Re: [PATCH locking/atomic 18/19] locking/atomic: Refrain from generating duplicate fallback kernel-doc

[Paul E. McKenney]

On Fri, May 12, 2023 at 06:03:26PM +0100, Mark Rutland wrote:
> On Fri, May 12, 2023 at 09:01:27AM -0700, Paul E. McKenney wrote:
> > On Fri, May 12, 2023 at 02:18:48PM +0100, Mark Rutland wrote:
> > > On Thu, May 11, 2023 at 12:12:16PM -0700, Paul E. McKenney wrote:
> > > > On Thu, May 11, 2023 at 06:10:00PM +0100, Mark Rutland wrote:
> > > > > I think that we can restructure the ifdeffery so that each ordering variant
> > > > > gets its own ifdeffery, and then we could place the kerneldoc immediately above
> > > > > that, e.g.
> > > > > 
> > > > > 	/**
> > > > > 	 * arch_atomic_inc_return_release()
> > > > > 	 *
> > > > > 	 * [ full kerneldoc block here ]
> > > > > 	 */
> > > > > 	#if defined(arch_atomic_inc_return_release)
> > > > > 	/* defined in arch code */
> > > > > 	#elif defined(arch_atomic_inc_return_relaxed)
> > > > > 	[ define in terms of arch_atomic_inc_return_relaxed ]
> > > > > 	#elif defined(arch_atomic_inc_return)
> > > > > 	[ define in terms of arch_atomic_inc_return ]
> > > > > 	#else
> > > > > 	[ define in terms of arch_atomic_fetch_inc_release ]
> > > > > 	#endif
> > > > > 
> > > > > ... with similar for the mandatory ops that each arch must provide, e.g.
> > > > > 
> > > > > 	/**
> > > > > 	 * arch_atomic_or()
> > > > > 	 *
> > > > > 	 * [ full kerneldoc block here ]
> > > > > 	 */
> > > > > 	/* arch_atomic_or() is mandatory -- architectures must define it! */
> > > > > 
> > > > > I had a go at that restructuring today, and while local build testing indicates
> > > > > I haven't got it quite right, I think it's possible:
> > > > > 
> > > > >   https://git.kernel.org/pub/scm/linux/kernel/git/mark/linux.git/log/?h=atomics/fallback-rework
> > > > > 
> > > > > Does that sound ok to you?
> > > > 
> > > > At first glance, it appears that your "TODO" locations have the same
> > > > information that I was using, so it should not be hard for me to adapt the
> > > > current kernel-doc generation to your new scheme.  (Famous last words!)
> > > 
> > > Great!
> > > 
> > > > Plus having the kernel-doc generation all in one place does have some
> > > > serious attractions.
> > > 
> > > :)
> > > 
> > > > I will continue maintaining my current stack, but would of course be
> > > > happy to port it on top of your refactoring.  If it turns out that
> > > > the refactoring will take a long time, we can discuss what to do in
> > > > the meantime.  But here is hoping that the refactoring goes smoothly!
> > > > That would be easier all around.  ;-)
> > > 
> > > FWIW, I think that's working now; every cross-build I've tried works.
> > > 
> > > I've updated the branch at:
> > > 
> > >   https://git.kernel.org/pub/scm/linux/kernel/git/mark/linux.git/log/?h=atomics/fallback-rework
> > > 
> > > Tagged as:
> > > 
> > >   atomics-fallback-rework-20230512
> > 
> > Thank you very much!
> > 
> > I expect to send v2 of my original late today on the perhaps unlikely
> > off-chance that someone might be interested in reviewing the verbiage.
> 
> I'll be more than happy to, though I suspect "late today" is far too late today
> for me in UK time terms, so I probably won't look until Monday.

Works for me!

> > More to the point, I have started porting my changes on top of your
> > stack.  My thought is to have a separate "."-included script that does
> > the kernel-doc work.
> 
> I was thinking that we'd have a gen_kerneldoc(...) shell function (probably in
> atomic-tbl.sh), but that's an easy thing to refactor after v2, so either way is
> fine for now!

Good point, will make that happen.  Easy to move the code, so might
as well be v1.  ;-)

> > I am also thinking in terms of putting the kernel-doc generation into
> > an "else" clause to the "is mandatory" check, and leaving the kernel-doc
> > for the mandatory functions in arch/x86/include/asm/atomic.h.
> 
> My thinking was that all the kernel-doc bits should live in the common header
> so that they're all easy to find when looking at the source code, and since if
> feels a bit weird to have to look into arch/x86/ to figure out the semantics of
> a function on !x86.
> 
> That said, if that's painful for some reason, please go with the easiest option
> for now and we can figure out how to attack it for v3. :)

I will give it a shot.

							Thanx, Paul