| Message ID | 1905098.zDJocX6404@kreacher |
|---|---|
| State | New |
| Headers | show |
| Series | [v2,1/4] cpufreq: Introduce CPUFREQ_NEED_UPDATE_LIMITS driver flag | expand |
Spelling mistake in $subject (driver) On 23-10-20, 17:36, Rafael J. Wysocki wrote: > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > Because sugov_update_next_freq() may skip a frequency update even if > the need_freq_update flag has been set for the policy at hand, policy > limits updates may not take effect as expected. > > For example, if the intel_pstate driver operates in the passive mode > with HWP enabled, it needs to update the HWP min and max limits when > the policy min and max limits change, respectively, but that may not > happen if the target frequency does not change along with the limit > at hand. In particular, if the policy min is changed first, causing > the target frequency to be adjusted to it, and the policy max limit > is changed later to the same value, the HWP max limit will not be > updated to follow it as expected, because the target frequency is > still equal to the policy min limit and it will not change until > that limit is updated. > > To address this issue, modify get_next_freq() to clear > need_freq_update only if the CPUFREQ_NEED_UPDATE_LIMITS flag is > not set for the cpufreq driver in use (and it should be set for all > potentially affected drivers) and make sugov_update_next_freq() > check need_freq_update and continue when it is set regardless of > whether or not the new target frequency is equal to the old one. > > Fixes: f6ebbcf08f37 ("cpufreq: intel_pstate: Implement passive mode with HWP enabled") > Reported-by: Zhang Rui <rui.zhang@intel.com> > Cc: 5.9+ <stable@vger.kernel.org> # 5.9+ > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > --- > > New patch in v2. > > --- > kernel/sched/cpufreq_schedutil.c | 8 ++++++-- > 1 file changed, 6 insertions(+), 2 deletions(-) > > Index: linux-pm/kernel/sched/cpufreq_schedutil.c > =================================================================== > --- linux-pm.orig/kernel/sched/cpufreq_schedutil.c > +++ linux-pm/kernel/sched/cpufreq_schedutil.c > @@ -102,11 +102,12 @@ static bool sugov_should_update_freq(str > static bool sugov_update_next_freq(struct sugov_policy *sg_policy, u64 time, > unsigned int next_freq) > { > - if (sg_policy->next_freq == next_freq) > + if (sg_policy->next_freq == next_freq && !sg_policy->need_freq_update) > return false; > > sg_policy->next_freq = next_freq; > sg_policy->last_freq_update_time = time; > + sg_policy->need_freq_update = false; > > return true; > } > @@ -164,7 +165,10 @@ static unsigned int get_next_freq(struct > if (freq == sg_policy->cached_raw_freq && !sg_policy->need_freq_update) > return sg_policy->next_freq; > > - sg_policy->need_freq_update = false; > + if (sg_policy->need_freq_update) > + sg_policy->need_freq_update = > + cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS); > + The behavior here is a bit different from what we did in cpufreq.c. In cpufreq core we are _always_ allowing the call to reach the driver's target() routine, but here we do it only if limits have changed. Wonder if we should have similar behavior here as well ? Over that the code here can be rewritten a bit like: if (sg_policy->need_freq_update) sg_policy->need_freq_update = cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS); else if (freq == sg_policy->cached_raw_freq) return sg_policy->next_freq; -- viresh
On Fri, 2020-10-23 at 17:36 +0200, Rafael J. Wysocki wrote: > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > Because sugov_update_next_freq() may skip a frequency update even if > the need_freq_update flag has been set for the policy at hand, policy > limits updates may not take effect as expected. > > For example, if the intel_pstate driver operates in the passive mode > with HWP enabled, it needs to update the HWP min and max limits when > the policy min and max limits change, respectively, but that may not > happen if the target frequency does not change along with the limit > at hand. In particular, if the policy min is changed first, causing > the target frequency to be adjusted to it, and the policy max limit > is changed later to the same value, the HWP max limit will not be > updated to follow it as expected, because the target frequency is > still equal to the policy min limit and it will not change until > that limit is updated. > > To address this issue, modify get_next_freq() to clear > need_freq_update only if the CPUFREQ_NEED_UPDATE_LIMITS flag is > not set for the cpufreq driver in use (and it should be set for all > potentially affected drivers) and make sugov_update_next_freq() > check need_freq_update and continue when it is set regardless of > whether or not the new target frequency is equal to the old one. > > Fixes: f6ebbcf08f37 ("cpufreq: intel_pstate: Implement passive mode > with HWP enabled") > Reported-by: Zhang Rui <rui.zhang@intel.com> > Cc: 5.9+ <stable@vger.kernel.org> # 5.9+ > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> I have confirmed that the problem is gone with this patch series applied. Tested-by: Zhang Rui <rui.zhang@intel.com> thanks, rui > --- > > New patch in v2. > > --- > kernel/sched/cpufreq_schedutil.c | 8 ++++++-- > 1 file changed, 6 insertions(+), 2 deletions(-) > > Index: linux-pm/kernel/sched/cpufreq_schedutil.c > =================================================================== > --- linux-pm.orig/kernel/sched/cpufreq_schedutil.c > +++ linux-pm/kernel/sched/cpufreq_schedutil.c > @@ -102,11 +102,12 @@ static bool sugov_should_update_freq(str > static bool sugov_update_next_freq(struct sugov_policy *sg_policy, > u64 time, > unsigned int next_freq) > { > - if (sg_policy->next_freq == next_freq) > + if (sg_policy->next_freq == next_freq && !sg_policy- > >need_freq_update) > return false; > > sg_policy->next_freq = next_freq; > sg_policy->last_freq_update_time = time; > + sg_policy->need_freq_update = false; > > return true; > } > @@ -164,7 +165,10 @@ static unsigned int get_next_freq(struct > if (freq == sg_policy->cached_raw_freq && !sg_policy- > >need_freq_update) > return sg_policy->next_freq; > > - sg_policy->need_freq_update = false; > + if (sg_policy->need_freq_update) > + sg_policy->need_freq_update = > + cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_L > IMITS); > + > sg_policy->cached_raw_freq = freq; > return cpufreq_driver_resolve_freq(policy, freq); > } > > >
On Tue, Oct 27, 2020 at 5:26 AM Viresh Kumar <viresh.kumar@linaro.org> wrote: > > Spelling mistake in $subject (driver) > > On 23-10-20, 17:36, Rafael J. Wysocki wrote: > > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > > > Because sugov_update_next_freq() may skip a frequency update even if > > the need_freq_update flag has been set for the policy at hand, policy > > limits updates may not take effect as expected. > > > > For example, if the intel_pstate driver operates in the passive mode > > with HWP enabled, it needs to update the HWP min and max limits when > > the policy min and max limits change, respectively, but that may not > > happen if the target frequency does not change along with the limit > > at hand. In particular, if the policy min is changed first, causing > > the target frequency to be adjusted to it, and the policy max limit > > is changed later to the same value, the HWP max limit will not be > > updated to follow it as expected, because the target frequency is > > still equal to the policy min limit and it will not change until > > that limit is updated. > > > > To address this issue, modify get_next_freq() to clear > > need_freq_update only if the CPUFREQ_NEED_UPDATE_LIMITS flag is > > not set for the cpufreq driver in use (and it should be set for all > > potentially affected drivers) and make sugov_update_next_freq() > > check need_freq_update and continue when it is set regardless of > > whether or not the new target frequency is equal to the old one. > > > > Fixes: f6ebbcf08f37 ("cpufreq: intel_pstate: Implement passive mode with HWP enabled") > > Reported-by: Zhang Rui <rui.zhang@intel.com> > > Cc: 5.9+ <stable@vger.kernel.org> # 5.9+ > > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > --- > > > > New patch in v2. > > > > --- > > kernel/sched/cpufreq_schedutil.c | 8 ++++++-- > > 1 file changed, 6 insertions(+), 2 deletions(-) > > > > Index: linux-pm/kernel/sched/cpufreq_schedutil.c > > =================================================================== > > --- linux-pm.orig/kernel/sched/cpufreq_schedutil.c > > +++ linux-pm/kernel/sched/cpufreq_schedutil.c > > @@ -102,11 +102,12 @@ static bool sugov_should_update_freq(str > > static bool sugov_update_next_freq(struct sugov_policy *sg_policy, u64 time, > > unsigned int next_freq) > > { > > - if (sg_policy->next_freq == next_freq) > > + if (sg_policy->next_freq == next_freq && !sg_policy->need_freq_update) > > return false; > > > > sg_policy->next_freq = next_freq; > > sg_policy->last_freq_update_time = time; > > + sg_policy->need_freq_update = false; > > > > return true; > > } > > @@ -164,7 +165,10 @@ static unsigned int get_next_freq(struct > > if (freq == sg_policy->cached_raw_freq && !sg_policy->need_freq_update) > > return sg_policy->next_freq; > > > > - sg_policy->need_freq_update = false; > > + if (sg_policy->need_freq_update) > > + sg_policy->need_freq_update = > > + cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS); > > + > > The behavior here is a bit different from what we did in cpufreq.c. In cpufreq > core we are _always_ allowing the call to reach the driver's target() routine, > but here we do it only if limits have changed. Wonder if we should have similar > behavior here as well ? I didn't think about that, but now that you mentioned it, I think that this is a good idea. Will send an updated patch with that implemented shortly. > Over that the code here can be rewritten a bit like: > > if (sg_policy->need_freq_update) > sg_policy->need_freq_update = cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS); > else if (freq == sg_policy->cached_raw_freq) > return sg_policy->next_freq; Right, but it will be somewhat different anyway. :-)
On 29-10-20, 12:12, Rafael J. Wysocki wrote: > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > Because sugov_update_next_freq() may skip a frequency update even if > the need_freq_update flag has been set for the policy at hand, policy > limits updates may not take effect as expected. > > For example, if the intel_pstate driver operates in the passive mode > with HWP enabled, it needs to update the HWP min and max limits when > the policy min and max limits change, respectively, but that may not > happen if the target frequency does not change along with the limit > at hand. In particular, if the policy min is changed first, causing > the target frequency to be adjusted to it, and the policy max limit > is changed later to the same value, the HWP max limit will not be > updated to follow it as expected, because the target frequency is > still equal to the policy min limit and it will not change until > that limit is updated. > > To address this issue, modify get_next_freq() to let the driver > callback run if the CPUFREQ_NEED_UPDATE_LIMITS cpufreq driver flag > is set regardless of whether or not the new frequency to set is > equal to the previous one. > > Fixes: f6ebbcf08f37 ("cpufreq: intel_pstate: Implement passive mode with HWP enabled") > Reported-by: Zhang Rui <rui.zhang@intel.com> > Tested-by: Zhang Rui <rui.zhang@intel.com> > Cc: 5.9+ <stable@vger.kernel.org> # 5.9+ > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > --- > > v2.1 -> v2.2: > * Instead of updating need_freq_update if CPUFREQ_NEED_UPDATE_LIMITS is set > in get_next_freq() and checking it again in sugov_update_next_freq(), > check CPUFREQ_NEED_UPDATE_LIMITS directly in sugov_update_next_freq(). > * Update the subject. > > v2 -> v2.1: > * Fix typo in the subject. > * Make get_next_freq() and sugov_update_next_freq() ignore the > sg_policy->next_freq == next_freq case when CPUFREQ_NEED_UPDATE_LIMITS > is set for the driver. > * Add Tested-by from Rui (this version lets the driver callback run more > often than the v2, so the behavior in the Rui's case doesn't change). > > --- > kernel/sched/cpufreq_schedutil.c | 6 ++++-- > 1 file changed, 4 insertions(+), 2 deletions(-) > > Index: linux-pm/kernel/sched/cpufreq_schedutil.c > =================================================================== > --- linux-pm.orig/kernel/sched/cpufreq_schedutil.c > +++ linux-pm/kernel/sched/cpufreq_schedutil.c > @@ -102,7 +102,8 @@ static bool sugov_should_update_freq(str > static bool sugov_update_next_freq(struct sugov_policy *sg_policy, u64 time, > unsigned int next_freq) > { > - if (sg_policy->next_freq == next_freq) > + if (sg_policy->next_freq == next_freq && > + !cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS)) > return false; Since sg_policy->next_freq is used elsewhere as well, this is the best we can do here. > sg_policy->next_freq = next_freq; > @@ -161,7 +162,8 @@ static unsigned int get_next_freq(struct > > freq = map_util_freq(util, freq, max); > > - if (freq == sg_policy->cached_raw_freq && !sg_policy->need_freq_update) > + if (freq == sg_policy->cached_raw_freq && !sg_policy->need_freq_update && > + !cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS)) > return sg_policy->next_freq; > > sg_policy->need_freq_update = false; But I was wondering if instead of this we just do this here: if (!cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS)) sg_policy->cached_raw_freq = freq; And so the above check will always fail. Acked-by: Viresh Kumar <viresh.kumar@linaro.org> -- viresh
On Thu, Oct 29, 2020 at 12:23 PM Viresh Kumar <viresh.kumar@linaro.org> wrote: > > On 29-10-20, 12:12, Rafael J. Wysocki wrote: > > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > > > Because sugov_update_next_freq() may skip a frequency update even if > > the need_freq_update flag has been set for the policy at hand, policy > > limits updates may not take effect as expected. > > > > For example, if the intel_pstate driver operates in the passive mode > > with HWP enabled, it needs to update the HWP min and max limits when > > the policy min and max limits change, respectively, but that may not > > happen if the target frequency does not change along with the limit > > at hand. In particular, if the policy min is changed first, causing > > the target frequency to be adjusted to it, and the policy max limit > > is changed later to the same value, the HWP max limit will not be > > updated to follow it as expected, because the target frequency is > > still equal to the policy min limit and it will not change until > > that limit is updated. > > > > To address this issue, modify get_next_freq() to let the driver > > callback run if the CPUFREQ_NEED_UPDATE_LIMITS cpufreq driver flag > > is set regardless of whether or not the new frequency to set is > > equal to the previous one. > > > > Fixes: f6ebbcf08f37 ("cpufreq: intel_pstate: Implement passive mode with HWP enabled") > > Reported-by: Zhang Rui <rui.zhang@intel.com> > > Tested-by: Zhang Rui <rui.zhang@intel.com> > > Cc: 5.9+ <stable@vger.kernel.org> # 5.9+ > > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > --- > > > > v2.1 -> v2.2: > > * Instead of updating need_freq_update if CPUFREQ_NEED_UPDATE_LIMITS is set > > in get_next_freq() and checking it again in sugov_update_next_freq(), > > check CPUFREQ_NEED_UPDATE_LIMITS directly in sugov_update_next_freq(). > > * Update the subject. > > > > v2 -> v2.1: > > * Fix typo in the subject. > > * Make get_next_freq() and sugov_update_next_freq() ignore the > > sg_policy->next_freq == next_freq case when CPUFREQ_NEED_UPDATE_LIMITS > > is set for the driver. > > * Add Tested-by from Rui (this version lets the driver callback run more > > often than the v2, so the behavior in the Rui's case doesn't change). > > > > --- > > kernel/sched/cpufreq_schedutil.c | 6 ++++-- > > 1 file changed, 4 insertions(+), 2 deletions(-) > > > > Index: linux-pm/kernel/sched/cpufreq_schedutil.c > > =================================================================== > > --- linux-pm.orig/kernel/sched/cpufreq_schedutil.c > > +++ linux-pm/kernel/sched/cpufreq_schedutil.c > > @@ -102,7 +102,8 @@ static bool sugov_should_update_freq(str > > static bool sugov_update_next_freq(struct sugov_policy *sg_policy, u64 time, > > unsigned int next_freq) > > { > > - if (sg_policy->next_freq == next_freq) > > + if (sg_policy->next_freq == next_freq && > > + !cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS)) > > return false; > > Since sg_policy->next_freq is used elsewhere as well, this is the best > we can do here. > > > sg_policy->next_freq = next_freq; > > @@ -161,7 +162,8 @@ static unsigned int get_next_freq(struct > > > > freq = map_util_freq(util, freq, max); > > > > - if (freq == sg_policy->cached_raw_freq && !sg_policy->need_freq_update) > > + if (freq == sg_policy->cached_raw_freq && !sg_policy->need_freq_update && > > + !cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS)) > > return sg_policy->next_freq; > > > > sg_policy->need_freq_update = false; > > But I was wondering if instead of this we just do this here: > > if (!cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS)) > sg_policy->cached_raw_freq = freq; > > And so the above check will always fail. I wrote it this way, because I want to avoid looking at the driver flags at all unless the update is going to be skipped. Otherwise we may end up fetching a new cache line here every time even if that is not needed. > Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Thanks!
Index: linux-pm/kernel/sched/cpufreq_schedutil.c =================================================================== --- linux-pm.orig/kernel/sched/cpufreq_schedutil.c +++ linux-pm/kernel/sched/cpufreq_schedutil.c @@ -102,11 +102,12 @@ static bool sugov_should_update_freq(str static bool sugov_update_next_freq(struct sugov_policy *sg_policy, u64 time, unsigned int next_freq) { - if (sg_policy->next_freq == next_freq) + if (sg_policy->next_freq == next_freq && !sg_policy->need_freq_update) return false; sg_policy->next_freq = next_freq; sg_policy->last_freq_update_time = time; + sg_policy->need_freq_update = false; return true; } @@ -164,7 +165,10 @@ static unsigned int get_next_freq(struct if (freq == sg_policy->cached_raw_freq && !sg_policy->need_freq_update) return sg_policy->next_freq; - sg_policy->need_freq_update = false; + if (sg_policy->need_freq_update) + sg_policy->need_freq_update = + cpufreq_driver_test_flags(CPUFREQ_NEED_UPDATE_LIMITS); + sg_policy->cached_raw_freq = freq; return cpufreq_driver_resolve_freq(policy, freq); }