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: <65b22cd4e8bb142c5b7b86bc33fb08de6f318089.1577017472.git.jstancek@redhat.com>
In-Reply-To: <65b22cd4e8bb142c5b7b86bc33fb08de6f318089.1577017472.git.jstancek@redhat.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Sun, 22 Dec 2019 09:50:53 -0800
Message-ID: <CAHk-=wiGB-Xt1CPbLQ3wY5KENq48Ws5WNwHz+aQp+gmZY+47EQ@mail.gmail.com>
Subject: Re: [PATCH] pipe: fix empty pipe check in pipe_write()
To:     Jan Stancek <jstancek@redhat.com>
Cc:     linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        David Howells <dhowells@redhat.com>,
        Al Viro <viro@zeniv.linux.org.uk>, rasibley@redhat.com
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sun, Dec 22, 2019 at 4:35 AM Jan Stancek <jstancek@redhat.com> wrote:
> Problem is that after pipe_write() reacquires pipe lock, it
> re-checks for empty pipe with potentially stale 'head' and
> doesn't wake up read side anymore. pipe->tail can advance
> beyond 'head', because there are multiple writers.

Thank you. Patch is obviously correct, applied.

I wonder how much that whole "cache head/tail/mask" really helps, and
if we should strive to get rid of it entirely (and just make
"pipe_emptuy()" and friends take a 'const struct pipe_inode_info *"
argument).

Oh well.  I've apple your one-liner, but next time I might decide the
cleverness and slight code generation advantage might not be worth it.

Hopefully there won't _be_ a next time, of course ;)

             Linus