patch "x86/cpufreq: relocate the driver register function" breaks cpu hot(un)plug

patch "x86/cpufreq: relocate the driver register function" breaks cpu hot(un)plug

[Dario Faggioli]

[-- Attachment #1.1: Type: text/plain, Size: 3872 bytes --]

Hey,

As far as my bisection goes, commit
49388f11d512bb92706ce046643bfbb3c1d963c9 "x86/cpufreq: relocate the
driver register function" prevents me from hot unplugging pCPUs.

Xen does not crash or anything, but dom0 is stalled. In fact, with
current staging, here's what I see:

root@Zhaman:~# echo 0 > /sys/devices/system/xen_cpu/xen_cpu6/online 
[   81.583001] INFO: rcu_sched detected stalls on CPUs/tasks: { 12} (detected by 3, t=5252 jiffies, g=1691, c=1690, q=76)
[   81.583036] Task dump for CPU 12:
[   81.583044] bash            R  running task        0  1347   1094 0x00000008
[   81.583056]  ffffffff00000000 0000000000000000 0000000000000000 ffff8800192c2e38
[   81.583070]  ffff8800008472e8 0000000000000002 ffff8800008472e8 ffff880013817858
[   81.583082]  0000000000000000 00000000000081a4 ffffffff811e8137 ffff8800192c2e38
[   81.583095] Call Trace:
[   81.583110]  [<ffffffff811e8137>] ? notify_change+0x2f7/0x390
[   81.583148]  [<ffffffff811c8c74>] ? do_truncate+0x74/0x90
[   81.583158]  [<ffffffff811e2866>] ? dput+0x26/0x230
[   81.583167]  [<ffffffff811d53c5>] ? terminate_walk+0x35/0x40
[   81.583176]  [<ffffffff811d92b1>] ? do_last+0x621/0x12c0
[   81.583188]  [<ffffffff8139f0e7>] ? xen_pcpu_down+0x47/0x70
[   81.583199]  [<ffffffff8156c64d>] ? store_online+0x9d/0xb0
[   81.583210]  [<ffffffff81240bfc>] ? kernfs_fop_write+0x12c/0x180
[   81.583220]  [<ffffffff811ca513>] ? __vfs_write+0x23/0xf0
[   81.583230]  [<ffffffff811cd142>] ? __sb_start_write+0x42/0xf0
[   81.583241]  [<ffffffff8125f711>] ? security_file_permission+0x21/0xa0
[   81.583250]  [<ffffffff811caea1>] ? vfs_write+0xa1/0x1c0
[   81.583259]  [<ffffffff811c828f>] ? filp_close+0x4f/0x70
[   81.583268]  [<ffffffff811cbb12>] ? SyS_write+0x42/0xb0
[   81.583277]  [<ffffffff811e9031>] ? __close_fd+0x71/0xb0
[   81.583287]  [<ffffffff815780f2>] ? system_call_fastpath+0x16/0x75
[  144.555020] INFO: rcu_sched detected stalls on CPUs/tasks: { 12} (detected by 4, t=21007 jiffies, g=1691, c=1690, q=244)
[  144.555046] Task dump for CPU 12:
[  144.555051] bash            R  running task        0  1347   1094 0x00000008
[  144.555059]  ffffffff00000000 0000000000000000 0000000000000000 ffff8800192c2e38
[  144.555068]  ffff8800008472e8 0000000000000002 ffff8800008472e8 ffff880013817858
[  144.555076]  0000000000000000 00000000000081a4 ffffffff811e8137 ffff8800192c2e38
[  144.555084] Call Trace:
[  144.555096]  [<ffffffff811e8137>] ? notify_change+0x2f7/0x390
[  144.555105]  [<ffffffff811c8c74>] ? do_truncate+0x74/0x90
[  144.555112]  [<ffffffff811e2866>] ? dput+0x26/0x230
[  144.555118]  [<ffffffff811d53c5>] ? terminate_walk+0x35/0x40
[  144.555124]  [<ffffffff811d92b1>] ? do_last+0x621/0x12c0
[  144.555164]  [<ffffffff8139f0e7>] ? xen_pcpu_down+0x47/0x70
[  144.555172]  [<ffffffff8156c64d>] ? store_online+0x9d/0xb0
[  144.555179]  [<ffffffff81240bfc>] ? kernfs_fop_write+0x12c/0x180
[  144.555186]  [<ffffffff811ca513>] ? __vfs_write+0x23/0xf0
[  144.555192]  [<ffffffff811cd142>] ? __sb_start_write+0x42/0xf0
[  144.555200]  [<ffffffff8125f711>] ? security_file_permission+0x21/0xa0
[  144.555206]  [<ffffffff811caea1>] ? vfs_write+0xa1/0x1c0
[  144.555212]  [<ffffffff811c828f>] ? filp_close+0x4f/0x70
[  144.555217]  [<ffffffff811cbb12>] ? SyS_write+0x42/0xb0
[  144.555223]  [<ffffffff811e9031>] ? __close_fd+0x71/0xb0
[  144.555230]  [<ffffffff815780f2>] ? system_call_fastpath+0x16/0x75

If I revert that patch, the issue goes away.

Any ideas? 

Regards,
Dario

PS. yes, I'll implement a cpu hotplug/unplug testcase ASAP. :-)

-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)


[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: patch "x86/cpufreq: relocate the driver register function" breaks cpu hot(un)plug

[Konrad Rzeszutek Wilk]

On Fri, Oct 09, 2015 at 06:48:23PM +0200, Dario Faggioli wrote:
> Hey,
> 
> As far as my bisection goes, commit
> 49388f11d512bb92706ce046643bfbb3c1d963c9 "x86/cpufreq: relocate the
> driver register function" prevents me from hot unplugging pCPUs.
> 
> Xen does not crash or anything, but dom0 is stalled. In fact, with
> current staging, here's what I see:
> 
> root@Zhaman:~# echo 0 > /sys/devices/system/xen_cpu/xen_cpu6/online 
> [   81.583001] INFO: rcu_sched detected stalls on CPUs/tasks: { 12} (detected by 3, t=5252 jiffies, g=1691, c=1690, q=76)
> [   81.583036] Task dump for CPU 12:
> [   81.583044] bash            R  running task        0  1347   1094 0x00000008
> [   81.583056]  ffffffff00000000 0000000000000000 0000000000000000 ffff8800192c2e38
> [   81.583070]  ffff8800008472e8 0000000000000002 ffff8800008472e8 ffff880013817858
> [   81.583082]  0000000000000000 00000000000081a4 ffffffff811e8137 ffff8800192c2e38
> [   81.583095] Call Trace:
> [   81.583110]  [<ffffffff811e8137>] ? notify_change+0x2f7/0x390
> [   81.583148]  [<ffffffff811c8c74>] ? do_truncate+0x74/0x90
> [   81.583158]  [<ffffffff811e2866>] ? dput+0x26/0x230
> [   81.583167]  [<ffffffff811d53c5>] ? terminate_walk+0x35/0x40
> [   81.583176]  [<ffffffff811d92b1>] ? do_last+0x621/0x12c0
> [   81.583188]  [<ffffffff8139f0e7>] ? xen_pcpu_down+0x47/0x70
> [   81.583199]  [<ffffffff8156c64d>] ? store_online+0x9d/0xb0
> [   81.583210]  [<ffffffff81240bfc>] ? kernfs_fop_write+0x12c/0x180
> [   81.583220]  [<ffffffff811ca513>] ? __vfs_write+0x23/0xf0
> [   81.583230]  [<ffffffff811cd142>] ? __sb_start_write+0x42/0xf0
> [   81.583241]  [<ffffffff8125f711>] ? security_file_permission+0x21/0xa0
> [   81.583250]  [<ffffffff811caea1>] ? vfs_write+0xa1/0x1c0
> [   81.583259]  [<ffffffff811c828f>] ? filp_close+0x4f/0x70
> [   81.583268]  [<ffffffff811cbb12>] ? SyS_write+0x42/0xb0
> [   81.583277]  [<ffffffff811e9031>] ? __close_fd+0x71/0xb0
> [   81.583287]  [<ffffffff815780f2>] ? system_call_fastpath+0x16/0x75
> [  144.555020] INFO: rcu_sched detected stalls on CPUs/tasks: { 12} (detected by 4, t=21007 jiffies, g=1691, c=1690, q=244)
> [  144.555046] Task dump for CPU 12:
> [  144.555051] bash            R  running task        0  1347   1094 0x00000008
> [  144.555059]  ffffffff00000000 0000000000000000 0000000000000000 ffff8800192c2e38
> [  144.555068]  ffff8800008472e8 0000000000000002 ffff8800008472e8 ffff880013817858
> [  144.555076]  0000000000000000 00000000000081a4 ffffffff811e8137 ffff8800192c2e38
> [  144.555084] Call Trace:
> [  144.555096]  [<ffffffff811e8137>] ? notify_change+0x2f7/0x390
> [  144.555105]  [<ffffffff811c8c74>] ? do_truncate+0x74/0x90
> [  144.555112]  [<ffffffff811e2866>] ? dput+0x26/0x230
> [  144.555118]  [<ffffffff811d53c5>] ? terminate_walk+0x35/0x40
> [  144.555124]  [<ffffffff811d92b1>] ? do_last+0x621/0x12c0
> [  144.555164]  [<ffffffff8139f0e7>] ? xen_pcpu_down+0x47/0x70
> [  144.555172]  [<ffffffff8156c64d>] ? store_online+0x9d/0xb0
> [  144.555179]  [<ffffffff81240bfc>] ? kernfs_fop_write+0x12c/0x180
> [  144.555186]  [<ffffffff811ca513>] ? __vfs_write+0x23/0xf0
> [  144.555192]  [<ffffffff811cd142>] ? __sb_start_write+0x42/0xf0
> [  144.555200]  [<ffffffff8125f711>] ? security_file_permission+0x21/0xa0
> [  144.555206]  [<ffffffff811caea1>] ? vfs_write+0xa1/0x1c0
> [  144.555212]  [<ffffffff811c828f>] ? filp_close+0x4f/0x70
> [  144.555217]  [<ffffffff811cbb12>] ? SyS_write+0x42/0xb0
> [  144.555223]  [<ffffffff811e9031>] ? __close_fd+0x71/0xb0
> [  144.555230]  [<ffffffff815780f2>] ? system_call_fastpath+0x16/0x75
> 
> If I revert that patch, the issue goes away.
> 
> Any ideas? 

I think it is due to xen-acpi-processor re-uploading the C and P states
whenever an CPU goes up. It also does this after S3 suspend.

Anyhow it may be due to the fact that cpufreq_register_driver in Xen is now
'__init' If you remove that little thing would it work?

> 
> Regards,
> Dario
> 
> PS. yes, I'll implement a cpu hotplug/unplug testcase ASAP. :-)
> 
> -- 
> <<This happens because I choose it to happen!>> (Raistlin Majere)
> -----------------------------------------------------------------
> Dario Faggioli, Ph.D, http://about.me/dario.faggioli
> Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)
> 



> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

Re: patch "x86/cpufreq: relocate the driver register function" breaks cpu hot(un)plug

[Wang, Wei W]

On 09/10/2015 04:01, Konrad Rzeszutek Wilk wrote:
> On Fri, Oct 09, 2015 at 06:48:23PM +0200, Dario Faggioli wrote:
> > Hey,
> >
> > As far as my bisection goes, commit
> > 49388f11d512bb92706ce046643bfbb3c1d963c9 "x86/cpufreq: relocate the
> > driver register function" prevents me from hot unplugging pCPUs.
> >
> > Xen does not crash or anything, but dom0 is stalled. In fact, with
> > current staging, here's what I see:
> >
> > root@Zhaman:~# echo 0 > /sys/devices/system/xen_cpu/xen_cpu6/online
> > [   81.583001] INFO: rcu_sched detected stalls on CPUs/tasks: { 12} (detected
> by 3, t=5252 jiffies, g=1691, c=1690, q=76)
> > [   81.583036] Task dump for CPU 12:
> > [   81.583044] bash            R  running task        0  1347   1094 0x00000008
> > [   81.583056]  ffffffff00000000 0000000000000000 0000000000000000
> ffff8800192c2e38
> > [   81.583070]  ffff8800008472e8 0000000000000002 ffff8800008472e8
> ffff880013817858
> > [   81.583082]  0000000000000000 00000000000081a4 ffffffff811e8137
> ffff8800192c2e38
> > [   81.583095] Call Trace:
> > [   81.583110]  [<ffffffff811e8137>] ? notify_change+0x2f7/0x390
> > [   81.583148]  [<ffffffff811c8c74>] ? do_truncate+0x74/0x90
> > [   81.583158]  [<ffffffff811e2866>] ? dput+0x26/0x230
> > [   81.583167]  [<ffffffff811d53c5>] ? terminate_walk+0x35/0x40
> > [   81.583176]  [<ffffffff811d92b1>] ? do_last+0x621/0x12c0
> > [   81.583188]  [<ffffffff8139f0e7>] ? xen_pcpu_down+0x47/0x70
> > [   81.583199]  [<ffffffff8156c64d>] ? store_online+0x9d/0xb0
> > [   81.583210]  [<ffffffff81240bfc>] ? kernfs_fop_write+0x12c/0x180
> > [   81.583220]  [<ffffffff811ca513>] ? __vfs_write+0x23/0xf0
> > [   81.583230]  [<ffffffff811cd142>] ? __sb_start_write+0x42/0xf0
> > [   81.583241]  [<ffffffff8125f711>] ? security_file_permission+0x21/0xa0
> > [   81.583250]  [<ffffffff811caea1>] ? vfs_write+0xa1/0x1c0
> > [   81.583259]  [<ffffffff811c828f>] ? filp_close+0x4f/0x70
> > [   81.583268]  [<ffffffff811cbb12>] ? SyS_write+0x42/0xb0
> > [   81.583277]  [<ffffffff811e9031>] ? __close_fd+0x71/0xb0
> > [   81.583287]  [<ffffffff815780f2>] ? system_call_fastpath+0x16/0x75
> > [  144.555020] INFO: rcu_sched detected stalls on CPUs/tasks: { 12}
> > (detected by 4, t=21007 jiffies, g=1691, c=1690, q=244) [  144.555046] Task
> dump for CPU 12:
> > [  144.555051] bash            R  running task        0  1347   1094 0x00000008
> > [  144.555059]  ffffffff00000000 0000000000000000 0000000000000000
> > ffff8800192c2e38 [  144.555068]  ffff8800008472e8 0000000000000002
> > ffff8800008472e8 ffff880013817858 [  144.555076]  0000000000000000
> > 00000000000081a4 ffffffff811e8137 ffff8800192c2e38 [  144.555084] Call
> Trace:
> > [  144.555096]  [<ffffffff811e8137>] ? notify_change+0x2f7/0x390 [
> > 144.555105]  [<ffffffff811c8c74>] ? do_truncate+0x74/0x90 [
> > 144.555112]  [<ffffffff811e2866>] ? dput+0x26/0x230 [  144.555118]
> > [<ffffffff811d53c5>] ? terminate_walk+0x35/0x40 [  144.555124]
> > [<ffffffff811d92b1>] ? do_last+0x621/0x12c0 [  144.555164]
> > [<ffffffff8139f0e7>] ? xen_pcpu_down+0x47/0x70 [  144.555172]
> > [<ffffffff8156c64d>] ? store_online+0x9d/0xb0 [  144.555179]
> > [<ffffffff81240bfc>] ? kernfs_fop_write+0x12c/0x180 [  144.555186]
> > [<ffffffff811ca513>] ? __vfs_write+0x23/0xf0 [  144.555192]
> > [<ffffffff811cd142>] ? __sb_start_write+0x42/0xf0 [  144.555200]
> > [<ffffffff8125f711>] ? security_file_permission+0x21/0xa0
> > [  144.555206]  [<ffffffff811caea1>] ? vfs_write+0xa1/0x1c0 [
> > 144.555212]  [<ffffffff811c828f>] ? filp_close+0x4f/0x70 [
> > 144.555217]  [<ffffffff811cbb12>] ? SyS_write+0x42/0xb0 [  144.555223]
> > [<ffffffff811e9031>] ? __close_fd+0x71/0xb0 [  144.555230]
> > [<ffffffff815780f2>] ? system_call_fastpath+0x16/0x75
> >
> > If I revert that patch, the issue goes away.
> >
> > Any ideas?

Hi Dario,

Please also remove "register_cpu_notifier(&cpu_nfb)" in the cpufreq_register_driver function as well. (found that it has already been included in cpufreq_presmp_nfb()).

Best,
Wei

> I think it is due to xen-acpi-processor re-uploading the C and P states whenever
> an CPU goes up. It also does this after S3 suspend.
> 
> Anyhow it may be due to the fact that cpufreq_register_driver in Xen is now
> '__init' If you remove that little thing would it work?
> 
> >
> > Regards,
> > Dario
> >
> > PS. yes, I'll implement a cpu hotplug/unplug testcase ASAP. :-)
> >
> > --
> > <<This happens because I choose it to happen!>> (Raistlin Majere)
> > -----------------------------------------------------------------
> > Dario Faggioli, Ph.D, http://about.me/dario.faggioli Senior Software
> > Engineer, Citrix Systems R&D Ltd., Cambridge (UK)
> >
> 
> 
> 
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xen.org
> > http://lists.xen.org/xen-devel

Re: patch "x86/cpufreq: relocate the driver register function" breaks cpu hot(un)plug

[Jan Beulich]

>>> On 09.10.15 at 22:00, <konrad.wilk@oracle.com> wrote:
> Anyhow it may be due to the fact that cpufreq_register_driver in Xen is now
> '__init' If you remove that little thing would it work?

Before adding this annotation I carefully check all callers, and both
which I could find are themselves __init. Did I overlook any?

Jan

Re: patch "x86/cpufreq: relocate the driver register function" breaks cpu hot(un)plug

[Jan Beulich]

>>> On 10.10.15 at 03:38, <wei.w.wang@intel.com> wrote:
> On 09/10/2015 04:01, Konrad Rzeszutek Wilk wrote:
>> On Fri, Oct 09, 2015 at 06:48:23PM +0200, Dario Faggioli wrote:
>> > Hey,
>> >
>> > As far as my bisection goes, commit
>> > 49388f11d512bb92706ce046643bfbb3c1d963c9 "x86/cpufreq: relocate the
>> > driver register function" prevents me from hot unplugging pCPUs.
>> >
>> > Xen does not crash or anything, but dom0 is stalled. In fact, with
>> > current staging, here's what I see:
>> >
>> > root@Zhaman:~# echo 0 > /sys/devices/system/xen_cpu/xen_cpu6/online
>> > [   81.583001] INFO: rcu_sched detected stalls on CPUs/tasks: { 12} 
> (detected
>> by 3, t=5252 jiffies, g=1691, c=1690, q=76)
>> > [   81.583036] Task dump for CPU 12:
>> > [   81.583044] bash            R  running task        0  1347   1094 
> 0x00000008
>> > [   81.583056]  ffffffff00000000 0000000000000000 0000000000000000
>> ffff8800192c2e38
>> > [   81.583070]  ffff8800008472e8 0000000000000002 ffff8800008472e8
>> ffff880013817858
>> > [   81.583082]  0000000000000000 00000000000081a4 ffffffff811e8137
>> ffff8800192c2e38
>> > [   81.583095] Call Trace:
>> > [   81.583110]  [<ffffffff811e8137>] ? notify_change+0x2f7/0x390
>> > [   81.583148]  [<ffffffff811c8c74>] ? do_truncate+0x74/0x90
>> > [   81.583158]  [<ffffffff811e2866>] ? dput+0x26/0x230
>> > [   81.583167]  [<ffffffff811d53c5>] ? terminate_walk+0x35/0x40
>> > [   81.583176]  [<ffffffff811d92b1>] ? do_last+0x621/0x12c0
>> > [   81.583188]  [<ffffffff8139f0e7>] ? xen_pcpu_down+0x47/0x70
>> > [   81.583199]  [<ffffffff8156c64d>] ? store_online+0x9d/0xb0
>> > [   81.583210]  [<ffffffff81240bfc>] ? kernfs_fop_write+0x12c/0x180
>> > [   81.583220]  [<ffffffff811ca513>] ? __vfs_write+0x23/0xf0
>> > [   81.583230]  [<ffffffff811cd142>] ? __sb_start_write+0x42/0xf0
>> > [   81.583241]  [<ffffffff8125f711>] ? security_file_permission+0x21/0xa0
>> > [   81.583250]  [<ffffffff811caea1>] ? vfs_write+0xa1/0x1c0
>> > [   81.583259]  [<ffffffff811c828f>] ? filp_close+0x4f/0x70
>> > [   81.583268]  [<ffffffff811cbb12>] ? SyS_write+0x42/0xb0
>> > [   81.583277]  [<ffffffff811e9031>] ? __close_fd+0x71/0xb0
>> > [   81.583287]  [<ffffffff815780f2>] ? system_call_fastpath+0x16/0x75
>> > [  144.555020] INFO: rcu_sched detected stalls on CPUs/tasks: { 12}
>> > (detected by 4, t=21007 jiffies, g=1691, c=1690, q=244) [  144.555046] Task
>> dump for CPU 12:
>> > [  144.555051] bash            R  running task        0  1347   1094 
> 0x00000008
>> > [  144.555059]  ffffffff00000000 0000000000000000 0000000000000000
>> > ffff8800192c2e38 [  144.555068]  ffff8800008472e8 0000000000000002
>> > ffff8800008472e8 ffff880013817858 [  144.555076]  0000000000000000
>> > 00000000000081a4 ffffffff811e8137 ffff8800192c2e38 [  144.555084] Call
>> Trace:
>> > [  144.555096]  [<ffffffff811e8137>] ? notify_change+0x2f7/0x390 [
>> > 144.555105]  [<ffffffff811c8c74>] ? do_truncate+0x74/0x90 [
>> > 144.555112]  [<ffffffff811e2866>] ? dput+0x26/0x230 [  144.555118]
>> > [<ffffffff811d53c5>] ? terminate_walk+0x35/0x40 [  144.555124]
>> > [<ffffffff811d92b1>] ? do_last+0x621/0x12c0 [  144.555164]
>> > [<ffffffff8139f0e7>] ? xen_pcpu_down+0x47/0x70 [  144.555172]
>> > [<ffffffff8156c64d>] ? store_online+0x9d/0xb0 [  144.555179]
>> > [<ffffffff81240bfc>] ? kernfs_fop_write+0x12c/0x180 [  144.555186]
>> > [<ffffffff811ca513>] ? __vfs_write+0x23/0xf0 [  144.555192]
>> > [<ffffffff811cd142>] ? __sb_start_write+0x42/0xf0 [  144.555200]
>> > [<ffffffff8125f711>] ? security_file_permission+0x21/0xa0
>> > [  144.555206]  [<ffffffff811caea1>] ? vfs_write+0xa1/0x1c0 [
>> > 144.555212]  [<ffffffff811c828f>] ? filp_close+0x4f/0x70 [
>> > 144.555217]  [<ffffffff811cbb12>] ? SyS_write+0x42/0xb0 [  144.555223]
>> > [<ffffffff811e9031>] ? __close_fd+0x71/0xb0 [  144.555230]
>> > [<ffffffff815780f2>] ? system_call_fastpath+0x16/0x75
>> >
>> > If I revert that patch, the issue goes away.
>> >
>> > Any ideas?
> 
> Please also remove "register_cpu_notifier(&cpu_nfb)" in the 
> cpufreq_register_driver function as well. (found that it has already been 
> included in cpufreq_presmp_nfb()).

Yes, I think this is it, and I seem to recall having complained about
this in an earlier version of the series (and I overlooked it now).
Dario, if could you let us know whether that helps?

Jan

Re: patch "x86/cpufreq: relocate the driver register function" breaks cpu hot(un)plug

[Dario Faggioli]

[-- Attachment #1.1: Type: text/plain, Size: 1003 bytes --]

On Mon, 2015-10-12 at 07:19 -0600, Jan Beulich wrote:
> > > > On 09.10.15 at 22:00, <konrad.wilk@oracle.com> wrote:
> > Anyhow it may be due to the fact that cpufreq_register_driver in
> > Xen is now
> > '__init' If you remove that little thing would it work?
> 
> Before adding this annotation I carefully check all callers, and both
> which I could find are themselves __init. Did I overlook any?
> 
No, I've just checked, and I also think __init is fine. And in fact, I
tried removing it, and the issue is still there.

I think the issue is the fact that we register the cpu notifier twice,
as an effect of 49388f11d512bb92706ce046643bfbb3c1d963c9, as Wei noted.

I'm just about try and confirm that...

Regards,
Dario
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)


[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: patch "x86/cpufreq: relocate the driver register function" breaks cpu hot(un)plug

[Andrew Cooper]

On 12/10/15 14:19, Jan Beulich wrote:
>>>> On 09.10.15 at 22:00, <konrad.wilk@oracle.com> wrote:
>> Anyhow it may be due to the fact that cpufreq_register_driver in Xen is now
>> '__init' If you remove that little thing would it work?
> Before adding this annotation I carefully check all callers, and both
> which I could find are themselves __init. Did I overlook any?

I agree that the __init status of the function is correct.

Other than the register_cpu_notifier() change, there is another change
which checks that !driver_data->target == !driver_data->setpolicy which
might be having an unintended side effect.

~Andrew

Re: patch "x86/cpufreq: relocate the driver register function" breaks cpu hot(un)plug

[Jan Beulich]

>>> On 12.10.15 at 15:25, <andrew.cooper3@citrix.com> wrote:
> On 12/10/15 14:19, Jan Beulich wrote:
>>>>> On 09.10.15 at 22:00, <konrad.wilk@oracle.com> wrote:
>>> Anyhow it may be due to the fact that cpufreq_register_driver in Xen is now
>>> '__init' If you remove that little thing would it work?
>> Before adding this annotation I carefully check all callers, and both
>> which I could find are themselves __init. Did I overlook any?
> 
> I agree that the __init status of the function is correct.
> 
> Other than the register_cpu_notifier() change, there is another change
> which checks that !driver_data->target == !driver_data->setpolicy which
> might be having an unintended side effect.

Hardly, without anyone setting setpolicy to non-NULL so far.

Jan

Re: patch "x86/cpufreq: relocate the driver register function" breaks cpu hot(un)plug

[Dario Faggioli]

[-- Attachment #1.1: Type: text/plain, Size: 1298 bytes --]

On Mon, 2015-10-12 at 07:22 -0600, Jan Beulich wrote:
> > > > On 10.10.15 at 03:38, <wei.w.wang@intel.com> wrote:
> > On 09/10/2015 04:01, Konrad Rzeszutek Wilk wrote:

> > Please also remove "register_cpu_notifier(&cpu_nfb)" in the 
> > cpufreq_register_driver function as well. (found that it has
> > already been 
> > included in cpufreq_presmp_nfb()).
> 
> Yes, I think this is it, and I seem to recall having complained about
> this in an earlier version of the series (and I overlooked it now).
>
In fact, I did not spot it with `git diff' either (and the changelog
says "Move the driver register function"!).

> Dario, if could you let us know whether that helps?
> 
Yep, I confirm that it was it. Avoiding calling register_cpu_notifier()
from cpufreq_register_driver(), I am able to put CPUs offline and then
back online again.

I've got to go AFK for a bit now, but I'm fine sending a patch later in
the afternoon. If someone wants to try beating me, no hard feelings.
:-)

Thanks and Regards,
Dario
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)


[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel