Konrad,
We''re seeing a failure where c-states are not parsed and loaded to the
hypervisor.
What is happening is when the following calls happen, there are no c-states in
the per CPU data.
[ 1.954368] [<ffffffff8157840e>] push_cxx_to_hypervisor+0x1a5/0x2d5
[ 1.954374] [<ffffffff815785ab>] upload_pm_data+0x6d/0xa2
[ 1.954380] [<ffffffff81afe101>] xen_acpi_processor_init+0x3c3/0x488
It looks like the c-state parsing should have happened in
acpi_processor_power_init but didn''t because of the following check at
the top:
if (disabled_by_idle_boot_param())
return 0;
The preceding is true because boot_option_idle_override is set to IDLE_HALT by
arch/x86/xen/setup.c xen_arch_setup.
Is the Xen version of acpi_processor supposed to do it''s own c-state
parsing or rely on the base acpi_processor c-state parsing? I''m
assuming Xen sets IDLE_HALT to stop runtime usage of Linux c-state management,
but it also stops parsing c-states.
Linux: 3.2.23
Xen: 4.0.4
Thanks for any help,
Tom Goetz
On Thu, Jan 03, 2013 at 01:53:04PM -0500, Tom Goetz wrote:> Konrad, > > We''re seeing a failure where c-states are not parsed and loaded to the hypervisor. > > What is happening is when the following calls happen, there are no c-states in the per CPU data. > > [ 1.954368] [<ffffffff8157840e>] push_cxx_to_hypervisor+0x1a5/0x2d5 > [ 1.954374] [<ffffffff815785ab>] upload_pm_data+0x6d/0xa2 > [ 1.954380] [<ffffffff81afe101>] xen_acpi_processor_init+0x3c3/0x488 > > It looks like the c-state parsing should have happened in acpi_processor_power_init but didn''t because of the following check at the top: > > if (disabled_by_idle_boot_param()) > return 0; > > The preceding is true because boot_option_idle_override is set to IDLE_HALT by arch/x86/xen/setup.c xen_arch_setup. > > Is the Xen version of acpi_processor supposed to do it''s own c-state parsing or rely on the base acpi_processor c-state parsing? I''m assuming Xen sets IDLE_HALT to stop runtime usage of Linux c-state management, but it also stops parsing c-states. > > Linux: 3.2.23I believe you are missing some patches to cpufreq.c and setup.c, which are: t 48cdd8287f47a3cdad5b9273a5ef81bf605f7826 Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Date: Tue Mar 13 20:06:57 2012 -0400 xen/cpufreq: Disable the cpu frequency scaling drivers from loading. commit cc7335b2f6acc0f24c7fac80ce536301f7d52214 Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Date: Mon Jan 23 10:53:57 2012 -0500 xen/setup/pm/acpi: Remove the call to boot_option_idle_override. commit a7b422cda5084db7265c3b23310a959b43b47529 Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Date: Tue Mar 13 19:18:39 2012 -0400 provide disable_cpufreq() function to disable the API. commit e5fd47bfab2df0c2184cc0bf4245d8e1bb7724fb Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Date: Mon Nov 21 18:02:02 2011 -0500 xen/pm_idle: Make pm_idle be default_idle under Xen.> Xen: 4.0.4 > > Thanks for any help, > > Tom Goetz > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xen.org > http://lists.xen.org/xen-devel >
On 01/04/2013 04:16 PM, Konrad Rzeszutek Wilk wrote:> > commit cc7335b2f6acc0f24c7fac80ce536301f7d52214 > Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> > Date: Mon Jan 23 10:53:57 2012 -0500 > > xen/setup/pm/acpi: Remove the call to boot_option_idle_override. >Turns out I botched a merge, and missed this change. Thanks, once again for helping identify this in short order. I owe you another beer at the next summit! Ben
On Jan 4, 2013, at 4:16 PM, Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> wrote:> On Thu, Jan 03, 2013 at 01:53:04PM -0500, Tom Goetz wrote: >> Konrad, >> >> We''re seeing a failure where c-states are not parsed and loaded to the hypervisor. >> >> What is happening is when the following calls happen, there are no c-states in the per CPU data. >> >> [ 1.954368] [<ffffffff8157840e>] push_cxx_to_hypervisor+0x1a5/0x2d5 >> [ 1.954374] [<ffffffff815785ab>] upload_pm_data+0x6d/0xa2 >> [ 1.954380] [<ffffffff81afe101>] xen_acpi_processor_init+0x3c3/0x488 >> >> It looks like the c-state parsing should have happened in acpi_processor_power_init but didn''t because of the following check at the top: >> >> if (disabled_by_idle_boot_param()) >> return 0; >> >> The preceding is true because boot_option_idle_override is set to IDLE_HALT by arch/x86/xen/setup.c xen_arch_setup. >> >> Is the Xen version of acpi_processor supposed to do it''s own c-state parsing or rely on the base acpi_processor c-state parsing? I''m assuming Xen sets IDLE_HALT to stop runtime usage of Linux c-state management, but it also stops parsing c-states. >> >> Linux: 3.2.23 > > I believe you are missing some patches to cpufreq.c and setup.c, which are: >...> > > commit cc7335b2f6acc0f24c7fac80ce536301f7d52214 > Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> > Date: Mon Jan 23 10:53:57 2012 -0500 > > xen/setup/pm/acpi: Remove the call to boot_option_idle_override.Missing this one. As Ben mentioned in his email, we were missing one of the patches you pointed us to. Thanks for all of the help. I''ll buy you a beer too! -Tom
> > > > commit cc7335b2f6acc0f24c7fac80ce536301f7d52214 > > Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> > > Date: Mon Jan 23 10:53:57 2012 -0500 > > > > xen/setup/pm/acpi: Remove the call to boot_option_idle_override. > > Missing this one. > > As Ben mentioned in his email, we were missing one of the patches you pointed us to. Thanks for all of the help. I''ll buy you a beer too!Two beers! Allright, who is next? Maybe I can folks to buy more than two beers.
Wednesday, January 9, 2013, 10:55:12 PM, you wrote:>> > >> > commit cc7335b2f6acc0f24c7fac80ce536301f7d52214 >> > Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> >> > Date: Mon Jan 23 10:53:57 2012 -0500 >> > >> > xen/setup/pm/acpi: Remove the call to boot_option_idle_override. >> >> Missing this one. >> >> As Ben mentioned in his email, we were missing one of the patches you pointed us to. Thanks for all of the help. I''ll buy you a beer too!> Two beers! Allright, who is next? Maybe I can folks to buy more than two beers.*grin* are you going to get lars to do a xen-oktoberfest-mini-summit in Munich ? :-p
On Wed, Jan 09, 2013 at 11:00:37PM +0100, Sander Eikelenboom wrote:> > Wednesday, January 9, 2013, 10:55:12 PM, you wrote: > > >> > > >> > commit cc7335b2f6acc0f24c7fac80ce536301f7d52214 > >> > Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> > >> > Date: Mon Jan 23 10:53:57 2012 -0500 > >> > > >> > xen/setup/pm/acpi: Remove the call to boot_option_idle_override. > >> > >> Missing this one. > >> > >> As Ben mentioned in his email, we were missing one of the patches you pointed us to. Thanks for all of the help. I''ll buy you a beer too! > > > Two beers! Allright, who is next? Maybe I can folks to buy more than two beers. > > > > *grin* are you going to get lars to do a xen-oktoberfest-mini-summit in Munich ? :-pHa! Originally I was hoping to see if I can get a six-pack from them :-)