linux-kernel - Re: [PATCH 1/5] locking: Introduce range reader/writer lock

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20170329161056.GI27446@linux-80c1.suse>
Date:   Wed, 29 Mar 2017 09:10:56 -0700
From:   Davidlohr Bueso <dave@...olabs.net>
To:     "Kirill A. Shutemov" <kirill@...temov.name>
Cc:     Laurent Dufour <ldufour@...ux.vnet.ibm.com>, mingo@...nel.org,
        peterz@...radead.org, akpm@...ux-foundation.org, jack@...e.cz,
        kirill.shutemov@...ux.intel.com, mhocko@...e.com,
        mgorman@...hsingularity.net, linux-kernel@...r.kernel.org,
        Davidlohr Bueso <dbueso@...e.de>
Subject: Re: [PATCH 1/5] locking: Introduce range reader/writer lock

On Wed, 29 Mar 2017, Kirill A. Shutemov wrote:

>On Wed, Mar 29, 2017 at 08:31:33AM -0700, Davidlohr Bueso wrote:
>> On Wed, 29 Mar 2017, Laurent Dufour wrote:
>>
>> > On 28/03/2017 18:58, Kirill A. Shutemov wrote:
>> > > On Tue, Mar 28, 2017 at 09:39:18AM -0700, Davidlohr Bueso wrote:
>> > > > I'll wait to see if there are any more concerns and send a v2 with your corrections.
>> > >
>> > > Have you tried drop-in replacement of mmap_sem with full range lock?
>> > > It would be interesting to see performance implication for this.
>> > >
>> >
>> > I've a patch that replace the mmap_sem with a full range lock, it seems
>> > to work fine for x86 and ppc64 for now. I'll send it soon.
>> > But I didn't yet check for performance. What is the best way to that ?
>>
>> I expect performance to take a measurable hit if we simply use full range
>> lock as a drop in replacement. My rwsem vs range lock measurements were
>> done with this in mind. We only win with range locks when improving the
>> level of parallelism.
>
>It would be hard sell if we would see performance degradation simple
>single-threaded workload.

Yeah, that's why I included very low contention in the lock comparison.
Deltas are very much within the noise region, it is with high contention
where things go south performance wise.