* Re: [PATCH 1/2] thermal: Add support for thermal sensor for Orion SoC
From: Zhang Rui @ 2013-01-04 7:33 UTC (permalink / raw)
To: Andrew Lunn
Cc: Nobuhiro Iwamatsu, linux ARM, linux-pm, Jason Cooper,
Sebastian Hesselbarth, Thomas Petazzoni, jgunthorpe
In-Reply-To: <20121214221159.GE7717@lunn.ch>
On Fri, 2012-12-14 at 23:11 +0100, Andrew Lunn wrote:
> On Sat, Dec 15, 2012 at 06:54:17AM +0900, Nobuhiro Iwamatsu wrote:
> > Hi,
> >
> > Thanks you for your work.
> > Sorry, I dont hava a time at this week about this.
>
> Its not a problem. We have plenty of time before the next merge
> window. I was just interested in seeing it work on my QNAP device, so
> did some of the cleanup work.
>
> > On Fri, Dec 14, 2012 at 8:03 PM, Andrew Lunn <andrew@lunn.ch> wrote:
> > > From: Nobuhiro Iwamatsu <iwamatsu@nigauri.org>
> > >
> > > Some Orion SoC has thermal sensor.
> > > This patch adds support for 88F6282 and 88F6283.
> > >
> > > Signed-off-by: Nobuhiro Iwamatsu <iwamatsu@nigauri.org>
> > > Signed-off-by: Andrew Lunn <andrew@lunn.ch>
> > > ---
> > > .../devicetree/bindings/thermal/orion-thermal.txt | 16 +++
> > > drivers/thermal/Kconfig | 7 ++
> > > drivers/thermal/Makefile | 1 +
> > > drivers/thermal/orion_thermal.c | 133 ++++++++++++++++++++
> > > 4 files changed, 157 insertions(+)
> > > create mode 100644 Documentation/devicetree/bindings/thermal/orion-thermal.txt
> > > create mode 100644 drivers/thermal/orion_thermal.c
> > >
> > > diff --git a/Documentation/devicetree/bindings/thermal/orion-thermal.txt b/Documentation/devicetree/bindings/thermal/orion-thermal.txt
> > > new file mode 100644
> > > index 0000000..5ce925d
> > > --- /dev/null
> > > +++ b/Documentation/devicetree/bindings/thermal/orion-thermal.txt
> > > @@ -0,0 +1,16 @@
> > > +* Orion Thermal
> > > +
> > > +This initial version is for Kirkwood 88F8262 & 88F6283 SoCs, however
> > > +it is expected the driver will sometime in the future be expanded to
> > > +also support Dove, using a different compatibility string.
> > > +
> > > +Required properties:
> > > +- compatible : "marvell,kirkwood-thermal"
> > > +- reg : Address range of the thermal registers
> > > +
> > > +Example:
> > > +
> > > + thermal@10078 {
> > > + compatible = "marvell,kirkwood";
> >
> > compatible = "marvell,kirkwood-thermal"; ?
>
> Yep, my error.
>
> > > + reg = <0x10078 0x4>;
> > > + };
> > > diff --git a/drivers/thermal/Kconfig b/drivers/thermal/Kconfig
> > > index e1cb6bd..3bba13f 100644
> > > --- a/drivers/thermal/Kconfig
> > > +++ b/drivers/thermal/Kconfig
> > > @@ -55,3 +55,10 @@ config EXYNOS_THERMAL
> > > help
> > > If you say yes here you get support for TMU (Thermal Managment
> > > Unit) on SAMSUNG EXYNOS series of SoC.
> > > +
> > > +config ORION_THERMAL
> > > + tristate "Temperature sensor on Marvel Orion SoCs"
> >
> > Marvel -> Marvell
>
> Missed that one, thanks.
>
> Thanks for the Tested-by. I will add it to the next version. I
> started work on Dove support, so i will probably repost when i have
> that ready for testing.
sorry for the late response, can you resend the refreshed version on top
of the thermal next tree?
thanks,
rui
^ permalink raw reply
* Re: [PATCH] drivers/thermal/spear_thermal.c: use devm_clk_get
From: Zhang Rui @ 2013-01-04 7:27 UTC (permalink / raw)
To: Julia Lawall; +Cc: kernel-janitors, linux-pm, linux-kernel
In-Reply-To: <1354876172-1634-1-git-send-email-Julia.Lawall@lip6.fr>
On Fri, 2012-12-07 at 11:29 +0100, Julia Lawall wrote:
> From: Julia Lawall <Julia.Lawall@lip6.fr>
>
> devm_clk_get allocates a resource that is released when a driver detaches.
> This patch uses devm_clk_get for data that is allocated in the probe
> function of a platform device and is only released in the remove function.
>
> Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
>
applied to thermal-next.
thanks,
rui
> ---
> I was not able to compile this code. At one point, devm_clk_get was not
> supported for all architectures. If that is still the case, and the code
> doesn't compile, then just ignore the patch.
>
> drivers/thermal/spear_thermal.c | 7 ++-----
> 1 file changed, 2 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/thermal/spear_thermal.c b/drivers/thermal/spear_thermal.c
> index 6b2d8b2..3c5ee56 100644
> --- a/drivers/thermal/spear_thermal.c
> +++ b/drivers/thermal/spear_thermal.c
> @@ -131,7 +131,7 @@ static int spear_thermal_probe(struct platform_device *pdev)
> return -ENOMEM;
> }
>
> - stdev->clk = clk_get(&pdev->dev, NULL);
> + stdev->clk = devm_clk_get(&pdev->dev, NULL);
> if (IS_ERR(stdev->clk)) {
> dev_err(&pdev->dev, "Can't get clock\n");
> return PTR_ERR(stdev->clk);
> @@ -140,7 +140,7 @@ static int spear_thermal_probe(struct platform_device *pdev)
> ret = clk_enable(stdev->clk);
> if (ret) {
> dev_err(&pdev->dev, "Can't enable clock\n");
> - goto put_clk;
> + return ret;
> }
>
> stdev->flags = val;
> @@ -163,8 +163,6 @@ static int spear_thermal_probe(struct platform_device *pdev)
>
> disable_clk:
> clk_disable(stdev->clk);
> -put_clk:
> - clk_put(stdev->clk);
>
> return ret;
> }
> @@ -183,7 +181,6 @@ static int spear_thermal_exit(struct platform_device *pdev)
> writel_relaxed(actual_mask & ~stdev->flags, stdev->thermal_base);
>
> clk_disable(stdev->clk);
> - clk_put(stdev->clk);
>
> return 0;
> }
>
^ permalink raw reply
* Re: [PATCH v2] thermal: rcar: add .get_trip_type/temp and .notify support
From: Zhang Rui @ 2013-01-04 7:23 UTC (permalink / raw)
To: Kuninori Morimoto; +Cc: Simon, Magnus, linux-pm, Kuninori Morimoto
In-Reply-To: <87fw3nrgu1.wl%kuninori.morimoto.gx@renesas.com>
On Sun, 2012-12-02 at 18:48 -0800, Kuninori Morimoto wrote:
> This patch adds .get_trip_type(), .get_trip_temp(), and .notify()
> on rcar_thermal_zone_ops.
> Driver will try platform power OFF if it reached to
> critical temperature.
>
> Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
applied to thermal-next tree.
thanks,
rui
> ---
> v1 -> v2
>
> - fixup (90 - 1) -> (90)
> - fixup comment /* +90 <= temp <= +135 */ -> /* +90 <= temp */
>
> drivers/thermal/rcar_thermal.c | 68 ++++++++++++++++++++++++++++++++++++++--
> 1 file changed, 65 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/thermal/rcar_thermal.c b/drivers/thermal/rcar_thermal.c
> index 90db951..89979ff 100644
> --- a/drivers/thermal/rcar_thermal.c
> +++ b/drivers/thermal/rcar_thermal.c
> @@ -22,10 +22,13 @@
> #include <linux/io.h>
> #include <linux/module.h>
> #include <linux/platform_device.h>
> +#include <linux/reboot.h>
> #include <linux/slab.h>
> #include <linux/spinlock.h>
> #include <linux/thermal.h>
>
> +#define IDLE_INTERVAL 5000
> +
> #define THSCR 0x2c
> #define THSSR 0x30
>
> @@ -176,8 +179,66 @@ static int rcar_thermal_get_temp(struct thermal_zone_device *zone,
> return 0;
> }
>
> +static int rcar_thermal_get_trip_type(struct thermal_zone_device *zone,
> + int trip, enum thermal_trip_type *type)
> +{
> + struct rcar_thermal_priv *priv = rcar_zone_to_priv(zone);
> +
> + /* see rcar_thermal_get_temp() */
> + switch (trip) {
> + case 0: /* +90 <= temp */
> + *type = THERMAL_TRIP_CRITICAL;
> + break;
> + default:
> + dev_err(priv->dev, "rcar driver trip error\n");
> + return -EINVAL;
> + }
> +
> + return 0;
> +}
> +
> +static int rcar_thermal_get_trip_temp(struct thermal_zone_device *zone,
> + int trip, unsigned long *temp)
> +{
> + struct rcar_thermal_priv *priv = rcar_zone_to_priv(zone);
> +
> + /* see rcar_thermal_get_temp() */
> + switch (trip) {
> + case 0: /* +90 <= temp */
> + *temp = MCELSIUS(90);
> + break;
> + default:
> + dev_err(priv->dev, "rcar driver trip error\n");
> + return -EINVAL;
> + }
> +
> + return 0;
> +}
> +
> +static int rcar_thermal_notify(struct thermal_zone_device *zone,
> + int trip, enum thermal_trip_type type)
> +{
> + struct rcar_thermal_priv *priv = rcar_zone_to_priv(zone);
> +
> + switch (type) {
> + case THERMAL_TRIP_CRITICAL:
> + /* FIXME */
> + dev_warn(priv->dev,
> + "Thermal reached to critical temperature\n");
> + machine_power_off();
> + break;
> + default:
> + break;
> + }
> +
> + return 0;
> +}
> +
> static struct thermal_zone_device_ops rcar_thermal_zone_ops = {
> - .get_temp = rcar_thermal_get_temp,
> + .get_temp = rcar_thermal_get_temp,
> + .get_trip_type = rcar_thermal_get_trip_type,
> + .get_trip_temp = rcar_thermal_get_trip_temp,
> + .notify = rcar_thermal_notify,
> };
>
> /*
> @@ -211,8 +272,9 @@ static int rcar_thermal_probe(struct platform_device *pdev)
> return -ENOMEM;
> }
>
> - zone = thermal_zone_device_register("rcar_thermal", 0, 0, priv,
> - &rcar_thermal_zone_ops, NULL, 0, 0);
> + zone = thermal_zone_device_register("rcar_thermal", 1, 0, priv,
> + &rcar_thermal_zone_ops, NULL, 0,
> + IDLE_INTERVAL);
> if (IS_ERR(zone)) {
> dev_err(&pdev->dev, "thermal zone device is NULL\n");
> return PTR_ERR(zone);
^ permalink raw reply
* Re: [PATCH] cpuidle - fix lock contention in the idle path
From: Daniel Lezcano @ 2013-01-04 6:27 UTC (permalink / raw)
To: rafael.j.wysocki
Cc: Russ Anderson, linux-pm, pdeschrijver, akpm, linux-kernel, rja
In-Reply-To: <20130102211314.GA29447@sgi.com>
On 01/02/2013 10:13 PM, Russ Anderson wrote:
> On Wed, Dec 26, 2012 at 11:01:48AM +0100, Daniel Lezcano wrote:
>> The commit bf4d1b5ddb78f86078ac6ae0415802d5f0c68f92 introduces
>> a lock in the cpuidle_get_cpu_driver function. This function
>> is used in the idle_call function.
>>
>> The problem is the contention with a large number of cpus because
>> they try to access the idle routine at the same time.
>>
>> The lock could be safely removed because of how is used the
>> cpuidle api. The cpuidle_register_driver is called first but
>> until the cpuidle_register_device is not called we don't
>> enter in the cpuidle idle call function because the device
>> is not enabled.
>>
>> The cpuidle_unregister_driver function, leading the a NULL driver,
>> is not called before the cpuidle_unregister_device.
>>
>> This is how is used the cpuidle api from the different drivers.
>>
>> However, a cleanup around the lock and a proper refcounting
>> mechanism should be used to ensure the consistency in the api,
>> like cpuidle_unregister_driver should failed if its refcounting
>> is not 0.
>>
>> These modifications will need some code reorganization and rewrite
>> which does not fit with a fix.
>
> I agree.
>
>> The following patch is a hot fix by returning to the initial behavior
>> by removing the lock when getting the driver.
>
> The patch fixes the problem. Verified on a system with 1024 cpus.
> Thanks.
>
>> Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
>
> Reported-by: Russ Anderson <rja@sgi.com>
> Acked-by: Russ Anderson <rja@sgi.com>
Hi Rafael,
could you consider this patch for merging ?
Thanks
-- Daniel
--
<http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs
Follow Linaro: <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog
^ permalink raw reply
* Re: [PATCH 3/3] cpufreq: Don't use cpu removed during cpufreq_driver_unregister
From: Srivatsa S. Bhat @ 2013-01-04 6:03 UTC (permalink / raw)
To: Viresh Kumar
Cc: rjw, rafael.j.wysocki, linaro-dev, nicolas.pitre, amit.kucheria,
mathieu.poirier, linux-kernel, cpufreq, pdsw-power-team, linux-pm
In-Reply-To: <CAKohpo=Z=nA2Mr23ior8c0Qmt75+bOVLvZwNdc7OXq-+rH1dWQ@mail.gmail.com>
On 01/04/2013 10:49 AM, Viresh Kumar wrote:
> On 3 January 2013 19:55, Srivatsa S. Bhat
> <srivatsa.bhat@linux.vnet.ibm.com> wrote:
>> I took a quick look at the problem you described above, and the cpufreq code..
>> If we cannot avoid calling cpufreq_add_dev() from cpufreq_remove_dev(), then I can't
>> think of anything better than what your patch does.
>
> Good :)
>
>> BTW, off-topic, while going through that path, I think I found a memory leak
>> in __cpufreq_remove_dev():
>>
>> if (unlikely(cpumask_weight(data->cpus) > 1)) {
>> for_each_cpu(j, data->cpus) {
>> if (j == cpu)
>> continue;
>> per_cpu(cpufreq_cpu_data, j) = NULL;
>> }
>> }
>>
>> We are assigning NULL without freeing that memory.
>
> Not really. All cpus in affected_cpus (data->cpus), share the same
> policy structure.
> We have already taken backup of cpufreq_cpu_data for the first cpu in "data" and
> are freeing it here:
>
> kfree(data);
>
Ah, ok, got it. Thanks!
Regards,
Srivatsa S. Bhat
^ permalink raw reply
* Re: [PATCH 3/3] cpufreq: Don't use cpu removed during cpufreq_driver_unregister
From: Viresh Kumar @ 2013-01-04 5:19 UTC (permalink / raw)
To: Srivatsa S. Bhat
Cc: rjw, rafael.j.wysocki, linaro-dev, nicolas.pitre, amit.kucheria,
mathieu.poirier, linux-kernel, cpufreq, pdsw-power-team, linux-pm
In-Reply-To: <50E594F2.4000800@linux.vnet.ibm.com>
On 3 January 2013 19:55, Srivatsa S. Bhat
<srivatsa.bhat@linux.vnet.ibm.com> wrote:
> I took a quick look at the problem you described above, and the cpufreq code..
> If we cannot avoid calling cpufreq_add_dev() from cpufreq_remove_dev(), then I can't
> think of anything better than what your patch does.
Good :)
> BTW, off-topic, while going through that path, I think I found a memory leak
> in __cpufreq_remove_dev():
>
> if (unlikely(cpumask_weight(data->cpus) > 1)) {
> for_each_cpu(j, data->cpus) {
> if (j == cpu)
> continue;
> per_cpu(cpufreq_cpu_data, j) = NULL;
> }
> }
>
> We are assigning NULL without freeing that memory.
Not really. All cpus in affected_cpus (data->cpus), share the same
policy structure.
We have already taken backup of cpufreq_cpu_data for the first cpu in "data" and
are freeing it here:
kfree(data);
^ permalink raw reply
* Re: [PATCH 1/3] cpufreq: Manage only online cpus
From: Viresh Kumar @ 2013-01-04 5:14 UTC (permalink / raw)
To: Rafael J. Wysocki
Cc: linaro-dev, nicolas.pitre, amit.kucheria, mathieu.poirier,
linux-kernel, cpufreq, pdsw-power-team, linux-pm
In-Reply-To: <6008147.vspAd29d9L@vostro.rjw.lan>
On 3 January 2013 17:32, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> True, but have those bugs been introduced recently (ie. in v3.8-rc1 or later)?
Don't know... I feel they were always there, its just that nobody
tested it that way :)
^ permalink raw reply
* Re: [PATCH v5 3/3] PM: Introduce Intel PowerClamp Driver
From: Joe Perches @ 2013-01-04 3:51 UTC (permalink / raw)
To: Jacob Pan
Cc: Linux PM, LKML, Rafael Wysocki, Len Brown, Thomas Gleixner,
H. Peter Anvin, Ingo Molnar, Zhang Rui, Rob Landley,
Arjan van de Ven, Paul McKenney
In-Reply-To: <1357225804-27191-4-git-send-email-jacob.jun.pan@linux.intel.com>
On Thu, 2013-01-03 at 07:10 -0800, Jacob Pan wrote:
> Intel PowerClamp driver performs synchronized idle injection across
> all online CPUs. The goal is to maintain a given package level C-state
> ratio.
[]
> +static int window_size_set(const char *arg, const struct kernel_param *kp)
> +{
> + int ret = 0;
> + unsigned long new_window_size;
> +
> + ret = kstrtoul(arg, 10, &new_window_size);
> + if (ret)
> + goto exit_win;
> + if (new_window_size > 10 || new_window_size < 2) {
> + pr_err("Invalid window size %lu, between 2-10\n",
> + new_window_size);
> + ret = -EINVAL;
> + }
> +
> + window_size = new_window_size;
Possible assignment of known invalid windows size?
Maybe you should add
goto exit;
after
ret = -EINVAL;
or add
new_window_size = clamp(new_window_size, 2ul, 10ul);
^ permalink raw reply
* Re: [PATCH v9 06/10] ata: zpodd: check zero power ready status
From: Aaron Lu @ 2013-01-04 1:04 UTC (permalink / raw)
To: Tejun Heo
Cc: James Bottomley, Rafael J. Wysocki, linux-pm, Jeff Garzik,
Alan Stern, Jeff Wu, Aaron Lu, linux-ide, linux-scsi, linux-acpi
In-Reply-To: <20121228211635.GB3062@mtj.dyndns.org>
On 12/29/2012 05:16 AM, Tejun Heo wrote:
> Hello,
>
> On Wed, Dec 26, 2012 at 09:42:14AM +0800, Aaron Lu wrote:
>>> This is really a round-about way to find out the matching device and
>>> it wouldn't work if the disk device nests deeper. Doesn't really look
>>> like a good idea to me.
>>
>> I don't quite understand the 'disk device nests deeper' case, can you
>> please elaborate? My understanding is, as long as the disk's part0
>> device has a parent, this function should work. For LLDs want to take
>
> Hmmm, maybe I misread but it looked like it wouldn't work if there are
> intermediate nodes between the parent device and part0. It might not
> happen for sata but I don't think it's a good idea to assume that the
> part0 and hardware device are connected directly.
>
> In general, it's a quite roundabout way to do it. Let's just push it
> through SCSI. That's how everything else is routed after all. It's
> confusing to do this one differently.
OK, thanks for the suggestion.
I'll prepare v11 using the previous demonstrated way(adding the
event_driven flag to scsi_device) to silence the poll.
Best regards,
Aaron
^ permalink raw reply
* [PATCH v7u1 31/31] x86, 64bit, mm: hibernate use generic mapping_init
From: Yinghai Lu @ 2013-01-04 0:48 UTC (permalink / raw)
To: Thomas Gleixner, Ingo Molnar, H. Peter Anvin
Cc: Eric W. Biederman, Andrew Morton, Borislav Petkov, Jan Kiszka,
Jason Wessel, linux-kernel, Yinghai Lu, Pavel Machek,
Rafael J. Wysocki, linux-pm
In-Reply-To: <1357260531-11115-1-git-send-email-yinghai@kernel.org>
Make it only map range in pfn_mapped array.
and it has kernel mapping with EXEC.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Cc: linux-pm@vger.kernel.org
---
arch/x86/power/hibernate_64.c | 66 ++++++++++++++---------------------------
1 file changed, 22 insertions(+), 44 deletions(-)
diff --git a/arch/x86/power/hibernate_64.c b/arch/x86/power/hibernate_64.c
index 460f314..a0fde91 100644
--- a/arch/x86/power/hibernate_64.c
+++ b/arch/x86/power/hibernate_64.c
@@ -11,6 +11,8 @@
#include <linux/gfp.h>
#include <linux/smp.h>
#include <linux/suspend.h>
+
+#include <asm/init.h>
#include <asm/proto.h>
#include <asm/page.h>
#include <asm/pgtable.h>
@@ -39,41 +41,21 @@ pgd_t *temp_level4_pgt;
void *relocated_restore_code;
-static int res_phys_pud_init(pud_t *pud, unsigned long address, unsigned long end)
+static void *alloc_pgt_page(void *context)
{
- long i, j;
-
- i = pud_index(address);
- pud = pud + i;
- for (; i < PTRS_PER_PUD; pud++, i++) {
- unsigned long paddr;
- pmd_t *pmd;
-
- paddr = address + i*PUD_SIZE;
- if (paddr >= end)
- break;
-
- pmd = (pmd_t *)get_safe_page(GFP_ATOMIC);
- if (!pmd)
- return -ENOMEM;
- set_pud(pud, __pud(__pa(pmd) | _KERNPG_TABLE));
- for (j = 0; j < PTRS_PER_PMD; pmd++, j++, paddr += PMD_SIZE) {
- unsigned long pe;
-
- if (paddr >= end)
- break;
- pe = __PAGE_KERNEL_LARGE_EXEC | paddr;
- pe &= __supported_pte_mask;
- set_pmd(pmd, __pmd(pe));
- }
- }
- return 0;
+ return (void *)get_safe_page(GFP_ATOMIC);
}
static int set_up_temporary_mappings(void)
{
- unsigned long start, end, next;
- int error;
+ struct x86_mapping_info info = {
+ .alloc_pgt_page = alloc_pgt_page,
+ .pmd_flag = __PAGE_KERNEL_LARGE_EXEC,
+ .kernel_mapping = true,
+ };
+ unsigned long mstart, mend;
+ int result;
+ int i;
temp_level4_pgt = (pgd_t *)get_safe_page(GFP_ATOMIC);
if (!temp_level4_pgt)
@@ -84,21 +66,17 @@ static int set_up_temporary_mappings(void)
init_level4_pgt[pgd_index(__START_KERNEL_map)]);
/* Set up the direct mapping from scratch */
- start = (unsigned long)pfn_to_kaddr(0);
- end = (unsigned long)pfn_to_kaddr(max_pfn);
-
- for (; start < end; start = next) {
- pud_t *pud = (pud_t *)get_safe_page(GFP_ATOMIC);
- if (!pud)
- return -ENOMEM;
- next = start + PGDIR_SIZE;
- if (next > end)
- next = end;
- if ((error = res_phys_pud_init(pud, __pa(start), __pa(next))))
- return error;
- set_pgd(temp_level4_pgt + pgd_index(start),
- mk_kernel_pgd(__pa(pud)));
+ for (i = 0; i < nr_pfn_mapped; i++) {
+ mstart = pfn_mapped[i].start << PAGE_SHIFT;
+ mend = pfn_mapped[i].end << PAGE_SHIFT;
+
+ result = kernel_ident_mapping_init(&info, temp_level4_pgt,
+ mstart, mend);
+
+ if (result)
+ return result;
}
+
return 0;
}
--
1.7.10.4
^ permalink raw reply related
* [PATCH v5 0/3] PM: Intel PowerClamp driver
From: Jacob Pan @ 2013-01-03 15:10 UTC (permalink / raw)
To: Linux PM, LKML, Rafael Wysocki
Cc: Len Brown, Thomas Gleixner, H. Peter Anvin, Ingo Molnar,
Zhang Rui, Rob Landley, Arjan van de Ven, Paul McKenney,
Joe Perches, Jacob Pan
v5 changes:
- use bool for clamping flag.
- removed duplicated #define MSR_PKG_Cx_RESIDENCY
- adjust window_size range test to match comments, which is between
2 and 10.
We have done some experiment with idle injection on Intel platforms.
The idea is to use the increasingly power efficient package level
C-states for power capping and passive thermal control.
Documentation is included in the patch to explain the theory of
operation, performance implication, calibration, scalability, and user
interface. Please refer to the following file for more details.
Documentation/thermal/intel_powerclamp.txt
Arjan van de Ven created the original idea and driver, I have been
refining driver in hope that they can be to be useful beyond a proof
of concept.
Jacob Pan (3):
tick: export nohz tick idle symbols for module use
x86/nmi: export local_touch_nmi() symbol for modules
PM: Introduce Intel PowerClamp Driver
Documentation/thermal/intel_powerclamp.txt | 307 +++++++++++
arch/x86/kernel/nmi.c | 1 +
drivers/thermal/Kconfig | 10 +
drivers/thermal/Makefile | 2 +
drivers/thermal/intel_powerclamp.c | 758 ++++++++++++++++++++++++++++
kernel/time/tick-sched.c | 2 +
6 files changed, 1080 insertions(+)
create mode 100644 Documentation/thermal/intel_powerclamp.txt
create mode 100644 drivers/thermal/intel_powerclamp.c
--
1.7.9.5
^ permalink raw reply
* [PATCH v5 1/3] tick: export nohz tick idle symbols for module use
From: Jacob Pan @ 2013-01-03 15:10 UTC (permalink / raw)
To: Linux PM, LKML, Rafael Wysocki
Cc: Len Brown, Thomas Gleixner, H. Peter Anvin, Ingo Molnar,
Zhang Rui, Rob Landley, Arjan van de Ven, Paul McKenney,
Joe Perches, Jacob Pan
In-Reply-To: <1357225804-27191-1-git-send-email-jacob.jun.pan@linux.intel.com>
Allow drivers such as intel_powerclamp to use these apis for
turning on/off ticks during idle.
Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com>
---
kernel/time/tick-sched.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index d58e552..a767757 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -553,6 +553,7 @@ void tick_nohz_idle_enter(void)
local_irq_enable();
}
+EXPORT_SYMBOL_GPL(tick_nohz_idle_enter);
/**
* tick_nohz_irq_exit - update next tick event from interrupt exit
@@ -681,6 +682,7 @@ void tick_nohz_idle_exit(void)
local_irq_enable();
}
+EXPORT_SYMBOL_GPL(tick_nohz_idle_exit);
static int tick_nohz_reprogram(struct tick_sched *ts, ktime_t now)
{
--
1.7.9.5
^ permalink raw reply related
* Re: [PATCH v4 3/3] PM: Introduce Intel PowerClamp Driver
From: Jacob Pan @ 2013-01-03 15:03 UTC (permalink / raw)
To: Joe Perches
Cc: Linux PM, LKML, Rafael Wysocki, Len Brown, Thomas Gleixner,
H. Peter Anvin, Ingo Molnar, Zhang Rui, Rob Landley,
Arjan van de Ven, Paul McKenney, Peter Zijlstra
In-Reply-To: <1357241654.5452.16.camel@joe-AO722>
On Thu, 03 Jan 2013 11:34:14 -0800
Joe Perches <joe@perches.com> wrote:
> On Thu, 2013-01-03 at 03:01 -0800, Jacob Pan wrote:
> > Intel PowerClamp driver performs synchronized idle injection across
> > all online CPUs. The goal is to maintain a given package level
> > C-state ratio.
>
> just trivia:
>
> > diff --git a/drivers/thermal/intel_powerclamp.c
> > b/drivers/thermal/intel_powerclamp.c
>
> > +static int clamping;
>
> bool and use true/false?
sure. will change in v5
>
> > +static int window_size_set(const char *arg, const struct
> > kernel_param *kp)
> []
> > + if (new_window_size >= 10 || new_window_size < 2) {
> > + pr_err("Invalid window size %lu, between 2-10\n",
> > + new_window_size);
>
> This is a misleading message or a bad test,
> It's from 2 to 9
>
> This would more normally be tested/written as
>
> val < low || val > high
>
will change to make the range match error message, between 2 to 10.
Thanks,
Jacob
^ permalink raw reply
* Re: [PATCH v4 3/3] PM: Introduce Intel PowerClamp Driver
From: Joe Perches @ 2013-01-03 19:34 UTC (permalink / raw)
To: Jacob Pan
Cc: Linux PM, LKML, Rafael Wysocki, Len Brown, Thomas Gleixner,
H. Peter Anvin, Ingo Molnar, Zhang Rui, Rob Landley,
Arjan van de Ven, Paul McKenney, Peter Zijlstra
In-Reply-To: <1357210875-5025-4-git-send-email-jacob.jun.pan@linux.intel.com>
On Thu, 2013-01-03 at 03:01 -0800, Jacob Pan wrote:
> Intel PowerClamp driver performs synchronized idle injection across
> all online CPUs. The goal is to maintain a given package level C-state
> ratio.
just trivia:
> diff --git a/drivers/thermal/intel_powerclamp.c b/drivers/thermal/intel_powerclamp.c
> +static int clamping;
bool and use true/false?
> +static int window_size_set(const char *arg, const struct kernel_param *kp)
[]
> + if (new_window_size >= 10 || new_window_size < 2) {
> + pr_err("Invalid window size %lu, between 2-10\n",
> + new_window_size);
This is a misleading message or a bad test,
It's from 2 to 9
This would more normally be tested/written as
val < low || val > high
^ permalink raw reply
* [PATCH v4 1/3] tick: export nohz tick idle symbols for module use
From: Jacob Pan @ 2013-01-03 11:01 UTC (permalink / raw)
To: Linux PM, LKML, Rafael Wysocki
Cc: Len Brown, Thomas Gleixner, H. Peter Anvin, Ingo Molnar,
Zhang Rui, Rob Landley, Arjan van de Ven, Paul McKenney,
Peter Zijlstra, Joe Perches, Jacob Pan
In-Reply-To: <1357210875-5025-1-git-send-email-jacob.jun.pan@linux.intel.com>
Allow drivers such as intel_powerclamp to use these apis for
turning on/off ticks during idle.
Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com>
---
kernel/time/tick-sched.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index d58e552..a767757 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -553,6 +553,7 @@ void tick_nohz_idle_enter(void)
local_irq_enable();
}
+EXPORT_SYMBOL_GPL(tick_nohz_idle_enter);
/**
* tick_nohz_irq_exit - update next tick event from interrupt exit
@@ -681,6 +682,7 @@ void tick_nohz_idle_exit(void)
local_irq_enable();
}
+EXPORT_SYMBOL_GPL(tick_nohz_idle_exit);
static int tick_nohz_reprogram(struct tick_sched *ts, ktime_t now)
{
--
1.7.9.5
^ permalink raw reply related
* [PATCH v4 0/3] PM: Intel PowerClamp driver
From: Jacob Pan @ 2013-01-03 11:01 UTC (permalink / raw)
To: Linux PM, LKML, Rafael Wysocki
Cc: Len Brown, Thomas Gleixner, H. Peter Anvin, Ingo Molnar,
Zhang Rui, Rob Landley, Arjan van de Ven, Paul McKenney,
Peter Zijlstra, Joe Perches, Jacob Pan
v4 changes:
More cosmetic changes since v3 based on
https://lkml.org/lkml/2012/11/26/778
We have done some experiment with idle injection on Intel platforms.
The idea is to use the increasingly power efficient package level
C-states for power capping and passive thermal control.
Documentation is included in the patch to explain the theory of
operation, performance implication, calibration, scalability, and user
interface. Please refer to the following file for more details.
Documentation/thermal/intel_powerclamp.txt
Arjan van de Ven created the original idea and driver, I have been
refining driver in hope that they can be to be useful beyond a proof
of concept.
Jacob Pan (3):
tick: export nohz tick idle symbols for module use
x86/nmi: export local_touch_nmi() symbol for modules
PM: Introduce Intel PowerClamp Driver
Documentation/thermal/intel_powerclamp.txt | 307 +++++++++++
arch/x86/kernel/nmi.c | 1 +
drivers/thermal/Kconfig | 10 +
drivers/thermal/Makefile | 2 +
drivers/thermal/intel_powerclamp.c | 762 ++++++++++++++++++++++++++++
kernel/time/tick-sched.c | 2 +
6 files changed, 1084 insertions(+)
create mode 100644 Documentation/thermal/intel_powerclamp.txt
create mode 100644 drivers/thermal/intel_powerclamp.c
--
1.7.9.5
^ permalink raw reply
* [PATCH v5 3/3] PM: Introduce Intel PowerClamp Driver
From: Jacob Pan @ 2013-01-03 15:10 UTC (permalink / raw)
To: Linux PM, LKML, Rafael Wysocki
Cc: Len Brown, Thomas Gleixner, H. Peter Anvin, Ingo Molnar,
Zhang Rui, Rob Landley, Arjan van de Ven, Paul McKenney,
Joe Perches, Jacob Pan
In-Reply-To: <1357225804-27191-1-git-send-email-jacob.jun.pan@linux.intel.com>
Intel PowerClamp driver performs synchronized idle injection across
all online CPUs. The goal is to maintain a given package level C-state
ratio.
Compared to other throttling methods already exist in the kernel,
such as ACPI PAD (taking CPUs offline) and clock modulation, this is often
more efficient in terms of performance per watt.
Please refer to Documentation/thermal/intel_powerclamp.txt for more details.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com>
---
Documentation/thermal/intel_powerclamp.txt | 307 +++++++++++
drivers/thermal/Kconfig | 10 +
drivers/thermal/Makefile | 2 +
drivers/thermal/intel_powerclamp.c | 758 ++++++++++++++++++++++++++++
4 files changed, 1077 insertions(+)
create mode 100644 Documentation/thermal/intel_powerclamp.txt
create mode 100644 drivers/thermal/intel_powerclamp.c
diff --git a/Documentation/thermal/intel_powerclamp.txt b/Documentation/thermal/intel_powerclamp.txt
new file mode 100644
index 0000000..332de4a
--- /dev/null
+++ b/Documentation/thermal/intel_powerclamp.txt
@@ -0,0 +1,307 @@
+ =======================
+ INTEL POWERCLAMP DRIVER
+ =======================
+By: Arjan van de Ven <arjan@linux.intel.com>
+ Jacob Pan <jacob.jun.pan@linux.intel.com>
+
+Contents:
+ (*) Introduction
+ - Goals and Objectives
+
+ (*) Theory of Operation
+ - Idle Injection
+ - Calibration
+
+ (*) Performance Analysis
+ - Effectiveness and Limitations
+ - Power vs Performance
+ - Scalability
+ - Calibration
+ - Comparison with Alternative Techniques
+
+ (*) Usage and Interfaces
+ - Generic Thermal Layer (sysfs)
+ - Kernel APIs (TBD)
+
+============
+INTRODUCTION
+============
+
+Consider the situation where a system’s power consumption must be
+reduced at runtime, due to power budget, thermal constraint, or noise
+level, and where active cooling is not preferred. Software managed
+passive power reduction must be performed to prevent the hardware
+actions that are designed for catastrophic scenarios.
+
+Currently, P-states, T-states (clock modulation), and CPU offlining
+are used for CPU throttling.
+
+On Intel CPUs, C-states provide effective power reduction, but so far
+they’re only used opportunistically, based on workload. With the
+development of intel_powerclamp driver, the method of synchronizing
+idle injection across all online CPU threads was introduced. The goal
+is to achieve forced and controllable C-state residency.
+
+Test/Analysis has been made in the areas of power, performance,
+scalability, and user experience. In many cases, clear advantage is
+shown over taking the CPU offline or modulating the CPU clock.
+
+
+===================
+THEORY OF OPERATION
+===================
+
+Idle Injection
+--------------
+
+On modern Intel processors (Nehalem or later), package level C-state
+residency is available in MSRs, thus also available to the kernel.
+
+These MSRs are:
+ #define MSR_PKG_C2_RESIDENCY 0x60D
+ #define MSR_PKG_C3_RESIDENCY 0x3F8
+ #define MSR_PKG_C6_RESIDENCY 0x3F9
+ #define MSR_PKG_C7_RESIDENCY 0x3FA
+
+If the kernel can also inject idle time to the system, then a
+closed-loop control system can be established that manages package
+level C-state. The intel_powerclamp driver is conceived as such a
+control system, where the target set point is a user-selected idle
+ratio (based on power reduction), and the error is the difference
+between the actual package level C-state residency ratio and the target idle
+ratio.
+
+Injection is controlled by high priority kernel threads, spawned for
+each online CPU.
+
+These kernel threads, with SCHED_FIFO class, are created to perform
+clamping actions of controlled duty ratio and duration. Each per-CPU
+thread synchronizes its idle time and duration, based on the rounding
+of jiffies, so accumulated errors can be prevented to avoid a jittery
+effect. Threads are also bound to the CPU such that they cannot be
+migrated, unless the CPU is taken offline. In this case, threads
+belong to the offlined CPUs will be terminated immediately.
+
+Running as SCHED_FIFO and relatively high priority, also allows such
+scheme to work for both preemptable and non-preemptable kernels.
+Alignment of idle time around jiffies ensures scalability for HZ
+values. This effect can be better visualized using a Perf timechart.
+The following diagram shows the behavior of kernel thread
+kidle_inject/cpu. During idle injection, it runs monitor/mwait idle
+for a given "duration", then relinquishes the CPU to other tasks,
+until the next time interval.
+
+The NOHZ schedule tick is disabled during idle time, but interrupts
+are not masked. Tests show that the extra wakeups from scheduler tick
+have a dramatic impact on the effectiveness of the powerclamp driver
+on large scale systems (Westmere system with 80 processors).
+
+CPU0
+ ____________ ____________
+kidle_inject/0 | sleep | mwait | sleep |
+ _________| |________| |_______
+ duration
+CPU1
+ ____________ ____________
+kidle_inject/1 | sleep | mwait | sleep |
+ _________| |________| |_______
+ ^
+ |
+ |
+ roundup(jiffies, interval)
+
+Only one CPU is allowed to collect statistics and update global
+control parameters. This CPU is referred to as the controlling CPU in
+this document. The controlling CPU is elected at runtime, with a
+policy that favors BSP, taking into account the possibility of a CPU
+hot-plug.
+
+In terms of dynamics of the idle control system, package level idle
+time is considered largely as a non-causal system where its behavior
+cannot be based on the past or current input. Therefore, the
+intel_powerclamp driver attempts to enforce the desired idle time
+instantly as given input (target idle ratio). After injection,
+powerclamp moniors the actual idle for a given time window and adjust
+the next injection accordingly to avoid over/under correction.
+
+When used in a causal control system, such as a temperature control,
+it is up to the user of this driver to implement algorithms where
+past samples and outputs are included in the feedback. For example, a
+PID-based thermal controller can use the powerclamp driver to
+maintain a desired target temperature, based on integral and
+derivative gains of the past samples.
+
+
+
+Calibration
+-----------
+During scalability testing, it is observed that synchronized actions
+among CPUs become challenging as the number of cores grows. This is
+also true for the ability of a system to enter package level C-states.
+
+To make sure the intel_powerclamp driver scales well, online
+calibration is implemented. The goals for doing such a calibration
+are:
+
+a) determine the effective range of idle injection ratio
+b) determine the amount of compensation needed at each target ratio
+
+Compensation to each target ratio consists of two parts:
+
+ a) steady state error compensation
+ This is to offset the error occurring when the system can
+ enter idle without extra wakeups (such as external interrupts).
+
+ b) dynamic error compensation
+ When an excessive amount of wakeups occurs during idle, an
+ additional idle ratio can be added to quiet interrupts, by
+ slowing down CPU activities.
+
+A debugfs file is provided for the user to examine compensation
+progress and results, such as on a Westmere system.
+[jacob@nex01 ~]$ cat
+/sys/kernel/debug/intel_powerclamp/powerclamp_calib
+controlling cpu: 0
+pct confidence steady dynamic (compensation)
+0 0 0 0
+1 1 0 0
+2 1 1 0
+3 3 1 0
+4 3 1 0
+5 3 1 0
+6 3 1 0
+7 3 1 0
+8 3 1 0
+...
+30 3 2 0
+31 3 2 0
+32 3 1 0
+33 3 2 0
+34 3 1 0
+35 3 2 0
+36 3 1 0
+37 3 2 0
+38 3 1 0
+39 3 2 0
+40 3 3 0
+41 3 1 0
+42 3 2 0
+43 3 1 0
+44 3 1 0
+45 3 2 0
+46 3 3 0
+47 3 0 0
+48 3 2 0
+49 3 3 0
+
+Calibration occurs during runtime. No offline method is available.
+Steady state compensation is used only when confidence levels of all
+adjacent ratios have reached satisfactory level. A confidence level
+is accumulated based on clean data collected at runtime. Data
+collected during a period without extra interrupts is considered
+clean.
+
+To compensate for excessive amounts of wakeup during idle, additional
+idle time is injected when such a condition is detected. Currently,
+we have a simple algorithm to double the injection ratio. A possible
+enhancement might be to throttle the offending IRQ, such as delaying
+EOI for level triggered interrupts. But it is a challenge to be
+non-intrusive to the scheduler or the IRQ core code.
+
+
+CPU Online/Offline
+------------------
+Per-CPU kernel threads are started/stopped upon receiving
+notifications of CPU hotplug activities. The intel_powerclamp driver
+keeps track of clamping kernel threads, even after they are migrated
+to other CPUs, after a CPU offline event.
+
+
+=====================
+Performance Analysis
+=====================
+This section describes the general performance data collected on
+multiple systems, including Westmere (80P) and Ivy Bridge (4P, 8P).
+
+Effectiveness and Limitations
+-----------------------------
+The maximum range that idle injection is allowed is capped at 50
+percent. As mentioned earlier, since interrupts are allowed during
+forced idle time, excessive interrupts could result in less
+effectiveness. The extreme case would be doing a ping -f to generated
+flooded network interrupts without much CPU acknowledgement. In this
+case, little can be done from the idle injection threads. In most
+normal cases, such as scp a large file, applications can be throttled
+by the powerclamp driver, since slowing down the CPU also slows down
+network protocol processing, which in turn reduces interrupts.
+
+When control parameters change at runtime by the controlling CPU, it
+may take an additional period for the rest of the CPUs to catch up
+with the changes. During this time, idle injection is out of sync,
+thus not able to enter package C- states at the expected ratio. But
+this effect is minor, in that in most cases change to the target
+ratio is updated much less frequently than the idle injection
+frequency.
+
+Scalability
+-----------
+Tests also show a minor, but measurable, difference between the 4P/8P
+Ivy Bridge system and the 80P Westmere server under 50% idle ratio.
+More compensation is needed on Westmere for the same amount of
+target idle ratio. The compensation also increases as the idle ratio
+gets larger. The above reason constitutes the need for the
+calibration code.
+
+On the IVB 8P system, compared to an offline CPU, powerclamp can
+achieve up to 40% better performance per watt. (measured by a spin
+counter summed over per CPU counting threads spawned for all running
+CPUs).
+
+====================
+Usage and Interfaces
+====================
+The powerclamp driver is registered to the generic thermal layer as a
+cooling device. Currently, it’s not bound to any thermal zones.
+
+jacob@chromoly:/sys/class/thermal/cooling_device14$ grep . *
+cur_state:0
+max_state:50
+type:intel_powerclamp
+
+Example usage:
+- To inject 25% idle time
+$ sudo sh -c "echo 25 > /sys/class/thermal/cooling_device80/cur_state
+"
+
+If the system is not busy and has more than 25% idle time already,
+then the powerclamp driver will not start idle injection. Using Top
+will not show idle injection kernel threads.
+
+If the system is busy (spin test below) and has less than 25% natural
+idle time, powerclamp kernel threads will do idle injection, which
+appear running to the scheduler. But the overall system idle is still
+reflected. In this example, 24.1% idle is shown. This helps the
+system admin or user determine the cause of slowdown, when a
+powerclamp driver is in action.
+
+
+Tasks: 197 total, 1 running, 196 sleeping, 0 stopped, 0 zombie
+Cpu(s): 71.2%us, 4.7%sy, 0.0%ni, 24.1%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
+Mem: 3943228k total, 1689632k used, 2253596k free, 74960k buffers
+Swap: 4087804k total, 0k used, 4087804k free, 945336k cached
+
+ PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
+ 3352 jacob 20 0 262m 644 428 S 286 0.0 0:17.16 spin
+ 3341 root -51 0 0 0 0 D 25 0.0 0:01.62 kidle_inject/0
+ 3344 root -51 0 0 0 0 D 25 0.0 0:01.60 kidle_inject/3
+ 3342 root -51 0 0 0 0 D 25 0.0 0:01.61 kidle_inject/1
+ 3343 root -51 0 0 0 0 D 25 0.0 0:01.60 kidle_inject/2
+ 2935 jacob 20 0 696m 125m 35m S 5 3.3 0:31.11 firefox
+ 1546 root 20 0 158m 20m 6640 S 3 0.5 0:26.97 Xorg
+ 2100 jacob 20 0 1223m 88m 30m S 3 2.3 0:23.68 compiz
+
+Tests have shown that by using the powerclamp driver as a cooling
+device, a PID based userspace thermal controller can manage to
+control CPU temperature effectively, when no other thermal influence
+is added. For example, a UltraBook user can compile the kernel under
+certain temperature (below most active trip points).
diff --git a/drivers/thermal/Kconfig b/drivers/thermal/Kconfig
index c2c77d1..7d90ab8 100644
--- a/drivers/thermal/Kconfig
+++ b/drivers/thermal/Kconfig
@@ -122,4 +122,14 @@ config DB8500_CPUFREQ_COOLING
bound cpufreq cooling device turns active to set CPU frequency low to
cool down the CPU.
+config INTEL_POWERCLAMP
+ tristate "Intel PowerClamp idle injection driver"
+ depends on THERMAL
+ depends on X86
+ depends on CPU_SUP_INTEL
+ help
+ Enable this to enable Intel PowerClamp idle injection driver. This
+ enforce idle time which results in more package C-state residency. The
+ user interface is exposed via generic thermal framework.
+
endif
diff --git a/drivers/thermal/Makefile b/drivers/thermal/Makefile
index d8da683..574f5f5 100644
--- a/drivers/thermal/Makefile
+++ b/drivers/thermal/Makefile
@@ -18,3 +18,5 @@ obj-$(CONFIG_RCAR_THERMAL) += rcar_thermal.o
obj-$(CONFIG_EXYNOS_THERMAL) += exynos_thermal.o
obj-$(CONFIG_DB8500_THERMAL) += db8500_thermal.o
obj-$(CONFIG_DB8500_CPUFREQ_COOLING) += db8500_cpufreq_cooling.o
+obj-$(CONFIG_INTEL_POWERCLAMP) += intel_powerclamp.o
+
diff --git a/drivers/thermal/intel_powerclamp.c b/drivers/thermal/intel_powerclamp.c
new file mode 100644
index 0000000..198cafa
--- /dev/null
+++ b/drivers/thermal/intel_powerclamp.c
@@ -0,0 +1,758 @@
+/*
+ * intel_powerclamp.c - package c-state idle injection
+ *
+ * Copyright (c) 2012, Intel Corporation.
+ *
+ * Authors:
+ * Arjan van de Ven <arjan@linux.intel.com>
+ * Jacob Pan <jacob.jun.pan@linux.intel.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ *
+ * TODO:
+ * 1. better handle wakeup from external interrupts, currently a fixed
+ * compensation is added to clamping duration when excessive amount
+ * of wakeups are observed during idle time. the reason is that in
+ * case of external interrupts without need for ack, clamping down
+ * cpu in non-irq context does not reduce irq. for majority of the
+ * cases, clamping down cpu does help reduce irq as well, we should
+ * be able to differenciate the two cases and give a quantitative
+ * solution for the irqs that we can control. perhaps based on
+ * get_cpu_iowait_time_us()
+ *
+ * 2. synchronization with other hw blocks
+ *
+ *
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/kthread.h>
+#include <linux/freezer.h>
+#include <linux/cpu.h>
+#include <linux/thermal.h>
+#include <linux/slab.h>
+#include <linux/tick.h>
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#include <linux/nmi.h>
+
+#include <asm/msr.h>
+#include <asm/mwait.h>
+#include <asm/cpu_device_id.h>
+#include <asm/idle.h>
+#include <asm/hardirq.h>
+
+#define MAX_TARGET_RATIO (50U)
+/* For each undisturbed clamping period (no extra wake ups during idle time),
+ * we increment the confidence counter for the given target ratio.
+ * CONFIDENCE_OK defines the level where runtime calibration results are
+ * valid.
+ */
+#define CONFIDENCE_OK (3)
+/* Default idle injection duration, driver adjust sleep time to meet target
+ * idle ratio. Similar to frequency modulation.
+ */
+#define DEFAULT_DURATION_JIFFIES (6)
+
+static unsigned int target_mwait;
+static struct dentry *debug_dir;
+
+/* user selected target */
+static unsigned int set_target_ratio;
+static unsigned int current_ratio;
+static bool should_skip;
+static bool reduce_irq;
+static atomic_t idle_wakeup_counter;
+static unsigned int control_cpu; /* The cpu assigned to collect stat and update
+ * control parameters. default to BSP but BSP
+ * can be offlined.
+ */
+static bool clamping;
+
+
+static struct task_struct __percpu **powerclamp_thread;
+static struct thermal_cooling_device *cooling_dev;
+static unsigned long *cpu_clamping_mask; /* bit map for tracking per cpu
+ * clamping thread
+ */
+static int duration;
+module_param(duration, int, 0600);
+MODULE_PARM_DESC(duration, "forced idle time for each attempt in msec.");
+
+static unsigned int pkg_cstate_ratio_cur;
+static unsigned int window_size;
+
+struct powerclamp_calibration_data {
+ unsigned long confidence; /* used for calibration, basically a counter
+ * gets incremented each time a clamping
+ * period is completed without extra wakeups
+ * once that counter is reached given level,
+ * compensation is deemed usable.
+ */
+ unsigned long steady_comp; /* steady state compensation used when
+ * no extra wakeups occurred.
+ */
+ unsigned long dynamic_comp; /* compensate excessive wakeup from idle
+ * mostly from external interrupts.
+ */
+};
+
+static struct powerclamp_calibration_data cal_data[MAX_TARGET_RATIO];
+
+static int window_size_set(const char *arg, const struct kernel_param *kp)
+{
+ int ret = 0;
+ unsigned long new_window_size;
+
+ ret = kstrtoul(arg, 10, &new_window_size);
+ if (ret)
+ goto exit_win;
+ if (new_window_size > 10 || new_window_size < 2) {
+ pr_err("Invalid window size %lu, between 2-10\n",
+ new_window_size);
+ ret = -EINVAL;
+ }
+
+ window_size = new_window_size;
+ smp_mb();
+
+exit_win:
+
+ return ret;
+}
+static struct kernel_param_ops window_size_ops = {
+ .set = window_size_set,
+ .get = param_get_int,
+};
+
+module_param_cb(window_size, &window_size_ops, &window_size, 0644);
+MODULE_PARM_DESC(window_size, "sliding window in number of clamping cycles\n"
+ "\tpowerclamp controls idle ratio within this window. larger\n"
+ "\twindow size results in slower response time but more smooth\n"
+ "\tclamping results. default to 2.");
+
+static void find_target_mwait(void)
+{
+ unsigned int eax, ebx, ecx, edx;
+ unsigned int highest_cstate = 0;
+ unsigned int highest_subcstate = 0;
+ int i;
+
+ if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
+ return;
+
+ cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx);
+
+ if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) ||
+ !(ecx & CPUID5_ECX_INTERRUPT_BREAK))
+ return;
+
+ edx >>= MWAIT_SUBSTATE_SIZE;
+ for (i = 0; i < 7 && edx; i++, edx >>= MWAIT_SUBSTATE_SIZE) {
+ if (edx & MWAIT_SUBSTATE_MASK) {
+ highest_cstate = i;
+ highest_subcstate = edx & MWAIT_SUBSTATE_MASK;
+ }
+ }
+ target_mwait = (highest_cstate << MWAIT_SUBSTATE_SIZE) |
+ (highest_subcstate - 1);
+
+}
+
+static u64 pkg_state_counter(void)
+{
+ u64 val;
+ u64 count = 0;
+
+ static bool skip_c2;
+ static bool skip_c3;
+ static bool skip_c6;
+ static bool skip_c7;
+
+ if (!skip_c2) {
+ if (!rdmsrl_safe(MSR_PKG_C2_RESIDENCY, &val))
+ count += val;
+ else
+ skip_c2 = true;
+ }
+
+ if (!skip_c3) {
+ if (!rdmsrl_safe(MSR_PKG_C3_RESIDENCY, &val))
+ count += val;
+ else
+ skip_c3 = true;
+ }
+
+ if (!skip_c6) {
+ if (!rdmsrl_safe(MSR_PKG_C6_RESIDENCY, &val))
+ count += val;
+ else
+ skip_c6 = true;
+ }
+
+ if (!skip_c7) {
+ if (!rdmsrl_safe(MSR_PKG_C7_RESIDENCY, &val))
+ count += val;
+ else
+ skip_c7 = true;
+ }
+
+ return count;
+}
+
+static void noop_timer(unsigned long foo)
+{
+ /* empty... just the fact that we get the interrupt wakes us up */
+}
+
+static unsigned int get_compensation(int ratio)
+{
+ unsigned int comp = 0;
+
+ /* we only use compensation if all adjacent ones are good */
+ if (ratio == 1 &&
+ cal_data[ratio].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio + 1].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio + 2].confidence >= CONFIDENCE_OK) {
+ comp = (cal_data[ratio].steady_comp +
+ cal_data[ratio + 1].steady_comp +
+ cal_data[ratio + 2].steady_comp) / 3;
+ } else if (ratio == MAX_TARGET_RATIO - 1 &&
+ cal_data[ratio].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio - 1].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio - 2].confidence >= CONFIDENCE_OK) {
+ comp = (cal_data[ratio].steady_comp +
+ cal_data[ratio - 1].steady_comp +
+ cal_data[ratio - 2].steady_comp) / 3;
+ } else if (cal_data[ratio].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio - 1].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio + 1].confidence >= CONFIDENCE_OK) {
+ comp = (cal_data[ratio].steady_comp +
+ cal_data[ratio - 1].steady_comp +
+ cal_data[ratio + 1].steady_comp) / 3;
+ }
+
+ /* REVISIT: simple penalty of double idle injection */
+ if (reduce_irq)
+ comp = ratio;
+ /* do not exceed limit */
+ if (comp + ratio >= MAX_TARGET_RATIO)
+ comp = MAX_TARGET_RATIO - ratio - 1;
+
+ return comp;
+}
+
+static void adjust_compensation(int target_ratio, unsigned int win)
+{
+ int delta;
+ struct powerclamp_calibration_data *d = &cal_data[target_ratio];
+
+ /*
+ * adjust compensations if confidence level has not been reached or
+ * there are too many wakeups during the last idle injection period, we
+ * cannot trust the data for compensation.
+ */
+ if (d->confidence >= CONFIDENCE_OK ||
+ atomic_read(&idle_wakeup_counter) >
+ win * num_online_cpus())
+ return;
+
+ delta = set_target_ratio - current_ratio;
+ /* filter out bad data */
+ if (delta >= 0 && delta <= (1+target_ratio/10)) {
+ if (d->steady_comp)
+ d->steady_comp =
+ roundup(delta+d->steady_comp, 2)/2;
+ else
+ d->steady_comp = delta;
+ d->confidence++;
+ }
+}
+
+static bool powerclamp_adjust_controls(unsigned int target_ratio,
+ unsigned int guard, unsigned int win)
+{
+ static u64 msr_last, tsc_last;
+ u64 msr_now, tsc_now;
+
+ /* check result for the last window */
+ msr_now = pkg_state_counter();
+ rdtscll(tsc_now);
+
+ /* calculate pkg cstate vs tsc ratio */
+ if (!msr_last || !tsc_last)
+ current_ratio = 1;
+ else if (tsc_now-tsc_last)
+ current_ratio = 100*(msr_now-msr_last)/
+ (tsc_now-tsc_last);
+
+ /* update record */
+ msr_last = msr_now;
+ tsc_last = tsc_now;
+
+ adjust_compensation(target_ratio, win);
+ /*
+ * too many external interrupts, set flag such
+ * that we can take measure later.
+ */
+ reduce_irq = atomic_read(&idle_wakeup_counter) >=
+ 2 * win * num_online_cpus();
+
+ atomic_set(&idle_wakeup_counter, 0);
+ /* if we are above target+guard, skip */
+ return set_target_ratio + guard <= current_ratio;
+}
+
+static int clamp_thread(void *arg)
+{
+ int cpunr = (unsigned long)arg;
+ DEFINE_TIMER(wakeup_timer, noop_timer, 0, 0);
+ static const struct sched_param param = {
+ .sched_priority = MAX_USER_RT_PRIO/2,
+ };
+ unsigned int count = 0;
+ unsigned int target_ratio;
+
+ set_bit(cpunr, cpu_clamping_mask);
+ set_freezable();
+ init_timer_on_stack(&wakeup_timer);
+ sched_setscheduler(current, SCHED_FIFO, ¶m);
+
+ while (true == clamping && !kthread_should_stop() &&
+ cpu_online(cpunr)) {
+ int sleeptime;
+ unsigned long target_jiffies;
+ unsigned int guard;
+ unsigned int compensation = 0;
+ int interval; /* jiffies to sleep for each attempt */
+ unsigned int duration_jiffies = msecs_to_jiffies(duration);
+ unsigned int window_size_now;
+
+ try_to_freeze();
+ /*
+ * make sure user selected ratio does not take effect until
+ * the next round. adjust target_ratio if user has changed
+ * target such that we can converge quickly.
+ */
+ target_ratio = set_target_ratio;
+ guard = 1 + target_ratio/20;
+ window_size_now = window_size;
+ count++;
+
+ /*
+ * systems may have different ability to enter package level
+ * c-states, thus we need to compensate the injected idle ratio
+ * to achieve the actual target reported by the HW.
+ */
+ compensation = get_compensation(target_ratio);
+ interval = duration_jiffies*100/(target_ratio+compensation);
+
+ /* align idle time */
+ target_jiffies = roundup(jiffies, interval);
+ sleeptime = target_jiffies - jiffies;
+ if (sleeptime <= 0)
+ sleeptime = 1;
+ schedule_timeout_interruptible(sleeptime);
+ /*
+ * only elected controlling cpu can collect stats and update
+ * control parameters.
+ */
+ if (cpunr == control_cpu && !(count%window_size_now)) {
+ should_skip =
+ powerclamp_adjust_controls(target_ratio,
+ guard, window_size_now);
+ smp_mb();
+ }
+
+ if (should_skip)
+ continue;
+
+ target_jiffies = jiffies + duration_jiffies;
+ mod_timer(&wakeup_timer, target_jiffies);
+ if (unlikely(local_softirq_pending()))
+ continue;
+ /*
+ * stop tick sched during idle time, interrupts are still
+ * allowed. thus jiffies are updated properly.
+ */
+ preempt_disable();
+ tick_nohz_idle_enter();
+ /* mwait until target jiffies is reached */
+ while (time_before(jiffies, target_jiffies)) {
+ unsigned long ecx = 1;
+ unsigned long eax = target_mwait;
+
+ /*
+ * REVISIT: may call enter_idle() to notify drivers who
+ * can save power during cpu idle. same for exit_idle()
+ */
+ local_touch_nmi();
+ stop_critical_timings();
+ __monitor((void *)¤t_thread_info()->flags, 0, 0);
+ cpu_relax(); /* allow HT sibling to run */
+ __mwait(eax, ecx);
+ start_critical_timings();
+ atomic_inc(&idle_wakeup_counter);
+ }
+ tick_nohz_idle_exit();
+ preempt_enable_no_resched();
+ }
+ del_timer_sync(&wakeup_timer);
+ clear_bit(cpunr, cpu_clamping_mask);
+
+ return 0;
+}
+
+/*
+ * 1 HZ polling while clamping is active, useful for userspace
+ * to monitor actual idle ratio.
+ */
+static void poll_pkg_cstate(struct work_struct *dummy);
+static DECLARE_DELAYED_WORK(poll_pkg_cstate_work, poll_pkg_cstate);
+static void poll_pkg_cstate(struct work_struct *dummy)
+{
+ static u64 msr_last;
+ static u64 tsc_last;
+ static unsigned long jiffies_last;
+
+ u64 msr_now;
+ unsigned long jiffies_now;
+ u64 tsc_now;
+
+ msr_now = pkg_state_counter();
+ rdtscll(tsc_now);
+ jiffies_now = jiffies;
+
+ /* calculate pkg cstate vs tsc ratio */
+ if (!msr_last || !tsc_last)
+ pkg_cstate_ratio_cur = 1;
+ else {
+ if (tsc_now - tsc_last)
+ pkg_cstate_ratio_cur = 100 * (msr_now - msr_last)/
+ (tsc_now - tsc_last);
+ }
+
+ /* update record */
+ msr_last = msr_now;
+ jiffies_last = jiffies_now;
+ tsc_last = tsc_now;
+
+ if (true == clamping)
+ schedule_delayed_work(&poll_pkg_cstate_work, HZ);
+}
+
+static int start_power_clamp(void)
+{
+ unsigned long cpu;
+ struct task_struct *thread;
+
+ /* check if pkg cstate counter is completely 0, abort in this case */
+ if (!pkg_state_counter()) {
+ pr_err("pkg cstate counter not functional, abort\n");
+ return -EINVAL;
+ }
+
+ set_target_ratio = clamp(set_target_ratio, 0U, MAX_TARGET_RATIO);
+ /* prevent cpu hotplug */
+ get_online_cpus();
+
+ /* prefer BSP */
+ control_cpu = 0;
+ if (!cpu_online(control_cpu))
+ control_cpu = smp_processor_id();
+
+ clamping = true;
+ schedule_delayed_work(&poll_pkg_cstate_work, 0);
+
+ /* start one thread per online cpu */
+ for_each_online_cpu(cpu) {
+ struct task_struct **p =
+ per_cpu_ptr(powerclamp_thread, cpu);
+
+ thread = kthread_create_on_node(clamp_thread,
+ (void *) cpu,
+ cpu_to_node(cpu),
+ "kidle_inject/%ld", cpu);
+ /* bind to cpu here */
+ if (likely(!IS_ERR(thread))) {
+ kthread_bind(thread, cpu);
+ wake_up_process(thread);
+ *p = thread;
+ }
+
+ }
+ put_online_cpus();
+
+ return 0;
+}
+
+static void end_power_clamp(void)
+{
+ int i;
+ struct task_struct *thread;
+
+ clamping = false;
+ /*
+ * make clamping visible to other cpus and give per cpu clamping threads
+ * sometime to exit, or gets killed later.
+ */
+ smp_mb();
+ msleep(20);
+ if (bitmap_weight(cpu_clamping_mask, num_possible_cpus())) {
+ for_each_set_bit(i, cpu_clamping_mask, num_possible_cpus()) {
+ pr_debug("clamping thread for cpu %d alive, kill\n", i);
+ thread = *per_cpu_ptr(powerclamp_thread, i);
+ kthread_stop(thread);
+ }
+ }
+}
+
+static int powerclamp_cpu_callback(struct notifier_block *nfb,
+ unsigned long action, void *hcpu)
+{
+ unsigned long cpu = (unsigned long)hcpu;
+ struct task_struct *thread;
+ struct task_struct **percpu_thread =
+ per_cpu_ptr(powerclamp_thread, cpu);
+
+ if (false == clamping)
+ goto exit_ok;
+
+ switch (action) {
+ case CPU_ONLINE:
+ thread = kthread_create_on_node(clamp_thread,
+ (void *) cpu,
+ cpu_to_node(cpu),
+ "kidle_inject/%lu", cpu);
+ if (likely(!IS_ERR(thread))) {
+ kthread_bind(thread, cpu);
+ wake_up_process(thread);
+ *percpu_thread = thread;
+ }
+ /* prefer BSP as controlling CPU */
+ if (cpu == 0) {
+ control_cpu = 0;
+ smp_mb();
+ }
+ break;
+ case CPU_DEAD:
+ if (test_bit(cpu, cpu_clamping_mask)) {
+ pr_err("cpu %lu dead but powerclamping thread is not\n",
+ cpu);
+ kthread_stop(*percpu_thread);
+ }
+ if (cpu == control_cpu) {
+ control_cpu = smp_processor_id();
+ smp_mb();
+ }
+ }
+
+exit_ok:
+ return NOTIFY_OK;
+}
+
+static struct notifier_block powerclamp_cpu_notifier = {
+ .notifier_call = powerclamp_cpu_callback,
+};
+
+static int powerclamp_get_max_state(struct thermal_cooling_device *cdev,
+ unsigned long *state)
+{
+ *state = MAX_TARGET_RATIO;
+
+ return 0;
+}
+
+static int powerclamp_get_cur_state(struct thermal_cooling_device *cdev,
+ unsigned long *state)
+{
+ if (true == clamping)
+ *state = pkg_cstate_ratio_cur;
+ else
+ /* to save power, do not poll idle ratio while not clamping */
+ *state = -1; /* indicates invalid state */
+
+ return 0;
+}
+
+static int powerclamp_set_cur_state(struct thermal_cooling_device *cdev,
+ unsigned long new_target_ratio)
+{
+ int ret = 0;
+
+ new_target_ratio = clamp(new_target_ratio, 0UL,
+ (unsigned long) (MAX_TARGET_RATIO-1));
+ if (set_target_ratio == 0 && new_target_ratio > 0) {
+ pr_info("Start idle injection to reduce power\n");
+ set_target_ratio = new_target_ratio;
+ ret = start_power_clamp();
+ goto exit_set;
+ } else if (set_target_ratio > 0 && new_target_ratio == 0) {
+ pr_info("Stop forced idle injection\n");
+ set_target_ratio = 0;
+ end_power_clamp();
+ } else /* adjust currently running */ {
+ set_target_ratio = new_target_ratio;
+ /* make new set_target_ratio visible to other cpus */
+ smp_mb();
+ }
+
+exit_set:
+ return ret;
+}
+
+/* bind to generic thermal layer as cooling device*/
+static struct thermal_cooling_device_ops powerclamp_cooling_ops = {
+ .get_max_state = powerclamp_get_max_state,
+ .get_cur_state = powerclamp_get_cur_state,
+ .set_cur_state = powerclamp_set_cur_state,
+};
+
+/* runs on Nehalem and later */
+static const struct x86_cpu_id intel_powerclamp_ids[] = {
+ { X86_VENDOR_INTEL, 6, 0x1a},
+ { X86_VENDOR_INTEL, 6, 0x1c},
+ { X86_VENDOR_INTEL, 6, 0x1e},
+ { X86_VENDOR_INTEL, 6, 0x1f},
+ { X86_VENDOR_INTEL, 6, 0x25},
+ { X86_VENDOR_INTEL, 6, 0x26},
+ { X86_VENDOR_INTEL, 6, 0x2a},
+ { X86_VENDOR_INTEL, 6, 0x2c},
+ { X86_VENDOR_INTEL, 6, 0x2d},
+ { X86_VENDOR_INTEL, 6, 0x2e},
+ { X86_VENDOR_INTEL, 6, 0x2f},
+ { X86_VENDOR_INTEL, 6, 0x3a},
+ {}
+};
+MODULE_DEVICE_TABLE(x86cpu, intel_powerclamp_ids);
+
+static int powerclamp_probe(void)
+{
+ if (!x86_match_cpu(intel_powerclamp_ids)) {
+ pr_err("Intel powerclamp does not run on family %d model %d\n",
+ boot_cpu_data.x86, boot_cpu_data.x86_model);
+ return -ENODEV;
+ }
+ if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC) ||
+ !boot_cpu_has(X86_FEATURE_CONSTANT_TSC) ||
+ !boot_cpu_has(X86_FEATURE_MWAIT) ||
+ !boot_cpu_has(X86_FEATURE_ARAT))
+ return -ENODEV;
+
+ /* find the deepest mwait value */
+ find_target_mwait();
+
+ return 0;
+}
+
+static int powerclamp_debug_show(struct seq_file *m, void *unused)
+{
+ int i = 0;
+
+ seq_printf(m, "controlling cpu: %d\n", control_cpu);
+ seq_printf(m, "pct confidence steady dynamic (compensation)\n");
+ for (i = 0; i < MAX_TARGET_RATIO; i++) {
+ seq_printf(m, "%d\t%lu\t%lu\t%lu\n",
+ i,
+ cal_data[i].confidence,
+ cal_data[i].steady_comp,
+ cal_data[i].dynamic_comp);
+ }
+
+ return 0;
+}
+
+static int powerclamp_debug_open(struct inode *inode,
+ struct file *file)
+{
+ return single_open(file, powerclamp_debug_show, inode->i_private);
+}
+
+static const struct file_operations powerclamp_debug_fops = {
+ .open = powerclamp_debug_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+ .owner = THIS_MODULE,
+};
+
+static inline void powerclamp_create_debug_files(void)
+{
+ debug_dir = debugfs_create_dir("intel_powerclamp", NULL);
+ if (!debug_dir)
+ return;
+
+ if (!debugfs_create_file("powerclamp_calib", S_IRUGO, debug_dir,
+ cal_data, &powerclamp_debug_fops))
+ goto file_error;
+
+ return;
+
+file_error:
+ debugfs_remove_recursive(debug_dir);
+}
+
+static int powerclamp_init(void)
+{
+ int retval;
+ int bitmap_size;
+
+ bitmap_size = BITS_TO_LONGS(num_possible_cpus()) * sizeof(long);
+ cpu_clamping_mask = kzalloc(bitmap_size, GFP_KERNEL);
+ if (!cpu_clamping_mask)
+ return -ENOMEM;
+
+ /* probe cpu features and ids here */
+ retval = powerclamp_probe();
+ if (retval)
+ return retval;
+ /* set default limit, maybe adjusted during runtime based on feedback */
+ window_size = 2;
+ register_hotcpu_notifier(&powerclamp_cpu_notifier);
+ powerclamp_thread = alloc_percpu(struct task_struct *);
+ cooling_dev = thermal_cooling_device_register("intel_powerclamp", NULL,
+ &powerclamp_cooling_ops);
+ if (IS_ERR(cooling_dev))
+ return -ENODEV;
+
+ if (!duration)
+ duration = jiffies_to_msecs(DEFAULT_DURATION_JIFFIES);
+ powerclamp_create_debug_files();
+
+ return 0;
+}
+module_init(powerclamp_init);
+
+static void powerclamp_exit(void)
+{
+ unregister_hotcpu_notifier(&powerclamp_cpu_notifier);
+ end_power_clamp();
+ free_percpu(powerclamp_thread);
+ thermal_cooling_device_unregister(cooling_dev);
+ kfree(cpu_clamping_mask);
+
+ cancel_delayed_work_sync(&poll_pkg_cstate_work);
+ debugfs_remove_recursive(debug_dir);
+}
+module_exit(powerclamp_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Arjan van de Ven <arjan@linux.intel.com>");
+MODULE_AUTHOR("Jacob Pan <jacob.jun.pan@linux.intel.com>");
+MODULE_DESCRIPTION("Package Level C-state Idle Injection for Intel CPUs");
--
1.7.9.5
^ permalink raw reply related
* [PATCH v5 2/3] x86/nmi: export local_touch_nmi() symbol for modules
From: Jacob Pan @ 2013-01-03 15:10 UTC (permalink / raw)
To: Linux PM, LKML, Rafael Wysocki
Cc: Len Brown, Thomas Gleixner, H. Peter Anvin, Ingo Molnar,
Zhang Rui, Rob Landley, Arjan van de Ven, Paul McKenney,
Joe Perches, Jacob Pan
In-Reply-To: <1357225804-27191-1-git-send-email-jacob.jun.pan@linux.intel.com>
Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com>
---
arch/x86/kernel/nmi.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c
index f84f5c5..6030805 100644
--- a/arch/x86/kernel/nmi.c
+++ b/arch/x86/kernel/nmi.c
@@ -509,3 +509,4 @@ void local_touch_nmi(void)
{
__this_cpu_write(last_nmi_rip, 0);
}
+EXPORT_SYMBOL_GPL(local_touch_nmi);
--
1.7.9.5
^ permalink raw reply related
* Re: [PATCH 3/3] cpufreq: Don't use cpu removed during cpufreq_driver_unregister
From: Srivatsa S. Bhat @ 2013-01-03 14:25 UTC (permalink / raw)
To: Viresh Kumar
Cc: rjw, rafael.j.wysocki, linaro-dev, nicolas.pitre, amit.kucheria,
mathieu.poirier, linux-kernel, cpufreq, pdsw-power-team, linux-pm
In-Reply-To: <1fe21314c2e17585e22c546e2cac12544f8f9733.1355636864.git.viresh.kumar@linaro.org>
Hi Viresh,
On 12/16/2012 11:20 AM, Viresh Kumar wrote:
> This is how the core works:
> cpufreq_driver_unregister()
> - subsys_interface_unregister()
> - for_each_cpu() call cpufreq_remove_dev(), i.e. 0,1,2,3,4 when we
> unregister.
>
> cpufreq_remove_dev():
> - Remove policy node
> - Call cpufreq_add_dev() for next cpu, sharing mask with removed cpu.
> i.e. When cpu 0 is removed, we call it for cpu 1. And when called for cpu 2,
> we call it for cpu 3.
> - cpufreq_add_dev() would call cpufreq_driver->init()
> - init would return mask as AND of 2, 3 and 4 for cluster A7.
> - cpufreq core would do online_cpu && policy->cpus
> Here is the BUG(). Because cpu hasn't died but we have just unregistered
> the cpufreq driver, online cpu would still have cpu 2 in it. And so thing
> go bad again.
>
> Solution: Keep cpumask of cpus that are registered with cpufreq core and clear
> cpus when we get a call from subsys_interface_unregister() via
> cpufreq_remove_dev().
>
I took a quick look at the problem you described above, and the cpufreq code..
If we cannot avoid calling cpufreq_add_dev() from cpufreq_remove_dev(), then I can't
think of anything better than what your patch does.
BTW, off-topic, while going through that path, I think I found a memory leak
in __cpufreq_remove_dev():
if (unlikely(cpumask_weight(data->cpus) > 1)) {
for_each_cpu(j, data->cpus) {
if (j == cpu)
continue;
per_cpu(cpufreq_cpu_data, j) = NULL;
}
}
We are assigning NULL without freeing that memory.
Regards,
Srivatsa S. Bhat
> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
> ---
> drivers/cpufreq/cpufreq.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> index a0a33bd..271d3be 100644
> --- a/drivers/cpufreq/cpufreq.c
> +++ b/drivers/cpufreq/cpufreq.c
> @@ -47,6 +47,9 @@ static DEFINE_PER_CPU(char[CPUFREQ_NAME_LEN], cpufreq_cpu_governor);
> #endif
> static DEFINE_SPINLOCK(cpufreq_driver_lock);
>
> +/* Used when we unregister cpufreq driver */
> +struct cpumask cpufreq_online_mask;
> +
> /*
> * cpu_policy_rwsem is a per CPU reader-writer semaphore designed to cure
> * all cpufreq/hotplug/workqueue/etc related lock issues.
> @@ -981,6 +984,7 @@ static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif)
> * managing offline cpus here.
> */
> cpumask_and(policy->cpus, policy->cpus, cpu_online_mask);
> + cpumask_and(policy->cpus, policy->cpus, &cpufreq_online_mask);
>
> policy->user_policy.min = policy->min;
> policy->user_policy.max = policy->max;
> @@ -1064,7 +1068,6 @@ static int __cpufreq_remove_dev(struct device *dev, struct subsys_interface *sif
> }
> per_cpu(cpufreq_cpu_data, cpu) = NULL;
>
> -
> #ifdef CONFIG_SMP
> /* if this isn't the CPU which is the parent of the kobj, we
> * only need to unlink, put and exit
> @@ -1185,6 +1188,7 @@ static int cpufreq_remove_dev(struct device *dev, struct subsys_interface *sif)
> if (unlikely(lock_policy_rwsem_write(cpu)))
> BUG();
>
> + cpumask_clear_cpu(cpu, &cpufreq_online_mask);
> retval = __cpufreq_remove_dev(dev, sif);
> return retval;
> }
> @@ -1903,6 +1907,8 @@ int cpufreq_register_driver(struct cpufreq_driver *driver_data)
> cpufreq_driver = driver_data;
> spin_unlock_irqrestore(&cpufreq_driver_lock, flags);
>
> + cpumask_setall(&cpufreq_online_mask);
> +
> ret = subsys_interface_register(&cpufreq_interface);
> if (ret)
> goto err_null_driver;
>
^ permalink raw reply
* Re: [PATCH 1/3] cpufreq: Manage only online cpus
From: Rafael J. Wysocki @ 2013-01-03 12:02 UTC (permalink / raw)
To: Viresh Kumar
Cc: linaro-dev, nicolas.pitre, amit.kucheria, mathieu.poirier,
linux-kernel, cpufreq, pdsw-power-team, linux-pm
In-Reply-To: <CAKohponGtLNBnWWX1miV3zjO4pJPumbWoPwoQ2+adCE5t2w5Dw@mail.gmail.com>
On Thursday, January 03, 2013 09:02:22 AM Viresh Kumar wrote:
> On 3 January 2013 06:43, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> >> BTW, i consider them as fixes and so would make sense to get them in next rc.
> >> What do you think?
> >
> > Yes, if somebody tells me "yes, this fixes a problem for me". Otherwise,
> > I don't quite see the reason.
>
> I don't know how much people test HOTPLUG, but there are clear bugs related
> to hotplug of cpus on a multiple cpu system :)
True, but have those bugs been introduced recently (ie. in v3.8-rc1 or later)?
Rafael
--
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
^ permalink raw reply
* [PATCH v4 3/3] PM: Introduce Intel PowerClamp Driver
From: Jacob Pan @ 2013-01-03 11:01 UTC (permalink / raw)
To: Linux PM, LKML, Rafael Wysocki
Cc: Len Brown, Thomas Gleixner, H. Peter Anvin, Ingo Molnar,
Zhang Rui, Rob Landley, Arjan van de Ven, Paul McKenney,
Peter Zijlstra, Joe Perches, Jacob Pan
In-Reply-To: <1357210875-5025-1-git-send-email-jacob.jun.pan@linux.intel.com>
Intel PowerClamp driver performs synchronized idle injection across
all online CPUs. The goal is to maintain a given package level C-state
ratio.
Compared to other throttling methods already exist in the kernel,
such as ACPI PAD (taking CPUs offline) and clock modulation, this is often
more efficient in terms of performance per watt.
Please refer to Documentation/thermal/intel_powerclamp.txt for more details.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com>
---
Documentation/thermal/intel_powerclamp.txt | 307 +++++++++++
drivers/thermal/Kconfig | 10 +
drivers/thermal/Makefile | 2 +
drivers/thermal/intel_powerclamp.c | 762 ++++++++++++++++++++++++++++
4 files changed, 1081 insertions(+)
create mode 100644 Documentation/thermal/intel_powerclamp.txt
create mode 100644 drivers/thermal/intel_powerclamp.c
diff --git a/Documentation/thermal/intel_powerclamp.txt b/Documentation/thermal/intel_powerclamp.txt
new file mode 100644
index 0000000..332de4a
--- /dev/null
+++ b/Documentation/thermal/intel_powerclamp.txt
@@ -0,0 +1,307 @@
+ =======================
+ INTEL POWERCLAMP DRIVER
+ =======================
+By: Arjan van de Ven <arjan@linux.intel.com>
+ Jacob Pan <jacob.jun.pan@linux.intel.com>
+
+Contents:
+ (*) Introduction
+ - Goals and Objectives
+
+ (*) Theory of Operation
+ - Idle Injection
+ - Calibration
+
+ (*) Performance Analysis
+ - Effectiveness and Limitations
+ - Power vs Performance
+ - Scalability
+ - Calibration
+ - Comparison with Alternative Techniques
+
+ (*) Usage and Interfaces
+ - Generic Thermal Layer (sysfs)
+ - Kernel APIs (TBD)
+
+============
+INTRODUCTION
+============
+
+Consider the situation where a system’s power consumption must be
+reduced at runtime, due to power budget, thermal constraint, or noise
+level, and where active cooling is not preferred. Software managed
+passive power reduction must be performed to prevent the hardware
+actions that are designed for catastrophic scenarios.
+
+Currently, P-states, T-states (clock modulation), and CPU offlining
+are used for CPU throttling.
+
+On Intel CPUs, C-states provide effective power reduction, but so far
+they’re only used opportunistically, based on workload. With the
+development of intel_powerclamp driver, the method of synchronizing
+idle injection across all online CPU threads was introduced. The goal
+is to achieve forced and controllable C-state residency.
+
+Test/Analysis has been made in the areas of power, performance,
+scalability, and user experience. In many cases, clear advantage is
+shown over taking the CPU offline or modulating the CPU clock.
+
+
+===================
+THEORY OF OPERATION
+===================
+
+Idle Injection
+--------------
+
+On modern Intel processors (Nehalem or later), package level C-state
+residency is available in MSRs, thus also available to the kernel.
+
+These MSRs are:
+ #define MSR_PKG_C2_RESIDENCY 0x60D
+ #define MSR_PKG_C3_RESIDENCY 0x3F8
+ #define MSR_PKG_C6_RESIDENCY 0x3F9
+ #define MSR_PKG_C7_RESIDENCY 0x3FA
+
+If the kernel can also inject idle time to the system, then a
+closed-loop control system can be established that manages package
+level C-state. The intel_powerclamp driver is conceived as such a
+control system, where the target set point is a user-selected idle
+ratio (based on power reduction), and the error is the difference
+between the actual package level C-state residency ratio and the target idle
+ratio.
+
+Injection is controlled by high priority kernel threads, spawned for
+each online CPU.
+
+These kernel threads, with SCHED_FIFO class, are created to perform
+clamping actions of controlled duty ratio and duration. Each per-CPU
+thread synchronizes its idle time and duration, based on the rounding
+of jiffies, so accumulated errors can be prevented to avoid a jittery
+effect. Threads are also bound to the CPU such that they cannot be
+migrated, unless the CPU is taken offline. In this case, threads
+belong to the offlined CPUs will be terminated immediately.
+
+Running as SCHED_FIFO and relatively high priority, also allows such
+scheme to work for both preemptable and non-preemptable kernels.
+Alignment of idle time around jiffies ensures scalability for HZ
+values. This effect can be better visualized using a Perf timechart.
+The following diagram shows the behavior of kernel thread
+kidle_inject/cpu. During idle injection, it runs monitor/mwait idle
+for a given "duration", then relinquishes the CPU to other tasks,
+until the next time interval.
+
+The NOHZ schedule tick is disabled during idle time, but interrupts
+are not masked. Tests show that the extra wakeups from scheduler tick
+have a dramatic impact on the effectiveness of the powerclamp driver
+on large scale systems (Westmere system with 80 processors).
+
+CPU0
+ ____________ ____________
+kidle_inject/0 | sleep | mwait | sleep |
+ _________| |________| |_______
+ duration
+CPU1
+ ____________ ____________
+kidle_inject/1 | sleep | mwait | sleep |
+ _________| |________| |_______
+ ^
+ |
+ |
+ roundup(jiffies, interval)
+
+Only one CPU is allowed to collect statistics and update global
+control parameters. This CPU is referred to as the controlling CPU in
+this document. The controlling CPU is elected at runtime, with a
+policy that favors BSP, taking into account the possibility of a CPU
+hot-plug.
+
+In terms of dynamics of the idle control system, package level idle
+time is considered largely as a non-causal system where its behavior
+cannot be based on the past or current input. Therefore, the
+intel_powerclamp driver attempts to enforce the desired idle time
+instantly as given input (target idle ratio). After injection,
+powerclamp moniors the actual idle for a given time window and adjust
+the next injection accordingly to avoid over/under correction.
+
+When used in a causal control system, such as a temperature control,
+it is up to the user of this driver to implement algorithms where
+past samples and outputs are included in the feedback. For example, a
+PID-based thermal controller can use the powerclamp driver to
+maintain a desired target temperature, based on integral and
+derivative gains of the past samples.
+
+
+
+Calibration
+-----------
+During scalability testing, it is observed that synchronized actions
+among CPUs become challenging as the number of cores grows. This is
+also true for the ability of a system to enter package level C-states.
+
+To make sure the intel_powerclamp driver scales well, online
+calibration is implemented. The goals for doing such a calibration
+are:
+
+a) determine the effective range of idle injection ratio
+b) determine the amount of compensation needed at each target ratio
+
+Compensation to each target ratio consists of two parts:
+
+ a) steady state error compensation
+ This is to offset the error occurring when the system can
+ enter idle without extra wakeups (such as external interrupts).
+
+ b) dynamic error compensation
+ When an excessive amount of wakeups occurs during idle, an
+ additional idle ratio can be added to quiet interrupts, by
+ slowing down CPU activities.
+
+A debugfs file is provided for the user to examine compensation
+progress and results, such as on a Westmere system.
+[jacob@nex01 ~]$ cat
+/sys/kernel/debug/intel_powerclamp/powerclamp_calib
+controlling cpu: 0
+pct confidence steady dynamic (compensation)
+0 0 0 0
+1 1 0 0
+2 1 1 0
+3 3 1 0
+4 3 1 0
+5 3 1 0
+6 3 1 0
+7 3 1 0
+8 3 1 0
+...
+30 3 2 0
+31 3 2 0
+32 3 1 0
+33 3 2 0
+34 3 1 0
+35 3 2 0
+36 3 1 0
+37 3 2 0
+38 3 1 0
+39 3 2 0
+40 3 3 0
+41 3 1 0
+42 3 2 0
+43 3 1 0
+44 3 1 0
+45 3 2 0
+46 3 3 0
+47 3 0 0
+48 3 2 0
+49 3 3 0
+
+Calibration occurs during runtime. No offline method is available.
+Steady state compensation is used only when confidence levels of all
+adjacent ratios have reached satisfactory level. A confidence level
+is accumulated based on clean data collected at runtime. Data
+collected during a period without extra interrupts is considered
+clean.
+
+To compensate for excessive amounts of wakeup during idle, additional
+idle time is injected when such a condition is detected. Currently,
+we have a simple algorithm to double the injection ratio. A possible
+enhancement might be to throttle the offending IRQ, such as delaying
+EOI for level triggered interrupts. But it is a challenge to be
+non-intrusive to the scheduler or the IRQ core code.
+
+
+CPU Online/Offline
+------------------
+Per-CPU kernel threads are started/stopped upon receiving
+notifications of CPU hotplug activities. The intel_powerclamp driver
+keeps track of clamping kernel threads, even after they are migrated
+to other CPUs, after a CPU offline event.
+
+
+=====================
+Performance Analysis
+=====================
+This section describes the general performance data collected on
+multiple systems, including Westmere (80P) and Ivy Bridge (4P, 8P).
+
+Effectiveness and Limitations
+-----------------------------
+The maximum range that idle injection is allowed is capped at 50
+percent. As mentioned earlier, since interrupts are allowed during
+forced idle time, excessive interrupts could result in less
+effectiveness. The extreme case would be doing a ping -f to generated
+flooded network interrupts without much CPU acknowledgement. In this
+case, little can be done from the idle injection threads. In most
+normal cases, such as scp a large file, applications can be throttled
+by the powerclamp driver, since slowing down the CPU also slows down
+network protocol processing, which in turn reduces interrupts.
+
+When control parameters change at runtime by the controlling CPU, it
+may take an additional period for the rest of the CPUs to catch up
+with the changes. During this time, idle injection is out of sync,
+thus not able to enter package C- states at the expected ratio. But
+this effect is minor, in that in most cases change to the target
+ratio is updated much less frequently than the idle injection
+frequency.
+
+Scalability
+-----------
+Tests also show a minor, but measurable, difference between the 4P/8P
+Ivy Bridge system and the 80P Westmere server under 50% idle ratio.
+More compensation is needed on Westmere for the same amount of
+target idle ratio. The compensation also increases as the idle ratio
+gets larger. The above reason constitutes the need for the
+calibration code.
+
+On the IVB 8P system, compared to an offline CPU, powerclamp can
+achieve up to 40% better performance per watt. (measured by a spin
+counter summed over per CPU counting threads spawned for all running
+CPUs).
+
+====================
+Usage and Interfaces
+====================
+The powerclamp driver is registered to the generic thermal layer as a
+cooling device. Currently, it’s not bound to any thermal zones.
+
+jacob@chromoly:/sys/class/thermal/cooling_device14$ grep . *
+cur_state:0
+max_state:50
+type:intel_powerclamp
+
+Example usage:
+- To inject 25% idle time
+$ sudo sh -c "echo 25 > /sys/class/thermal/cooling_device80/cur_state
+"
+
+If the system is not busy and has more than 25% idle time already,
+then the powerclamp driver will not start idle injection. Using Top
+will not show idle injection kernel threads.
+
+If the system is busy (spin test below) and has less than 25% natural
+idle time, powerclamp kernel threads will do idle injection, which
+appear running to the scheduler. But the overall system idle is still
+reflected. In this example, 24.1% idle is shown. This helps the
+system admin or user determine the cause of slowdown, when a
+powerclamp driver is in action.
+
+
+Tasks: 197 total, 1 running, 196 sleeping, 0 stopped, 0 zombie
+Cpu(s): 71.2%us, 4.7%sy, 0.0%ni, 24.1%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
+Mem: 3943228k total, 1689632k used, 2253596k free, 74960k buffers
+Swap: 4087804k total, 0k used, 4087804k free, 945336k cached
+
+ PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
+ 3352 jacob 20 0 262m 644 428 S 286 0.0 0:17.16 spin
+ 3341 root -51 0 0 0 0 D 25 0.0 0:01.62 kidle_inject/0
+ 3344 root -51 0 0 0 0 D 25 0.0 0:01.60 kidle_inject/3
+ 3342 root -51 0 0 0 0 D 25 0.0 0:01.61 kidle_inject/1
+ 3343 root -51 0 0 0 0 D 25 0.0 0:01.60 kidle_inject/2
+ 2935 jacob 20 0 696m 125m 35m S 5 3.3 0:31.11 firefox
+ 1546 root 20 0 158m 20m 6640 S 3 0.5 0:26.97 Xorg
+ 2100 jacob 20 0 1223m 88m 30m S 3 2.3 0:23.68 compiz
+
+Tests have shown that by using the powerclamp driver as a cooling
+device, a PID based userspace thermal controller can manage to
+control CPU temperature effectively, when no other thermal influence
+is added. For example, a UltraBook user can compile the kernel under
+certain temperature (below most active trip points).
diff --git a/drivers/thermal/Kconfig b/drivers/thermal/Kconfig
index c2c77d1..7d90ab8 100644
--- a/drivers/thermal/Kconfig
+++ b/drivers/thermal/Kconfig
@@ -122,4 +122,14 @@ config DB8500_CPUFREQ_COOLING
bound cpufreq cooling device turns active to set CPU frequency low to
cool down the CPU.
+config INTEL_POWERCLAMP
+ tristate "Intel PowerClamp idle injection driver"
+ depends on THERMAL
+ depends on X86
+ depends on CPU_SUP_INTEL
+ help
+ Enable this to enable Intel PowerClamp idle injection driver. This
+ enforce idle time which results in more package C-state residency. The
+ user interface is exposed via generic thermal framework.
+
endif
diff --git a/drivers/thermal/Makefile b/drivers/thermal/Makefile
index d8da683..574f5f5 100644
--- a/drivers/thermal/Makefile
+++ b/drivers/thermal/Makefile
@@ -18,3 +18,5 @@ obj-$(CONFIG_RCAR_THERMAL) += rcar_thermal.o
obj-$(CONFIG_EXYNOS_THERMAL) += exynos_thermal.o
obj-$(CONFIG_DB8500_THERMAL) += db8500_thermal.o
obj-$(CONFIG_DB8500_CPUFREQ_COOLING) += db8500_cpufreq_cooling.o
+obj-$(CONFIG_INTEL_POWERCLAMP) += intel_powerclamp.o
+
diff --git a/drivers/thermal/intel_powerclamp.c b/drivers/thermal/intel_powerclamp.c
new file mode 100644
index 0000000..46b3674
--- /dev/null
+++ b/drivers/thermal/intel_powerclamp.c
@@ -0,0 +1,762 @@
+/*
+ * intel_powerclamp.c - package c-state idle injection
+ *
+ * Copyright (c) 2012, Intel Corporation.
+ *
+ * Authors:
+ * Arjan van de Ven <arjan@linux.intel.com>
+ * Jacob Pan <jacob.jun.pan@linux.intel.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ *
+ * TODO:
+ * 1. better handle wakeup from external interrupts, currently a fixed
+ * compensation is added to clamping duration when excessive amount
+ * of wakeups are observed during idle time. the reason is that in
+ * case of external interrupts without need for ack, clamping down
+ * cpu in non-irq context does not reduce irq. for majority of the
+ * cases, clamping down cpu does help reduce irq as well, we should
+ * be able to differenciate the two cases and give a quantitative
+ * solution for the irqs that we can control. perhaps based on
+ * get_cpu_iowait_time_us()
+ *
+ * 2. synchronization with other hw blocks
+ *
+ *
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/kthread.h>
+#include <linux/freezer.h>
+#include <linux/cpu.h>
+#include <linux/thermal.h>
+#include <linux/slab.h>
+#include <linux/tick.h>
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#include <linux/nmi.h>
+
+#include <asm/msr.h>
+#include <asm/mwait.h>
+#include <asm/cpu_device_id.h>
+#include <asm/idle.h>
+#include <asm/hardirq.h>
+
+#define MSR_PKG_C2_RESIDENCY 0x60D
+#define MSR_PKG_C3_RESIDENCY 0x3F8
+#define MSR_PKG_C6_RESIDENCY 0x3F9
+#define MSR_PKG_C7_RESIDENCY 0x3FA
+
+#define MAX_TARGET_RATIO (50U)
+/* For each undisturbed clamping period (no extra wake ups during idle time),
+ * we increment the confidence counter for the given target ratio.
+ * CONFIDENCE_OK defines the level where runtime calibration results are
+ * valid.
+ */
+#define CONFIDENCE_OK (3)
+/* Default idle injection duration, driver adjust sleep time to meet target
+ * idle ratio. Similar to frequency modulation.
+ */
+#define DEFAULT_DURATION_JIFFIES (6)
+
+static unsigned int target_mwait;
+static struct dentry *debug_dir;
+
+/* user selected target */
+static unsigned int set_target_ratio;
+static unsigned int current_ratio;
+static bool should_skip;
+static bool reduce_irq;
+static atomic_t idle_wakeup_counter;
+static unsigned int control_cpu; /* The cpu assigned to collect stat and update
+ * control parameters. default to BSP but BSP
+ * can be offlined.
+ */
+static int clamping;
+
+
+static struct task_struct __percpu **powerclamp_thread;
+static struct thermal_cooling_device *cooling_dev;
+static unsigned long *cpu_clamping_mask; /* bit map for tracking per cpu
+ * clamping thread
+ */
+static int duration;
+module_param(duration, int, 0600);
+MODULE_PARM_DESC(duration, "forced idle time for each attempt in msec.");
+
+static unsigned int pkg_cstate_ratio_cur;
+static unsigned int window_size;
+
+struct powerclamp_calibration_data {
+ unsigned long confidence; /* used for calibration, basically a counter
+ * gets incremented each time a clamping
+ * period is completed without extra wakeups
+ * once that counter is reached given level,
+ * compensation is deemed usable.
+ */
+ unsigned long steady_comp; /* steady state compensation used when
+ * no extra wakeups occurred.
+ */
+ unsigned long dynamic_comp; /* compensate excessive wakeup from idle
+ * mostly from external interrupts.
+ */
+};
+
+static struct powerclamp_calibration_data cal_data[MAX_TARGET_RATIO];
+
+static int window_size_set(const char *arg, const struct kernel_param *kp)
+{
+ int ret = 0;
+ unsigned long new_window_size;
+
+ ret = kstrtoul(arg, 10, &new_window_size);
+ if (ret)
+ goto exit_win;
+ if (new_window_size >= 10 || new_window_size < 2) {
+ pr_err("Invalid window size %lu, between 2-10\n",
+ new_window_size);
+ ret = -EINVAL;
+ }
+
+ window_size = new_window_size;
+ smp_mb();
+
+exit_win:
+
+ return ret;
+}
+static struct kernel_param_ops window_size_ops = {
+ .set = window_size_set,
+ .get = param_get_int,
+};
+
+module_param_cb(window_size, &window_size_ops, &window_size, 0644);
+MODULE_PARM_DESC(window_size, "sliding window in number of clamping cycles\n"
+ "\tpowerclamp controls idle ratio within this window. larger\n"
+ "\twindow size results in slower response time but more smooth\n"
+ "\tclamping results. default to 2.");
+
+static void find_target_mwait(void)
+{
+ unsigned int eax, ebx, ecx, edx;
+ unsigned int highest_cstate = 0;
+ unsigned int highest_subcstate = 0;
+ int i;
+
+ if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
+ return;
+
+ cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx);
+
+ if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) ||
+ !(ecx & CPUID5_ECX_INTERRUPT_BREAK))
+ return;
+
+ edx >>= MWAIT_SUBSTATE_SIZE;
+ for (i = 0; i < 7 && edx; i++, edx >>= MWAIT_SUBSTATE_SIZE) {
+ if (edx & MWAIT_SUBSTATE_MASK) {
+ highest_cstate = i;
+ highest_subcstate = edx & MWAIT_SUBSTATE_MASK;
+ }
+ }
+ target_mwait = (highest_cstate << MWAIT_SUBSTATE_SIZE) |
+ (highest_subcstate - 1);
+
+}
+
+static u64 pkg_state_counter(void)
+{
+ u64 val;
+ u64 count = 0;
+
+ static bool skip_c2;
+ static bool skip_c3;
+ static bool skip_c6;
+ static bool skip_c7;
+
+ if (!skip_c2) {
+ if (!rdmsrl_safe(MSR_PKG_C2_RESIDENCY, &val))
+ count += val;
+ else
+ skip_c2 = true;
+ }
+
+ if (!skip_c3) {
+ if (!rdmsrl_safe(MSR_PKG_C3_RESIDENCY, &val))
+ count += val;
+ else
+ skip_c3 = true;
+ }
+
+ if (!skip_c6) {
+ if (!rdmsrl_safe(MSR_PKG_C6_RESIDENCY, &val))
+ count += val;
+ else
+ skip_c6 = true;
+ }
+
+ if (!skip_c7) {
+ if (!rdmsrl_safe(MSR_PKG_C7_RESIDENCY, &val))
+ count += val;
+ else
+ skip_c7 = true;
+ }
+
+ return count;
+}
+
+static void noop_timer(unsigned long foo)
+{
+ /* empty... just the fact that we get the interrupt wakes us up */
+}
+
+static unsigned int get_compensation(int ratio)
+{
+ unsigned int comp = 0;
+
+ /* we only use compensation if all adjacent ones are good */
+ if (ratio == 1 &&
+ cal_data[ratio].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio + 1].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio + 2].confidence >= CONFIDENCE_OK) {
+ comp = (cal_data[ratio].steady_comp +
+ cal_data[ratio + 1].steady_comp +
+ cal_data[ratio + 2].steady_comp) / 3;
+ } else if (ratio == MAX_TARGET_RATIO - 1 &&
+ cal_data[ratio].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio - 1].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio - 2].confidence >= CONFIDENCE_OK) {
+ comp = (cal_data[ratio].steady_comp +
+ cal_data[ratio - 1].steady_comp +
+ cal_data[ratio - 2].steady_comp) / 3;
+ } else if (cal_data[ratio].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio - 1].confidence >= CONFIDENCE_OK &&
+ cal_data[ratio + 1].confidence >= CONFIDENCE_OK) {
+ comp = (cal_data[ratio].steady_comp +
+ cal_data[ratio - 1].steady_comp +
+ cal_data[ratio + 1].steady_comp) / 3;
+ }
+
+ /* REVISIT: simple penalty of double idle injection */
+ if (reduce_irq)
+ comp = ratio;
+ /* do not exceed limit */
+ if (comp + ratio >= MAX_TARGET_RATIO)
+ comp = MAX_TARGET_RATIO - ratio - 1;
+
+ return comp;
+}
+
+static void adjust_compensation(int target_ratio, unsigned int win)
+{
+ int delta;
+ struct powerclamp_calibration_data *d = &cal_data[target_ratio];
+
+ /*
+ * adjust compensations if confidence level has not been reached or
+ * there are too many wakeups during the last idle injection period, we
+ * cannot trust the data for compensation.
+ */
+ if (d->confidence >= CONFIDENCE_OK ||
+ atomic_read(&idle_wakeup_counter) >
+ win * num_online_cpus())
+ return;
+
+ delta = set_target_ratio - current_ratio;
+ /* filter out bad data */
+ if (delta >= 0 && delta <= (1+target_ratio/10)) {
+ if (d->steady_comp)
+ d->steady_comp =
+ roundup(delta+d->steady_comp, 2)/2;
+ else
+ d->steady_comp = delta;
+ d->confidence++;
+ }
+}
+
+static bool powerclamp_adjust_controls(unsigned int target_ratio,
+ unsigned int guard, unsigned int win)
+{
+ static u64 msr_last, tsc_last;
+ u64 msr_now, tsc_now;
+
+ /* check result for the last window */
+ msr_now = pkg_state_counter();
+ rdtscll(tsc_now);
+
+ /* calculate pkg cstate vs tsc ratio */
+ if (!msr_last || !tsc_last)
+ current_ratio = 1;
+ else if (tsc_now-tsc_last)
+ current_ratio = 100*(msr_now-msr_last)/
+ (tsc_now-tsc_last);
+
+ /* update record */
+ msr_last = msr_now;
+ tsc_last = tsc_now;
+
+ adjust_compensation(target_ratio, win);
+ /*
+ * too many external interrupts, set flag such
+ * that we can take measure later.
+ */
+ reduce_irq = atomic_read(&idle_wakeup_counter) >=
+ 2 * win * num_online_cpus();
+
+ atomic_set(&idle_wakeup_counter, 0);
+ /* if we are above target+guard, skip */
+ return set_target_ratio + guard <= current_ratio;
+}
+
+static int clamp_thread(void *arg)
+{
+ int cpunr = (unsigned long)arg;
+ DEFINE_TIMER(wakeup_timer, noop_timer, 0, 0);
+ static const struct sched_param param = {
+ .sched_priority = MAX_USER_RT_PRIO/2,
+ };
+ unsigned int count = 0;
+ unsigned int target_ratio;
+
+ set_bit(cpunr, cpu_clamping_mask);
+ set_freezable();
+ init_timer_on_stack(&wakeup_timer);
+ sched_setscheduler(current, SCHED_FIFO, ¶m);
+
+ while (clamping && !kthread_should_stop() && cpu_online(cpunr)) {
+ int sleeptime;
+ unsigned long target_jiffies;
+ unsigned int guard;
+ unsigned int compensation = 0;
+ int interval; /* jiffies to sleep for each attempt */
+ unsigned int duration_jiffies = msecs_to_jiffies(duration);
+ unsigned int window_size_now;
+
+ try_to_freeze();
+ /*
+ * make sure user selected ratio does not take effect until
+ * the next round. adjust target_ratio if user has changed
+ * target such that we can converge quickly.
+ */
+ target_ratio = set_target_ratio;
+ guard = 1 + target_ratio/20;
+ window_size_now = window_size;
+ count++;
+
+ /*
+ * systems may have different ability to enter package level
+ * c-states, thus we need to compensate the injected idle ratio
+ * to achieve the actual target reported by the HW.
+ */
+ compensation = get_compensation(target_ratio);
+ interval = duration_jiffies*100/(target_ratio+compensation);
+
+ /* align idle time */
+ target_jiffies = roundup(jiffies, interval);
+ sleeptime = target_jiffies - jiffies;
+ if (sleeptime <= 0)
+ sleeptime = 1;
+ schedule_timeout_interruptible(sleeptime);
+ /*
+ * only elected controlling cpu can collect stats and update
+ * control parameters.
+ */
+ if (cpunr == control_cpu && !(count%window_size_now)) {
+ should_skip =
+ powerclamp_adjust_controls(target_ratio,
+ guard, window_size_now);
+ smp_mb();
+ }
+
+ if (should_skip)
+ continue;
+
+ target_jiffies = jiffies + duration_jiffies;
+ mod_timer(&wakeup_timer, target_jiffies);
+ if (unlikely(local_softirq_pending()))
+ continue;
+ /*
+ * stop tick sched during idle time, interrupts are still
+ * allowed. thus jiffies are updated properly.
+ */
+ preempt_disable();
+ tick_nohz_idle_enter();
+ /* mwait until target jiffies is reached */
+ while (time_before(jiffies, target_jiffies)) {
+ unsigned long ecx = 1;
+ unsigned long eax = target_mwait;
+
+ /*
+ * REVISIT: may call enter_idle() to notify drivers who
+ * can save power during cpu idle. same for exit_idle()
+ */
+ local_touch_nmi();
+ stop_critical_timings();
+ __monitor((void *)¤t_thread_info()->flags, 0, 0);
+ cpu_relax(); /* allow HT sibling to run */
+ __mwait(eax, ecx);
+ start_critical_timings();
+ atomic_inc(&idle_wakeup_counter);
+ }
+ tick_nohz_idle_exit();
+ preempt_enable_no_resched();
+ }
+ del_timer_sync(&wakeup_timer);
+ clear_bit(cpunr, cpu_clamping_mask);
+
+ return 0;
+}
+
+/*
+ * 1 HZ polling while clamping is active, useful for userspace
+ * to monitor actual idle ratio.
+ */
+static void poll_pkg_cstate(struct work_struct *dummy);
+static DECLARE_DELAYED_WORK(poll_pkg_cstate_work, poll_pkg_cstate);
+static void poll_pkg_cstate(struct work_struct *dummy)
+{
+ static u64 msr_last;
+ static u64 tsc_last;
+ static unsigned long jiffies_last;
+
+ u64 msr_now;
+ unsigned long jiffies_now;
+ u64 tsc_now;
+
+ msr_now = pkg_state_counter();
+ rdtscll(tsc_now);
+ jiffies_now = jiffies;
+
+ /* calculate pkg cstate vs tsc ratio */
+ if (!msr_last || !tsc_last)
+ pkg_cstate_ratio_cur = 1;
+ else {
+ if (tsc_now - tsc_last)
+ pkg_cstate_ratio_cur = 100 * (msr_now - msr_last)/
+ (tsc_now - tsc_last);
+ }
+
+ /* update record */
+ msr_last = msr_now;
+ jiffies_last = jiffies_now;
+ tsc_last = tsc_now;
+
+ if (clamping)
+ schedule_delayed_work(&poll_pkg_cstate_work, HZ);
+}
+
+static int start_power_clamp(void)
+{
+ unsigned long cpu;
+ struct task_struct *thread;
+
+ /* check if pkg cstate counter is completely 0, abort in this case */
+ if (!pkg_state_counter()) {
+ pr_err("pkg cstate counter not functional, abort\n");
+ return -EINVAL;
+ }
+
+ set_target_ratio = clamp(set_target_ratio, 0U, MAX_TARGET_RATIO);
+ /* prevent cpu hotplug */
+ get_online_cpus();
+
+ /* prefer BSP */
+ control_cpu = 0;
+ if (!cpu_online(control_cpu))
+ control_cpu = smp_processor_id();
+
+ clamping = 1;
+ schedule_delayed_work(&poll_pkg_cstate_work, 0);
+
+ /* start one thread per online cpu */
+ for_each_online_cpu(cpu) {
+ struct task_struct **p =
+ per_cpu_ptr(powerclamp_thread, cpu);
+
+ thread = kthread_create_on_node(clamp_thread,
+ (void *) cpu,
+ cpu_to_node(cpu),
+ "kidle_inject/%ld", cpu);
+ /* bind to cpu here */
+ if (likely(!IS_ERR(thread))) {
+ kthread_bind(thread, cpu);
+ wake_up_process(thread);
+ *p = thread;
+ }
+
+ }
+ put_online_cpus();
+
+ return 0;
+}
+
+static void end_power_clamp(void)
+{
+ int i;
+ struct task_struct *thread;
+
+ clamping = 0;
+ /*
+ * make clamping visible to other cpus and give per cpu clamping threads
+ * sometime to exit, or gets killed later.
+ */
+ smp_mb();
+ msleep(20);
+ if (bitmap_weight(cpu_clamping_mask, num_possible_cpus())) {
+ for_each_set_bit(i, cpu_clamping_mask, num_possible_cpus()) {
+ pr_debug("clamping thread for cpu %d alive, kill\n", i);
+ thread = *per_cpu_ptr(powerclamp_thread, i);
+ kthread_stop(thread);
+ }
+ }
+}
+
+static int powerclamp_cpu_callback(struct notifier_block *nfb,
+ unsigned long action, void *hcpu)
+{
+ unsigned long cpu = (unsigned long)hcpu;
+ struct task_struct *thread;
+ struct task_struct **percpu_thread =
+ per_cpu_ptr(powerclamp_thread, cpu);
+
+ if (!clamping)
+ goto exit_ok;
+
+ switch (action) {
+ case CPU_ONLINE:
+ thread = kthread_create_on_node(clamp_thread,
+ (void *) cpu,
+ cpu_to_node(cpu),
+ "kidle_inject/%lu", cpu);
+ if (likely(!IS_ERR(thread))) {
+ kthread_bind(thread, cpu);
+ wake_up_process(thread);
+ *percpu_thread = thread;
+ }
+ /* prefer BSP as controlling CPU */
+ if (cpu == 0) {
+ control_cpu = 0;
+ smp_mb();
+ }
+ break;
+ case CPU_DEAD:
+ if (test_bit(cpu, cpu_clamping_mask)) {
+ pr_err("cpu %lu dead but powerclamping thread is not\n",
+ cpu);
+ kthread_stop(*percpu_thread);
+ }
+ if (cpu == control_cpu) {
+ control_cpu = smp_processor_id();
+ smp_mb();
+ }
+ }
+
+exit_ok:
+ return NOTIFY_OK;
+}
+
+static struct notifier_block powerclamp_cpu_notifier = {
+ .notifier_call = powerclamp_cpu_callback,
+};
+
+static int powerclamp_get_max_state(struct thermal_cooling_device *cdev,
+ unsigned long *state)
+{
+ *state = MAX_TARGET_RATIO;
+
+ return 0;
+}
+
+static int powerclamp_get_cur_state(struct thermal_cooling_device *cdev,
+ unsigned long *state)
+{
+ if (clamping)
+ *state = pkg_cstate_ratio_cur;
+ else
+ /* to save power, do not poll idle ratio while not clamping */
+ *state = -1; /* indicates invalid state */
+
+ return 0;
+}
+
+static int powerclamp_set_cur_state(struct thermal_cooling_device *cdev,
+ unsigned long new_target_ratio)
+{
+ int ret = 0;
+
+ new_target_ratio = clamp(new_target_ratio, 0UL,
+ (unsigned long) (MAX_TARGET_RATIO-1));
+ if (set_target_ratio == 0 && new_target_ratio > 0) {
+ pr_info("Start idle injection to reduce power\n");
+ set_target_ratio = new_target_ratio;
+ ret = start_power_clamp();
+ goto exit_set;
+ } else if (set_target_ratio > 0 && new_target_ratio == 0) {
+ pr_info("Stop forced idle injection\n");
+ set_target_ratio = 0;
+ end_power_clamp();
+ } else /* adjust currently running */ {
+ set_target_ratio = new_target_ratio;
+ /* make new set_target_ratio visible to other cpus */
+ smp_mb();
+ }
+
+exit_set:
+ return ret;
+}
+
+/* bind to generic thermal layer as cooling device*/
+static struct thermal_cooling_device_ops powerclamp_cooling_ops = {
+ .get_max_state = powerclamp_get_max_state,
+ .get_cur_state = powerclamp_get_cur_state,
+ .set_cur_state = powerclamp_set_cur_state,
+};
+
+/* runs on Nehalem and later */
+static const struct x86_cpu_id intel_powerclamp_ids[] = {
+ { X86_VENDOR_INTEL, 6, 0x1a},
+ { X86_VENDOR_INTEL, 6, 0x1c},
+ { X86_VENDOR_INTEL, 6, 0x1e},
+ { X86_VENDOR_INTEL, 6, 0x1f},
+ { X86_VENDOR_INTEL, 6, 0x25},
+ { X86_VENDOR_INTEL, 6, 0x26},
+ { X86_VENDOR_INTEL, 6, 0x2a},
+ { X86_VENDOR_INTEL, 6, 0x2c},
+ { X86_VENDOR_INTEL, 6, 0x2d},
+ { X86_VENDOR_INTEL, 6, 0x2e},
+ { X86_VENDOR_INTEL, 6, 0x2f},
+ { X86_VENDOR_INTEL, 6, 0x3a},
+ {}
+};
+MODULE_DEVICE_TABLE(x86cpu, intel_powerclamp_ids);
+
+static int powerclamp_probe(void)
+{
+ if (!x86_match_cpu(intel_powerclamp_ids)) {
+ pr_err("Intel powerclamp does not run on family %d model %d\n",
+ boot_cpu_data.x86, boot_cpu_data.x86_model);
+ return -ENODEV;
+ }
+ if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC) ||
+ !boot_cpu_has(X86_FEATURE_CONSTANT_TSC) ||
+ !boot_cpu_has(X86_FEATURE_MWAIT) ||
+ !boot_cpu_has(X86_FEATURE_ARAT))
+ return -ENODEV;
+
+ /* find the deepest mwait value */
+ find_target_mwait();
+
+ return 0;
+}
+
+static int powerclamp_debug_show(struct seq_file *m, void *unused)
+{
+ int i = 0;
+
+ seq_printf(m, "controlling cpu: %d\n", control_cpu);
+ seq_printf(m, "pct confidence steady dynamic (compensation)\n");
+ for (i = 0; i < MAX_TARGET_RATIO; i++) {
+ seq_printf(m, "%d\t%lu\t%lu\t%lu\n",
+ i,
+ cal_data[i].confidence,
+ cal_data[i].steady_comp,
+ cal_data[i].dynamic_comp);
+ }
+
+ return 0;
+}
+
+static int powerclamp_debug_open(struct inode *inode,
+ struct file *file)
+{
+ return single_open(file, powerclamp_debug_show, inode->i_private);
+}
+
+static const struct file_operations powerclamp_debug_fops = {
+ .open = powerclamp_debug_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+ .owner = THIS_MODULE,
+};
+
+static inline void powerclamp_create_debug_files(void)
+{
+ debug_dir = debugfs_create_dir("intel_powerclamp", NULL);
+ if (!debug_dir)
+ return;
+
+ if (!debugfs_create_file("powerclamp_calib", S_IRUGO, debug_dir,
+ cal_data, &powerclamp_debug_fops))
+ goto file_error;
+
+ return;
+
+file_error:
+ debugfs_remove_recursive(debug_dir);
+}
+
+static int powerclamp_init(void)
+{
+ int retval;
+ int bitmap_size;
+
+ bitmap_size = BITS_TO_LONGS(num_possible_cpus()) * sizeof(long);
+ cpu_clamping_mask = kzalloc(bitmap_size, GFP_KERNEL);
+ if (!cpu_clamping_mask)
+ return -ENOMEM;
+
+ /* probe cpu features and ids here */
+ retval = powerclamp_probe();
+ if (retval)
+ return retval;
+ /* set default limit, maybe adjusted during runtime based on feedback */
+ window_size = 2;
+ register_hotcpu_notifier(&powerclamp_cpu_notifier);
+ powerclamp_thread = alloc_percpu(struct task_struct *);
+ cooling_dev = thermal_cooling_device_register("intel_powerclamp", NULL,
+ &powerclamp_cooling_ops);
+ if (IS_ERR(cooling_dev))
+ return -ENODEV;
+
+ if (!duration)
+ duration = jiffies_to_msecs(DEFAULT_DURATION_JIFFIES);
+ powerclamp_create_debug_files();
+
+ return 0;
+}
+module_init(powerclamp_init);
+
+static void powerclamp_exit(void)
+{
+ unregister_hotcpu_notifier(&powerclamp_cpu_notifier);
+ end_power_clamp();
+ free_percpu(powerclamp_thread);
+ thermal_cooling_device_unregister(cooling_dev);
+ kfree(cpu_clamping_mask);
+
+ cancel_delayed_work_sync(&poll_pkg_cstate_work);
+ debugfs_remove_recursive(debug_dir);
+}
+module_exit(powerclamp_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Arjan van de Ven <arjan@linux.intel.com>");
+MODULE_AUTHOR("Jacob Pan <jacob.jun.pan@linux.intel.com>");
+MODULE_DESCRIPTION("Package Level C-state Idle Injection for Intel CPUs");
--
1.7.9.5
^ permalink raw reply related
* [PATCH v4 2/3] x86/nmi: export local_touch_nmi() symbol for modules
From: Jacob Pan @ 2013-01-03 11:01 UTC (permalink / raw)
To: Linux PM, LKML, Rafael Wysocki
Cc: Len Brown, Thomas Gleixner, H. Peter Anvin, Ingo Molnar,
Zhang Rui, Rob Landley, Arjan van de Ven, Paul McKenney,
Peter Zijlstra, Joe Perches, Jacob Pan
In-Reply-To: <1357210875-5025-1-git-send-email-jacob.jun.pan@linux.intel.com>
Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com>
---
arch/x86/kernel/nmi.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c
index f84f5c5..6030805 100644
--- a/arch/x86/kernel/nmi.c
+++ b/arch/x86/kernel/nmi.c
@@ -509,3 +509,4 @@ void local_touch_nmi(void)
{
__this_cpu_write(last_nmi_rip, 0);
}
+EXPORT_SYMBOL_GPL(local_touch_nmi);
--
1.7.9.5
^ permalink raw reply related
* Re: [PATCH 2/2] thermal: db8500: Use of_match_ptr() macro in db8500_cpufreq_cooling.c
From: Hongbo Zhang @ 2013-01-03 9:07 UTC (permalink / raw)
To: Sachin Kamat; +Cc: linux-pm, rui.zhang, patches, Hongbo Zhang
In-Reply-To: <1355907059-28720-2-git-send-email-sachin.kamat@linaro.org>
On 19 December 2012 16:50, Sachin Kamat <sachin.kamat@linaro.org> wrote:
> This eliminates having an #ifdef returning NULL for the case
> when OF is disabled.
>
> Cc: Hongbo Zhang <hongbo.zhang@stericsson.com>
> Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
> ---
> Compile tested on linux-next.
> ---
> drivers/thermal/db8500_cpufreq_cooling.c | 5 ++---
> 1 files changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/thermal/db8500_cpufreq_cooling.c b/drivers/thermal/db8500_cpufreq_cooling.c
> index 4cf8e72..2141985 100644
> --- a/drivers/thermal/db8500_cpufreq_cooling.c
> +++ b/drivers/thermal/db8500_cpufreq_cooling.c
> @@ -21,6 +21,7 @@
> #include <linux/cpufreq.h>
> #include <linux/err.h>
> #include <linux/module.h>
> +#include <linux/of.h>
> #include <linux/platform_device.h>
> #include <linux/slab.h>
>
> @@ -73,15 +74,13 @@ static const struct of_device_id db8500_cpufreq_cooling_match[] = {
> { .compatible = "stericsson,db8500-cpufreq-cooling" },
> {},
> };
> -#else
> -#define db8500_cpufreq_cooling_match NULL
> #endif
>
> static struct platform_driver db8500_cpufreq_cooling_driver = {
> .driver = {
> .owner = THIS_MODULE,
> .name = "db8500-cpufreq-cooling",
> - .of_match_table = db8500_cpufreq_cooling_match,
> + .of_match_table = of_match_ptr(db8500_cpufreq_cooling_match),
> },
> .probe = db8500_cpufreq_cooling_probe,
> .suspend = db8500_cpufreq_cooling_suspend,
> --
> 1.7.4.1
>
Good one.
Reviewed-by: hongbo.zhang@linaro.org
Thanks.
^ permalink raw reply
* Re: [PATCH 1/2] thermal: db8500: Use of_match_ptr() macro in db8500_thermal.c
From: Hongbo Zhang @ 2013-01-03 9:05 UTC (permalink / raw)
To: Sachin Kamat; +Cc: linux-pm, rui.zhang, patches, Hongbo Zhang
In-Reply-To: <1355907059-28720-1-git-send-email-sachin.kamat@linaro.org>
On 19 December 2012 16:50, Sachin Kamat <sachin.kamat@linaro.org> wrote:
> This eliminates having an #ifdef returning NULL for the case
> when OF is disabled.
>
> Cc: Hongbo Zhang <hongbo.zhang@stericsson.com>
> Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
> ---
> Compile tested on linux-next.
> ---
> drivers/thermal/db8500_thermal.c | 4 +---
> 1 files changed, 1 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/thermal/db8500_thermal.c b/drivers/thermal/db8500_thermal.c
> index ec71ade..61ce60a 100644
> --- a/drivers/thermal/db8500_thermal.c
> +++ b/drivers/thermal/db8500_thermal.c
> @@ -508,15 +508,13 @@ static const struct of_device_id db8500_thermal_match[] = {
> { .compatible = "stericsson,db8500-thermal" },
> {},
> };
> -#else
> -#define db8500_thermal_match NULL
> #endif
>
> static struct platform_driver db8500_thermal_driver = {
> .driver = {
> .owner = THIS_MODULE,
> .name = "db8500-thermal",
> - .of_match_table = db8500_thermal_match,
> + .of_match_table = of_match_ptr(db8500_thermal_match),
> },
> .probe = db8500_thermal_probe,
> .suspend = db8500_thermal_suspend,
> --
> 1.7.4.1
>
Good one.
Reviewed-by: hongbo.zhang@linaro.org
^ permalink raw reply
* Re: [PATCH 1/3] cpufreq: Manage only online cpus
From: Viresh Kumar @ 2013-01-03 3:32 UTC (permalink / raw)
To: Rafael J. Wysocki
Cc: linaro-dev, nicolas.pitre, amit.kucheria, mathieu.poirier,
linux-kernel, cpufreq, pdsw-power-team, linux-pm
In-Reply-To: <7569403.CvHy03LdiU@vostro.rjw.lan>
On 3 January 2013 06:43, Rafael J. Wysocki <rjw@sisk.pl> wrote:
>> BTW, i consider them as fixes and so would make sense to get them in next rc.
>> What do you think?
>
> Yes, if somebody tells me "yes, this fixes a problem for me". Otherwise,
> I don't quite see the reason.
I don't know how much people test HOTPLUG, but there are clear bugs related
to hotplug of cpus on a multiple cpu system :)
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox