __always_inline to functions that use the register keyword
in a function currently annotated inline?
I'm happy to look into this if there is likely to be some benefit in
turning on CONFIG_OPTIMIZE_INLINING.
Thanks,
Andrew Murray
>
> Will