Date: Tue, 22 Aug 2017 22:53:16 +0200
From: Peter Zijlstra <peterz@infradead.org>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: "Liang, Kan" <kan.liang@intel.com>,
        Mel Gorman <mgorman@techsingularity.net>, Mel Gorman <mgorman@suse.de>,
        "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
        Tim Chen <tim.c.chen@linux.intel.com>, Ingo Molnar <mingo@elte.hu>,
        Andi Kleen <ak@linux.intel.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Johannes Weiner <hannes@cmpxchg.org>, Jan Kara <jack@suse.cz>,
        linux-mm <linux-mm@kvack.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 1/2] sched/wait: Break up long wake list walk
Message-ID: <20170822205316.GS32112@worktop.programming.kicks-ass.net>
References: <37D7C6CF3E00A74B8858931C1DB2F07753787AE4@SHSMSX103.ccr.corp.intel.com>
 <CA+55aFxZjjqUM4kPvNEeZahPovBHFATiwADj-iPTDN0-jnU67Q@mail.gmail.com>
 <20170818185455.qol3st2nynfa47yc@techsingularity.net>
 <CA+55aFwX0yrUPULrDxTWVCg5c6DKh-yCG84NXVxaptXNQ4O_kA@mail.gmail.com>
 <20170821183234.kzennaaw2zt2rbwz@techsingularity.net>
 <37D7C6CF3E00A74B8858931C1DB2F07753788B58@SHSMSX103.ccr.corp.intel.com>
 <37D7C6CF3E00A74B8858931C1DB2F0775378A24A@SHSMSX103.ccr.corp.intel.com>
 <CA+55aFy=4y0fq9nL2WR1x8vwzJrDOdv++r036LXpR=6Jx8jpzg@mail.gmail.com>
 <37D7C6CF3E00A74B8858931C1DB2F0775378A377@SHSMSX103.ccr.corp.intel.com>
 <CA+55aFwavpFfKNW9NVgNhLggqhii-guc5aX1X5fxrPK+==id0g@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CA+55aFwavpFfKNW9NVgNhLggqhii-guc5aX1X5fxrPK+==id0g@mail.gmail.com>
User-Agent: Mutt/1.5.22.1 (2013-10-16)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 559
Lines: 18

On Tue, Aug 22, 2017 at 01:42:13PM -0700, Linus Torvalds wrote:
> +void wait_on_page_bit_or_yield(struct page *page, int bit_nr)
> +{
> +	if (PageWaiters(page)) {
> +		yield();
> +		return;
> +	}
> +	wait_on_page_bit(page, bit_nr);
> +}

So _the_ problem with yield() is when you hit this with a RT task it
will busy spin and possibly not allow the task that actually has the
lock to make progress at all.

So ideally there'd be a timeout or other limit on the amount of yield().

This being bit-spinlocks leaves us very short on state to play with
though :/