| Age | Commit message (Collapse) | Author |
|---|
|
To be future-proof and for better readability the time comparisons are
modified to use time_before() instead of plain, error-prone math.
Signed-off-by: Manuel Schölling <manuel.schoelling@gmx.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
The purpose of the acpi_pad driver is to implement the "processor power
aggregator" device as described in the ACPI 4.0 spec section 8.5. It
takes requests from the BIOS (via ACPI) to put a specified number of
CPUs into idle, in order to save power, until further notice.
It does this by creating high-priority threads that try to keep the CPUs
in a high C-state (using the monitor/mwait CPU instructions). The
mwait() call is in a loop that checks periodically if the thread should
end and a few other things.
It was discovered through testing that the power_saving threads were
causing the system to consume more power than the system was consuming
before the threads were created. A counter in the main loop of
power_saving_thread() revealed that it was spinning. The mwait()
instruction was not keeping the CPU in a high C state very much if at
all.
Here is a simplification of the loop in function power_saving_thread() in
drivers/acpi/acpi_pad.c
while (!kthread_should_stop()) {
:
try_to_freeze()
:
while (!need_resched()) {
:
if (!need_resched())
__mwait(power_saving_mwait_eax, 1);
:
if (jiffies > expire_time) {
do_sleep = 1;
break;
}
}
}
If need_resched() returns true, then mwait() is not called. It was
returning true because of things like timer interrupts, as in the
following sequence.
hrtimer_interrupt->__run_hrtimer->tick_sched_timer-> update_process_times->
rcu_check_callbacks->rcu_pending->__rcu_pending->set_need_resched
Kernels 3.5.0-rc2+ do not exhibit this problem, because a patch to
try_to_freeze() in include/linux/freezer.h introduces a call to
might_sleep(), which ultimately calls schedule() to clear the reschedule
flag and allows the the loop to execute the call to mwait().
However, the changes to try_to_freeze are unrelated to acpi_pad, and it
does not seem like a good idea to rely on an unrelated patch in a
function that could later be changed and reintroduce this bug.
Therefore, it seems better to make an explicit call to schedule() in the
outer loop when the need_resched flag is set.
Reported-and-tested-by: Stuart Hayes <stuart_hayes@dell.com>
Signed-off-by: Tony Camuso <tcamuso@redhat.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Use public function acpi_evaluate_ost() to replace open-coded
version of evaluating ACPI _OST method.
Signed-off-by: Jiang Liu <jiang.liu@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull ACPI and power management updates from Rafael Wysocki:
"As far as the number of commits goes, the top spot belongs to ACPI
this time with cpufreq in the second position and a handful of PM
core, PNP and cpuidle updates. They are fixes and cleanups mostly, as
usual, with a couple of new features in the mix.
The most visible change is probably that we will create struct
acpi_device objects (visible in sysfs) for all devices represented in
the ACPI tables regardless of their status and there will be a new
sysfs attribute under those objects allowing user space to check that
status via _STA.
Consequently, ACPI device eject or generally hot-removal will not
delete those objects, unless the table containing the corresponding
namespace nodes is unloaded, which is extremely rare. Also ACPI
container hotplug will be handled quite a bit differently and cpufreq
will support CPU boost ("turbo") generically and not only in the
acpi-cpufreq driver.
Specifics:
- ACPI core changes to make it create a struct acpi_device object for
every device represented in the ACPI tables during all namespace
scans regardless of the current status of that device. In
accordance with this, ACPI hotplug operations will not delete those
objects, unless the underlying ACPI tables go away.
- On top of the above, new sysfs attribute for ACPI device objects
allowing user space to check device status by triggering the
execution of _STA for its ACPI object. From Srinivas Pandruvada.
- ACPI core hotplug changes reducing code duplication, integrating
the PCI root hotplug with the core and reworking container hotplug.
- ACPI core simplifications making it use ACPI_COMPANION() in the
code "glueing" ACPI device objects to "physical" devices.
- ACPICA update to upstream version 20131218. This adds support for
the DBG2 and PCCT tables to ACPICA, fixes some bugs and improves
debug facilities. From Bob Moore, Lv Zheng and Betty Dall.
- Init code change to carry out the early ACPI initialization
earlier. That should allow us to use ACPI during the timekeeping
initialization and possibly to simplify the EFI initialization too.
From Chun-Yi Lee.
- Clenups of the inclusions of ACPI headers in many places all over
from Lv Zheng and Rashika Kheria (work in progress).
- New helper for ACPI _DSM execution and rework of the code in
drivers that uses _DSM to execute it via the new helper. From
Jiang Liu.
- New Win8 OSI blacklist entries from Takashi Iwai.
- Assorted ACPI fixes and cleanups from Al Stone, Emil Goode, Hanjun
Guo, Lan Tianyu, Masanari Iida, Oliver Neukum, Prarit Bhargava,
Rashika Kheria, Tang Chen, Zhang Rui.
- intel_pstate driver updates, including proper Baytrail support,
from Dirk Brandewie and intel_pstate documentation from Ramkumar
Ramachandra.
- Generic CPU boost ("turbo") support for cpufreq from Lukasz
Majewski.
- powernow-k6 cpufreq driver fixes from Mikulas Patocka.
- cpufreq core fixes and cleanups from Viresh Kumar, Jane Li, Mark
Brown.
- Assorted cpufreq drivers fixes and cleanups from Anson Huang, John
Tobias, Paul Bolle, Paul Walmsley, Sachin Kamat, Shawn Guo, Viresh
Kumar.
- cpuidle cleanups from Bartlomiej Zolnierkiewicz.
- Support for hibernation APM events from Bin Shi.
- Hibernation fix to avoid bringing up nonboot CPUs with ACPI EC
disabled during thaw transitions from Bjørn Mork.
- PM core fixes and cleanups from Ben Dooks, Leonardo Potenza, Ulf
Hansson.
- PNP subsystem fixes and cleanups from Dmitry Torokhov, Levente
Kurusa, Rashika Kheria.
- New tool for profiling system suspend from Todd E Brandt and a
cpupower tool cleanup from One Thousand Gnomes"
* tag 'pm+acpi-3.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (153 commits)
thermal: exynos: boost: Automatic enable/disable of BOOST feature (at Exynos4412)
cpufreq: exynos4x12: Change L0 driver data to CPUFREQ_BOOST_FREQ
Documentation: cpufreq / boost: Update BOOST documentation
cpufreq: exynos: Extend Exynos cpufreq driver to support boost
cpufreq / boost: Kconfig: Support for software-managed BOOST
acpi-cpufreq: Adjust the code to use the common boost attribute
cpufreq: Add boost frequency support in core
intel_pstate: Add trace point to report internal state.
cpufreq: introduce cpufreq_generic_get() routine
ARM: SA1100: Create dummy clk_get_rate() to avoid build failures
cpufreq: stats: create sysfs entries when cpufreq_stats is a module
cpufreq: stats: free table and remove sysfs entry in a single routine
cpufreq: stats: remove hotplug notifiers
cpufreq: stats: handle cpufreq_unregister_driver() and suspend/resume properly
cpufreq: speedstep: remove unused speedstep_get_state
platform: introduce OF style 'modalias' support for platform bus
PM / tools: new tool for suspend/resume performance optimization
ACPI: fix module autoloading for ACPI enumerated devices
ACPI: add module autoloading support for ACPI enumerated devices
ACPI: fix create_modalias() return value handling
...
|
|
People seem to delight in writing wrong and broken mwait idle routines;
collapse the lot.
This leaves mwait_play_dead() the sole remaining user of __mwait() and
new __mwait() users are probably doing it wrong.
Also remove __sti_mwait() as its unused.
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jacob Jun Pan <jacob.jun.pan@linux.intel.com>
Cc: Mike Galbraith <bitbucket@online.de>
Cc: Len Brown <lenb@kernel.org>
Cc: Rui Zhang <rui.zhang@intel.com>
Acked-by: Rafael Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131212141654.616820819@infradead.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
|
|
Replace direct inclusions of <acpi/acpi.h>, <acpi/acpi_bus.h> and
<acpi/acpi_drivers.h>, which are incorrect, with <linux/acpi.h>
inclusions and remove some inclusions of those files that aren't
necessary.
First of all, <acpi/acpi.h>, <acpi/acpi_bus.h> and <acpi/acpi_drivers.h>
should not be included directly from any files that are built for
CONFIG_ACPI unset, because that generally leads to build warnings about
undefined symbols in !CONFIG_ACPI builds. For CONFIG_ACPI set,
<linux/acpi.h> includes those files and for CONFIG_ACPI unset it
provides stub ACPI symbols to be used in that case.
Second, there are ordering dependencies between those files that always
have to be met. Namely, it is required that <acpi/acpi_bus.h> be included
prior to <acpi/acpi_drivers.h> so that the acpi_pci_root declarations the
latter depends on are always there. And <acpi/acpi.h> which provides
basic ACPICA type declarations should always be included prior to any other
ACPI headers in CONFIG_ACPI builds. That also is taken care of including
<linux/acpi.h> as appropriate.
Signed-off-by: Lv Zheng <lv.zheng@intel.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Matthew Garrett <mjg59@srcf.ucam.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com> (drivers/pci stuff)
Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> (Xen stuff)
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux
Pull PTR_RET() removal patches from Rusty Russell:
"PTR_RET() is a weird name, and led to some confusing usage. We ended
up with PTR_ERR_OR_ZERO(), and replacing or fixing all the usages.
This has been sitting in linux-next for a whole cycle"
[ There are still some PTR_RET users scattered about, with some of them
possibly being new, but most of them existing in Rusty's tree too. We
have that
#define PTR_RET(p) PTR_ERR_OR_ZERO(p)
thing in <linux/err.h>, so they continue to work for now - Linus ]
* tag 'PTR_RET-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux:
GFS2: Replace PTR_RET with PTR_ERR_OR_ZERO
Btrfs: volume: Replace PTR_RET with PTR_ERR_OR_ZERO
drm/cma: Replace PTR_RET with PTR_ERR_OR_ZERO
sh_veu: Replace PTR_RET with PTR_ERR_OR_ZERO
dma-buf: Replace PTR_RET with PTR_ERR_OR_ZERO
drivers/rtc: Replace PTR_RET with PTR_ERR_OR_ZERO
mm/oom_kill: remove weird use of ERR_PTR()/PTR_ERR().
staging/zcache: don't use PTR_RET().
remoteproc: don't use PTR_RET().
pinctrl: don't use PTR_RET().
acpi: Replace weird use of PTR_RET.
s390: Replace weird use of PTR_RET.
PTR_RET is now PTR_ERR_OR_ZERO(): Replace most.
PTR_RET is now PTR_ERR_OR_ZERO
|
|
It is quite some time that this one has been deprecated.
Get rid of it.
Should some really important user be overseen, it may be reverted and
the userspace program worked on first, but it is time to do something
to get rid of this old stuff...
Signed-off-by: Thomas Renninger <trenn@suse.de>
Acked-by: Matthew Garrett <matthew.garrett@nebula.com>
Acked-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
This functions is really weird. It sets rc to -ENOMEM, then overrides
it. It was converted to PTR_RET in a1458187 when it should have
simply been rewritten.
This version makes it more explicit, with a single IS_ERR() test.
Cc: Alexandru Gheorghiu <gheorghiuandru@gmail.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Use PTR_RET instead of explicit checking with IS_ERR.
Signed-off-by: Alexandru Gheorghiu <gheorghiuandru@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
The second argument of ACPI driver .remove() operation is only used
by the ACPI processor driver and the value passed to that driver
through it is always available from the given struct acpi_device
object's removal_type field. For this reason, the second ACPI driver
.remove() argument is in fact useless, so drop it.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Jiang Liu <jiang.liu@huawei.com>
Acked-by: Toshi Kani <toshi.kani@hp.com>
Acked-by: Yinghai Lu <yinghai@kernel.org>
|
|
Replace a few calls to strict_strtoul() in acpi_pad.c with kstrtoul()
and use pr_warn() instead of printk() in the same file.
[rjw: Modified the subject and changelog.]
Signed-off-by: Josh Taylor <joshua.taylor0@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
'cpuidle-coupled', 'cpuidle-tweaks', 'intel_idle-ivb', 'ost', 'red-hat-bz-772730', 'thermal', 'thermal-spear' and 'turbostat-v2' into release
|
|
In an effort to be fair to bound processes,
acpi_pad periodically moves its forced-idle threads.
The default interval for moving the threads is 10 seconds.
Measurements show that reducing this to 1 second has no
power or performance impact, so reduce default to 1 second.
Signed-off-by: Len Brown <len.brown@intel.com>
|
|
The acpi_pad driver can get stuck in destroy_power_saving_task()
waiting for kthread_stop() to stop a power_saving thread. The problem
is that the isolated_cpus_lock mutex is owned when
destroy_power_saving_task() calls kthread_stop(), which waits for a
power_saving thread to end, and the power_saving thread tries to
acquire the isolated_cpus_lock when it calls round_robin_cpu(). This
patch fixes the issue by making round_robin_cpu() use its own mutex.
https://bugzilla.kernel.org/show_bug.cgi?id=42981
Cc: stable@vger.kernel.org
Signed-off-by: Stuart Hayes <Stuart_Hayes@Dell.com>
Signed-off-by: Len Brown <len.brown@intel.com>
|
|
"acpi_pad/%d" is a better thread name than
generic "power_saving/%d" because users who see
these threads will know the name of the driver
that caused them.
Signed-off-by: Len Brown <len.brown@intel.com>
|
|
Minor format change.
Signed-off-by: Naga Chumbalkar <nagananda.chumbalkar@hp.com>
Signed-off-by: Len Brown <len.brown@intel.com>
|
|
Make /sys output from acpi_pad more readable.
Before the fix:
# cat idlecpus idlepct rrtime
00000000510
After the fix:
# cat idlecpus idlepct rrtime
00000000
5
10
Signed-off-by: Naga Chumbalkar <nagananda.chumbalkar@hp.com>
Signed-off-by: Len Brown <len.brown@intel.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
* 'x86-idle-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86, hotplug: In the MWAIT case of play_dead, CLFLUSH the cache line
x86, hotplug: Move WBINVD back outside the play_dead loop
x86, hotplug: Use mwait to offline a processor, fix the legacy case
x86, mwait: Move mwait constants to a common header file
|
|
acpi_pad.c:432: warning: ‘num_cpus’ may be used uninitialized in this function
gcc 4.4.4 was unable to notice that num_cpus is always set.
Re-arrange the code to un-confuse gcc, and also make
it easier for humans to read....
Signed-off-by: Len Brown <len.browns@intel.com>
|
|
We have MWAIT constants spread across three different .c files, for no
good reason. Move them all into a common header file.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Reviewed-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Len Brown <lenb@kernel.org>
LKML-Reference: <tip-*@git.kernel.org>
|
|
Now that all arches have been converted over to use generic time via
clocksources or arch_gettimeoffset(), we can remove the GENERIC_TIME
config option and simplify the generic code.
Signed-off-by: John Stultz <johnstul@us.ibm.com>
LKML-Reference: <1279068988-21864-4-git-send-email-johnstul@us.ibm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
|
|
As suggested in Venki's suggestion in the commit 0dc698b,
add LAPIC unstable detection in the acpi_pad drvier too.
Signed-off-by: Chen Gong <gong.chen@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-idle-2.6
* 'idle-release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-idle-2.6:
intel_idle: native hardware cpuidle driver for latest Intel processors
ACPI: acpi_idle: touch TS_POLLING only in the non-MWAIT case
acpi_pad: uses MONITOR/MWAIT, so it doesn't need to clear TS_POLLING
sched: clarify commment for TS_POLLING
ACPI: allow a native cpuidle driver to displace ACPI
cpuidle: make cpuidle_curr_driver static
cpuidle: add cpuidle_unregister_driver() error check
cpuidle: fail to register if !CONFIG_CPU_IDLE
|
|
acpi pad driver kind of aggressively marks TSC as unstable at init
time, on mwait capable and non X86_FEATURE_NONSTOP_TSC systems. This is
irrespective of whether pad driver is ever going to be used on the
system or deep C-states are supported/used. This will affect every user
who just happens to compile in (or get a kernel version which
compiles in) acpi pad driver.
Move mark_tsc_unstable() out of init to the actual idle invocation path
of the pad driver.
There is also another bug/missing_feature in the code that it does not
support 'always running apic timer' and switches to broadcast mode
unconditionally. Shaohua, can you take a look at that please.
Signed-off-by: Venkatesh Pallipadi <venki@google.com>
Signed-off-by: Len Brown <len.brown@intel.com>
|
|
api_pad exclusively uses MONITOR/MWAIT to sleep in idle,
so it does not need the wakeup IPI during idle sleep
that is provoked by clearing TS_POLLING.
Signed-off-by: Len Brown <len.brown@intel.com>
Cc: Shaohua Li <shaohua.li@intel.com>
|
|
cpi_device_class can only be 19 characters and a NULL terminator.
With the current name we get a buffer overflow in acpi_pad_add()
strcpy(acpi_device_class(device), ACPI_PROCESSOR_AGGREGATOR_CLASS);
[akpm@linux-foundation.org: call it acpi_pad, per Shaohua Li]
Signed-off-by: Dan Carpenter <error27@gmail.com>
Cc: walter harms <wharms@bfs.de>
Acked-by: Shaohua Li <shaohua.li@intel.com>
Cc: Len Brown <lenb@kernel.org>
Acked-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Len Brown <len.brown@intel.com>
|
|
implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
|
|
|
|
There are some fixes listed below:
1. When met a bogus BIOS, the return value of cpu number maybe is
a negative value so that acpi_pad_pur get an unexpected result.
2. the return value of function acpi_pad_idle_cpus is useless.
3. enhance the process of create_power_saving_task/destroy_power_saving_task
4. Add more error checks when evaluating _PUR object.
5. one typo fix
Signed-off-by: Chen Gong <gong.chen@linux.intel.com>
Acked-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
|
|
drivers/acpi/acpi_pad.c: In function 'power_saving_thread':
drivers/acpi/acpi_pad.c:103: warning: 'preferred_cpu' may be used uninitialized in this function
Cc: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Len Brown <len.brown@intel.com>
|
|
ACPI 4.0 created the logical "processor aggregator device" as
a mechinism for platforms to ask the OS to force otherwise busy
processors to enter (power saving) idle.
The intent is to lower power consumption to ride-out
transient electrical and thermal emergencies,
rather than powering off the server.
On platforms that can save more power/performance via P-states,
the platform will first exhaust P-states before forcing idle.
However, the relative benefit of P-states vs. idle states
is platform dependent, and thus this driver need not know
or care about it.
This driver does not use the kernel's CPU hot-plug mechanism
because after the transient emergency is over, the system must
be returned to its normal state, and hotplug would permanently
break both cpusets and binding.
So to force idle, the driver creates a power saving thread.
The scheduler will migrate the thread to the preferred CPU.
The thread has max priority and has SCHED_RR policy,
so it can occupy one CPU. To save power, the thread will
invoke the deep C-state entry instructions.
To avoid starvation, the thread will sleep 5% of the time
time for every second (current RT scheduler has threshold
to avoid starvation, but if other CPUs are idle,
the CPU can borrow CPU timer from other,
which makes the mechanism not work here)
Vaidyanathan Srinivasan has proposed scheduler enhancements
to allow injecting idle time into the system. This driver doesn't
depend on those enhancements, but could cut over to them
when they are available.
Peter Z. does not favor upstreaming this driver until
the those scheduler enhancements are in place. However,
we favor upstreaming this driver now because it is useful
now, and can be enhanced over time.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
NACKed-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Len Brown <len.brown@intel.com>
|
