When a new sampling rate is shorter than the current one, (e.g., 1 sec
--> 10 ms) regardless how short the new one is, the current ondemand
mechanism wait for the previously set timer to be expired.

For example, if the user has just expressed that the sampling rate
should be 10 ms from now and the previous was 1000 ms, the new rate may
become effective 999 ms later, which could be not acceptable for the
user if the user has intended to speed up sampling because the system is
expected to react to CPU load fluctuation quickly from __now__.

In order to address this issue, we need to cancel the previously set
timer (schedule_delayed_work) and reset the timer if resetting timer is
expected to trigger the delayed_work ealier.
Signed-off-by: NMyungJoo Ham <myungjoo.ham@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: NDave Jones <davej@redhat.com>

The S3C2416/S3C2450 SoCs support two sources for the armclk.

The first source is the so called armdiv which divides the msysclk down
to provide necessary cpu rates. In this mode the core voltage must be
always at 1.3V. The frequency from the armdiv is not allowed to be
lower than the hclk frequency.

In the second mode the armclk can be sourced directly from the hclk in
the so called "dynamic voltags scaling" (dvs) mode. Here the armdiv
isn't used at all. Also in this mode the core voltage may be lowered.
Existing hardware and tests with it suggest 1.0V as sufficient.

When changing the clock source to the armdiv from the hclk, the SoC
shows stability issues if the new frequency is higher than the current
hclk frequency. Hence the driver always forces the armdiv to the hclk
frequency before the source change and lets the cpufreq issue another
set_target call for higher frequencies.

To mark the hclk frequency as lower as the corresponding armdiv
frequency it is set 1MHz below the real frequency. This lets the cpufreq
framework change between 133MHz based on hclk and 133MHz based on armdiv
at will.
Signed-off-by: NHeiko Stuebner <heiko@sntech.de>
Tested-by: NAndrey Gusakov <dron0gus@gmail.com>
Signed-off-by: NDave Jones <davej@redhat.com>

exynos4210-cpufreq.c is not buildable on non-exynos builds, so it's
pointless allowing this option to be exposed.  Fix this by adding a
dependency on ARCH_EXYNOS.

drivers/cpufreq/exynos4210-cpufreq.c:20:29: error: mach/regs-clock.h: No such file or directory
drivers/cpufreq/exynos4210-cpufreq.c:21:26: error: mach/cpufreq.h: No such file or directory
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>
Cc: cpufreq@vger.kernel.org
Cc: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: NDave Jones <davej@redhat.com>

According to replacing the name of EXYNOS clock registers,
this patch updates exynos4210-cpufreq.c file where it is used.

Cc: Jaecheol Lee <jc.lee@samsung.com>
Signed-off-by: NKukjin Kim <kgene.kim@samsung.com>
Signed-off-by: NDave Jones <davej@redhat.com>

As per definition, locking_frequency is the initial frequency which is
set by boot-loader. Hence the value is updated with the initial value
during boot time init call.

This code was present in exynos210-cpufreq.c before this consolidation
patch.
- a125a17f ([CPUFREQ] EXYNOS: Make EXYNOS common cpufreq driver).
Signed-off-by: NTushar Behera <tushar.behera@linaro.org>
Signed-off-by: NInderpal Singh <inderpal.singh@linaro.org>
Signed-off-by: NDave Jones <davej@redhat.com>

We don't have any of the other code for VDDINT, including the variable
declaration, so don't try to get it as we can't build.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: NDave Jones <davej@redhat.com>

This patch removes no referencing header files and cleaned up
useless code.
Signed-off-by: NJaecheol Lee <jc.lee@samsung.com>
Signed-off-by: NKukjin Kim <kgene.kim@samsung.com>
Signed-off-by: NDave Jones <davej@redhat.com>

To support various EXYNOS series SoCs commonly,
added exynos common structure.
exynos-cpufreq.c => EXYNOS series common cpufreq driver
exynos4210-cpufreq.c => EXYNOS4210 support cpufreq driver
Signed-off-by: NJaecheol Lee <jc.lee@samsung.com>
Signed-off-by: NKukjin Kim <kgene.kim@samsung.com>
Signed-off-by: NDave Jones <davej@redhat.com>

Signed-off-by: NAndreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: NDave Jones <davej@redhat.com>

The driver uses the pstate number from the status register as index in
its table of ACPI pstates (powernow_table). This is wrong as this is
not a 1-to-1 mapping.

For example we can have _PSS information to just utilize Pstate 0 and
Pstate 4, ie.

  powernow-k8: Core Performance Boosting: on.
  powernow-k8:    0 : pstate 0 (2200 MHz)
  powernow-k8:    1 : pstate 4 (1400 MHz)

In this example the driver's powernow_table has just 2 entries. Using
the pstate number (4) as index into this table is just plain wrong.
Signed-off-by: NAndreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: NDave Jones <davej@redhat.com>

Due to CPB we can't directly map SW Pstates to Pstate MSRs. Get rid of
the paranoia check. (assuming that the ACPI Pstate information is
correct.)
Signed-off-by: NAndreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: NDave Jones <davej@redhat.com>

During scaling up of cpu frequency, loops_per_jiffy
is updated upon invoking PRECHANGE notifier.
If setting to new frequency fails in cpufreq driver,
lpj is left at incorrect value.

Hence update lpj only if cpu frequency is changed,
i.e. upon invoking POSTCHANGE notifier.

Penalty would be that during time period between
changing cpu frequency & invocation of POSTCHANGE
notifier, udelay(x) may not gurantee minimal delay
of 'x' us for frequency scaling up operation.

Perhaps a better solution would be to define
CPUFREQ_ABORTCHANGE & handle accordingly, but then
it would be more intrusive (using ABORTCHANGE may
help drivers also; if any has registered notifier
and expect POST for a PRECHANGE, their needs can
be taken care using ABORT)
Signed-off-by: NAfzal Mohammed <afzal@ti.com>
Signed-off-by: NDave Jones <davej@redhat.com>

CPU frequency is guranteed to be changed on notifier callback with
CPUFREQ_POSTCHANGE. Notifier callback with CPUFREQ_PRECHANGE does
not gurantee a change in frequency; after it, if cpufreq driver is
unable to change CPU to new frequency. This results in wrong
information being fed to user (if setting CPU frequency fails)
upon doing like,

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed

Hence in userspace governer update cpu_cur_freq only if notifier
has been called with POSTCHANGE.
Signed-off-by: NAfzal Mohammed <afzal@ti.com>
Signed-off-by: NDave Jones <davej@redhat.com>

This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem
and converts the devices to regular devices. The sysdev drivers are
implemented as subsystem interfaces now.

After all sysdev classes are ported to regular driver core entities, the
sysdev implementation will be entirely removed from the kernel.

Userspace relies on events and generic sysfs subsystem infrastructure
from sysdev devices, which are made available with this conversion.

Cc: Haavard Skinnemoen <hskinnemoen@gmail.com>
Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Borislav Petkov <bp@amd64.org>
Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk>
Cc: Len Brown <lenb@kernel.org>
Cc: Zhang Rui <rui.zhang@intel.com>
Cc: Dave Jones <davej@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: NKay Sievers <kay.sievers@vrfy.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Make cputime_t and cputime64_t nocast to enable sparse checking to
detect incorrect use of cputime. Drop the cputime macros for simple
scalar operations. The conversion macros are still needed.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

CPUFREQ Remove wall variable from cpufreq_gov_dbs_init()

Remove wall variable from cpufreq_gov_dbs_init() as
get_cpu_idle_time_us() no longer updates the last_update_time
unconditionally. Passing non-NULL last_update_time address
will result in accounting additional idle time with
update_ts_time_stats() before returning idle_sleeptime.
Signed-off-by: NKamalesh Babulal <kamalesh@linux.vnet.ibm.com>
Signed-off-by: NDave Jones <davej@redhat.com>
--
 drivers/cpufreq/cpufreq_ondemand.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)

This patch is modify code for stable working
1. Remove unused register access code
2. Change sequence for frequency changing
Signed-off-by: NJaecheol Lee <jc.lee@samsung.com>
Signed-off-by: NJonghwan Choi <jhbird.choi@samsung.com>
Signed-off-by: NJongpill Lee <boyko.lee@samsung.com>
Signed-off-by: NKukjin Kim <kgene.kim@samsung.com>
Signed-off-by: NDave Jones <davej@redhat.com>

This patch is changes frequency table for cpu divider for stable frequency.
Signed-off-by: NJaecheol Lee <jc.lee@samsung.com>
Signed-off-by: NJongpill Lee <boyko.lee@samsung.com>
Signed-off-by: NKukjin Kim <kgene.kim@samsung.com>
Signed-off-by: NDave Jones <davej@redhat.com>

This patch removes code for bus on cpufreq because the code
for bus frequency changing moves to busfreq driver.
So code about bus on cpufreq is not necessary.
Signed-off-by: NJaecheol Lee <jc.lee@samsung.com>
Signed-off-by: NJongpill Lee <boyko.lee@samsung.com>
Signed-off-by: NKukjin Kim <kgene.kim@samsung.com>
Signed-off-by: NDave Jones <davej@redhat.com>

They're already consistent but it saves remembering to do so.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: NDave Jones <davej@redhat.com>

This patch changes fields in cpustat from a structure, to an
u64 array. Math gets easier, and the code is more flexible.
Signed-off-by: NGlauber Costa <glommer@parallels.com>
Reviewed-by: NKAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Paul Tuner <pjt@google.com>
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1322498719-2255-2-git-send-email-glommer@parallels.comSigned-off-by: NIngo Molnar <mingo@elte.hu>

The variable i is removed by commit ded84337
"[CPUFREQ] db8500: remove unneeded for loop iteration over freq_table",
but current code to print available frequencies still uses the i variable.
Thus add the i variable back to fix below buld error:

  CC      drivers/cpufreq/db8500-cpufreq.o
drivers/cpufreq/db8500-cpufreq.c: In function 'db8500_cpufreq_init':
drivers/cpufreq/db8500-cpufreq.c:123: error: 'i' undeclared (first use in this function)
drivers/cpufreq/db8500-cpufreq.c:123: error: (Each undeclared identifier is reported only once
drivers/cpufreq/db8500-cpufreq.c:123: error: for each function it appears in.)
make[2]: *** [drivers/cpufreq/db8500-cpufreq.o] Error 1
make[1]: *** [drivers/cpufreq] Error 2
make: *** [drivers] Error 2

This patch also fixes using uninitialized i variable as array index.
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Acked-by: NLinus Walleij <linus.walleij@linaro.org>
Signed-off-by: NDave Jones <davej@redhat.com>

Minor fixups to work starting with v3.2:
- use the new omap_device API for getting a device by name.
- need to include <linux/module.h>
Signed-off-by: NKevin Hilman <khilman@ti.com>

We use a single frequency table for multiple CPUs. But, with
OMAP4, since we have multiple CPUs, the cpu_init call for CPU1
causes freq_table previously allocated for CPU0 to be overwritten.
In addition, we dont free the table on exit path.

We solve this by maintaining an atomic type counter to ensure
just a single table exists at a given time.
Signed-off-by: NNishanth Menon <nm@ti.com>
Signed-off-by: NKevin Hilman <khilman@ti.com>

Release the mpu_clk in fail paths.
Reported-by: NTodd Poynor <toddpoynor@google.com>
Signed-off-by: NNishanth Menon <nm@ti.com>
Signed-off-by: NKevin Hilman <khilman@ti.com>

OMAP2 is the only family using clk_[init|exit]_cpufreq_table, however,
the cpufreq code does not currently use clk_init_cpufreq_table. As a
result, it is unusuable for OMAP2 and only usable only on platforms
using OPP library.

Remove the unbalanced clk_exit_cpufreq_table().  Any platforms where
OPPs are not availble will fail on init because a freq table will not
be properly initialized.
Signed-off-by: NNishanth Menon <nm@ti.com>
[khilman@ti.com: changelog edits, and graceful failure mode changes]
Signed-off-by: NKevin Hilman <khilman@ti.com>

OMAP2+ all have frequency tables, hence the hacks we had for older
silicon do not need to be carried forward. As part of this change,
use cpufreq_frequency_table_target to find the best match for
frequency requested.
Signed-off-by: NNishanth Menon <nm@ti.com>
Signed-off-by: NKevin Hilman <khilman@ti.com>

if we do not have mpu_dev we normally fail in cpu_init. It is better
to fail driver registration if the devices are not available.
Signed-off-by: NNishanth Menon <nm@ti.com>
Signed-off-by: NKevin Hilman <khilman@ti.com>

Clk name does'nt need to dynamically detected during clk init.
move them off to driver initialization, if we dont have a clk name,
there is no point in registering the driver anyways. The actual clk
get and put is left at cpu_init and exit functions.
Signed-off-by: NNishanth Menon <nm@ti.com>
Signed-off-by: NKevin Hilman <khilman@ti.com>

Sometimes, bootloaders starts up with a frequency which is not
in the OPP table. At cpu_init, policy->cur contains the frequency
we pick at boot.  It is possible that system might have fixed
it's boot frequency later on as part of power initialization.
After this condition, the first call to omap_target results in the
following:

omap_getspeed(actual device frequency) != policy->cur(frequency that
cpufreq thinks that the system is at), and it is possible that
freqs.old == freqs.new (because the governor requested a scale down).

We exit without triggering the notifiers in the current code, which
does'nt let code which depends on cpufreq_notify_transition to have
accurate information as to what the system frequency is.

Instead, we do a normal transition if policy->cur is wrong, then,
freqs.old will be the actual cpu frequency, freqs.new will be the
actual new cpu frequency and all required notifiers have the accurate
information.
Acked-by: NNishanth Menon <nm@ti.com>
Signed-off-by: NColin Cross <ccross@google.com>
Signed-off-by: NKevin Hilman <khilman@ti.com>

Enable all CPUs in the shared policy in the CPU init callback.
Otherwise, the governor CPUFREQ_GOV_START event is invoked with
a policy that only includes the first CPU, leaving other CPUs
uninitialized by the governor.
Signed-off-by: NTodd Poynor <toddpoynor@google.com>
Acked-by: NSantosh Shilimkar <santosh.shilimkar@ti.com>
Signed-off-by: NKevin Hilman <khilman@ti.com>

On OMAP SMP configuartion, both processors share the voltage
and clock. So both CPUs needs to be scaled together and hence
needs software co-ordination.

Also, update lpj with reference value to avoid progressive error.

Adjust _both_ the per-cpu loops_per_jiffy and global lpj. Calibrate
them with with reference to the initial values to avoid a
progressively bigger and bigger error in the value over time.

While at this, re-use the notifiers for UP/SMP since on UP machine or
UP_ON_SMP policy->cpus mask would contain only the boot CPU.

Based on initial SMP support by Santosh Shilimkar.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: NSantosh Shilimkar <santosh.shilimkar@ti.com>
[khilman@ti.com: due to overlap/rework, combined original Santosh patch
                 and Russell's rework]
Signed-off-by: NKevin Hilman <khilman@ti.com>

Move OMAP cpufreq driver from arch/arm/mach-omap2 into
drivers/cpufreq, along with a few cleanups:

- generalize support for better handling of different SoCs in the OMAP
- use OPP layer instead of OMAP clock internals for frequency table init
Signed-off-by: NSantosh Shilimkar <santosh.shilimkar@ti.com>
[khilman@ti.com: move to drivers]
Signed-off-by: NKevin Hilman <khilman@ti.com>

The header change has removed an implicit include of module.h, breaking
the build due to the use of THIS_MODULE. Fix that.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

So that we can clean up the header files and not be relying
on implicit includes from device.h ---> module.h
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

This adds support for the 200 MHz frequency mode of the
DB8500 SoC, and prints the available frequencies at init
time.

Cc: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>
Signed-off-by: NDave Jones <davej@redhat.com>

Don't know why to do the loop iteration here. It looks unneeded.
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Signed-off-by: NDave Jones <davej@redhat.com>

We have various bootloaders for Exynos4210 machines. Some of they
set the ARM core frequency at boot time even when the boot is a resume
from suspend-to-RAM. Such changes may create inconsistency in the
data of CPUFREQ driver and have incurred hang issues with suspend-to-RAM.

This patch enables to save and restore CPU frequencies with pm-notifier and
sets the frequency at the initial (boot-time) value so that there wouldn't
be any inconsistency between bootloader and kernel. This patch does not
use CPUFREQ's suspend/resume callbacks because they are syscore-ops, which
do not allow to use mutex that is being used by regulators that are used by
the target function.

This also prevents any CPUFREQ transitions during suspend-resume context,
which could be dangerous at noirq-context along with regulator framework.
Signed-off-by: NMyungJoo Ham <myungjoo.ham@samsung.com>
Signed-off-by: NKyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: NDave Jones <davej@redhat.com>

The same clock is used for all cpus so we must notify the frequency change
for each one in order to update the configuration of all twd clockevents.

change since V1:
* use policy->cpus instead of cpu_online_mask
Signed-off-by: NVincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: NDave Jones <davej@redhat.com>

Add new module option "set_max_voltage".
One of the lessons learned from Adaptive Powersaver is that voltage values
returned by processor are for worst case scenario. But required voltage
is changing with CPU temperature. And even processors produced in the same
batch can have different minimum voltage necessary for stable work at
specified frequency.
On Elonex Webbook, once system starts, temperature never drops below
48 deg. C. Loading module after systems start allows user to lower CPU
voltage and still have stable system.
Sadly C7 doesn't allow code to set frequency or voltage from outside limits.
If you ask it to set voltage lower then minimum it will ignore you. Thats
why it isn't possible to change minimum voltage for minimum frequency too.
Changing maximum voltage on Elonex Webbook leads to very good results. Looks
like VIA C7 1.6GHz 1084mV can safetly run at 892mV. This means 83% of
orginal value. If same percentage applies to power generated it means 12.5W
in the place of 15W. Not much, but it is better then nothing.
Only C7-M makes it possible.
If voltage is too low by 16mV or more you will experience kernel panic.
If voltage is too low by 32mV or more you will experience system freeze.
Signed-off-by: NRafał Bilski <rafalbilski@interia.pl>
Signed-off-by: NDave Jones <davej@redhat.com>