Occasionally, the exit latency of the idle state selected by the menu
governor may exceed the PM QoS CPU wakeup latency limit.  Namely, if the
scheduler tick has been stopped already and predicted_ns is greater than
the tick period length, the governor may return an idle state whose exit
latency exceeds latency_req because that decision is made before
checking the current idle state's exit latency.

For instance, say that there are 3 idle states, 0, 1, and 2.  For idle
states 0 and 1, the exit latency is equal to the target residency and
the values are 0 and 5 us, respectively.  State 2 is deeper and has the
exit latency and target residency of 200 us and 2 ms (which is greater
than the tick period length), respectively.

Say that predicted_ns is equal to TICK_NSEC and the PM QoS latency
limit is 20 us.  After the first two iterations of the main loop in
menu_select(), idx becomes 1 and in the third iteration of it the target
residency of the current state (state 2) is greater than predicted_ns.
State 2 is not a polling one and predicted_ns is not less than TICK_NSEC,
so the check on whether or not the tick has been stopped is done.  Say
that the tick has been stopped already and there are no imminent timers
(that is, delta_tick is greater than the target residency of state 2).
In that case, idx becomes 2 and it is returned immediately, but the exit
latency of state 2 exceeds the latency limit.

Address this issue by modifying the code to compare the exit latency of
the current idle state (idle state i) with the latency limit before
comparing its target residency with predicted_ns, which allows one
more exit_latency_ns check that becomes redundant to be dropped.

However, after the above change, latency_req cannot take the predicted_ns
value any more, which takes place after commit 38f83090f515 ("cpuidle:
menu: Remove iowait influence"), because it may cause a polling state
to be returned prematurely.

In the context of the previous example say that predicted_ns is 3000 and
the PM QoS latency limit is still 20 us.  Additionally, say that idle
state 0 is a polling one.  Moving the exit_latency_ns check before the
target_residency_ns one causes the loop to terminate in the second
iteration, before the target_residency_ns check, so idle state 0 will be
returned even though previously state 1 would be returned if there were
no imminent timers.

For this reason, remove the assignment of the predicted_ns value to
latency_req from the code.

Fixes: 5ef499cd571c ("cpuidle: menu: Handle stopped tick more aggressively")
Cc: 4.17+ <stable@vger.kernel.org> # 4.17+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Christian Loehle <christian.loehle@arm.com>
Link: https://patch.msgid.link/5043159.31r3eYUQgx@rafael.j.wysocki

Marc has reported that commit 85975daeaa4d ("cpuidle: menu: Avoid
discarding useful information") caused the number of wakeup interrupts
to increase on an idle system [1], which was not expected to happen
after merely allowing shallower idle states to be selected by the
governor in some cases.

However, on the system in question, all of the idle states deeper than
WFI are rejected by the driver due to a firmware issue [2].  This causes
the governor to only consider the recent interval duriation data
corresponding to attempts to enter WFI that are successful and the
recent invervals table is filled with values lower than the scheduler
tick period.  Consequently, the governor predicts an idle duration
below the scheduler tick period length and avoids stopping the tick
more often which leads to the observed symptom.

Address it by modifying the governor to update the recent intervals
table also when entering the previously selected idle state fails, so
it knows that the short idle intervals might have been the minority
had the selected idle states been actually entered every time.

Fixes: 85975daeaa4d ("cpuidle: menu: Avoid discarding useful information")
Link: https://lore.kernel.org/linux-pm/86o6sv6n94.wl-maz@kernel.org/ [1]
Link: https://lore.kernel.org/linux-pm/7ffcb716-9a1b-48c2-aaa4-469d0df7c792@arm.com/ [2]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Christian Loehle <christian.loehle@arm.com>
Tested-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Christian Loehle <christian.loehle@arm.com>
Link: https://patch.msgid.link/2793874.mvXUDI8C0e@rafael.j.wysocki

Pull pmdomain updates from Ulf Hansson:
 "pmdomain core:
   - Leave powered-on genpds on until ->sync_state() or late_initcall_sync
   - Export a common ->sync_state() helper for genpd providers
   - Add generic ->sync_state() support
   - Add a bus/driver for genpd provider-devices
   - Introduce dev_pm_genpd_is_on() for consumers

  pmdomain providers:
   - cpuidle-psci: Drop redundant ->sync_state() support
   - cpuidle-riscv-sbi: Drop redundant ->sync_state() support
   - imx: Set ISI panic write for imx8m-blk-ctrl
   - qcom: Add support for Glymur and Milos RPMh power-domains
   - qcom: Use of_genpd_sync_state() for power-domains
   - rockchip: Add support for the RK3528 variant
   - samsung: Fix splash-screen handover by enforcing a ->sync_state()
   - sunxi: Add support for Allwinner A523's PCK600 power-controller
   - tegra: Opt-out from genpd's common ->sync_state() support for pmc
   - thead: Instantiate a GPU power sequencer via the auxiliary bus
   - renesas: Move init to postcore_initcalls
   - xilinx: Move ->sync_state() support to firmware driver
   - xilinx: Use of_genpd_sync_state() for power-domains

  pmdomain consumers:
   - remoteproc: imx_rproc: Fixup the detect/attach procedure for
     pre-booted cores"

* tag 'pmdomain-v6.17' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/linux-pm: (44 commits)
  pmdomain: qcom: rpmhpd: Add Glymur RPMh Power Domains
  dt-bindings: power: rpmpd: Add Glymur power domains
  remoteproc: imx_rproc: detect and attach to pre-booted remote cores
  remoteproc: imx_rproc: skip clock enable when M-core is managed by the SCU
  pmdomain: core: introduce dev_pm_genpd_is_on()
  pmdomain: ti: Select PM_GENERIC_DOMAINS
  pmdomain: sunxi: sun20i-ppu: change to tristate and enable for ARCH_SUNXI
  pmdomain: sunxi: add driver for Allwinner A523's PCK-600 power controller
  pmdomain: sunxi: sun20i-ppu: add A523 support
  pmdomain: samsung: Fix splash-screen handover by enforcing a sync_state
  cpuidle: riscv-sbi: Drop redundant sync_state support
  cpuidle: psci: Drop redundant sync_state support
  pmdomain: core: Leave powered-on genpds on until sync_state
  pmdomain: core: Leave powered-on genpds on until late_initcall_sync
  pmdomain: core: Default to use of_genpd_sync_state() for genpd providers
  driver core: Add dev_set_drv_sync_state()
  pmdomain: core: Add common ->sync_state() support for genpd providers
  driver core: Export get_dev_from_fwnode()
  firmware: xilinx: Use of_genpd_sync_state()
  firmware: xilinx: Don't share zynqmp_pm_init_finalize()
  ...

Pull power management updates from Rafael Wysocki:
 "As is tradition, cpufreq is the part with the largest number of
  updates that include core fixes and cleanups as well as updates of
  several assorted drivers, but there are also quite a few updates
  related to system sleep, mostly focused on asynchronous suspend and
  resume of devices and on making the integration of system suspend
  and resume with runtime PM easier.

  Runtime PM is also updated to allow some code duplication in drivers
  to be eliminated going forward and to work more consistently overall
  in some cases.

  Apart from that, there are some driver core updates related to PM
  domains that should help to address ordering issues with devm_ cleanup
  routines relying on PM domains, some assorted devfreq updates
  including core fixes and cleanups, tooling updates, and documentation
  and MAINTAINERS updates.

  Specifics:

   - Fix two initialization ordering issues in the cpufreq core and a
     governor initialization error path in it, and clean it up (Lifeng
     Zheng)

   - Add Granite Rapids support in no-HWP mode to the intel_pstate
     cpufreq driver (Li RongQing)

   - Make intel_pstate always use HWP_DESIRED_PERF when operating in the
     passive mode (Rafael Wysocki)

   - Allow building the tegra124 cpufreq driver as a module (Aaron
     Kling)

   - Do minor cleanups for Rust cpufreq and cpumask APIs and fix
     MAINTAINERS entry for cpu.rs (Abhinav Ananthu, Ritvik Gupta, Lukas
     Bulwahn)

   - Clean up assorted cpufreq drivers (Arnd Bergmann, Dan Carpenter,
     Krzysztof Kozlowski, Sven Peter, Svyatoslav Ryhel, Lifeng Zheng)

   - Add the NEED_UPDATE_LIMITS flag to the CPPC cpufreq driver
     (Prashant Malani)

   - Fix minimum performance state label error in the amd-pstate driver
     documentation (Shouye Liu)

   - Add the CPUFREQ_GOV_STRICT_TARGET flag to the userspace cpufreq
     governor and explain HW coordination influence on it in the
     documentation (Shashank Balaji)

   - Fix opencoded for_each_cpu() in idle_state_valid() in the DT
     cpuidle driver (Yury Norov)

   - Remove info about non-existing QoS interfaces from the PM QoS
     documentation (Ulf Hansson)

   - Use c_* types via kernel prelude in Rust for OPP (Abhinav Ananthu)

   - Add HiSilicon uncore frequency scaling driver to devfreq (Jie Zhan)

   - Allow devfreq drivers to add custom sysfs ABIs (Jie Zhan)

   - Simplify the sun8i-a33-mbus devfreq driver by using more devm
     functions (Uwe Kleine-König)

   - Fix an index typo in trans_stat() in devfreq (Chanwoo Choi)

   - Check devfreq governor before using governor->name (Lifeng Zheng)

   - Remove a redundant devfreq_get_freq_range() call from
     devfreq_add_device() (Lifeng Zheng)

   - Limit max_freq with scaling_min_freq in devfreq (Lifeng Zheng)

   - Replace sscanf() with kstrtoul() in set_freq_store() (Lifeng Zheng)

   - Extend the asynchronous suspend and resume of devices to handle
     suppliers like parents and consumers like children (Rafael Wysocki)

   - Make pm_runtime_force_resume() work for drivers that set the
     DPM_FLAG_SMART_SUSPEND flag and allow PCI drivers and drivers that
     collaborate with the general ACPI PM domain to set it (Rafael
     Wysocki)

   - Add kernel parameter to disable asynchronous suspend/resume of
     devices (Tudor Ambarus)

   - Drop redundant might_sleep() calls from some functions in the
     device suspend/resume core code (Zhongqiu Han)

   - Fix the handling of monitors connected right before waking up the
     system from sleep (tuhaowen)

   - Clean up MAINTAINERS entries for suspend and hibernation (Rafael
     Wysocki)

   - Fix error code path in the KEXEC_JUMP flow and drop a redundant
     pm_restore_gfp_mask() call from it (Rafael Wysocki)

   - Rearrange suspend/resume error handling in the core device suspend
     and resume code (Rafael Wysocki)

   - Fix up white space that does not follow coding style in the
     hibernation core code (Darshan Rathod)

   - Document return values of suspend-related API functions in the
     runtime PM framework (Sakari Ailus)

   - Mark last busy stamp in multiple autosuspend-related functions in
     the runtime PM framework and update its documentation (Sakari
     Ailus)

   - Take active children into account in pm_runtime_get_if_in_use() for
     consistency (Rafael Wysocki)

   - Fix NULL pointer dereference in get_pd_power_uw() in the dtpm_cpu
     power capping driver (Sivan Zohar-Kotzer)

   - Add support for the Bartlett Lake platform to the Intel RAPL power
     capping driver (Qiao Wei)

   - Add PL4 support for Panther Lake to the intel_rapl_msr power
     capping driver (Zhang Rui)

   - Update contact information in the PM ABI docs and maintainer
     information in the power domains DT binding (Rafael Wysocki)

   - Update PM header inclusions to follow the IWYU (Include What You
     Use) principle (Andy Shevchenko)

   - Add flags to specify power on attach/detach for PM domains, make
     the driver core detach PM domains in device_unbind_cleanup(), and
     drop the dev_pm_domain_detach() call from the platform bus type
     (Claudiu Beznea)

   - Improve Python binding's Makefile for cpupower (John B. Wyatt IV)

   - Fix printing of CORE, CPU fields in cpupower-monitor (Gautham
     Shenoy)"

* tag 'pm-6.17-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (75 commits)
  cpufreq: CPPC: Mark driver with NEED_UPDATE_LIMITS flag
  PM: docs: Use my kernel.org address in ABI docs and DT bindings
  PM: hibernate: Fix up white space that does not follow coding style
  PM: sleep: Rearrange suspend/resume error handling in the core
  Documentation: amd-pstate:fix minimum performance state label error
  PM: runtime: Take active children into account in pm_runtime_get_if_in_use()
  kexec_core: Drop redundant pm_restore_gfp_mask() call
  kexec_core: Fix error code path in the KEXEC_JUMP flow
  PM: sleep: Clean up MAINTAINERS entries for suspend and hibernation
  drivers: cpufreq: add Tegra114 support
  rust: cpumask: Replace `MaybeUninit` and `mem::zeroed` with `Opaque` APIs
  cpufreq: Exit governor when failed to start old governor
  cpufreq: Move the check of cpufreq_driver->get into cpufreq_verify_current_freq()
  cpufreq: Init policy->rwsem before it may be possibly used
  cpufreq: Initialize cpufreq-based frequency-invariance later
  cpufreq: Remove duplicate check in __cpufreq_offline()
  cpufreq: Contain scaling_cur_freq.attr in cpufreq_attrs
  cpufreq: intel_pstate: Add Granite Rapids support in no-HWP mode
  cpufreq: intel_pstate: Always use HWP_DESIRED_PERF in passive mode
  PM / devfreq: Add HiSilicon uncore frequency scaling driver
  ...

Merge the pmdomain fixes for v6.16-rc[n] into the next branch, to allow
them to get tested together with the new changes that are targeted for
v6.17.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

Currently cpu hotplug with the PREEMPT_RT option set in the kernel is
not supported because the underlying generic power domain functions
used in the cpu hotplug callbacks are incompatible from a lock point
of view. This situation prevents the suspend to idle to reach the
deepest idle state for the "cluster" as identified in the
undermentioned commit.

Use the compatible ones when PREEMPT_RT is enabled and remove the
boolean disabling the hotplug callbacks with this option.

With this change the platform can reach the deepest idle state
allowing at suspend time to consume less power.

Tested-on Lenovo T14s with the following script:

echo 0 > /sys/devices/system/cpu/cpu3/online
BEFORE=$(cat /sys/kernel/debug/pm_genpd/power-domain-cpu-cluster0/idle_states | grep S0 | awk '{ print $3 }') ;
rtcwake -s 1 -m mem;
AFTER=$(cat /sys/kernel/debug/pm_genpd/power-domain-cpu-cluster0/idle_states | grep S0 | awk '{ print $3 }');
if [ $BEFORE -lt $AFTER ]; then
    echo "Test successful"
else
    echo "Test failed"
fi
echo 1 > /sys/devices/system/cpu/cpu3/online

Fixes: 1c4b2932bd62 ("cpuidle: psci: Enable the hierarchical topology for s2idle on PREEMPT_RT")
Cc: Raghavendra Kakarla <quic_rkakarla@quicinc.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20250709154728.733920-1-daniel.lezcano@linaro.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

The function opencodes the for_each_cpu_from() by using an open for-loop.
Fix that in sake of readability.

While there, drop the 'valid' variable as it's pretty useless here.

Signed-off-by: Yury Norov [NVIDIA] <yury.norov@gmail.com>
Link: https://patch.msgid.link/20250604213908.27819-1-yury.norov@gmail.com
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

The recent updates to the genpd core, can entirely manage the sync_state
support for the cpuidle-riscv-sbi-domain. More precisely, genpd prevents
our ->power_off() callback from being invoked, until all of our consumers
are ready for it.

Let's therefore drop the sync_state support for the
cpuidle-riscv-sbi-domain as it has become redundant.

Cc: Anup Patel <anup@brainfault.org>
Cc: linux-riscv@lists.infradead.org
Tested-by: Hiago De Franco <hiago.franco@toradex.com> # Colibri iMX8X
Tested-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> # TI AM62A,Xilinx ZynqMP ZCU106
Reviewed-by: Rahul Pathak <rpathak@ventanamicro.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Link: https://lore.kernel.org/r/20250701114733.636510-25-ulf.hansson@linaro.org

The recent updates to the genpd core, can entirely manage the sync_state
support for the cpuidle-psci-domain. More precisely, genpd prevents our
->power_off() callback from being invoked, until all of our consumers are
ready for it.

Let's therefore drop the sync_state support for the cpuidle-psci-domain as
it has become redundant.

Tested-by: Hiago De Franco <hiago.franco@toradex.com> # Colibri iMX8X
Tested-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> # TI AM62A,Xilinx ZynqMP ZCU106
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Link: https://lore.kernel.org/r/20250701114733.636510-24-ulf.hansson@linaro.org

The riscv-sbi-domain implements its own specific ->sync_state() callback.
Let's set the GENPD_FLAG_NO_SYNC_STATE to inform genpd about it.

Moreover, let's call of_genpd_sync_state() to make sure genpd tries to
power off unused PM domains.

Cc: Anup Patel <anup@brainfault.org>
Cc: linux-riscv@lists.infradead.org
Tested-by: Hiago De Franco <hiago.franco@toradex.com> # Colibri iMX8X
Tested-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> # TI AM62A,Xilinx ZynqMP ZCU106
Reviewed-by: Anup Patel <anup@brainfault.org>
Reviewed-by: Rahul Pathak <rpathak@ventanamicro.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Link: https://lore.kernel.org/r/20250701114733.636510-12-ulf.hansson@linaro.org