kernel.git/drivers/cpufreq/intel_pstate.c, branch linux-6.16.y

PM: EM: Add function for registering a PD without capacity update

2025-09-19T14:37:30Z

commit e0423541477dfb684fbc6e6b5386054bc650f264 upstream. The intel_pstate driver manages CPU capacity changes itself and it does not need an update of the capacity of all CPUs in the system to be carried out after registering a PD. Moreover, in some configurations (for instance, an SMT-capable hybrid x86 system booted with nosmt in the kernel command line) the em_check_capacity_update() call at the end of em_dev_register_perf_domain() always fails and reschedules itself to run once again in 1 s, so effectively it runs in vain every 1 s forever. To address this, introduce a new variant of em_dev_register_perf_domain(), called em_dev_register_pd_no_update(), that does not invoke em_check_capacity_update(), and make intel_pstate use it instead of the original. Fixes: 7b010f9b9061 ("cpufreq: intel_pstate: EAS support for hybrid platforms") Closes: https://lore.kernel.org/linux-pm/40212796-734c-4140-8a85-854f72b8144d@panix.com/ Reported-by: Kenneth R. Crudup Tested-by: Kenneth R. Crudup Cc: 6.16+ # 6.16+ Signed-off-by: Rafael J. Wysocki Signed-off-by: Greg Kroah-Hartman

cpufreq: intel_pstate: Add Granite Rapids support in no-HWP mode

2025-08-20T16:40:58Z

[ Upstream commit fc64e0421598aaa87d61184f6777b52614a095be ] Users may disable HWP in firmware, in which case intel_pstate wouldn't load unless the CPU model is explicitly supported. Signed-off-by: Li RongQing Link: https://patch.msgid.link/20250623105601.3924-1-lirongqing@baidu.com Signed-off-by: Rafael J. Wysocki Signed-off-by: Sasha Levin

cpufreq: intel_pstate: Always use HWP_DESIRED_PERF in passive mode

2025-08-15T14:38:33Z

[ Upstream commit 1cefe495cacba5fb0417da3a75a1a76e3546d176 ] In the passive mode, intel_cpufreq_update_pstate() sets HWP_MIN_PERF in accordance with the target frequency to ensure delivering adequate performance, but it sets HWP_DESIRED_PERF to 0, so the processor has no indication that the desired performance level is actually equal to the floor one. This may cause it to choose a performance point way above the desired level. Moreover, this is inconsistent with intel_cpufreq_adjust_perf() which actually sets HWP_DESIRED_PERF in accordance with the target performance value. Address this by adjusting intel_cpufreq_update_pstate() to pass target_pstate as both the minimum and the desired performance levels to intel_cpufreq_hwp_update(). Fixes: a365ab6b9dfb ("cpufreq: intel_pstate: Implement the ->adjust_perf() callback") Signed-off-by: Rafael J. Wysocki Tested-by: Shashank Balaji Link: https://patch.msgid.link/6173276.lOV4Wx5bFT@rjwysocki.net Signed-off-by: Sasha Levin

Merge tag 'pm-6.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm

2025-05-27T23:48:47Z

Pull power management updates from Rafael Wysocki: "Once again, the changes are dominated by cpufreq updates, but this time the majority of them are cpufreq core changes, mostly related to the introduction of policy locking guards and __free() usage, and fixes related to boost handling. Still, there is also a significant update of the intel_pstate driver making it register an energy model when running on a hybrid platform which is used for enabling energy-aware scheduling (EAS) if the driver operates in the passive mode (and schedutil is used as the cpufreq governor for all CPUs which is the passive mode default). There are some amd-pstate driver updates too, for a good measure, including the "Requested CPU Min frequency" BIOS option support and new online/offline callbacks. In the cpuidle space, the most significant change is the addition of a C1 demotion on/off sysfs knob to intel_idle which should help some users to configure their systems more precisely. There is also the conversion of the PSCI cpuidle driver to a faux device one and there are two small updates of cpuidle governors. Device power management is also modified quite a bit, especially the handling of devices with asynchronous suspend and resume enabled during system transitions. They are now going to be handled more asynchronously during suspend transitions and somewhat less aggressively during resume transitions. Apart from the above, the operating performance points (OPP) library is now going to use mutex locking guards and scope-based cleanup helpers and there is the usual bunch of assorted fixes and code cleanups. Specifics: - Fix potential division-by-zero error in em_compute_costs() (Yaxiong Tian) - Fix typos in energy model documentation and example driver code (Moon Hee Lee, Atul Kumar Pant) - Rearrange the energy model management code and add a new function for adjusting a CPU energy model after adjusting the capacity of the given CPU to it (Rafael Wysocki) - Refactor cpufreq_online(), add and use cpufreq policy locking guards, use __free() in policy reference counting, and clean up core cpufreq code on top of that (Rafael Wysocki) - Fix boost handling on CPU suspend/resume and sysfs updates (Viresh Kumar) - Fix des_perf clamping with max_perf in amd_pstate_update() (Dhananjay Ugwekar) - Add offline, online and suspend callbacks to the amd-pstate driver, rename and use the existing amd_pstate_epp callbacks in it (Dhananjay Ugwekar) - Add support for the "Requested CPU Min frequency" BIOS option to the amd-pstate driver (Dhananjay Ugwekar) - Reset amd-pstate driver mode after running selftests (Swapnil Sapkal) - Avoid shadowing ret in amd_pstate_ut_check_driver() (Nathan Chancellor) - Add helper for governor checks to the schedutil cpufreq governor and move cpufreq-specific EAS checks to cpufreq (Rafael Wysocki) - Populate the cpu_capacity sysfs entries from the intel_pstate driver after registering asym capacity support (Ricardo Neri) - Add support for enabling Energy-aware scheduling (EAS) to the intel_pstate driver when operating in the passive mode on a hybrid platform (Rafael Wysocki) - Drop redundant cpus_read_lock() from store_local_boost() in the cpufreq core (Seyediman Seyedarab) - Replace sscanf() with kstrtouint() in the cpufreq code and use a symbol instead of a raw number in it (Bowen Yu) - Add support for autonomous CPU performance state selection to the CPPC cpufreq driver (Lifeng Zheng) - OPP: Add dev_pm_opp_set_level() (Praveen Talari) - Introduce scope-based cleanup headers and mutex locking guards in OPP core (Viresh Kumar) - Switch OPP to use kmemdup_array() (Zhang Enpei) - Optimize bucket assignment when next_timer_ns equals KTIME_MAX in the menu cpuidle governor (Zhongqiu Han) - Convert the cpuidle PSCI driver to a faux device one (Sudeep Holla) - Add C1 demotion on/off sysfs knob to the intel_idle driver (Artem Bityutskiy) - Fix typos in two comments in the teo cpuidle governor (Atul Kumar Pant) - Fix denying of auto suspend in pm_suspend_timer_fn() (Charan Teja Kalla) - Move debug runtime PM attributes to runtime_attrs[] (Rafael Wysocki) - Add new devm_ functions for enabling runtime PM and runtime PM reference counting (Bence Csókás) - Remove size arguments from strscpy() calls in the hibernation core code (Thorsten Blum) - Adjust the handling of devices with asynchronous suspend enabled during system suspend and resume to start resuming them immediately after resuming their parents and to start suspending such a device immediately after suspending its first child (Rafael Wysocki) - Adjust messages printed during tasks freezing to avoid using pr_cont() (Andrew Sayers, Paul Menzel) - Clean up unnecessary usage of !! in pm_print_times_init() (Zihuan Zhang) - Add missing wakeup source attribute relax_count to sysfs and remove the space character at the end ofi the string produced by pm_show_wakelocks() (Zijun Hu) - Add configurable pm_test delay for hibernation (Zihuan Zhang) - Disable asynchronous suspend in ucsi_ccg_probe() to prevent the cypd4226 device on Tegra boards from suspending prematurely (Jon Hunter) - Unbreak printing PM debug messages during hibernation and clean up some related code (Rafael Wysocki) - Add a systemd service to run cpupower and change cpupower binding's Makefile to use -lcpupower (John B. Wyatt IV, Francesco Poli)" * tag 'pm-6.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (72 commits) cpufreq: CPPC: Add support for autonomous selection cpufreq: Update sscanf() to kstrtouint() cpufreq: Replace magic number OPP: switch to use kmemdup_array() PM: freezer: Rewrite restarting tasks log to remove stray *done.* PM: runtime: fix denying of auto suspend in pm_suspend_timer_fn() cpufreq: drop redundant cpus_read_lock() from store_local_boost() cpupower: do not install files to /etc/default/ cpupower: do not call systemctl at install time cpupower: do not write DESTDIR to cpupower.service PM: sleep: Introduce pm_sleep_transition_in_progress() cpufreq/amd-pstate: Avoid shadowing ret in amd_pstate_ut_check_driver() cpufreq: intel_pstate: Document hybrid processor support cpufreq: intel_pstate: EAS: Increase cost for CPUs using L3 cache cpufreq: intel_pstate: EAS support for hybrid platforms PM: EM: Introduce em_adjust_cpu_capacity() PM: EM: Move CPU capacity check to em_adjust_new_capacity() PM: EM: Documentation: Fix typos in example driver code cpufreq: Drop policy locking from cpufreq_policy_is_good_for_eas() PM: sleep: Introduce pm_suspend_in_progress() ...

cpufreq: intel_pstate: EAS: Increase cost for CPUs using L3 cache

2025-05-13T12:35:25Z

On some hybrid platforms some efficient CPUs (E-cores) are not connected to the L3 cache, but there are no other differences between them and the other E-cores that use L3. In that case, it is generally more efficient to run "light" workloads on the E-cores that do not use L3 and allow all of the cores using L3, including P-cores, to go into idle states. For this reason, slightly increase the cost for all CPUs sharing the L3 cache to make EAS prefer CPUs that do not use it to the other CPUs of the same type (if any). Signed-off-by: Rafael J. Wysocki Link: https://patch.msgid.link/2032776.usQuhbGJ8B@rjwysocki.net

cpufreq: intel_pstate: EAS support for hybrid platforms

2025-05-13T12:35:25Z

Modify intel_pstate to register EM perf domains for CPUs on hybrid platforms without SMT which causes EAS to be enabled on them when schedutil is used as the cpufreq governor (which requires intel_pstate to operate in the passive mode). This change is targeting platforms (for example, Lunar Lake) where the "little" CPUs (E-cores) are always more energy-efficient than the "big" or "performance" CPUs (P-cores) when run at the same HWP performance level, so it is sufficient to tell EAS that E-cores are always preferred (so long as there is enough spare capacity on one of them to run the given task). However, migrating tasks between CPUs of the same type too often is not desirable because it may hurt both performance and energy efficiency due to leaving warm caches behind. For this reason, register a separate perf domain for each CPU and choose the cost values for them so that the cost mostly depends on the CPU type, but there is also a small component of it depending on the performance level (utilization) which helps to balance the load between CPUs of the same type. The cost component related to the CPU type is computed with the help of the observation that the IPC metric value for a given CPU is inversely proportional to its performance-to-frequency scaling factor and the cost of running code on it can be assumed to be roughly proportional to that IPC ratio (in principle, the higher the IPC ratio, the more resources are utilized when running at a given frequency, so the cost should be higher). For all CPUs that are online at the system initialization time, EM perf domains are registered when the driver starts up, after asymmetric capacity support has been enabled. For the CPUs that become online later, EM perf domains are registered after setting the asymmetric capacity for them. Signed-off-by: Rafael J. Wysocki Tested-by: Christian Loehle Reviewed-by: Dietmar Eggemann Link: https://patch.msgid.link/6057101.MhkbZ0Pkbq@rjwysocki.net

cpufreq: intel_pstate: Populate the cpu_capacity sysfs entries

2025-05-07T19:56:55Z

Intel hybrid processors have CPUs of different capacity. Populate the interface /sys/devices/system/cpu/cpuN/cpu_capacity. This interface uses the per-CPU variable `cpu_scale`. On x86 this variable has no other use besides feeding the sysfs entries. Initialize it when setting CPU capacity for the scheduler and scale-invariant code. Feed it with arch_scale_cpu_capacity() as it gives capacity normalized to the interval [0, SCHED_CAPACITY_SCALE]. Signed-off-by: Ricardo Neri Signed-off-by: Rafael J. Wysocki

Merge tag 'v6.15-rc5' into x86/msr, to pick up fixes and to resolve conflicts

2025-05-06T17:42:00Z

Conflicts: drivers/cpufreq/intel_pstate.c Signed-off-by: Ingo Molnar

Merge back cpufreq material for 6.16

2025-05-04T10:35:42Z

Merge tag 'v6.15-rc4' into x86/msr, to pick up fixes and resolve conflicts

2025-05-02T07:43:44Z

Signed-off-by: Ingo Molnar