From: "Iremonger, Bernard" <bernard.iremonger@intel.com>
To: "Ananyev, Konstantin" <konstantin.ananyev@intel.com>,
Stephen Hemminger <stephen@networkplumber.org>,
"Doherty, Declan" <declan.doherty@intel.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>,
"Iremonger, Bernard" <bernard.iremonger@intel.com>
Subject: Re: [dpdk-dev] [PATCH 1/5] bonding: replace spinlock with read/write lock
Date: Thu, 26 May 2016 16:24:17 +0000 [thread overview]
Message-ID: <8CEF83825BEC744B83065625E567D7C21A007EA4@IRSMSX108.ger.corp.intel.com> (raw)
In-Reply-To: <2601191342CEEE43887BDE71AB97725836B5098C@irsmsx105.ger.corp.intel.com>
Hi Konstantin,
<snip>
> > > > On 05/05/16 18:12, Stephen Hemminger wrote:
> > > > > On Thu, 5 May 2016 16:14:56 +0100 Bernard Iremonger
> > > > > <bernard.iremonger@intel.com> wrote:
> > > > >
> > > > >> Fixes: a45b288ef21a ("bond: support link status polling")
> > > > >> Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
> > > > >
> > > > > You know an uncontested reader/writer lock is significantly
> > > > > slower than a spinlock.
> > > > >
> > > >
> > > > As we can have multiple readers of the active slave list / primary
> > > > slave, basically any tx/rx burst call needs to protect against a
> > > > device being removed/closed during it's operation now that we
> > > > support hotplugging, in the worst case this could mean we have
> > > > 2(rx+tx) * queues possibly using the active slave list
> > > > simultaneously, in that case I would have thought that a spinlock
> > > > would have a much more significant affect on performance?
> > >
> > > Right, but the window where the shared variable is accessed is very
> > > small, and it is actually faster to use spinlock for that.
> >
> > I don't think that window we hold the lock is that small, let say if
> > we have a burst of 32 packets * (let say) 50 cycles/pkt = ~1500 cycles - each
> IO thread would stall.
> > For me that's long enough to justify rwlock usage here, especially
> > that DPDK rwlock price is not much bigger (as I remember) then
> > spinlock - it is basically 1 CAS operation.
>
> As another alternative we can have a spinlock per queue, then different IO
> threads doing RX/XTX over different queues will be uncontended at all.
> Though control thread would need to grab locks for all configured queues :)
>
> Konstantin
>
I am preparing a v2 patchset which uses a spinlock per queue.
Regards,
Bernard.
next prev parent reply other threads:[~2016-05-26 16:24 UTC|newest]
Thread overview: 42+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-05-05 15:14 [dpdk-dev] [PATCH 0/5] bonding: locks Bernard Iremonger
2016-05-05 15:14 ` [dpdk-dev] [PATCH 1/5] bonding: replace spinlock with read/write lock Bernard Iremonger
2016-05-05 17:12 ` Stephen Hemminger
2016-05-06 10:32 ` Declan Doherty
2016-05-06 15:55 ` Stephen Hemminger
2016-05-13 17:10 ` Ananyev, Konstantin
2016-05-13 17:18 ` Ananyev, Konstantin
2016-05-26 16:24 ` Iremonger, Bernard [this message]
2016-05-05 15:14 ` [dpdk-dev] [PATCH 2/5] bonding: add read/write lock to rx/tx burst functions Bernard Iremonger
2016-05-05 15:14 ` [dpdk-dev] [PATCH 3/5] bonding: remove memcopy of slaves from rx/tx burst function Bernard Iremonger
2016-05-05 15:14 ` [dpdk-dev] [PATCH 4/5] bonding: add read/write lock to stop function Bernard Iremonger
2016-05-05 15:15 ` [dpdk-dev] [PATCH 5/5] bonding: add read/write lock to the link_update function Bernard Iremonger
2016-05-26 16:38 ` [dpdk-dev] [PATCH v2 0/6] bonding: locks Bernard Iremonger
2016-05-26 16:38 ` [dpdk-dev] [PATCH v2 1/6] bonding: add spinlock to rx and tx queues Bernard Iremonger
2016-06-10 18:12 ` Ananyev, Konstantin
2016-06-12 17:11 ` [dpdk-dev] [PATCH v3 0/4] bonding: locks Bernard Iremonger
2016-06-12 17:11 ` [dpdk-dev] [PATCH v3 1/4] bonding: add spinlock to rx and tx queues Bernard Iremonger
2016-06-12 17:11 ` [dpdk-dev] [PATCH v3 2/4] bonding: grab queue spinlocks in slave add and remove Bernard Iremonger
2016-06-12 17:11 ` [dpdk-dev] [PATCH v3 3/4] bonding: take queue spinlock in rx/tx burst functions Bernard Iremonger
2016-06-13 9:18 ` Bruce Richardson
2016-06-13 12:28 ` Iremonger, Bernard
2016-06-16 14:32 ` Bruce Richardson
2016-06-16 15:00 ` Thomas Monjalon
2016-06-16 16:41 ` Iremonger, Bernard
2016-06-16 18:38 ` Thomas Monjalon
2017-02-15 18:01 ` Ferruh Yigit
2017-02-16 9:13 ` Bruce Richardson
2017-02-16 11:39 ` Iremonger, Bernard
2017-02-20 11:15 ` Ferruh Yigit
2016-09-09 11:29 ` Ferruh Yigit
2016-06-12 17:11 ` [dpdk-dev] [PATCH v3 4/4] bonding: remove memcpy from " Bernard Iremonger
2016-09-11 12:39 ` Yuanhan Liu
2016-05-26 16:38 ` [dpdk-dev] [PATCH v2 2/6] bonding: grab queue spinlocks in slave add and remove Bernard Iremonger
2016-06-10 18:14 ` Ananyev, Konstantin
2016-05-26 16:38 ` [dpdk-dev] [PATCH v2 3/6] bonding: take queue spinlock in rx/tx burst functions Bernard Iremonger
2016-06-10 18:14 ` Ananyev, Konstantin
2016-05-26 16:38 ` [dpdk-dev] [PATCH v2 4/6] bonding: add spinlock to stop function Bernard Iremonger
2016-05-26 16:38 ` [dpdk-dev] [PATCH v2 5/6] bonding: add spinlock to link update function Bernard Iremonger
2016-05-26 16:38 ` [dpdk-dev] [PATCH v2 6/6] bonding: remove memcpy from burst functions Bernard Iremonger
2016-06-10 18:15 ` Ananyev, Konstantin
2016-06-10 14:45 ` [dpdk-dev] [PATCH v2 0/6] bonding: locks Bruce Richardson
2016-06-10 18:24 ` Iremonger, Bernard
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=8CEF83825BEC744B83065625E567D7C21A007EA4@IRSMSX108.ger.corp.intel.com \
--to=bernard.iremonger@intel.com \
--cc=declan.doherty@intel.com \
--cc=dev@dpdk.org \
--cc=konstantin.ananyev@intel.com \
--cc=stephen@networkplumber.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).