Re: [PATCH 5/7] userfaultfd: switch to exclusive wakeup for blocking reads

From: Linus Torvalds <torvalds@linux-foundation.org>
To: Andrea Arcangeli <aarcange@redhat.com>
Cc: "Huangpeng (Peter)" <peter.huangpeng@huawei.com>,
	Paolo Bonzini <pbonzini@redhat.com>,
	qemu-devel@nongnu.org, Pavel Emelyanov <xemul@parallels.com>,
	Hugh Dickins <hughd@google.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	"Dr. David Alan Gilbert" <dgilbert@redhat.com>,
	Andres Lagar-Cavilla <andreslc@google.com>,
	Andy Lutomirski <luto@amacapital.net>,
	linux-mm@kvack.org, Johannes Weiner <hannes@cmpxchg.org>,
	Rik van Riel <riel@redhat.com>,
	"Kirill A. Shutemov" <kirill@shutemov.name>,
	linux-kernel@vger.kernel.org, zhang.zhanghailiang@huawei.com,
	Sanidhya Kashyap <sanidhya.gatech@gmail.com>,
	Dave Hansen <dave.hansen@intel.com>,
	Peter Feiner <pfeiner@google.com>, Mel Gorman <mgorman@suse.de>,
	kvm@vger.kernel.org
Subject: Re: [PATCH 5/7] userfaultfd: switch to exclusive wakeup for blocking reads
Date: Mon, 15 Jun 2015 08:19:07 -1000	[thread overview]
Message-ID: <CA+55aFxD8hakE9SjhAD1_vJ9PATK+90k7yHQ2cENqGqK8r3QhQ@mail.gmail.com> (raw)
In-Reply-To: <1434388931-24487-6-git-send-email-aarcange@redhat.com>

[-- Attachment #1: Type: text/plain, Size: 1033 bytes --]

On Jun 15, 2015 7:22 AM, "Andrea Arcangeli" <aarcange@redhat.com> wrote:
>
> Blocking reads can easily use exclusive wakeups. Poll in theory could
> too but there's no poll_wait_exclusive in common code yet.

NAK.

Tie while commit message is crap, and so us the comment

No, your really cannot "easily use exclusive waits", and no, using them for
polling isn't about a lack of interface, it's about the fact that it would
be buggy shit.

What if the process doing the polling never doors anything with the end
result? Maybe it meant to, but it got killed before it could? Are you going
to leave everybody else blocked, even though there are pending events?

The same us try of read() too. What if the reader only reads party of the
message? The wake didn't wake anybody else, so now people are (again)
blocked despite there being data.

So no, exclusive waiting is never "simple". You have to 100% guarantee that
you will consume all the data that caused the wake event (or perhaps wake
the next person up if you don't).

    Linus

[-- Attachment #2: Type: text/html, Size: 1301 bytes --]