Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20230629155433.4170837-1-dhowells@redhat.com>
In-Reply-To: <20230629155433.4170837-1-dhowells@redhat.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Thu, 29 Jun 2023 10:56:04 -0700
Message-ID: <CAHk-=wiDwfyj0CCupT-oEToqsNLcbsTQdcgDupF=ZETUjJQJtQ@mail.gmail.com>
Subject: Re: [RFC PATCH 0/4] splice: Fix corruption in data spliced to pipe
To:     David Howells <dhowells@redhat.com>
Cc:     netdev@vger.kernel.org, Matthew Wilcox <willy@infradead.org>,
        Dave Chinner <david@fromorbit.com>,
        Matt Whitlock <kernel@mattwhitlock.name>,
        Jens Axboe <axboe@kernel.dk>, linux-fsdevel@kvack.org,
        linux-mm@kvack.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, 29 Jun 2023 at 08:55, David Howells <dhowells@redhat.com> wrote:
>
> Matt Whitlock, Matthew Wilcox and Dave Chinner are of the opinion that data
> in the pipe must not be seen to change and that if it does, this is a bug.

I'm not convinced.

The whole *point* of vmsplice (and splicing from a file) is the zero-copy.

If you don't want the zero-copy, then you should use just "write()".

So I disagree violently. This is not a bug unless you can point to
some other correctness issues.

The "stableness" of the data is literally the *only* difference
between vmsplice() and write().

> Whilst this does allow the code to be somewhat simplified, it also results
> in a loss of performance: stolen pages have to be reloaded in accessed
> again; more data has to be copied.

No. It literally results in a loss of THE WHOLE POINT of vmsplice().

                    Linus