Re: [PATCH v2 03/13] filemap: align the index to mapping_min_order in the page cache

From: Kent Overstreet <kent.overstreet@linux.dev>
To: Matthew Wilcox <willy@infradead.org>
Cc: "Pankaj Raghav (Samsung)" <kernel@pankajraghav.com>,
	 linux-fsdevel@vger.kernel.org, linux-xfs@vger.kernel.org,
	djwong@kernel.org, mcgrof@kernel.org,  linux-mm@kvack.org,
	hare@suse.de, david@fromorbit.com, akpm@linux-foundation.org,
	 gost.dev@samsung.com, linux-kernel@vger.kernel.org,
	chandan.babu@oracle.com,  Pankaj Raghav <p.raghav@samsung.com>
Subject: Re: [PATCH v2 03/13] filemap: align the index to mapping_min_order in the page cache
Date: Fri, 1 Mar 2024 15:04:33 -0500	[thread overview]
Message-ID: <c5rw63nyg2tdkgeuvriu74jjv2vszy2luorhmv3gb4uz2z4msz@2ktshazjwc2n> (raw)
In-Reply-To: <ZeIr_2fiEpWLgmsv@casper.infradead.org>

On Fri, Mar 01, 2024 at 07:26:55PM +0000, Matthew Wilcox wrote:
> On Fri, Mar 01, 2024 at 05:44:34PM +0100, Pankaj Raghav (Samsung) wrote:
> > +#define DEFINE_READAHEAD_ALIGNED(ractl, f, r, m, i)			\
> > +	struct readahead_control ractl = {				\
> > +		.file = f,						\
> > +		.mapping = m,						\
> > +		.ra = r,						\
> > +		._index = mapping_align_start_index(m, i),		\
> > +	}
> 
> My point was that you didn't need to do any of this.
> 
> Look, I've tried to give constructive review, but I feel like I'm going
> to have to be blunt.  There is no evidence of design or understanding
> in these patches or their commit messages.  You don't have a coherent
> message about "These things have to be aligned; these things can be at
> arbitrary alignment".  If you have thought about it, it doesn't show.

Don't you think you might be going off a bit much? I looked over these
patches after we talked privately, and they looked pretty sensible to
me...

Yes, we _always_ want more thorough commit messages that properly
explain the motivations for changes, but in my experience that's the
thing that takes the longest to learn how to do well as an engineer...
ease up abit.

> So, let's start off: Is the index in ractl aligned or not, and why do
> you believe that's the right approach?  And review each of the patches
> in this series with the answer to that question in mind because you are
> currently inconsistent.

^ this is a real point though, DEFINE_READAHEAD_ALIGNED() feels off to
me.