Re: [PATCH] x86/syscalls: Mark expected switch fall-throughs

From: "Gustavo A. R. Silva" <garsilva@embeddedor.com>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>, "H. Peter Anvin" <hpa@zytor.com>,
	x86@kernel.org, LKML <linux-kernel@vger.kernel.org>,
	Kees Cook <keescook@chromium.org>,
	Linus Torvalds <torvalds@linux-foundation.org>
Subject: Re: [PATCH] x86/syscalls: Mark expected switch fall-throughs
Date: Tue, 28 Nov 2017 12:53:31 -0600	[thread overview]
Message-ID: <20171128125331.Horde.XUXRWEiZ7FMJv6yoDylZxN8@gator4166.hostgator.com> (raw)
In-Reply-To: <alpine.DEB.2.20.1711281944010.2222@nanos>

Quoting Thomas Gleixner <tglx@linutronix.de>:

> On Tue, 28 Nov 2017, Thomas Gleixner wrote:
>
> +CC Linus.
>
>> On Tue, 28 Nov 2017, Thomas Gleixner wrote:
>>
>> > On Tue, 28 Nov 2017, Gustavo A. R. Silva wrote:
>> > > Quoting Thomas Gleixner <tglx@linutronix.de>:
>> > > > > To be honest, such comments annoy me during a code review  
>> especially when
>> > > > > the fallthrough is so obvious as in this case. There might  
>> be cases where
>> > > > > its worth to document because it's non obvious, but documenting the
>> > > > > obvious
>> > > > > just for the sake of documenting it is just wrong.
>> > > >
>> > >
>> > > I understand that and I agree that in this particular case it  
>> is just obvious.
>> > > The thing is that if we want to benefit from having the  
>> compiler help us to
>> > > spot these kind of issues before committing our code, we have  
>> to address every
>> > > place in the whole code-base.
>> > >
>> > > > And _IF_ at all then you want a fixed macro for this and not a comment
>> > > > which will be formatted as people see it fit.
>> > > >
>> > > > GCC supports: __attribute__ ((fallthrough)) which we can wrap  
>> into a macro,
>> > > > e.g. falltrough()
>> > > >
>> > > > That'd be useful, but adding all these comments and then  
>> having to chase a
>> > > > gazillion of warning instances to figure out whether there is  
>> a comment or
>> > > > not is just backwards.
>> > > >
>> > >
>> > > I have run into this before and people find what you suggest  
>> even uglier.
>> >
>> > It's not about ugly. It's about _USEFULL_.
>> >
>> > The comments are ugly AND completely useless for the compiler and they are
>> > going to be malformatted so checker tools can't differentiate the false
>> > positives.
>> >
>> > The macro, in which more or less ugly form written, is both documentation
>> > and helps the compiler NOT to emit the same crap over and over.
>>
>> Just checked and GCC really supports analyzing the comment to some extent.
>>
>> But just look at
>>
>>     https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77817
>>
>>  " It is not really possible.  __attribute__((fallthrough)) has precise
>>    rules on where it can appear, while /* FALLTHRU */ comments, being
>>    comments, can appear anywhere.  Especially with -Wimplicit-fallthrough=1
>>    when all comments are considered fallthru comments... "
>>

This is what we want to add:

# Warn about missing switch break or fall-through comment.
KBUILD_CFLAGS  += $(call cc-option,-Wimplicit-fallthrough)

>> I have no idea who came up with that brilliant idea of parsing comments in
>> the code. It's so simple to make this parser completely fail that it's not
>> even funny anymore.
>>

I don't get why someone would want to do that to himself. :/

>> I don't care what other people prefer. The code base I'm responsible for
>> gets either proper annotations or nothing.
>
> And in fact we want ONE solution for the whole kernel. And comments are
> obviously the wrong one.
>

OK. I'll discuss this and see how we can come up with the best solution.

Thank you for your feedback
--
Gustavo A. R. Silva