WARNING - OLD ARCHIVES

This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-devel

[Xen-devel] Re: [PATCH 00/20] x86: ticket lock rewrite and paravirtualiz

To: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Subject: [Xen-devel] Re: [PATCH 00/20] x86: ticket lock rewrite and paravirtualization
From: Jeremy Fitzhardinge <jeremy@xxxxxxxx>
Date: Mon, 15 Nov 2010 13:02:09 -0800
Cc: Xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxx>, Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxx>, Srivatsa Vaddagiri <vatsa@xxxxxxxxxxxxxxxxxx>, Linux Kernel Mailing List <linux-kernel@xxxxxxxxxxxxxxx>, Jan Beulich <JBeulich@xxxxxxxxxx>, Linux Virtualization <virtualization@xxxxxxxxxxxxxxxxxxxxxxxxxx>, Avi Kivity <avi@xxxxxxxxxx>, "H. Peter Anvin" <hpa@xxxxxxxxx>
Delivery-date: Mon, 15 Nov 2010 13:03:09 -0800
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
In-reply-to: <1289852088.2109.553.camel@laptop>
List-help: <mailto:xen-devel-request@lists.xensource.com?subject=help>
List-id: Xen developer discussion <xen-devel.lists.xensource.com>
List-post: <mailto:xen-devel@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=unsubscribe>
References: <cover.1288794124.git.jeremy.fitzhardinge@xxxxxxxxxx> <4CDDBBD3.5050903@xxxxxxxxx> <4CDDBCE4.80906@xxxxxxxx> <4CDDBDB5.8000800@xxxxxxxxx> <4CE1915F.60507@xxxxxxxx> <4CE1920F.5000509@xxxxxxxxx> <1289852088.2109.553.camel@laptop>
Sender: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.12) Gecko/20101027 Fedora/3.1.6-1.fc13 Lightning/1.0b3pre Thunderbird/3.1.6
On 11/15/2010 12:14 PM, Peter Zijlstra wrote:
> On Mon, 2010-11-15 at 12:03 -0800, H. Peter Anvin wrote:
>> On 11/15/2010 12:00 PM, Jeremy Fitzhardinge wrote:
>>> Another approach I discussed with PeterZ and Mathieu is to steal the LSB
>>> of the ticket counters (halving the max CPU count) to use as a "there is
>>> someone in slowpath waiting on this lock".  But I haven't spent the time
>>> to work out an algorithm to maintain that flag (or flags, since there
>>> are bits available) in a correct and efficient way.
>>>
>> Definitely worth pondering.
> Right, so the idea was to make the ticket increment 2, which would leave
> the LSB of both the head and tail available. I think that if one were to
> set both (using a cmpxchg), the ticket fast-path wouldn't need any
> changes since head==tail is still the correct condition for acquisition.
>
> Then the unlock needs an added conditional:
>   if (tail & 1) 
>       unlock_slowpath()

The tricky part is knowing how to clear the bit(s) on the last person
dropping out of the slow path, and making that race-free with respect to
new lockers entering the slow path.  I guess you could leave it in
slowpath state until you're the last unlocker (ie, you're unlocking into
uncontended state), whereupon you also clear the bits; I guess that
would probably need a cmpxchg to make it safe WRT new lockers entering
slowpath.

As a heuristic, it shouldn't be too bad performancewise, since
(handwaving) if ticketholder N has entered the slowpath, then its likely
that N+1 will as well.

    J


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel