lkml.org 
[lkml]   [2014]   [Oct]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    From
    Date
    SubjectRE: [PATCH] drivers: base: update cpu offline info when do hotplug
    Dan,

    > -----Original Message-----
    > From: ddstreet@gmail.com [mailto:ddstreet@gmail.com] On Behalf Of Dan
    > Streetman
    > Sent: 2014年10月21日 1:03
    > To: Neil Zhang
    > Cc: Greg KH; linux-kernel@vger.kernel.org
    > Subject: Re: [PATCH] drivers: base: update cpu offline info when do
    > hotplug
    >
    > On Mon, Oct 20, 2014 at 3:40 AM, Neil Zhang <zhangwm@marvell.com> wrote:
    > > Greg,
    > >
    > >
    > > -----Original Message-----
    > > From: Greg KH [mailto:gregkh@linuxfoundation.org]
    > > Sent: 2014年10月20日 14:48
    > > To: Neil Zhang
    > > Cc: linux-kernel@vger.kernel.org
    > > Subject: Re: [PATCH] drivers: base: update cpu offline info when do
    > > hotplug
    > >
    > > On Sun, Oct 19, 2014 at 11:39:23PM -0700, Neil Zhang wrote:
    > >>> How much noise is this going to cause on a big/little system that
    > >>> constantly hot unplug/plugs processors all of the time?
    > >>
    > >> Can you explain more what kind of noise will be introduced on a
    > big/little system?
    > >
    > > Have you tested this on such a machine?
    > >
    > > I didn't have such kind of machine on hand.
    > > Can anyone has such machine to verify it?
    > > Thanks!
    >
    > I tested this on a ppc PowerVM system, using dlpar operations to
    > remove/add cpus.
    >
    > Without this patch the cpu online nodes get out of sync with the main
    > online node (and the actual state of the cpus), because they aren't
    > updated as the cpus are brought up/down:
    >
    > [root@br10p02 cpu]$ pwd
    > /sys/devices/system/cpu
    > [root@br10p02 cpu]$ cat online
    > 0-39
    > [root@br10p02 cpu]$ for n in {0..47} ; do test $( cat cpu$n/online ) -
    > eq 1 && echo -n "$n " ; done ; echo ""
    > 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
    > 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
    >
    >
    > While with the patch, the cpu online nodes are kept up to date as the
    > cpus are brought up/down:
    >
    > [root@br10p02 cpu]$ pwd
    > /sys/devices/system/cpu
    > [root@br10p02 cpu]$ cat online
    > 0-39
    > [root@br10p02 cpu]$ for n in {0..47} ; do test $( cat cpu$n/online ) -
    > eq 1 && echo -n "$n " ; done ; echo ""
    > 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
    > 27 28 29 30 31 32 33 34 35 36 37 38 39
    >
    >
    > Feel free to add

    Thanks for your kindly test!

    >
    > Tested-by: Dan Streetman <ddstreet@ieee.org>
    >
    > >
    > >> As I know IKS on arm will use cpu_suspend way to power down a core.
    > >
    > > Are you sure that it also doesn't use that same functionality to drop
    > a processor to save power?
    > >
    > > As I know it use cpu_suspend to switch out a processor in IKS and
    > there is no cpu hotplug notifier in this procedure.
    > >
    > >
    > > Why do you need/want this notification? What are you going to do
    > with this information that you don't already have?
    > >
    > > The offline won't be updated if an in kernel hotplug governor plug in
    > / out a core which cause the sysfs interface report a wrong status.
    > >
    > >
    > > thanks,
    > >
    > > greg k-h
    > >
    > >
    > > Best Regards,
    > > Neil Zhang


    Best Regards,
    Neil Zhang
    \
     
     \ /
      Last update: 2014-10-21 03:01    [W:3.170 / U:0.216 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site