| Message ID | 1408015178-21745-1-git-send-email-shilpa.bhat@linux.vnet.ibm.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Delegated to: | Michael Ellerman |
| Headers | show |
On 14 August 2014 16:49, Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com> wrote: > This patch ensures the cpus to kexec/reboot at nominal frequency. > Nominal frequency is the highest cpu frequency on PowerPC at > which the cores can run without getting throttled. > > If the host kernel had set the cpus to a low pstate and then it > kexecs/reboots to a cpufreq disabled kernel it would cause the target > kernel to perform poorly. It will also increase the boot up time of > the target kernel. So set the cpus to high pstate, in this case to > nominal frequency before rebooting to avoid such scenarios. > > The reboot notifier will suspend the cpufreq governor and enable > nominal frequency to be set during a reboot/kexec similar to the > suspend operartion. > > Signed-off-by: Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com> > Reviewed-by: Preeti U Murthy <preeti@linux.vnet.ibm.com> > --- > drivers/cpufreq/powernv-cpufreq.c | 16 ++++++++++++++++ > 1 file changed, 16 insertions(+) > > diff --git a/drivers/cpufreq/powernv-cpufreq.c b/drivers/cpufreq/powernv-cpufreq.c > index 379c083..e9f3d3a 100644 > --- a/drivers/cpufreq/powernv-cpufreq.c > +++ b/drivers/cpufreq/powernv-cpufreq.c > @@ -26,6 +26,7 @@ > #include <linux/cpufreq.h> > #include <linux/smp.h> > #include <linux/of.h> > +#include <linux/reboot.h> > > #include <asm/cputhreads.h> > #include <asm/firmware.h> > @@ -314,9 +315,21 @@ static int powernv_cpufreq_cpu_init(struct cpufreq_policy *policy) > for (i = 0; i < threads_per_core; i++) > cpumask_set_cpu(base + i, policy->cpus); > > + policy->suspend_freq = pstate_id_to_freq(powernv_pstate_info.nominal); > return cpufreq_table_validate_and_show(policy, powernv_freqs); > } > > +static int powernv_cpufreq_reboot_notifier(struct notifier_block *nb, > + unsigned long action, void *unused) > +{ > + cpufreq_suspend(); > + return NOTIFY_DONE; > +} > + > +static struct notifier_block powernv_cpufreq_reboot_nb = { > + .notifier_call = powernv_cpufreq_reboot_notifier, > +}; > + > static struct cpufreq_driver powernv_cpufreq_driver = { > .name = "powernv-cpufreq", > .flags = CPUFREQ_CONST_LOOPS, > @@ -325,6 +338,7 @@ static struct cpufreq_driver powernv_cpufreq_driver = { > .target_index = powernv_cpufreq_target_index, > .get = powernv_cpufreq_get, > .attr = powernv_cpu_freq_attr, > + .suspend = cpufreq_generic_suspend, I couldn't understand why you have added a notifier here. This callback by itself should be enough. Isn't it? And then you have called cpufreq_suspend(), which is absolutely wrong, from that notifier..
On 08/18/2014 01:16 PM, Viresh Kumar wrote: > On 14 August 2014 16:49, Shilpasri G Bhat > <shilpa.bhat@linux.vnet.ibm.com> wrote: >> This patch ensures the cpus to kexec/reboot at nominal frequency. >> Nominal frequency is the highest cpu frequency on PowerPC at >> which the cores can run without getting throttled. >> >> If the host kernel had set the cpus to a low pstate and then it >> kexecs/reboots to a cpufreq disabled kernel it would cause the target >> kernel to perform poorly. It will also increase the boot up time of >> the target kernel. So set the cpus to high pstate, in this case to >> nominal frequency before rebooting to avoid such scenarios. >> >> The reboot notifier will suspend the cpufreq governor and enable >> nominal frequency to be set during a reboot/kexec similar to the >> suspend operartion. >> >> Signed-off-by: Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com> >> Reviewed-by: Preeti U Murthy <preeti@linux.vnet.ibm.com> >> --- >> drivers/cpufreq/powernv-cpufreq.c | 16 ++++++++++++++++ >> 1 file changed, 16 insertions(+) >> >> diff --git a/drivers/cpufreq/powernv-cpufreq.c b/drivers/cpufreq/powernv-cpufreq.c >> index 379c083..e9f3d3a 100644 >> --- a/drivers/cpufreq/powernv-cpufreq.c >> +++ b/drivers/cpufreq/powernv-cpufreq.c >> @@ -26,6 +26,7 @@ >> #include <linux/cpufreq.h> >> #include <linux/smp.h> >> #include <linux/of.h> >> +#include <linux/reboot.h> >> >> #include <asm/cputhreads.h> >> #include <asm/firmware.h> >> @@ -314,9 +315,21 @@ static int powernv_cpufreq_cpu_init(struct cpufreq_policy *policy) >> for (i = 0; i < threads_per_core; i++) >> cpumask_set_cpu(base + i, policy->cpus); >> >> + policy->suspend_freq = pstate_id_to_freq(powernv_pstate_info.nominal); >> return cpufreq_table_validate_and_show(policy, powernv_freqs); >> } >> >> +static int powernv_cpufreq_reboot_notifier(struct notifier_block *nb, >> + unsigned long action, void *unused) >> +{ >> + cpufreq_suspend(); >> + return NOTIFY_DONE; >> +} >> + >> +static struct notifier_block powernv_cpufreq_reboot_nb = { >> + .notifier_call = powernv_cpufreq_reboot_notifier, >> +}; >> + >> static struct cpufreq_driver powernv_cpufreq_driver = { >> .name = "powernv-cpufreq", >> .flags = CPUFREQ_CONST_LOOPS, >> @@ -325,6 +338,7 @@ static struct cpufreq_driver powernv_cpufreq_driver = { >> .target_index = powernv_cpufreq_target_index, >> .get = powernv_cpufreq_get, >> .attr = powernv_cpu_freq_attr, >> + .suspend = cpufreq_generic_suspend, > I couldn't understand why you have added a notifier here. This callback > by itself should be enough. Isn't it? > > And then you have called cpufreq_suspend(), which is absolutely wrong, > from that notifier.. Hi Viresh, The intention here is stop the cpufreq governor and then to set the cpus to nominal frequency so as to ensure that the frequency won't be changed later. The .suspend callback of the driver is not called during reboot/kexec. So we need an explicit reboot notifier to call cpufreq-suspend() to suffice the requirement. Thanks and Regards, Shilpa
On 21 August 2014 10:36, Shilpasri G Bhat <shilpabhatppc@gmail.com> wrote: > The intention here is stop the cpufreq governor and then to set the cpus to > nominal frequency so as to ensure that the frequency won't be changed later. > > The .suspend callback of the driver is not called during reboot/kexec. > So we need an explicit reboot notifier to call cpufreq-suspend() to > suffice the requirement. Hi Shilpa, No, we can't allow any platform driver to misuse cpufreq_suspend(). Platform drivers aren't *allowed* to call this routine. Now the deal is how do we move to nominal frequency on reboot.. @Rafael: Any suggestions? How do we ensure that governors are stopped on these notifiers, or if there is some other solution here?
Hi Viresh, On 08/21/2014 11:56 AM, Viresh Kumar wrote: > On 21 August 2014 10:36, Shilpasri G Bhat <shilpabhatppc@gmail.com> wrote: >> The intention here is stop the cpufreq governor and then to set the cpus to >> nominal frequency so as to ensure that the frequency won't be changed later. >> >> The .suspend callback of the driver is not called during reboot/kexec. >> So we need an explicit reboot notifier to call cpufreq-suspend() to >> suffice the requirement. > > Hi Shilpa, > > No, we can't allow any platform driver to misuse cpufreq_suspend(). > Platform drivers aren't *allowed* to call this routine. At the moment this looks like the best way forward. We need to do this cleanly by ensuring that we stop the governors and then call into the driver to deal with the cpu frequency in its own way during reboot. The best way to do this would be by calling this routine. Either this or cpufreq_suspend() should be called in the reboot path generically. The latter might not be an enticing option for other platforms. Regards Preeti U Murthy > > Now the deal is how do we move to nominal frequency on reboot.. > @Rafael: Any suggestions? How do we ensure that governors > are stopped on these notifiers, or if there is some other solution here? > _______________________________________________ > Linuxppc-dev mailing list > Linuxppc-dev@lists.ozlabs.org > https://lists.ozlabs.org/listinfo/linuxppc-dev >
On 22 August 2014 08:18, Preeti U Murthy <preeti@linux.vnet.ibm.com> wrote: > At the moment this looks like the best way forward. We need to do this > cleanly by ensuring that we stop the governors and then call into the > driver to deal with the cpu frequency in its own way during reboot. The > best way to do this would be by calling this routine. Either this or > cpufreq_suspend() should be called in the reboot path generically. The > latter might not be an enticing option for other platforms. Its not that I am doubting if this will work or not. But this Hack is using routines not meant for this purpose. And that being a core routine, things aren't that straightforward anymore. @Rafael: Ping!!
diff --git a/drivers/cpufreq/powernv-cpufreq.c b/drivers/cpufreq/powernv-cpufreq.c index 379c083..e9f3d3a 100644 --- a/drivers/cpufreq/powernv-cpufreq.c +++ b/drivers/cpufreq/powernv-cpufreq.c @@ -26,6 +26,7 @@ #include <linux/cpufreq.h> #include <linux/smp.h> #include <linux/of.h> +#include <linux/reboot.h> #include <asm/cputhreads.h> #include <asm/firmware.h> @@ -314,9 +315,21 @@ static int powernv_cpufreq_cpu_init(struct cpufreq_policy *policy) for (i = 0; i < threads_per_core; i++) cpumask_set_cpu(base + i, policy->cpus); + policy->suspend_freq = pstate_id_to_freq(powernv_pstate_info.nominal); return cpufreq_table_validate_and_show(policy, powernv_freqs); } +static int powernv_cpufreq_reboot_notifier(struct notifier_block *nb, + unsigned long action, void *unused) +{ + cpufreq_suspend(); + return NOTIFY_DONE; +} + +static struct notifier_block powernv_cpufreq_reboot_nb = { + .notifier_call = powernv_cpufreq_reboot_notifier, +}; + static struct cpufreq_driver powernv_cpufreq_driver = { .name = "powernv-cpufreq", .flags = CPUFREQ_CONST_LOOPS, @@ -325,6 +338,7 @@ static struct cpufreq_driver powernv_cpufreq_driver = { .target_index = powernv_cpufreq_target_index, .get = powernv_cpufreq_get, .attr = powernv_cpu_freq_attr, + .suspend = cpufreq_generic_suspend, }; static int __init powernv_cpufreq_init(void) @@ -342,12 +356,14 @@ static int __init powernv_cpufreq_init(void) return rc; } + register_reboot_notifier(&powernv_cpufreq_reboot_nb); return cpufreq_register_driver(&powernv_cpufreq_driver); } module_init(powernv_cpufreq_init); static void __exit powernv_cpufreq_exit(void) { + unregister_reboot_notifier(&powernv_cpufreq_reboot_nb); cpufreq_unregister_driver(&powernv_cpufreq_driver); } module_exit(powernv_cpufreq_exit);