Update EPP default for balance_performance to 32. This will give better performance out of the box using Intel P-State powersave governor while still offering power savings compared to performance governor. This is in line with what has already been done for Emerald Rapids and Sapphire Rapids. Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Link: https://patch.msgid.link/20241112235946.368082-1-srinivas.pandruvada@linux.intel.com [ rjw: Subject and changelog edits ] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Notice that hybrid_init_cpu_capacity_scaling() only needs to hold
hybrid_capacity_lock around __hybrid_init_cpu_capacity_scaling()
calls, so introduce a "locked" wrapper around the latter and call
it from the former. This allows to drop a local variable and a
label that are not needed any more.
Also, rename __hybrid_init_cpu_capacity_scaling() to
__hybrid_refresh_cpu_capacity_scaling() for consistency.
Interestingly enough, this fixes a locking issue introduced by commit
929ebc93ccaa ("cpufreq: intel_pstate: Set asymmetric CPU capacity on
hybrid systems") that put an arch_enable_hybrid_capacity_scale() call
under hybrid_capacity_lock, which was a mistake because the latter is
acquired in CPU hotplug paths and so it cannot be held around
cpus_read_lock() calls.
Link: https://lore.kernel.org/linux-pm/SJ1PR11MB6129EDBF22F8A90FC3A3EDC8B9582@SJ1PR11MB6129.namprd11.prod.outlook.com/
Fixes: 929ebc93ccaa ("cpufreq: intel_pstate: Set asymmetric CPU capacity on hybrid systems")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reported-by: "Borah, Chaitanya Kumar" <chaitanya.kumar.borah@intel.com>
Link: https://patch.msgid.link/12554508.O9o76ZdvQC@rjwysocki.net
[ rjw: Changelog update ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit 929ebc93ccaa ("cpufreq: intel_pstate: Set asymmetric CPU
capacity on hybrid systems") overlooked a corner case in which some
CPUs may be offline to start with and brought back online later,
after the intel_pstate driver has been registered, so their asymmetric
capacity will not be set.
Address this by calling hybrid_update_capacity() in the CPU
initialization path that is executed instead of the online path
for those CPUs.
Note that this asymmetric capacity update will be skipped during
driver initialization and mode switches because hybrid_max_perf_cpu
is NULL in those cases.
Fixes: 929ebc93ccaa ("cpufreq: intel_pstate: Set asymmetric CPU capacity on hybrid systems")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://patch.msgid.link/1913414.tdWV9SEqCh@rjwysocki.net
Modify intel_pstate_register_driver() to clear hybrid_max_perf_cpu before calling cpufreq_register_driver(), so that asymmetric CPU capacity scaling is not updated until hybrid_init_cpu_capacity_scaling() runs down the road. This is done in preparation for a subsequent change adding asymmetric CPU capacity computation to the CPU init path to handle CPUs that are initially offline. The information on whether or not hybrid_max_perf_cpu was NULL before it has been cleared is passed to hybrid_init_cpu_capacity_scaling(), so full initialization of CPU capacity scaling can be skipped if it has been carried out already. No intentional functional impact. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Link: https://patch.msgid.link/4616631.LvFx2qVVIh@rjwysocki.net
notify_hwp_interrupt() is called via sysvec_thermal() -> smp_thermal_vector() -> intel_thermal_interrupt() in hard irq context. For this reason it must not use a simple spin_lock that sleeps with PREEMPT_RT enabled. So convert it to a raw spinlock. Reported-by: xiao sheng wen <atzlinux@sina.com> Link: https://bugs.debian.org/1076483 Signed-off-by: Uwe Kleine-König <ukleinek@debian.org> Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: xiao sheng wen <atzlinux@sina.com> Link: https://patch.msgid.link/20240919081121.10784-2-ukleinek@debian.org Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Make intel_pstate use the HWP_HIGHEST_PERF values from MSR_HWP_CAPABILITIES to set asymmetric CPU capacity information via the previously introduced arch_set_cpu_capacity() on hybrid systems without SMT. Setting asymmetric CPU capacity is generally necessary to allow the scheduler to compute task sizes in a consistent way across all CPUs in a system where they differ by capacity. That, in turn, should help to improve scheduling decisions. It is also necessary for the schedutil cpufreq governor to operate as expected on hybrid systems where tasks migrate between CPUs of different capacities. The underlying observation is that intel_pstate already uses MSR_HWP_CAPABILITIES to get CPU performance information which is exposed by it via sysfs and CPU performance scaling is based on it. Thus using this information for setting asymmetric CPU capacity is consistent with what the driver has been doing already. Moreover, HWP_HIGHEST_PERF reflects the maximum capacity of a given CPU including both the instructions-per-cycle (IPC) factor and the maximum turbo frequency and the units in which that value is expressed are the same for all CPUs in the system, so the maximum capacity ratio between two CPUs can be obtained by computing the ratio of their HWP_HIGHEST_PERF values. Of course, in principle that capacity ratio need not be directly applicable at lower frequencies, so using it for providing the asymmetric CPU capacity information to the scheduler is a rough approximation, but it is as good as it gets. Also, measurements indicate that this approximation is not too bad in practice. If the given system is hybrid and non-SMT, the new code disables ITMT support in the scheduler (because it may get in the way of asymmetric CPU capacity code in the scheduler that automatically gets enabled by setting asymmetric CPU capacity) after initializing all online CPUs and finds the one with the maximum HWP_HIGHEST_PERF value. Next, it computes the capacity number for each (online) CPU by dividing the product of its HWP_HIGHEST_PERF and SCHED_CAPACITY_SCALE by the maximum HWP_HIGHEST_PERF. When a CPU goes offline, its capacity is reset to SCHED_CAPACITY_SCALE and if it is the one with the maximum HWP_HIGHEST_PERF value, the capacity numbers for all of the other online CPUs are recomputed. This also takes care of a cleanup during driver operation mode changes. Analogously, when a new CPU goes online, its capacity number is updated and if its HWP_HIGHEST_PERF value is greater than the current maximum one, the capacity numbers for all of the other online CPUs are recomputed. The case when the driver is notified of a CPU capacity change, either through the HWP interrupt or through an ACPI notification, is handled similarly to the CPU online case above, except that if the target CPU is the current highest-capacity one and its capacity is reduced, the capacity numbers for all of the other online CPUs need to be recomputed either. If the driver's "no_trubo" sysfs attribute is updated, all of the CPU capacity information is computed from scratch to reflect the new turbo status. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reviewed-by: Ricardo Neri <ricardo.neri-calderon@linux.intel.com> Tested-by: Ricardo Neri <ricardo.neri-calderon@linux.intel.com> # scale invariance Link: https://patch.msgid.link/1979653.PYKUYFuaPT@rjwysocki.net [ rjw: Fixed a typo in the changelog ] [ rjw: Renamed 3 new functions and added a comment ] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Prevent intel_pstate from loading when OOB (Out Of Band) P-states mode is enabled. The OOB identifying bits are same as for the prior generation CPUs like Emerald Rapids servers. Add Granite Rapids and Sierra Forest CPU models to intel_pstate_cpu_oob_ids[]. Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Link: https://patch.msgid.link/20240802184839.1909091-1-srinivas.pandruvada@linux.intel.com Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
On Intel Emerald Rapids machines, we ship the Energy Performance Preference (EPP) default for balance_performance as 128. However, during an internal investigation together with Intel, we have determined that 32 is a more suitable value. This leads to significant improvements in both performance and energy: POV-Ray: 32% faster | 12% less energy OpenSSL: 12% faster | energy within 1% Build Linux Kernel: 29% faster | 18% less energy Therefore, we should move the default EPP for balance_performance to 32. This is in line with what has already been done for Sapphire Rapids. Signed-off-by: Pedro Henrique Kopper <pedro.kopper@canonical.com> Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Link: https://patch.msgid.link/Zqu6zjVMoiXwROBI@capivara Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Merge ARM cpufreq updates for 6.11 from Viresh Kumar: "- cpufreq: Add Loongson-3 CPUFreq driver support (Huacai Chen). - Make exit() callback return void (Lizhe and Viresh Kumar). - Minor cleanups and fixes in several drivers (Bryan Brattlof, Javier Carrasco, Jagadeesh Kona, Jeff Johnson, Nícolas F. R. A. Prado, Primoz Fiser, Raphael Gallais-Pou, and Riwen Lu)." * tag 'cpufreq-arm-updates-6.11' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/vireshk/pm: (21 commits) cpufreq: sti: fix build warning cpufreq: mediatek: Use dev_err_probe in every error path in probe cpufreq: Add Loongson-3 CPUFreq driver support cpufreq: Make cpufreq_driver->exit() return void cpufreq: pcc: Remove empty exit() callback cpufreq: loongson2: Remove empty exit() callback cpufreq: nforce2: Remove empty exit() callback cpufreq: sti: add missing MODULE_DEVICE_TABLE entry for stih418 cpufreq: ti: update OPP table for AM62Px SoCs cpufreq: ti: update OPP table for AM62Ax SoCs cpufreq: sun50i: add Allwinner H700 speed bin cpufreq/cppc: Don't compare desired_perf in target() OPP: ti: Fix ti_opp_supply_probe wrong return values cpufreq: ti-cpufreq: Handle deferred probe with dev_err_probe() cpufreq: dt-platdev: add missing MODULE_DESCRIPTION() macro cpufreq: longhaul: Fix kernel-doc param for longhaul_setstate cpufreq: qcom-nvmem: eliminate uses of of_node_put() cpufreq: qcom-nvmem: fix memory leaks in probe error paths cpufreq: scmi: Avoid overflow of target_freq in fast switch cpufreq: sun50i: replace of_node_put() with automatic cleanup handler ...