The specific rounding adds conditionally only 1/256 to fractional
part of core_pct.
We can safely remove it without any noticeable impact in
calculations.
Use div64_u64 instead of div_u64 to avoid possible overflow of
sample->mperf as divisor
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Simplify the code by removing the inline functions pstate_increase and
pstate_decrease and use directly the intel_pstate_set_pstate.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Acked-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Currently we shift right aperf and mperf variables by FRAC_BITS
to prevent overflow when we convert them to fix point numbers
(shift left by FRAC_BITS).
But this is not necessary, because we actually use delta aperf and mperf
which are much less than APERF and MPERF values.
So, use the unmodified APERF and MPERF values in calculation.
This also adds 8 bits in precision, although the gain is insignificant.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
According to Intel 64 and IA-32 Architectures SDM, Volume 3,
Chapter 14.2, "Software needs to exercise care to avoid delays
between the two RDMSRs (for example interrupts)".
So, disable interrupts during reading MSRs IA32_APERF and IA32_MPERF.
This should increase the accuracy of the calculations.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Suppress checkpatch.pl --strict warnings:
CHECK: Alignment should match open parenthesis
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Remove the unnecessary intermediate assignment and use directly the
pid_params.sample_rate_ms variable.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Remove unnecessary parentheses.
Also, add parentheses in one case for better readability.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
We can fit these lines in a single one, under the 80 characters
limit.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
div_s64() accepts the divisor parameter as s32. Helper div_fp()
also accepts divisor as int32_t.
So, remove the unnecessary int64_t type casting.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Since we never remove sysfs entry and debugfs files, we can make
the intel_pstate_kobject and debugfs_parent locals.
Also, annotate with __init intel_pstate_sysfs_expose_params()
and intel_pstate_debug_expose_params() in order to be freed
after bootstrap.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Ensure that cpu->cpu is set before writing MSR_IA32_PERF_CTL during CPU
initialization. Otherwise only cpu0 has its P-state set and all other
cores are left with their values unchanged.
In most cases, this is not too serious because the P-states will be set
correctly when the timer function is run. But when the default governor
is set to performance, the per-CPU current_pstate stays the same forever
and no attempts are made to write the MSRs again.
Signed-off-by: Vincent Minet <vincent@vincent-minet.net>
Cc: All applicable <stable@vger.kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
If turbo is disabled in the BIOS bit 38 should be set in
MSR_IA32_MISC_ENABLE register per section 14.3.2.1 of the SDM Vol 3
document 325384-050US Feb 2014. If this bit is set do *not* attempt
to disable trubo via the MSR_IA32_PERF_CTL register. On some systems
trying to disable turbo via MSR_IA32_PERF_CTL will cause subsequent
writes to MSR_IA32_PERF_CTL not take affect, in fact reading
MSR_IA32_PERF_CTL will not show the IDA/Turbo DISENGAGE bit(32) as
set. A write of bit 32 to zero returns to normal operation.
Also deal with the case where the processor does not support
turbo and the BIOS does not report the fact in MSR_IA32_MISC_ENABLE
but does report the max and turbo P states as the same value.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=64251
Cc: 3.13+ <stable@vger.kernel.org> # 3.13+
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit 21855ff5 (intel_pstate: Set turbo VID for BayTrail) introduced
setting the turbo VID which is required to prevent a machine check on
some Baytrail SKUs under heavy graphics based workloads. The
docmumentation update that brought the requirement to light also
changed the bit mask used for enumerating P state and VID values from
0x7f to 0x3f.
This change returns the mask value to 0x7f.
Tested with the Intel NUC DN2820FYK,
BIOS version FYBYT10H.86A.0034.2014.0513.1413 with v3.16-rc1 and
v3.14.8 kernel versions.
Fixes: 21855ff5 (intel_pstate: Set turbo VID for BayTrail)
Link: https://bugzilla.kernel.org/show_bug.cgi?id=77951
Reported-and-tested-by: Rune Reterson <rune@megahurts.dk>
Reported-and-tested-by: Eric Eickmeyer <erich@ericheickmeyer.com>
Cc: 3.13+ <stable@vger.kernel.org> # 3.13+
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
There was a mistake in the actual rounding portion this previous patch:
f0fe3cd7e1 (intel_pstate: Correct rounding in busy calculation) such that
the rounding was asymetric and incorrect.
Severity: Not very serious, but can increase target pstate by one extra value.
For real world work flows the issue should self correct (but I have no proof).
It is the equivalent of different PID gains for positive and negative numbers.
Examples:
-3.000000 used to round to -4, rounds to -3 with this patch.
-3.503906 used to round to -5, rounds to -4 with this patch.
Fixes: f0fe3cd7e1 (intel_pstate: Correct rounding in busy calculation)
Signed-off-by: Doug Smythies <dsmythies@telus.net>
Cc: 3.14+ <stable@vger.kernel.org> # 3.14+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
We check the CPU ID during driver init. There is no need
to do it again per logical CPU initialization.
So, remove the duplicate check.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
This change makes the busy calculation using 64 bit math which prevents
overflow for large values of aperf/mperf.
Cc: 3.14+ <stable@vger.kernel.org> # 3.14+
Signed-off-by: Doug Smythies <dsmythies@telus.net>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The PID assumes that samples are of equal time, which for a deferable
timers this is not true when the system goes idle. This causes the
PID to take a long time to converge to the min P state and depending
on the pattern of the idle load can make the P state appear stuck.
The hold-off value of three sample times before using the scaling is
to give a grace period for applications that have high performance
requirements and spend a lot of time idle, The poster child for this
behavior is the ffmpeg benchmark in the Phoronix test suite.
Cc: 3.14+ <stable@vger.kernel.org> # 3.14+
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Changing to fixed point math throughout the busy calculation in
commit e66c1768 (Change busy calculation to use fixed point
math.) Introduced some inaccuracies by rounding the busy value at two
points in the calculation. This change removes roundings and moves
the rounding to the output of the PID where the calculations are
complete and the value returned as an integer.
Fixes: e66c176837 (intel_pstate: Change busy calculation to use fixed point math.)
Reported-by: Doug Smythies <dsmythies@telus.net>
Cc: 3.14+ <stable@vger.kernel.org> # 3.14+
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit fcb6a15c (intel_pstate: Take core C0 time into account for core
busy calculation) introduced a regression referenced below. The issue
with "lockup" after suspend that this commit was addressing is now dealt
with in the suspend path.
Fixes: fcb6a15c2e (intel_pstate: Take core C0 time into account for core busy calculation)
Link: https://bugzilla.kernel.org/show_bug.cgi?id=66581
Link: https://bugzilla.kernel.org/show_bug.cgi?id=75121
Reported-by: Doug Smythies <dsmythies@telus.net>
Cc: 3.14+ <stable@vger.kernel.org> # 3.14+
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Although, a value is assigned to member name of struct cpudata,
it is never used.
We can safely remove it.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Setting the P state of the core to max at init time is a hold over
from early implementation of intel_pstate where intel_pstate disabled
cpufreq and loaded VERY early in the boot sequence. This was to
ensure that intel_pstate did not affect boot time. This in not needed
now that intel_pstate is a cpufreq driver.
Removing this covers the case where a CPU has gone through a manual
CPU offline/online cycle and the P state is set to MAX on init and the
CPU immediately goes idle. Due to HW coordination the P state request
on the idle CPU will drag all cores to MAX P state until the load is
reevaluated when to core goes non-idle.
Reported-by: Patrick Marlier <patrick.marlier@gmail.com>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Cc: 3.14+ <stable@vger.kernel.org> # 3.14+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Add support for Broadwell processors.
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
A documentation update exposed that there is a separate set of VID
values that must be used in the turbo/boost P state range. Add
enumerating and setting the correct VID for P states in the turbo
range.
Cc: v3.13+ <stable@vger.kernel.org> # v3.13+
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Since commit d37e2b7644 ("intel_pstate: remove unneeded sample buffers")
we use only one sample. So, there is no need to pass the sample
pointer to intel_pstate_calc_busy. Instead, get the pointer from
cpudata. Also, remove the unused SAMPLE_COUNT macro.
While at it, reformat the first line in this function.
Signed-off-by: Stratos Karafotis <stratosk@semaphore.gr>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Ensure that no timer callback is running since we are about to free
the timer structure. We cannot guarantee that the call back is called
on the CPU where the timer is running.
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Change to use the new ->stop_cpu() callback to do clean up during CPU
hotplug. The requested P state for an offline core will be used by the
hardware coordination function to select the package P state. If the
core is under load when it is offlined it will fix the package P state
floor to the requested P state of offline core.
Reported-by: Patrick Marlier <patrick.marlier@gmail.com>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
commit d253d2a526 (Improve accuracy by not truncating until final
result), changed internal variables of the PID to be fixed point
numbers. Update the pid_reset() to reflect this change.
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Remove unneeded sample buffers, intel_pstate operates on the most
recent sample only. This save some memory and make the code more
readable.
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Commit fcb6a15c2e (intel_pstate: Take core C0 time into account for
core busy calculation) introduced a regression on some processor SKUs
supported by intel_pstate. This was due to the truncation caused by
using integer math to calculate core busy and C0 percentages.
On a i7-4770K processor operating at 800Mhz going to 100% utilization
the percent busy of the CPU using integer math is 22%, but it actually
is 22.85%. This value scaled to the current frequency returned 97
which the PID interpreted as no error and did not adjust the P state.
Tested on i7-4770K, i7-2600, i5-3230M.
Fixes: fcb6a15c2e (intel_pstate: Take core C0 time into account for core busy calculation)
References: https://lkml.org/lkml/2014/2/19/626
References: https://bugzilla.kernel.org/show_bug.cgi?id=70941
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
A documentation update exposed the existance of the turbo ratio
register. Update baytrail support to use the turbo range.
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Cc: 3.13+ <stable@vger.kernel.org> # 3.13+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
LFM (max efficiency ratio) is the max frequency at minimum voltage
supported by the processor. Using LFM as the minimum P state
increases performmance without affecting power. By not using P states
below LFM we avoid using P states that are less power efficient.
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Cc: 3.13+ <stable@vger.kernel.org> # 3.13+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Remove the reporting of energy since it does not provide any useful
information about the state of the driver and will be a maintainance
headache going forward since the RAPL energy units register is not
architectural and subject to change between micro-architectures
References: https://bugzilla.kernel.org/show_bug.cgi?id=69831
Fixes: b69880f9cc (intel_pstate: Add trace point to report internal state.)
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Take non-idle time into account when calculating core busy time.
This ensures that intel_pstate will notice a decrease in load.
References: https://bugzilla.kernel.org/show_bug.cgi?id=66581
Cc: 3.10+ <stable@vger.kernel.org> # 3.10+
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Add perf trace event "power:pstate_sample" to report driver state to
aid in diagnosing issues reported against intel_pstate.
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
KVM environments do not support APERF/MPERF MSRs. intel_pstate cannot
operate without these registers.
The previous validity checks in intel_pstate_msrs_not_valid() are
insufficent in nested KVMs.
References: https://bugzilla.redhat.com/show_bug.cgi?id=1046317
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Cc: All applicable <stable@vger.kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Remove the periodic P state boost. This code required for some corner
case benchmark tests. The calculation of the required P state was
incorrect/inaccurate and would not allow P state increase.
This was fixed by a combination of commits:
2134ed4 cpufreq / intel_pstate: Change to scale off of max P-state
d253d2a intel_pstate: Improve accuracy by not truncating until final result
References: https://bugzilla.kernel.org/show_bug.cgi?id=64271
Reported-by: Doug Smythies <dsmythies@telus.net>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Baytrail requires setting P state and voltage pairs when adjusting the
requested P state. Add function for retrieving the valid voltage
values and modify *_set_pstate() functions to caluclate the
appropriate voltage for the requested P state.
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
* pm-cpufreq:
intel_pstate: skip the driver if ACPI has power mgmt option
cpufreq: ondemand: Remove redundant return statement
cpufreq: move freq change notifications to cpufreq core
cpufreq: distinguish drivers that do asynchronous notifications
cpufreq/intel_pstate: Add static declarations to internal functions
cpufreq: arm_big_little: reconfigure switcher behavior at run time
cpufreq: arm_big_little: add in-kernel switching (IKS) support
ARM: vexpress/TC2: register vexpress-spc cpufreq device
cpufreq: arm_big_little: add vexpress SPC interface driver
ARM: vexpress/TC2: add cpu clock support
ARM: vexpress/TC2: add support for CPU DVFS
Do not load the Intel pstate driver if the platform firmware
(ACPI BIOS) supports the power management alternatives.
The ACPI BIOS indicates that the OS control mode can be used
if the _PSS (Performance Supported States) is defined in ACPI
table. For the OS control mode, the Intel pstate driver will be
loaded.
HP BIOS has several power management modes (firmware, OS-control and
so on). For the OS control mode in HP BIOS, the Intel p-state driver
will be loaded. When the customer chooses the firmware power
management in HP BIOS, the Intel p-state driver will be ignored.
I put hw_vendor_info vendor_info in case other vendors (Dell, Lenovo...)
have their firmware power management. Vendors should make sure their
firmware power management works properly, and they can go for adding
their vendor info to the variable.
I have verified the patch on HP ProLiant servers. The patch worked
correctly.
Signed-off-by: Adrian Huang <adrianhuang0701@gmail.com>
[rjw: Fixed up !CONFIG_ACPI build]
[Linda Knippers: As Adrian has recently left HP, I retested the
updated patch on an HP ProLiant server and verified that it is
behaving correctly. When the BIOS is configured for OS control for
power management, the intel_pstate driver loads as expected. When
the BIOS is configured to provide the power management, the
intel_pstate driver does not load and we get the pcc_cpufreq driver
instead.]
Signed-off-by: Linda Knippers <linda.knippers@hp.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Fixes warnings reported by kbuild test robot
sparse warnings: (new ones prefixed by >>)
drivers/cpufreq/intel_pstate.c:729:6: sparse: symbol 'copy_pid_params' was not declared. Should it be static?
drivers/cpufreq/intel_pstate.c:739:6: sparse: symbol 'copy_cpu_funcs' was not declared. Should it be static?
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
* pm-cpufreq: (167 commits)
cpufreq: create per policy rwsem instead of per CPU cpu_policy_rwsem
intel_pstate: Add Baytrail support
intel_pstate: Refactor driver to support CPUs with different MSR layouts
cpufreq: Implement light weight ->target_index() routine
PM / OPP: rename header to linux/pm_opp.h
PM / OPP: rename data structures to dev_pm equivalents
PM / OPP: rename functions to dev_pm_opp*
cpufreq / governor: Remove fossil comment
cpufreq: exynos4210: Use the common clock framework to set APLL clock rate
cpufreq: exynos4x12: Use the common clock framework to set APLL clock rate
cpufreq: Detect spurious invocations of update_policy_cpu()
cpufreq: pmac64: enable cpufreq on iMac G5 (iSight) model
cpufreq: pmac64: provide cpufreq transition latency for older G5 models
cpufreq: pmac64: speed up frequency switch
cpufreq: highbank-cpufreq: Enable Midway/ECX-2000
exynos-cpufreq: fix false return check from "regulator_set_voltage"
speedstep-centrino: Remove unnecessary braces
acpi-cpufreq: Add comment under ACPI_ADR_SPACE_SYSTEM_IO case
cpufreq: arm-big-little: use clk_get instead of clk_get_sys
cpufreq: exynos: Show a list of available frequencies
...
Conflicts:
drivers/devfreq/exynos/exynos5_bus.c
Add support for the Baytrail processor.
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Non-core processors have a different MSR layout to commumicate P state
information. Refactor the driver to use CPU dependent accessors to
P state information.
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The minimum pstate is supposed to be a percentage of the maximum P
state available. Calculate min using max pstate and not the
current max which may have been limited by the user
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
