android_kernel_oneplus_msm8998/drivers/cpuidle/cpuidle-powernv.c

341 lines
8.4 KiB
C
Raw Normal View History

/*
* cpuidle-powernv - idle state cpuidle driver.
* Adapted from drivers/cpuidle/cpuidle-pseries
*
*/
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/moduleparam.h>
#include <linux/cpuidle.h>
#include <linux/cpu.h>
#include <linux/notifier.h>
#include <linux/clockchips.h>
#include <linux/of.h>
#include <linux/slab.h>
#include <asm/machdep.h>
#include <asm/firmware.h>
#include <asm/opal.h>
#include <asm/runlatch.h>
#define MAX_POWERNV_IDLE_STATES 8
struct cpuidle_driver powernv_idle_driver = {
.name = "powernv_idle",
.owner = THIS_MODULE,
};
static int max_idle_state;
static struct cpuidle_state *cpuidle_state_table;
cpuidle: powernv: Fix promotion from snooze if next state disabled commit 0a4ec6aa035a52c422eceb2ed51ed88392a3d6c2 upstream. The commit 78eaa10f027c ("cpuidle: powernv/pseries: Auto-promotion of snooze to deeper idle state") introduced a timeout for the snooze idle state so that it could be eventually be promoted to a deeper idle state. The snooze timeout value is static and set to the target residency of the next idle state, which would train the cpuidle governor to pick the next idle state eventually. The unfortunate side-effect of this is that if the next idle state(s) is disabled, the CPU will forever remain in snooze, despite the fact that the system is completely idle, and other deeper idle states are available. This patch fixes the issue by dynamically setting the snooze timeout to the target residency of the next enabled state on the device. Before Patch: POWER8 : Only nap disabled. $ cpupower monitor sleep 30 sleep took 30.01297 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | Nap | Fast 0| 8| 0| 96.41| 0.00| 0.00 0| 8| 1| 96.43| 0.00| 0.00 0| 8| 2| 96.47| 0.00| 0.00 0| 8| 3| 96.35| 0.00| 0.00 0| 8| 4| 96.37| 0.00| 0.00 0| 8| 5| 96.37| 0.00| 0.00 0| 8| 6| 96.47| 0.00| 0.00 0| 8| 7| 96.47| 0.00| 0.00 POWER9: Shallow states (stop0lite, stop1lite, stop2lite, stop0, stop1, stop2) disabled: $ cpupower monitor sleep 30 sleep took 30.05033 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | stop | stop | stop | stop | stop | stop | stop | stop 0| 16| 0| 89.79| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 1| 90.12| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 2| 90.21| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 3| 90.29| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 After Patch: POWER8 : Only nap disabled. $ cpupower monitor sleep 30 sleep took 30.01200 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | Nap | Fast 0| 8| 0| 16.58| 0.00| 77.21 0| 8| 1| 18.42| 0.00| 75.38 0| 8| 2| 4.70| 0.00| 94.09 0| 8| 3| 17.06| 0.00| 81.73 0| 8| 4| 3.06| 0.00| 95.73 0| 8| 5| 7.00| 0.00| 96.80 0| 8| 6| 1.00| 0.00| 98.79 0| 8| 7| 5.62| 0.00| 94.17 POWER9: Shallow states (stop0lite, stop1lite, stop2lite, stop0, stop1, stop2) disabled: $ cpupower monitor sleep 30 sleep took 30.02110 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | stop | stop | stop | stop | stop | stop | stop | stop 0| 0| 0| 0.69| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 9.39| 89.70 0| 0| 1| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.05| 93.21 0| 0| 2| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 89.93 0| 0| 3| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 93.26 Fixes: 78eaa10f027c ("cpuidle: powernv/pseries: Auto-promotion of snooze to deeper idle state") Cc: stable@vger.kernel.org # v4.2+ Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com> Reviewed-by: Balbir Singh <bsingharora@gmail.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-05-31 17:45:09 +05:30
static u64 default_snooze_timeout;
static bool snooze_timeout_en;
cpuidle: powernv: Fix promotion from snooze if next state disabled commit 0a4ec6aa035a52c422eceb2ed51ed88392a3d6c2 upstream. The commit 78eaa10f027c ("cpuidle: powernv/pseries: Auto-promotion of snooze to deeper idle state") introduced a timeout for the snooze idle state so that it could be eventually be promoted to a deeper idle state. The snooze timeout value is static and set to the target residency of the next idle state, which would train the cpuidle governor to pick the next idle state eventually. The unfortunate side-effect of this is that if the next idle state(s) is disabled, the CPU will forever remain in snooze, despite the fact that the system is completely idle, and other deeper idle states are available. This patch fixes the issue by dynamically setting the snooze timeout to the target residency of the next enabled state on the device. Before Patch: POWER8 : Only nap disabled. $ cpupower monitor sleep 30 sleep took 30.01297 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | Nap | Fast 0| 8| 0| 96.41| 0.00| 0.00 0| 8| 1| 96.43| 0.00| 0.00 0| 8| 2| 96.47| 0.00| 0.00 0| 8| 3| 96.35| 0.00| 0.00 0| 8| 4| 96.37| 0.00| 0.00 0| 8| 5| 96.37| 0.00| 0.00 0| 8| 6| 96.47| 0.00| 0.00 0| 8| 7| 96.47| 0.00| 0.00 POWER9: Shallow states (stop0lite, stop1lite, stop2lite, stop0, stop1, stop2) disabled: $ cpupower monitor sleep 30 sleep took 30.05033 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | stop | stop | stop | stop | stop | stop | stop | stop 0| 16| 0| 89.79| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 1| 90.12| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 2| 90.21| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 3| 90.29| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 After Patch: POWER8 : Only nap disabled. $ cpupower monitor sleep 30 sleep took 30.01200 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | Nap | Fast 0| 8| 0| 16.58| 0.00| 77.21 0| 8| 1| 18.42| 0.00| 75.38 0| 8| 2| 4.70| 0.00| 94.09 0| 8| 3| 17.06| 0.00| 81.73 0| 8| 4| 3.06| 0.00| 95.73 0| 8| 5| 7.00| 0.00| 96.80 0| 8| 6| 1.00| 0.00| 98.79 0| 8| 7| 5.62| 0.00| 94.17 POWER9: Shallow states (stop0lite, stop1lite, stop2lite, stop0, stop1, stop2) disabled: $ cpupower monitor sleep 30 sleep took 30.02110 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | stop | stop | stop | stop | stop | stop | stop | stop 0| 0| 0| 0.69| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 9.39| 89.70 0| 0| 1| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.05| 93.21 0| 0| 2| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 89.93 0| 0| 3| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 93.26 Fixes: 78eaa10f027c ("cpuidle: powernv/pseries: Auto-promotion of snooze to deeper idle state") Cc: stable@vger.kernel.org # v4.2+ Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com> Reviewed-by: Balbir Singh <bsingharora@gmail.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-05-31 17:45:09 +05:30
static u64 get_snooze_timeout(struct cpuidle_device *dev,
struct cpuidle_driver *drv,
int index)
{
int i;
if (unlikely(!snooze_timeout_en))
return default_snooze_timeout;
for (i = index + 1; i < drv->state_count; i++) {
struct cpuidle_state *s = &drv->states[i];
struct cpuidle_state_usage *su = &dev->states_usage[i];
if (s->disabled || su->disable)
continue;
return s->target_residency * tb_ticks_per_usec;
}
return default_snooze_timeout;
}
static int snooze_loop(struct cpuidle_device *dev,
struct cpuidle_driver *drv,
int index)
{
u64 snooze_exit_time;
local_irq_enable();
set_thread_flag(TIF_POLLING_NRFLAG);
cpuidle: powernv: Fix promotion from snooze if next state disabled commit 0a4ec6aa035a52c422eceb2ed51ed88392a3d6c2 upstream. The commit 78eaa10f027c ("cpuidle: powernv/pseries: Auto-promotion of snooze to deeper idle state") introduced a timeout for the snooze idle state so that it could be eventually be promoted to a deeper idle state. The snooze timeout value is static and set to the target residency of the next idle state, which would train the cpuidle governor to pick the next idle state eventually. The unfortunate side-effect of this is that if the next idle state(s) is disabled, the CPU will forever remain in snooze, despite the fact that the system is completely idle, and other deeper idle states are available. This patch fixes the issue by dynamically setting the snooze timeout to the target residency of the next enabled state on the device. Before Patch: POWER8 : Only nap disabled. $ cpupower monitor sleep 30 sleep took 30.01297 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | Nap | Fast 0| 8| 0| 96.41| 0.00| 0.00 0| 8| 1| 96.43| 0.00| 0.00 0| 8| 2| 96.47| 0.00| 0.00 0| 8| 3| 96.35| 0.00| 0.00 0| 8| 4| 96.37| 0.00| 0.00 0| 8| 5| 96.37| 0.00| 0.00 0| 8| 6| 96.47| 0.00| 0.00 0| 8| 7| 96.47| 0.00| 0.00 POWER9: Shallow states (stop0lite, stop1lite, stop2lite, stop0, stop1, stop2) disabled: $ cpupower monitor sleep 30 sleep took 30.05033 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | stop | stop | stop | stop | stop | stop | stop | stop 0| 16| 0| 89.79| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 1| 90.12| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 2| 90.21| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 3| 90.29| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 After Patch: POWER8 : Only nap disabled. $ cpupower monitor sleep 30 sleep took 30.01200 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | Nap | Fast 0| 8| 0| 16.58| 0.00| 77.21 0| 8| 1| 18.42| 0.00| 75.38 0| 8| 2| 4.70| 0.00| 94.09 0| 8| 3| 17.06| 0.00| 81.73 0| 8| 4| 3.06| 0.00| 95.73 0| 8| 5| 7.00| 0.00| 96.80 0| 8| 6| 1.00| 0.00| 98.79 0| 8| 7| 5.62| 0.00| 94.17 POWER9: Shallow states (stop0lite, stop1lite, stop2lite, stop0, stop1, stop2) disabled: $ cpupower monitor sleep 30 sleep took 30.02110 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | stop | stop | stop | stop | stop | stop | stop | stop 0| 0| 0| 0.69| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 9.39| 89.70 0| 0| 1| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.05| 93.21 0| 0| 2| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 89.93 0| 0| 3| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 93.26 Fixes: 78eaa10f027c ("cpuidle: powernv/pseries: Auto-promotion of snooze to deeper idle state") Cc: stable@vger.kernel.org # v4.2+ Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com> Reviewed-by: Balbir Singh <bsingharora@gmail.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-05-31 17:45:09 +05:30
snooze_exit_time = get_tb() + get_snooze_timeout(dev, drv, index);
ppc64_runlatch_off();
while (!need_resched()) {
HMT_low();
HMT_very_low();
if (snooze_timeout_en && get_tb() > snooze_exit_time)
break;
}
HMT_medium();
ppc64_runlatch_on();
clear_thread_flag(TIF_POLLING_NRFLAG);
smp_mb();
return index;
}
static int nap_loop(struct cpuidle_device *dev,
struct cpuidle_driver *drv,
int index)
{
ppc64_runlatch_off();
power7_idle();
ppc64_runlatch_on();
return index;
}
tick/idle/powerpc: Do not register idle states with CPUIDLE_FLAG_TIMER_STOP set in periodic mode On some archs, the local clockevent device stops in deep cpuidle states. The broadcast framework is used to wakeup cpus in these idle states, in which either an external clockevent device is used to send wakeup ipis or the hrtimer broadcast framework kicks in in the absence of such a device. One cpu is nominated as the broadcast cpu and this cpu sends wakeup ipis to sleeping cpus at the appropriate time. This is the implementation in the oneshot mode of broadcast. In periodic mode of broadcast however, the presence of such cpuidle states results in the cpuidle driver calling tick_broadcast_enable() which shuts down the local clockevent devices of all the cpus and appoints the tick broadcast device as the clockevent device for each of them. This works on those archs where the tick broadcast device is a real clockevent device. But on archs which depend on the hrtimer mode of broadcast, the tick broadcast device hapens to be a pseudo device. The consequence is that the local clockevent devices of all cpus are shutdown and the kernel hangs at boot time in periodic mode. Let us thus not register the cpuidle states which have CPUIDLE_FLAG_TIMER_STOP flag set, on archs which depend on the hrtimer mode of broadcast in periodic mode. This patch takes care of doing this on powerpc. The cpus would not have entered into such deep cpuidle states in periodic mode on powerpc anyway. So there is no loss here. Signed-off-by: Preeti U Murthy <preeti@linux.vnet.ibm.com> Cc: 3.19+ <stable@vger.kernel.org> # 3.19+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-24 01:48:01 -05:00
/* Register for fastsleep only in oneshot mode of broadcast */
#ifdef CONFIG_TICK_ONESHOT
static int fastsleep_loop(struct cpuidle_device *dev,
struct cpuidle_driver *drv,
int index)
{
unsigned long old_lpcr = mfspr(SPRN_LPCR);
unsigned long new_lpcr;
if (unlikely(system_state < SYSTEM_RUNNING))
return index;
new_lpcr = old_lpcr;
/* Do not exit powersave upon decrementer as we've setup the timer
* offload.
*/
new_lpcr &= ~LPCR_PECE1;
mtspr(SPRN_LPCR, new_lpcr);
power7_sleep();
mtspr(SPRN_LPCR, old_lpcr);
return index;
}
tick/idle/powerpc: Do not register idle states with CPUIDLE_FLAG_TIMER_STOP set in periodic mode On some archs, the local clockevent device stops in deep cpuidle states. The broadcast framework is used to wakeup cpus in these idle states, in which either an external clockevent device is used to send wakeup ipis or the hrtimer broadcast framework kicks in in the absence of such a device. One cpu is nominated as the broadcast cpu and this cpu sends wakeup ipis to sleeping cpus at the appropriate time. This is the implementation in the oneshot mode of broadcast. In periodic mode of broadcast however, the presence of such cpuidle states results in the cpuidle driver calling tick_broadcast_enable() which shuts down the local clockevent devices of all the cpus and appoints the tick broadcast device as the clockevent device for each of them. This works on those archs where the tick broadcast device is a real clockevent device. But on archs which depend on the hrtimer mode of broadcast, the tick broadcast device hapens to be a pseudo device. The consequence is that the local clockevent devices of all cpus are shutdown and the kernel hangs at boot time in periodic mode. Let us thus not register the cpuidle states which have CPUIDLE_FLAG_TIMER_STOP flag set, on archs which depend on the hrtimer mode of broadcast in periodic mode. This patch takes care of doing this on powerpc. The cpus would not have entered into such deep cpuidle states in periodic mode on powerpc anyway. So there is no loss here. Signed-off-by: Preeti U Murthy <preeti@linux.vnet.ibm.com> Cc: 3.19+ <stable@vger.kernel.org> # 3.19+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-24 01:48:01 -05:00
#endif
/*
* States for dedicated partition case.
*/
static struct cpuidle_state powernv_states[MAX_POWERNV_IDLE_STATES] = {
{ /* Snooze */
.name = "snooze",
.desc = "snooze",
.exit_latency = 0,
.target_residency = 0,
.enter = &snooze_loop },
};
static int powernv_cpuidle_add_cpu_notifier(struct notifier_block *n,
unsigned long action, void *hcpu)
{
int hotcpu = (unsigned long)hcpu;
struct cpuidle_device *dev =
per_cpu(cpuidle_devices, hotcpu);
if (dev && cpuidle_get_driver()) {
switch (action) {
case CPU_ONLINE:
case CPU_ONLINE_FROZEN:
cpuidle_pause_and_lock();
cpuidle_enable_device(dev);
cpuidle_resume_and_unlock();
break;
case CPU_DEAD:
case CPU_DEAD_FROZEN:
cpuidle_pause_and_lock();
cpuidle_disable_device(dev);
cpuidle_resume_and_unlock();
break;
default:
return NOTIFY_DONE;
}
}
return NOTIFY_OK;
}
static struct notifier_block setup_hotplug_notifier = {
.notifier_call = powernv_cpuidle_add_cpu_notifier,
};
/*
* powernv_cpuidle_driver_init()
*/
static int powernv_cpuidle_driver_init(void)
{
int idle_state;
struct cpuidle_driver *drv = &powernv_idle_driver;
drv->state_count = 0;
for (idle_state = 0; idle_state < max_idle_state; ++idle_state) {
/* Is the state not enabled? */
if (cpuidle_state_table[idle_state].enter == NULL)
continue;
drv->states[drv->state_count] = /* structure copy */
cpuidle_state_table[idle_state];
drv->state_count += 1;
}
cpuidle: powernv: Pass correct drv->cpumask for registration [ Upstream commit 293d264f13cbde328d5477f49e3103edbc1dc191 ] drv->cpumask defaults to cpu_possible_mask in __cpuidle_driver_init(). On PowerNV platform cpu_present could be less than cpu_possible in cases where firmware detects the cpu, but it is not available to the OS. When CONFIG_HOTPLUG_CPU=n, such cpus are not hotplugable at runtime and hence we skip creating cpu_device. This breaks cpuidle on powernv where register_cpu() is not called for cpus in cpu_possible_mask that cannot be hot-added at runtime. Trying cpuidle_register_device() on cpu without cpu_device will cause crash like this: cpu 0xf: Vector: 380 (Data SLB Access) at [c000000ff1503490] pc: c00000000022c8bc: string+0x34/0x60 lr: c00000000022ed78: vsnprintf+0x284/0x42c sp: c000000ff1503710 msr: 9000000000009033 dar: 6000000060000000 current = 0xc000000ff1480000 paca = 0xc00000000fe82d00 softe: 0 irq_happened: 0x01 pid = 1, comm = swapper/8 Linux version 4.11.0-rc2 (sv@sagarika) (gcc version 4.9.4 (Buildroot 2017.02-00004-gc28573e) ) #15 SMP Fri Mar 17 19:32:02 IST 2017 enter ? for help [link register ] c00000000022ed78 vsnprintf+0x284/0x42c [c000000ff1503710] c00000000022ebb8 vsnprintf+0xc4/0x42c (unreliable) [c000000ff1503800] c00000000022ef40 vscnprintf+0x20/0x44 [c000000ff1503830] c0000000000ab61c vprintk_emit+0x94/0x2cc [c000000ff15038a0] c0000000000acc9c vprintk_func+0x60/0x74 [c000000ff15038c0] c000000000619694 printk+0x38/0x4c [c000000ff15038e0] c000000000224950 kobject_get+0x40/0x60 [c000000ff1503950] c00000000022507c kobject_add_internal+0x60/0x2c4 [c000000ff15039e0] c000000000225350 kobject_init_and_add+0x70/0x78 [c000000ff1503a60] c00000000053c288 cpuidle_add_sysfs+0x9c/0xe0 [c000000ff1503ae0] c00000000053aeac cpuidle_register_device+0xd4/0x12c [c000000ff1503b30] c00000000053b108 cpuidle_register+0x98/0xcc [c000000ff1503bc0] c00000000085eaf0 powernv_processor_idle_init+0x140/0x1e0 [c000000ff1503c60] c00000000000cd60 do_one_initcall+0xc0/0x15c [c000000ff1503d20] c000000000833e84 kernel_init_freeable+0x1a0/0x25c [c000000ff1503dc0] c00000000000d478 kernel_init+0x24/0x12c [c000000ff1503e30] c00000000000b564 ret_from_kernel_thread+0x5c/0x78 This patch fixes the bug by passing correct cpumask from powernv-cpuidle driver. Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com> Acked-by: Michael Ellerman <mpe@ellerman.id.au> [ rjw: Comment massage ] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <alexander.levin@verizon.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-03-23 20:52:46 +05:30
/*
* On the PowerNV platform cpu_present may be less than cpu_possible in
* cases when firmware detects the CPU, but it is not available to the
* OS. If CONFIG_HOTPLUG_CPU=n, then such CPUs are not hotplugable at
* run time and hence cpu_devices are not created for those CPUs by the
* generic topology_init().
*
* drv->cpumask defaults to cpu_possible_mask in
* __cpuidle_driver_init(). This breaks cpuidle on PowerNV where
* cpu_devices are not created for CPUs in cpu_possible_mask that
* cannot be hot-added later at run time.
*
* Trying cpuidle_register_device() on a CPU without a cpu_device is
* incorrect, so pass a correct CPU mask to the generic cpuidle driver.
*/
drv->cpumask = (struct cpumask *)cpu_present_mask;
return 0;
}
static int powernv_add_idle_states(void)
{
struct device_node *power_mgt;
int nr_idle_states = 1; /* Snooze */
int dt_idle_states;
u32 *latency_ns, *residency_ns, *flags;
int i, rc;
/* Currently we have snooze statically defined */
power_mgt = of_find_node_by_path("/ibm,opal/power-mgt");
if (!power_mgt) {
pr_warn("opal: PowerMgmt Node not found\n");
goto out;
}
/* Read values of any property to determine the num of idle states */
dt_idle_states = of_property_count_u32_elems(power_mgt, "ibm,cpu-idle-state-flags");
if (dt_idle_states < 0) {
pr_warn("cpuidle-powernv: no idle states found in the DT\n");
goto out;
}
flags = kzalloc(sizeof(*flags) * dt_idle_states, GFP_KERNEL);
if (of_property_read_u32_array(power_mgt,
"ibm,cpu-idle-state-flags", flags, dt_idle_states)) {
pr_warn("cpuidle-powernv : missing ibm,cpu-idle-state-flags in DT\n");
goto out_free_flags;
}
latency_ns = kzalloc(sizeof(*latency_ns) * dt_idle_states, GFP_KERNEL);
rc = of_property_read_u32_array(power_mgt,
"ibm,cpu-idle-state-latencies-ns", latency_ns, dt_idle_states);
if (rc) {
pr_warn("cpuidle-powernv: missing ibm,cpu-idle-state-latencies-ns in DT\n");
goto out_free_latency;
}
residency_ns = kzalloc(sizeof(*residency_ns) * dt_idle_states, GFP_KERNEL);
rc = of_property_read_u32_array(power_mgt,
"ibm,cpu-idle-state-residency-ns", residency_ns, dt_idle_states);
for (i = 0; i < dt_idle_states; i++) {
/*
* Cpuidle accepts exit_latency and target_residency in us.
* Use default target_residency values if f/w does not expose it.
*/
if (flags[i] & OPAL_PM_NAP_ENABLED) {
/* Add NAP state */
strcpy(powernv_states[nr_idle_states].name, "Nap");
strcpy(powernv_states[nr_idle_states].desc, "Nap");
powernv_states[nr_idle_states].flags = 0;
powernv_states[nr_idle_states].target_residency = 100;
powernv_states[nr_idle_states].enter = &nap_loop;
tick/idle/powerpc: Do not register idle states with CPUIDLE_FLAG_TIMER_STOP set in periodic mode On some archs, the local clockevent device stops in deep cpuidle states. The broadcast framework is used to wakeup cpus in these idle states, in which either an external clockevent device is used to send wakeup ipis or the hrtimer broadcast framework kicks in in the absence of such a device. One cpu is nominated as the broadcast cpu and this cpu sends wakeup ipis to sleeping cpus at the appropriate time. This is the implementation in the oneshot mode of broadcast. In periodic mode of broadcast however, the presence of such cpuidle states results in the cpuidle driver calling tick_broadcast_enable() which shuts down the local clockevent devices of all the cpus and appoints the tick broadcast device as the clockevent device for each of them. This works on those archs where the tick broadcast device is a real clockevent device. But on archs which depend on the hrtimer mode of broadcast, the tick broadcast device hapens to be a pseudo device. The consequence is that the local clockevent devices of all cpus are shutdown and the kernel hangs at boot time in periodic mode. Let us thus not register the cpuidle states which have CPUIDLE_FLAG_TIMER_STOP flag set, on archs which depend on the hrtimer mode of broadcast in periodic mode. This patch takes care of doing this on powerpc. The cpus would not have entered into such deep cpuidle states in periodic mode on powerpc anyway. So there is no loss here. Signed-off-by: Preeti U Murthy <preeti@linux.vnet.ibm.com> Cc: 3.19+ <stable@vger.kernel.org> # 3.19+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-24 01:48:01 -05:00
}
/*
* All cpuidle states with CPUIDLE_FLAG_TIMER_STOP set must come
* within this config dependency check.
*/
#ifdef CONFIG_TICK_ONESHOT
if (flags[i] & OPAL_PM_SLEEP_ENABLED ||
flags[i] & OPAL_PM_SLEEP_ENABLED_ER1) {
/* Add FASTSLEEP state */
strcpy(powernv_states[nr_idle_states].name, "FastSleep");
strcpy(powernv_states[nr_idle_states].desc, "FastSleep");
powernv_states[nr_idle_states].flags = CPUIDLE_FLAG_TIMER_STOP;
powernv_states[nr_idle_states].target_residency = 300000;
powernv_states[nr_idle_states].enter = &fastsleep_loop;
}
tick/idle/powerpc: Do not register idle states with CPUIDLE_FLAG_TIMER_STOP set in periodic mode On some archs, the local clockevent device stops in deep cpuidle states. The broadcast framework is used to wakeup cpus in these idle states, in which either an external clockevent device is used to send wakeup ipis or the hrtimer broadcast framework kicks in in the absence of such a device. One cpu is nominated as the broadcast cpu and this cpu sends wakeup ipis to sleeping cpus at the appropriate time. This is the implementation in the oneshot mode of broadcast. In periodic mode of broadcast however, the presence of such cpuidle states results in the cpuidle driver calling tick_broadcast_enable() which shuts down the local clockevent devices of all the cpus and appoints the tick broadcast device as the clockevent device for each of them. This works on those archs where the tick broadcast device is a real clockevent device. But on archs which depend on the hrtimer mode of broadcast, the tick broadcast device hapens to be a pseudo device. The consequence is that the local clockevent devices of all cpus are shutdown and the kernel hangs at boot time in periodic mode. Let us thus not register the cpuidle states which have CPUIDLE_FLAG_TIMER_STOP flag set, on archs which depend on the hrtimer mode of broadcast in periodic mode. This patch takes care of doing this on powerpc. The cpus would not have entered into such deep cpuidle states in periodic mode on powerpc anyway. So there is no loss here. Signed-off-by: Preeti U Murthy <preeti@linux.vnet.ibm.com> Cc: 3.19+ <stable@vger.kernel.org> # 3.19+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-24 01:48:01 -05:00
#endif
powernv_states[nr_idle_states].exit_latency =
((unsigned int)latency_ns[i]) / 1000;
if (!rc) {
powernv_states[nr_idle_states].target_residency =
((unsigned int)residency_ns[i]) / 1000;
}
nr_idle_states++;
}
kfree(residency_ns);
out_free_latency:
kfree(latency_ns);
out_free_flags:
kfree(flags);
out:
return nr_idle_states;
}
/*
* powernv_idle_probe()
* Choose state table for shared versus dedicated partition
*/
static int powernv_idle_probe(void)
{
if (cpuidle_disable != IDLE_NO_OVERRIDE)
return -ENODEV;
if (firmware_has_feature(FW_FEATURE_OPAL)) {
cpuidle_state_table = powernv_states;
/* Device tree can indicate more idle states */
max_idle_state = powernv_add_idle_states();
cpuidle: powernv: Fix promotion from snooze if next state disabled commit 0a4ec6aa035a52c422eceb2ed51ed88392a3d6c2 upstream. The commit 78eaa10f027c ("cpuidle: powernv/pseries: Auto-promotion of snooze to deeper idle state") introduced a timeout for the snooze idle state so that it could be eventually be promoted to a deeper idle state. The snooze timeout value is static and set to the target residency of the next idle state, which would train the cpuidle governor to pick the next idle state eventually. The unfortunate side-effect of this is that if the next idle state(s) is disabled, the CPU will forever remain in snooze, despite the fact that the system is completely idle, and other deeper idle states are available. This patch fixes the issue by dynamically setting the snooze timeout to the target residency of the next enabled state on the device. Before Patch: POWER8 : Only nap disabled. $ cpupower monitor sleep 30 sleep took 30.01297 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | Nap | Fast 0| 8| 0| 96.41| 0.00| 0.00 0| 8| 1| 96.43| 0.00| 0.00 0| 8| 2| 96.47| 0.00| 0.00 0| 8| 3| 96.35| 0.00| 0.00 0| 8| 4| 96.37| 0.00| 0.00 0| 8| 5| 96.37| 0.00| 0.00 0| 8| 6| 96.47| 0.00| 0.00 0| 8| 7| 96.47| 0.00| 0.00 POWER9: Shallow states (stop0lite, stop1lite, stop2lite, stop0, stop1, stop2) disabled: $ cpupower monitor sleep 30 sleep took 30.05033 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | stop | stop | stop | stop | stop | stop | stop | stop 0| 16| 0| 89.79| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 1| 90.12| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 2| 90.21| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 0| 16| 3| 90.29| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00 After Patch: POWER8 : Only nap disabled. $ cpupower monitor sleep 30 sleep took 30.01200 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | Nap | Fast 0| 8| 0| 16.58| 0.00| 77.21 0| 8| 1| 18.42| 0.00| 75.38 0| 8| 2| 4.70| 0.00| 94.09 0| 8| 3| 17.06| 0.00| 81.73 0| 8| 4| 3.06| 0.00| 95.73 0| 8| 5| 7.00| 0.00| 96.80 0| 8| 6| 1.00| 0.00| 98.79 0| 8| 7| 5.62| 0.00| 94.17 POWER9: Shallow states (stop0lite, stop1lite, stop2lite, stop0, stop1, stop2) disabled: $ cpupower monitor sleep 30 sleep took 30.02110 seconds and exited with status 0 |Idle_Stats PKG |CORE|CPU | snoo | stop | stop | stop | stop | stop | stop | stop | stop 0| 0| 0| 0.69| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 9.39| 89.70 0| 0| 1| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.05| 93.21 0| 0| 2| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 89.93 0| 0| 3| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 0.00| 93.26 Fixes: 78eaa10f027c ("cpuidle: powernv/pseries: Auto-promotion of snooze to deeper idle state") Cc: stable@vger.kernel.org # v4.2+ Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com> Reviewed-by: Balbir Singh <bsingharora@gmail.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2018-05-31 17:45:09 +05:30
default_snooze_timeout = TICK_USEC * tb_ticks_per_usec;
if (max_idle_state > 1)
snooze_timeout_en = true;
} else
return -ENODEV;
return 0;
}
static int __init powernv_processor_idle_init(void)
{
int retval;
retval = powernv_idle_probe();
if (retval)
return retval;
powernv_cpuidle_driver_init();
retval = cpuidle_register(&powernv_idle_driver, NULL);
if (retval) {
printk(KERN_DEBUG "Registration of powernv driver failed.\n");
return retval;
}
register_cpu_notifier(&setup_hotplug_notifier);
printk(KERN_DEBUG "powernv_idle_driver registered\n");
return 0;
}
device_initcall(powernv_processor_idle_init);