Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190813224620.31005-1-dave@stgolabs.net> <20190813224620.31005-2-dave@stgolabs.net>
 <20190821215707.GA99147@google.com> <20190822044936.qusm5zgjdf6n5fds@linux-r8p5>
 <20190822181701.zhfdkjbwjh56i3ax@linux-r8p5>
In-Reply-To: <20190822181701.zhfdkjbwjh56i3ax@linux-r8p5>
From:   Michel Lespinasse <walken@google.com>
Date:   Thu, 22 Aug 2019 13:10:37 -0700
Message-ID: <CANN689E+xsZWOKFBuv1pkpXO-i4i=Yhg3ebnD++ujz7yfDqwuQ@mail.gmail.com>
Subject: Re: [PATCH 1/3] x86,mm/pat: Use generic interval trees
To:     Davidlohr Bueso <dave@stgolabs.net>
Cc:     Ingo Molnar <mingo@kernel.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Peter Zijlstra <peterz@infradead.org>,
        Andrew Morton <akpm@linux-foundation.org>, x86@kernel.org,
        LKML <linux-kernel@vger.kernel.org>,
        Davidlohr Bueso <dbueso@suse.de>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

I think vma_interval_tree is a bit of a mixed bag, but mostly leans
towards using half closed intervals.

Right now vma_last_pgoff() has to do -1 because of the interval tree
using closed intervals. Similarly, rmap_walk_file(), which I consider
to be the main user of the vma_interval_tree, also has to do -1 when
computing pgoff_end because of the interval tree closed intervals. So,
I think overall vma_interval_tree would also more naturally use
half-open intervals.

But, that's not a 100% thing for vma_interval_tree, as it also has
uses that do stabbing queries (in arch specific code, in hugetlb
cases, and in dax code).

On Thu, Aug 22, 2019 at 11:17 AM Davidlohr Bueso <dave@stgolabs.net> wrote:
>
> >On Wed, 21 Aug 2019, Michel Lespinasse wrote:
> >>As I had commented some time ago, I wish the interval trees used [start,end)
> >>intervals instead of [start,last] - it would be a better fit for basically
> >>all of the current interval tree users.
>
> So the vma_interval_tree (which is a pretty important user) tends to break this
> pattern, as most lookups are [a,a]. We would have to update most of the
> vma_interval_tree_foreach calls, for example, to now do [a,a+1[ such that we
> don't break things. Some cases for the anon_vma_tree as well (ie memory-failure).
>
> I'm not sure anymore it's worth going down this path as we end up exchanging one
> hack for another (and the vma_interval_tree is a pretty big user); but I'm sure
> you're aware of this and thus disagree.
>
> Thanks,
> Davidlohr


-- 
Michel "Walken" Lespinasse
A program is never fully debugged until the last user dies.