Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210701204246.2037142-1-agruenba@redhat.com> <CAHk-=wjk6KP3vSLFNPMjoaZ4xY4u=DjTM+C+hCS3QKt+XAE6OA@mail.gmail.com>
 <CAHc6FU5XMfTOBW1a0OnMaPSmgWECvO-KYAo8HeVNsCyd-HyV3g@mail.gmail.com>
In-Reply-To: <CAHc6FU5XMfTOBW1a0OnMaPSmgWECvO-KYAo8HeVNsCyd-HyV3g@mail.gmail.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Thu, 1 Jul 2021 17:30:28 -0700
Message-ID: <CAHk-=whaRsSuUeVBn_rLECWyG6h9RGuF9idsT_S+==vnsT8org@mail.gmail.com>
Subject: Re: [PATCH] gfs2: Fix mmap + page fault deadlocks
To:     Andreas Gruenbacher <agruenba@redhat.com>
Cc:     Alexander Viro <viro@zeniv.linux.org.uk>,
        cluster-devel <cluster-devel@redhat.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Jan Kara <jack@suse.cz>, Matthew Wilcox <willy@infradead.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Jul 1, 2021 at 5:20 PM Andreas Gruenbacher <agruenba@redhat.com> wrote:
>
> On Thu, Jul 1, 2021 at 11:41 PM Linus Torvalds
> <torvalds@linux-foundation.org> wrote:
> > Also, I have to say that I think the direct-IO code is fundamentally
> > mis-designed. Why it is doing the page lookup _during_ the IO is a
> > complete mystery to me. Why wasn't that done ahead of time before the
> > filesystem took the locks it needed?
>
> That would be inconvenient for reads, when the number of bytes read is
> much smaller than the buffer size and we won't need to page in the
> entire buffer.

What?

A file read will READ THE WHOLE BUFFER.

We're not talking pipes or ttys here. If you ask for X bytes, you'll
get X bytes.

Of course, if you ask for more data than the file has, that's another
thing, but who really does that with direct-IO? And if they do, why
should we care about their silly behavior?

Face it, right now direct-IO is *BUGGY* because of this, and you can
deadlock filesystems with it.

So tell me again how it's "inconvenient" to fix this bug, and fix the
bad direct-IO design?

              Linus