| Message ID | f963d09e57115969dae32827ade5558b0467d3a0.1624266901.git.viresh.kumar@linaro.org |
|---|---|
| State | Accepted |
| Commit | 1eb5dde674f57b1a1918dab33f09e35cdd64eb07 |
| Headers | show |
| Series | cpufreq: cppc: Add support for frequency invariance | expand |
On 24-06-21, 10:48, Ionela Voinescu wrote: > On Monday 21 Jun 2021 at 14:49:37 (+0530), Viresh Kumar wrote: > > The Frequency Invariance Engine (FIE) is providing a frequency scaling > > correction factor that helps achieve more accurate load-tracking. > [..] > > +static void cppc_cpufreq_cpu_fie_exit(struct cpufreq_policy *policy) > > +{ > > + struct cppc_freq_invariance *cppc_fi; > > + int cpu; > > + > > + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) > > + return; > > + > > + /* policy->cpus will be empty here, use related_cpus instead */ > > + topology_clear_scale_freq_source(SCALE_FREQ_SOURCE_CPPC, policy->related_cpus); > > + > > + for_each_cpu(cpu, policy->related_cpus) { > > + cppc_fi = &per_cpu(cppc_freq_inv, cpu); > > Do you think it might be worth having here something like: > > if (!cppc_fi->cpu_data) > continue; > > This would be to protect against cases where the platform does not boot > with all CPUs or the module is loaded after some have already been > offlined. Unlikely, but.. Even in that case policy->cpus will contain all offline+online CPUs (at ->init() time), isn't it ? > > + irq_work_sync(&cppc_fi->irq_work); > > + kthread_cancel_work_sync(&cppc_fi->work); > > + } > > +} > > The rest of the code is almost the same as the original, so that is all > from me :). > > Thanks, > Ionela. -- viresh
Hey, On Thursday 24 Jun 2021 at 18:34:18 (+0530), Viresh Kumar wrote: > On 24-06-21, 10:48, Ionela Voinescu wrote: > > On Monday 21 Jun 2021 at 14:49:37 (+0530), Viresh Kumar wrote: > > > The Frequency Invariance Engine (FIE) is providing a frequency scaling > > > correction factor that helps achieve more accurate load-tracking. > > [..] > > > +static void cppc_cpufreq_cpu_fie_exit(struct cpufreq_policy *policy) > > > +{ > > > + struct cppc_freq_invariance *cppc_fi; > > > + int cpu; > > > + > > > + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) > > > + return; > > > + > > > + /* policy->cpus will be empty here, use related_cpus instead */ > > > + topology_clear_scale_freq_source(SCALE_FREQ_SOURCE_CPPC, policy->related_cpus); > > > + > > > + for_each_cpu(cpu, policy->related_cpus) { > > > + cppc_fi = &per_cpu(cppc_freq_inv, cpu); > > > > Do you think it might be worth having here something like: > > > > if (!cppc_fi->cpu_data) > > continue; > > > > This would be to protect against cases where the platform does not boot > > with all CPUs or the module is loaded after some have already been > > offlined. Unlikely, but.. > > Even in that case policy->cpus will contain all offline+online CPUs (at ->init() > time), isn't it ? > Right, my bad. I missed cpumask_and(policy->cpus, policy->cpus, cpu_online_mask) being done after init(). It logically seems a bit wrong, but drivers are in control of setting policy->cpus and acting on it, and in this case the driver does the right thing. Thanks, Ionela. > > > + irq_work_sync(&cppc_fi->irq_work); > > > + kthread_cancel_work_sync(&cppc_fi->work); > > > + } > > > +} > > > > The rest of the code is almost the same as the original, so that is all > > from me :). > > > > Thanks, > > Ionela. > > -- > viresh
On 25-06-21, 09:54, Ionela Voinescu wrote: > Hey, > > On Thursday 24 Jun 2021 at 18:34:18 (+0530), Viresh Kumar wrote: > > On 24-06-21, 10:48, Ionela Voinescu wrote: > > > On Monday 21 Jun 2021 at 14:49:37 (+0530), Viresh Kumar wrote: > > > > The Frequency Invariance Engine (FIE) is providing a frequency scaling > > > > correction factor that helps achieve more accurate load-tracking. > > > [..] > > > > +static void cppc_cpufreq_cpu_fie_exit(struct cpufreq_policy *policy) > > > > +{ > > > > + struct cppc_freq_invariance *cppc_fi; > > > > + int cpu; > > > > + > > > > + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) > > > > + return; > > > > + > > > > + /* policy->cpus will be empty here, use related_cpus instead */ > > > > + topology_clear_scale_freq_source(SCALE_FREQ_SOURCE_CPPC, policy->related_cpus); > > > > + > > > > + for_each_cpu(cpu, policy->related_cpus) { > > > > + cppc_fi = &per_cpu(cppc_freq_inv, cpu); > > > > > > Do you think it might be worth having here something like: > > > > > > if (!cppc_fi->cpu_data) > > > continue; > > > > > > This would be to protect against cases where the platform does not boot > > > with all CPUs or the module is loaded after some have already been > > > offlined. Unlikely, but.. > > > > Even in that case policy->cpus will contain all offline+online CPUs (at ->init() > > time), isn't it ? > > > > Right, my bad. I missed cpumask_and(policy->cpus, policy->cpus, > cpu_online_mask) being done after init(). It logically seems a bit > wrong, but drivers are in control of setting policy->cpus and acting on > it, and in this case the driver does the right thing. Do you want me to re-add your Reviewed-by here ? -- viresh
On Friday 25 Jun 2021 at 22:24:18 (+0530), Viresh Kumar wrote: > On 25-06-21, 09:54, Ionela Voinescu wrote: > > Hey, > > > > On Thursday 24 Jun 2021 at 18:34:18 (+0530), Viresh Kumar wrote: > > > On 24-06-21, 10:48, Ionela Voinescu wrote: > > > > On Monday 21 Jun 2021 at 14:49:37 (+0530), Viresh Kumar wrote: > > > > > The Frequency Invariance Engine (FIE) is providing a frequency scaling > > > > > correction factor that helps achieve more accurate load-tracking. > > > > [..] > > > > > +static void cppc_cpufreq_cpu_fie_exit(struct cpufreq_policy *policy) > > > > > +{ > > > > > + struct cppc_freq_invariance *cppc_fi; > > > > > + int cpu; > > > > > + > > > > > + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) > > > > > + return; > > > > > + > > > > > + /* policy->cpus will be empty here, use related_cpus instead */ > > > > > + topology_clear_scale_freq_source(SCALE_FREQ_SOURCE_CPPC, policy->related_cpus); > > > > > + > > > > > + for_each_cpu(cpu, policy->related_cpus) { > > > > > + cppc_fi = &per_cpu(cppc_freq_inv, cpu); > > > > > > > > Do you think it might be worth having here something like: > > > > > > > > if (!cppc_fi->cpu_data) > > > > continue; > > > > > > > > This would be to protect against cases where the platform does not boot > > > > with all CPUs or the module is loaded after some have already been > > > > offlined. Unlikely, but.. > > > > > > Even in that case policy->cpus will contain all offline+online CPUs (at ->init() > > > time), isn't it ? > > > > > > > Right, my bad. I missed cpumask_and(policy->cpus, policy->cpus, > > cpu_online_mask) being done after init(). It logically seems a bit > > wrong, but drivers are in control of setting policy->cpus and acting on > > it, and in this case the driver does the right thing. > > Do you want me to re-add your Reviewed-by here ? > To be honest I would like to have more time on this before you merge the set, to better understand Qian's results and some observations I have for Thunder X2 (I will share in a bit). For the code, I think it's fine. I have a single observation regarding the following code: > +static void cppc_cpufreq_cpu_fie_init(struct cpufreq_policy *policy) > +{ > + struct cppc_freq_invariance *cppc_fi; > + int cpu, ret; > + > + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) > + return; > + > + for_each_cpu(cpu, policy->cpus) { > + cppc_fi = &per_cpu(cppc_freq_inv, cpu); > + cppc_fi->cpu = cpu; > + cppc_fi->cpu_data = policy->driver_data; > + kthread_init_work(&cppc_fi->work, cppc_scale_freq_workfn); > + init_irq_work(&cppc_fi->irq_work, cppc_irq_work); > + > + ret = cppc_get_perf_ctrs(cpu, &cppc_fi->prev_perf_fb_ctrs); > + if (ret) { > + pr_warn("%s: failed to read perf counters for cpu:%d: %d\n", > + __func__, cpu, ret); > + return; > + } For this condition above, think about a scenario where reading counters for offline CPUs returns an error. I'm not sure if that can happen, to be honest. That would mean here that you will never initialise the freq source unless all CPUs in the policy are online at policy creation. My recommendation is to warn about the failed read of perf counters but only return from this function if the target CPU is online as well when reading counters fails. This is probably a nit, so I'll let you decide if you want to do something about this. Thanks, Ionela. > + } > + > + /* Register for freq-invariance */ > + topology_set_scale_freq_source(&cppc_sftd, policy->cpus); > +} > -- > viresh
On 28-06-21, 11:49, Ionela Voinescu wrote: > To be honest I would like to have more time on this before you merge the > set, to better understand Qian's results and some observations I have > for Thunder X2 (I will share in a bit). Ideally, this code was already merged in 5.13 and would have required us to fix any problems as we encounter them. I did revert it because it caused a kernel crash and I wasn't sure if there was a sane/easy way of fixing that so late in the release cycle. That was the right thing to do then. All those issues are gone now, we may have an issue around rounding of counters or some hardware specific issues, it isn't clear yet. But the stuff works fine otherwise, doesn't make the kernel crash and it is controlled with a CONFIG_ option, so those who don't want to use it can still disable it. The merge window is here now, if we don't merge it now, it gets delayed by a full cycle (roughly two months) and if we merge it now and are able to narrow down the rounding issues, if there are any, we will have full two months to make a fix for that and still push it in 5.14 itself. And so I would like to get it merged in this merge window itself, it also makes sure more people would get to test it, like Qian was able to figure out a problem here for us. > For the code, I think it's fine. I have a single observation regarding > the following code: > > > +static void cppc_cpufreq_cpu_fie_init(struct cpufreq_policy *policy) > > +{ > > + struct cppc_freq_invariance *cppc_fi; > > + int cpu, ret; > > + > > + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) > > + return; > > + > > + for_each_cpu(cpu, policy->cpus) { > > + cppc_fi = &per_cpu(cppc_freq_inv, cpu); > > + cppc_fi->cpu = cpu; > > + cppc_fi->cpu_data = policy->driver_data; > > + kthread_init_work(&cppc_fi->work, cppc_scale_freq_workfn); > > + init_irq_work(&cppc_fi->irq_work, cppc_irq_work); > > + > > + ret = cppc_get_perf_ctrs(cpu, &cppc_fi->prev_perf_fb_ctrs); > > + if (ret) { > > + pr_warn("%s: failed to read perf counters for cpu:%d: %d\n", > > + __func__, cpu, ret); > > + return; > > + } > > For this condition above, think about a scenario where reading counters > for offline CPUs returns an error. I'm not sure if that can happen, to > be honest. That would mean here that you will never initialise the freq > source unless all CPUs in the policy are online at policy creation. > > My recommendation is to warn about the failed read of perf counters but > only return from this function if the target CPU is online as well when > reading counters fails. > > This is probably a nit, so I'll let you decide if you want to do something > about this. That is a very good observation actually. Thanks for that. This is how I fixed it. diff --git a/drivers/cpufreq/cppc_cpufreq.c b/drivers/cpufreq/cppc_cpufreq.c index d688877e8fbe..f6540068d0fe 100644 --- a/drivers/cpufreq/cppc_cpufreq.c +++ b/drivers/cpufreq/cppc_cpufreq.c @@ -171,7 +171,13 @@ static void cppc_cpufreq_cpu_fie_init(struct cpufreq_policy *policy) if (ret) { pr_warn("%s: failed to read perf counters for cpu:%d: %d\n", __func__, cpu, ret); - return; + + /* + * Don't abort if the CPU was offline while the driver + * was getting registered. + */ + if (cpu_online(cpu)) + return; } } -- viresh
Hey, On Tuesday 29 Jun 2021 at 10:02:44 (+0530), Viresh Kumar wrote: > On 28-06-21, 11:49, Ionela Voinescu wrote: > > To be honest I would like to have more time on this before you merge the > > set, to better understand Qian's results and some observations I have > > for Thunder X2 (I will share in a bit). > > Ideally, this code was already merged in 5.13 and would have required > us to fix any problems as we encounter them. I did revert it because > it caused a kernel crash and I wasn't sure if there was a sane/easy > way of fixing that so late in the release cycle. That was the right > thing to do then. > > All those issues are gone now, we may have an issue around rounding of > counters or some hardware specific issues, it isn't clear yet. > > But the stuff works fine otherwise, doesn't make the kernel crash and > it is controlled with a CONFIG_ option, so those who don't want to use > it can still disable it. > > The merge window is here now, if we don't merge it now, it gets > delayed by a full cycle (roughly two months) and if we merge it now > and are able to narrow down the rounding issues, if there are any, we > will have full two months to make a fix for that and still push it in > 5.14 itself. > > And so I would like to get it merged in this merge window itself, it > also makes sure more people would get to test it, like Qian was able > to figure out a problem here for us. > Okay, makes sense. I have not seen this code actually do anything wrong so far, and the issues I see on ThunderX2 point more to misbehaving counters for this purpose. This being said, I would have probably preferred for this feature to be disabled by default, until we've tested more, but that won't give the chance to anyone else to test. > > For the code, I think it's fine. I have a single observation regarding > > the following code: > > > > > +static void cppc_cpufreq_cpu_fie_init(struct cpufreq_policy *policy) > > > +{ > > > + struct cppc_freq_invariance *cppc_fi; > > > + int cpu, ret; > > > + > > > + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) > > > + return; > > > + > > > + for_each_cpu(cpu, policy->cpus) { > > > + cppc_fi = &per_cpu(cppc_freq_inv, cpu); > > > + cppc_fi->cpu = cpu; > > > + cppc_fi->cpu_data = policy->driver_data; > > > + kthread_init_work(&cppc_fi->work, cppc_scale_freq_workfn); > > > + init_irq_work(&cppc_fi->irq_work, cppc_irq_work); > > > + > > > + ret = cppc_get_perf_ctrs(cpu, &cppc_fi->prev_perf_fb_ctrs); > > > + if (ret) { > > > + pr_warn("%s: failed to read perf counters for cpu:%d: %d\n", > > > + __func__, cpu, ret); > > > + return; > > > + } > > > > For this condition above, think about a scenario where reading counters > > for offline CPUs returns an error. I'm not sure if that can happen, to > > be honest. That would mean here that you will never initialise the freq > > source unless all CPUs in the policy are online at policy creation. > > > > My recommendation is to warn about the failed read of perf counters but > > only return from this function if the target CPU is online as well when > > reading counters fails. > > > > This is probably a nit, so I'll let you decide if you want to do something > > about this. > > That is a very good observation actually. Thanks for that. This is how > I fixed it. > > diff --git a/drivers/cpufreq/cppc_cpufreq.c b/drivers/cpufreq/cppc_cpufreq.c > index d688877e8fbe..f6540068d0fe 100644 > --- a/drivers/cpufreq/cppc_cpufreq.c > +++ b/drivers/cpufreq/cppc_cpufreq.c > @@ -171,7 +171,13 @@ static void cppc_cpufreq_cpu_fie_init(struct cpufreq_policy *policy) > if (ret) { > pr_warn("%s: failed to read perf counters for cpu:%d: %d\n", > __func__, cpu, ret); > - return; > + > + /* > + * Don't abort if the CPU was offline while the driver > + * was getting registered. > + */ > + if (cpu_online(cpu)) > + return; > } > } > > -- Thanks! Reviewed-by: Ionela Voinescu <ionela.voinescu@arm.com> Ionela. > viresh
On 29-06-21, 09:47, Ionela Voinescu wrote: > Okay, makes sense. I have not seen this code actually do anything wrong > so far, and the issues I see on ThunderX2 point more to misbehaving > counters for this purpose. This being said, I would have probably > preferred for this feature to be disabled by default, until we've tested > more, but that won't give the chance to anyone else to test. > > Thanks! > > Reviewed-by: Ionela Voinescu <ionela.voinescu@arm.com> Thanks for understanding Ionela. -- viresh
diff --git a/drivers/cpufreq/Kconfig.arm b/drivers/cpufreq/Kconfig.arm index e65e0a43be64..a5c5f70acfc9 100644 --- a/drivers/cpufreq/Kconfig.arm +++ b/drivers/cpufreq/Kconfig.arm @@ -19,6 +19,16 @@ config ACPI_CPPC_CPUFREQ If in doubt, say N. +config ACPI_CPPC_CPUFREQ_FIE + bool "Frequency Invariance support for CPPC cpufreq driver" + depends on ACPI_CPPC_CPUFREQ && GENERIC_ARCH_TOPOLOGY + default y + help + This extends frequency invariance support in the CPPC cpufreq driver, + by using CPPC delivered and reference performance counters. + + If in doubt, say N. + config ARM_ALLWINNER_SUN50I_CPUFREQ_NVMEM tristate "Allwinner nvmem based SUN50I CPUFreq driver" depends on ARCH_SUNXI diff --git a/drivers/cpufreq/cppc_cpufreq.c b/drivers/cpufreq/cppc_cpufreq.c index 490175d65082..db550fc92931 100644 --- a/drivers/cpufreq/cppc_cpufreq.c +++ b/drivers/cpufreq/cppc_cpufreq.c @@ -10,14 +10,18 @@ #define pr_fmt(fmt) "CPPC Cpufreq:" fmt +#include <linux/arch_topology.h> #include <linux/kernel.h> #include <linux/module.h> #include <linux/delay.h> #include <linux/cpu.h> #include <linux/cpufreq.h> #include <linux/dmi.h> +#include <linux/irq_work.h> +#include <linux/kthread.h> #include <linux/time.h> #include <linux/vmalloc.h> +#include <uapi/linux/sched/types.h> #include <asm/unaligned.h> @@ -57,6 +61,210 @@ static struct cppc_workaround_oem_info wa_info[] = { } }; +#ifdef CONFIG_ACPI_CPPC_CPUFREQ_FIE + +/* Frequency invariance support */ +struct cppc_freq_invariance { + int cpu; + struct irq_work irq_work; + struct kthread_work work; + struct cppc_perf_fb_ctrs prev_perf_fb_ctrs; + struct cppc_cpudata *cpu_data; +}; + +static DEFINE_PER_CPU(struct cppc_freq_invariance, cppc_freq_inv); +static struct kthread_worker *kworker_fie; + +static struct cpufreq_driver cppc_cpufreq_driver; +static unsigned int hisi_cppc_cpufreq_get_rate(unsigned int cpu); +static int cppc_perf_from_fbctrs(struct cppc_cpudata *cpu_data, + struct cppc_perf_fb_ctrs *fb_ctrs_t0, + struct cppc_perf_fb_ctrs *fb_ctrs_t1); + +/** + * cppc_scale_freq_workfn - CPPC arch_freq_scale updater for frequency invariance + * @work: The work item. + * + * The CPPC driver register itself with the topology core to provide its own + * implementation (cppc_scale_freq_tick()) of topology_scale_freq_tick() which + * gets called by the scheduler on every tick. + * + * Note that the arch specific counters have higher priority than CPPC counters, + * if available, though the CPPC driver doesn't need to have any special + * handling for that. + * + * On an invocation of cppc_scale_freq_tick(), we schedule an irq work (since we + * reach here from hard-irq context), which then schedules a normal work item + * and cppc_scale_freq_workfn() updates the per_cpu arch_freq_scale variable + * based on the counter updates since the last tick. + */ +static void cppc_scale_freq_workfn(struct kthread_work *work) +{ + struct cppc_freq_invariance *cppc_fi; + struct cppc_perf_fb_ctrs fb_ctrs = {0}; + struct cppc_cpudata *cpu_data; + unsigned long local_freq_scale; + u64 perf; + + cppc_fi = container_of(work, struct cppc_freq_invariance, work); + cpu_data = cppc_fi->cpu_data; + + if (cppc_get_perf_ctrs(cppc_fi->cpu, &fb_ctrs)) { + pr_warn("%s: failed to read perf counters\n", __func__); + return; + } + + perf = cppc_perf_from_fbctrs(cpu_data, &cppc_fi->prev_perf_fb_ctrs, + &fb_ctrs); + cppc_fi->prev_perf_fb_ctrs = fb_ctrs; + + perf <<= SCHED_CAPACITY_SHIFT; + local_freq_scale = div64_u64(perf, cpu_data->perf_caps.highest_perf); + + /* This can happen due to counter's overflow */ + if (unlikely(local_freq_scale > 1024)) + local_freq_scale = 1024; + + per_cpu(arch_freq_scale, cppc_fi->cpu) = local_freq_scale; +} + +static void cppc_irq_work(struct irq_work *irq_work) +{ + struct cppc_freq_invariance *cppc_fi; + + cppc_fi = container_of(irq_work, struct cppc_freq_invariance, irq_work); + kthread_queue_work(kworker_fie, &cppc_fi->work); +} + +static void cppc_scale_freq_tick(void) +{ + struct cppc_freq_invariance *cppc_fi = &per_cpu(cppc_freq_inv, smp_processor_id()); + + /* + * cppc_get_perf_ctrs() can potentially sleep, call that from the right + * context. + */ + irq_work_queue(&cppc_fi->irq_work); +} + +static struct scale_freq_data cppc_sftd = { + .source = SCALE_FREQ_SOURCE_CPPC, + .set_freq_scale = cppc_scale_freq_tick, +}; + +static void cppc_cpufreq_cpu_fie_init(struct cpufreq_policy *policy) +{ + struct cppc_freq_invariance *cppc_fi; + int cpu, ret; + + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) + return; + + for_each_cpu(cpu, policy->cpus) { + cppc_fi = &per_cpu(cppc_freq_inv, cpu); + cppc_fi->cpu = cpu; + cppc_fi->cpu_data = policy->driver_data; + kthread_init_work(&cppc_fi->work, cppc_scale_freq_workfn); + init_irq_work(&cppc_fi->irq_work, cppc_irq_work); + + ret = cppc_get_perf_ctrs(cpu, &cppc_fi->prev_perf_fb_ctrs); + if (ret) { + pr_warn("%s: failed to read perf counters for cpu:%d: %d\n", + __func__, cpu, ret); + return; + } + } + + /* Register for freq-invariance */ + topology_set_scale_freq_source(&cppc_sftd, policy->cpus); +} + +/* + * We free all the resources on policy's removal and not on CPU removal as the + * irq-work are per-cpu and the hotplug core takes care of flushing the pending + * irq-works (hint: smpcfd_dying_cpu()) on CPU hotplug. Even if the kthread-work + * fires on another CPU after the concerned CPU is removed, it won't harm. + * + * We just need to make sure to remove them all on policy->exit(). + */ +static void cppc_cpufreq_cpu_fie_exit(struct cpufreq_policy *policy) +{ + struct cppc_freq_invariance *cppc_fi; + int cpu; + + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) + return; + + /* policy->cpus will be empty here, use related_cpus instead */ + topology_clear_scale_freq_source(SCALE_FREQ_SOURCE_CPPC, policy->related_cpus); + + for_each_cpu(cpu, policy->related_cpus) { + cppc_fi = &per_cpu(cppc_freq_inv, cpu); + irq_work_sync(&cppc_fi->irq_work); + kthread_cancel_work_sync(&cppc_fi->work); + } +} + +static void __init cppc_freq_invariance_init(void) +{ + struct sched_attr attr = { + .size = sizeof(struct sched_attr), + .sched_policy = SCHED_DEADLINE, + .sched_nice = 0, + .sched_priority = 0, + /* + * Fake (unused) bandwidth; workaround to "fix" + * priority inheritance. + */ + .sched_runtime = 1000000, + .sched_deadline = 10000000, + .sched_period = 10000000, + }; + int ret; + + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) + return; + + kworker_fie = kthread_create_worker(0, "cppc_fie"); + if (IS_ERR(kworker_fie)) + return; + + ret = sched_setattr_nocheck(kworker_fie->task, &attr); + if (ret) { + pr_warn("%s: failed to set SCHED_DEADLINE: %d\n", __func__, + ret); + kthread_destroy_worker(kworker_fie); + return; + } +} + +static void cppc_freq_invariance_exit(void) +{ + if (cppc_cpufreq_driver.get == hisi_cppc_cpufreq_get_rate) + return; + + kthread_destroy_worker(kworker_fie); + kworker_fie = NULL; +} + +#else +static inline void cppc_cpufreq_cpu_fie_init(struct cpufreq_policy *policy) +{ +} + +static inline void cppc_cpufreq_cpu_fie_exit(struct cpufreq_policy *policy) +{ +} + +static inline void cppc_freq_invariance_init(void) +{ +} + +static inline void cppc_freq_invariance_exit(void) +{ +} +#endif /* CONFIG_ACPI_CPPC_CPUFREQ_FIE */ + /* Callback function used to retrieve the max frequency from DMI */ static void cppc_find_dmi_mhz(const struct dmi_header *dm, void *private) { @@ -335,8 +543,10 @@ static int cppc_cpufreq_cpu_init(struct cpufreq_policy *policy) cpu_data->perf_ctrls.desired_perf = caps->highest_perf; ret = cppc_set_perf(cpu, &cpu_data->perf_ctrls); - if (!ret) + if (!ret) { + cppc_cpufreq_cpu_fie_init(policy); return 0; + } pr_debug("Err setting perf value:%d on CPU:%d. ret:%d\n", caps->highest_perf, cpu, ret); @@ -353,6 +563,8 @@ static int cppc_cpufreq_cpu_exit(struct cpufreq_policy *policy) unsigned int cpu = policy->cpu; int ret; + cppc_cpufreq_cpu_fie_exit(policy); + cpu_data->perf_ctrls.desired_perf = caps->lowest_perf; ret = cppc_set_perf(cpu, &cpu_data->perf_ctrls); @@ -372,12 +584,12 @@ static inline u64 get_delta(u64 t1, u64 t0) return (u32)t1 - (u32)t0; } -static int cppc_get_rate_from_fbctrs(struct cppc_cpudata *cpu_data, - struct cppc_perf_fb_ctrs *fb_ctrs_t0, - struct cppc_perf_fb_ctrs *fb_ctrs_t1) +static int cppc_perf_from_fbctrs(struct cppc_cpudata *cpu_data, + struct cppc_perf_fb_ctrs *fb_ctrs_t0, + struct cppc_perf_fb_ctrs *fb_ctrs_t1) { u64 delta_reference, delta_delivered; - u64 reference_perf, delivered_perf; + u64 reference_perf; reference_perf = fb_ctrs_t0->reference_perf; @@ -386,14 +598,11 @@ static int cppc_get_rate_from_fbctrs(struct cppc_cpudata *cpu_data, delta_delivered = get_delta(fb_ctrs_t1->delivered, fb_ctrs_t0->delivered); - /* Check to avoid divide-by zero */ - if (delta_reference || delta_delivered) - delivered_perf = (reference_perf * delta_delivered) / - delta_reference; - else - delivered_perf = cpu_data->perf_ctrls.desired_perf; + /* Check to avoid divide-by zero and invalid delivered_perf */ + if (!delta_reference || !delta_delivered) + return cpu_data->perf_ctrls.desired_perf; - return cppc_cpufreq_perf_to_khz(cpu_data, delivered_perf); + return (reference_perf * delta_delivered) / delta_reference; } static unsigned int cppc_cpufreq_get_rate(unsigned int cpu) @@ -401,6 +610,7 @@ static unsigned int cppc_cpufreq_get_rate(unsigned int cpu) struct cppc_perf_fb_ctrs fb_ctrs_t0 = {0}, fb_ctrs_t1 = {0}; struct cpufreq_policy *policy = cpufreq_cpu_get(cpu); struct cppc_cpudata *cpu_data = policy->driver_data; + u64 delivered_perf; int ret; cpufreq_cpu_put(policy); @@ -415,7 +625,10 @@ static unsigned int cppc_cpufreq_get_rate(unsigned int cpu) if (ret) return ret; - return cppc_get_rate_from_fbctrs(cpu_data, &fb_ctrs_t0, &fb_ctrs_t1); + delivered_perf = cppc_perf_from_fbctrs(cpu_data, &fb_ctrs_t0, + &fb_ctrs_t1); + + return cppc_cpufreq_perf_to_khz(cpu_data, delivered_perf); } static int cppc_cpufreq_set_boost(struct cpufreq_policy *policy, int state) @@ -516,14 +729,21 @@ static void cppc_check_hisi_workaround(void) static int __init cppc_cpufreq_init(void) { + int ret; + if ((acpi_disabled) || !acpi_cpc_valid()) return -ENODEV; INIT_LIST_HEAD(&cpu_data_list); cppc_check_hisi_workaround(); + cppc_freq_invariance_init(); - return cpufreq_register_driver(&cppc_cpufreq_driver); + ret = cpufreq_register_driver(&cppc_cpufreq_driver); + if (ret) + cppc_freq_invariance_exit(); + + return ret; } static inline void free_cpu_data(void) @@ -541,6 +761,7 @@ static inline void free_cpu_data(void) static void __exit cppc_cpufreq_exit(void) { cpufreq_unregister_driver(&cppc_cpufreq_driver); + cppc_freq_invariance_exit(); free_cpu_data(); } diff --git a/include/linux/arch_topology.h b/include/linux/arch_topology.h index 11e555cfaecb..f180240dc95f 100644 --- a/include/linux/arch_topology.h +++ b/include/linux/arch_topology.h @@ -37,6 +37,7 @@ bool topology_scale_freq_invariant(void); enum scale_freq_source { SCALE_FREQ_SOURCE_CPUFREQ = 0, SCALE_FREQ_SOURCE_ARCH, + SCALE_FREQ_SOURCE_CPPC, }; struct scale_freq_data { diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 4ca80df205ce..5226cc26a095 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -6389,6 +6389,7 @@ int sched_setattr_nocheck(struct task_struct *p, const struct sched_attr *attr) { return __sched_setscheduler(p, attr, false, true); } +EXPORT_SYMBOL_GPL(sched_setattr_nocheck); /** * sched_setscheduler_nocheck - change the scheduling policy and/or RT priority of a thread from kernelspace.