RE: [PATCH] drivers: base: update cpu offline info when do hotplug

[Neil Zhang]

> -----Original Message-----
> From: Greg KH [mailto:gregkh@xxxxxxxxxxxxxxxxxxx]
> Sent: 2014å10æ27æ 9:59
> To: Neil Zhang
> Cc: Dan Streetman; linux-kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH] drivers: base: update cpu offline info when do hotplug
> 
> On Sun, Oct 26, 2014 at 06:43:11PM -0700, Neil Zhang wrote:
> > Greg,
> >
> >
> > > -----Original Message-----
> > > From: ddstreet@xxxxxxxxx [mailto:ddstreet@xxxxxxxxx] On Behalf Of
> > > Dan Streetman
> > > Sent: 2014å10æ21æ 1:03
> > > To: Neil Zhang
> > > Cc: Greg KH; linux-kernel@xxxxxxxxxxxxxxx
> > > Subject: Re: [PATCH] drivers: base: update cpu offline info when do
> > > hotplug
> > >
> > > On Mon, Oct 20, 2014 at 3:40 AM, Neil Zhang <zhangwm@xxxxxxxxxxx> wrote:
> > > > Greg,
> > > >
> > > >
> > > > -----Original Message-----
> > > > From: Greg KH [mailto:gregkh@xxxxxxxxxxxxxxxxxxx]
> > > > Sent: 2014å10æ20æ 14:48
> > > > To: Neil Zhang
> > > > Cc: linux-kernel@xxxxxxxxxxxxxxx
> > > > 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
> > >
> > > Tested-by: Dan Streetman <ddstreet@xxxxxxxx>
> > >
> >
> > It's a real bug in the kernel.
> 
> As this has been this way for many years, I tend to think it's not all that
> important...

Actually this bug was introduced by the following patch.

commit 0902a9044fa5b7a0456ea4daacec2c2b3189ba8c
Author: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
Date:   Fri May 3 00:25:49 2013 +0200

    Driver core: Use generic offline/online for CPU offline/online

So seems not that long :)

> 
> > What's your comments about this patch?
> 
> It's one of 1800+ patches that need my comments, please be patient.

Got it, thanks!

> 
> greg k-h

Best Regards,
Neil Zhang