Greg,
> -----Original Message----- > From: ddstr...@gmail.com [mailto:ddstr...@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 <zhan...@marvell.com> wrote: > > Greg, > > > > > > -----Original Message----- > > From: Greg KH [mailto:gre...@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 > > Tested-by: Dan Streetman <ddstr...@ieee.org> > It's a real bug in the kernel. What's your comments about this patch? > > > >> 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