Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
Date:   Wed, 2 Mar 2022 18:43:57 +0000
From:   Matthew Wilcox <willy@infradead.org>
To:     Kees Cook <keescook@chromium.org>
Cc:     Ingo Molnar <mingo@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Juri Lelli <juri.lelli@redhat.com>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Ben Segall <bsegall@google.com>, Mel Gorman <mgorman@suse.de>,
        Daniel Bristot de Oliveira <bristot@redhat.com>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH 00/19] Enable -Wshadow=local for kernel/sched
Message-ID: <Yh+67VF0/OFSp15y@casper.infradead.org>
References: <20220302043451.2441320-1-willy@infradead.org>
 <202203021030.EEEF58C2@keescook>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <202203021030.EEEF58C2@keescook>
Precedence: bulk

On Wed, Mar 02, 2022 at 10:32:23AM -0800, Kees Cook wrote:
> On Wed, Mar 02, 2022 at 04:34:32AM +0000, Matthew Wilcox (Oracle) wrote:
> > I thought I'd choose one of the more core parts of the kernel to
> > demonstrate the value of -Wshadow.  It found two places where there are
> > shadowed variables that are at least confusing.  For all I know they're
> > buggy and my resolution of these warnings is wrong.
> > 
> > The first 12 patches just untangle the unclean uses of __ret in wait.h
> > & friends.  Then 4 patches to fix problems in headers that are noticed
> > by kernel/sched.  Two patches fix the two places in kernel/sched/
> > with shadowed variables and the final patch adds -Wshadow=local to
> > the Makefile.
> 
> You are my hero. I was pulling my hair out trying to figure out how
> to deal with this a few months ago, and the use of UNIQUE_ID was the
> key. Yay!
> 
> > I'm quite certain this patch series isn't going in as-is.  But maybe
> > it'll inspire some patches that can go in.
> 
> I think it's pretty darn close. One thing that can be done to test the
> results for the first 12 patches is to do a binary comparison -- these
> changes _should_ have no impact on the final machine code. (It'll
> totally change the debug sections, etc, but the machine code should be
> the same.)

Peter pointed out that I got confused about which __ret was being
referred to:

<peterz> +#define __wait_event_freezable_timeout(wq_head, condition, timeout, __ret) \
<peterz> +       ___wait_event(wq_head, ___wait_cond_timeout(condition, __ret),  \
<peterz> +                     TASK_INTERRUPTIBLE, 0, timeout,                   \
<peterz> +                     __ret = freezable_schedule_timeout(__ret), UNIQUE_ID)
<peterz> so now that internal variable is UNIQUE_ID, whatever that is
<peterz> but the condition argument was supposed to look at that
<peterz> but you explicitly pulled that out

ie "__ret = freezable_schedule_timeout(__ret)" is supposed to refer to
the inner __ret, not the outer __ret.  Which was the opposite of what
I thought was supposed to happen.

We can fix this, of course.  Something like ...

#define ___wait_event_freezable_timeout(wq_head, condition, timeout, ret) \
	___wait_event(wq_head, ___wait_cond_timeout(condition, ret),      \
		TASK_INTERRUPTIBLE, 0, timeout,				  \
		ret = freezable_schedule_timeout(ret), ret)

#define __wait_event_freezable_timeout(wq_head, condition, timeout) \
	___wait_event_freezable_timeout(wq_head, condition, timeout, UNIQUE_ID)

... and now all the 'ret' refer to the thing that they look like they're
referring to.