Messages in this thread | | | Date | Tue, 21 Oct 2014 11:57:24 +0900 | From | Yasuaki Ishimatsu <> | Subject | Re: [PATCH] drivers: base: update cpu offline info when do hotplug |
| |
Hi Neil and Dan,
(2014/10/21 2:02), Dan Streetman wrote: > 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
How is the issue reproduced.
Here is a result on my x86 box with linux-3.18-rc1.
- before offline CPU # cd /sys/devices/system/cpu/ # cat online 0-59 # for n in {0..59} ; do test $( cat cpu$n/online ) -eq 1 && echo -n "$n " ; done ; echo "" 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
- after offline CPU{1..59} # for n in {1..59} ; do echo 0 > cpu$n/online; done # cat online 0 # for n in {0..59} ; do test $( cat cpu$n/online ) -eq 1 && echo -n "$n " ; done ; echo "" 0
It seems that dev->offline is set to correct valute.
Thanks, Yasuaki Ishimatsu
> > > While with the patch, the cpu online nodes are kept up to date as the > cpus are brought up/down: > > [root@br10p02 cpu]$ 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@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 > -- > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > Please read the FAQ at http://www.tux.org/lkml/ >
-- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |