Re: [PATCH RFC v2 2/2] mm: Rework swap handling of zap_pte_range

From: John Hubbard <jhubbard@nvidia.com>
To: Matthew Wilcox <willy@infradead.org>
Cc: Peter Xu <peterx@redhat.com>,
	linux-kernel@vger.kernel.org, linux-mm@kvack.org,
	David Hildenbrand <david@redhat.com>,
	Andrea Arcangeli <aarcange@redhat.com>,
	Yang Shi <shy828301@gmail.com>, Vlastimil Babka <vbabka@suse.cz>,
	Hugh Dickins <hughd@google.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Alistair Popple <apopple@nvidia.com>,
	"Kirill A . Shutemov" <kirill@shutemov.name>
Subject: Re: [PATCH RFC v2 2/2] mm: Rework swap handling of zap_pte_range
Date: Tue, 16 Nov 2021 11:06:18 -0800	[thread overview]
Message-ID: <b055cf62-c957-db35-2bf9-630345d63cf6@nvidia.com> (raw)
In-Reply-To: <YZOt7qD6yeSXJgv4@casper.infradead.org>

On 11/16/21 05:11, Matthew Wilcox wrote:
> On Tue, Nov 16, 2021 at 12:51:13AM -0800, John Hubbard wrote:
>> On 11/15/21 05:57, Matthew Wilcox wrote:
>>> On Mon, Nov 15, 2021 at 09:49:51PM +0800, Peter Xu wrote:
>>>> Clean the code up by merging the device private/exclusive swap entry handling
>>>> with the rest, then we merge the pte clear operation too.
>>>>
>>>> struct* page is defined in multiple places in the function, move it upward.
>>>
>>> Is that actually a good thing?  There was a time when declaring
>>
>> Yes. It is a very good thing. Having multiple cases of shadowed variables
>> (in this case I'm using programming language terminology, or what I
>> remember it as, anyway) provides lots of opportunities to create
>> hard-to-spot bugs.
> 
> I think you're misremembering.  These are shadowed variables:

OK, I remembered correctly, but read the diffs a little too quickly, and...

> 
> int a;
> 
> int b(void)
> {
> 	int a;
> 	if (c) {
> 		int a;
> 	}
> }
> 
> This is not:
> 
> int b(void)
> {

...missed that there is no longer a "int a" at the top level. But it does
still present a small land mine, in that just adding a top level "int a"
creates all these shadowed variables (not necessarily bugs, yet, I know).

It's less of an issue here, then I first thought. Generally, it's probably best
to either use "int a" throughout, or differently named variables at lower
levels...or make smaller functions. Because if a variable name is reused
a lot in the same function then there is likely a relationship of sorts
between the instances, and it's worth deciding what that is.

> 	if (c) {
> 		int a;
> 	} else {
> 		int a;
> 	}
> }
> 
> I really wish we could turn on -Wshadow, but we get compilation warnings
> from header files right now.  Or we did last time I checked.
> 

...and as you say, it would be nice if the programmer could just let the
compiler figure out if there is a real problem. The elaborate rituals to
stay out of harm's way are not as good as a tool that checks. :)

thanks,
-- 
John Hubbard
NVIDIA