linux-stable/drivers/cpuidle/cpuidle-powernv.c
Gautham R. Shenoy 443004a666 cpuidle: powernv: Fix promotion from snooze if next state disabled
commit 0a4ec6aa03 upstream.

The commit 78eaa10f02 ("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: 78eaa10f02 ("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>
2018-07-03 11:23:08 +02:00

398 lines
10 KiB
C

/*
* 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 POWERNV_THRESHOLD_LATENCY_NS 200000
struct cpuidle_driver powernv_idle_driver = {
.name = "powernv_idle",
.owner = THIS_MODULE,
};
static int max_idle_state;
static struct cpuidle_state *cpuidle_state_table;
static u64 stop_psscr_table[CPUIDLE_STATE_MAX];
static u64 default_snooze_timeout;
static bool snooze_timeout_en;
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);
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;
}
/* 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;
}
#endif
static int stop_loop(struct cpuidle_device *dev,
struct cpuidle_driver *drv,
int index)
{
ppc64_runlatch_off();
power9_idle_stop(stop_psscr_table[index]);
ppc64_runlatch_on();
return index;
}
/*
* States for dedicated partition case.
*/
static struct cpuidle_state powernv_states[CPUIDLE_STATE_MAX] = {
{ /* Snooze */
.name = "snooze",
.desc = "snooze",
.exit_latency = 0,
.target_residency = 0,
.enter = snooze_loop },
};
static int powernv_cpuidle_cpu_online(unsigned int cpu)
{
struct cpuidle_device *dev = per_cpu(cpuidle_devices, cpu);
if (dev && cpuidle_get_driver()) {
cpuidle_pause_and_lock();
cpuidle_enable_device(dev);
cpuidle_resume_and_unlock();
}
return 0;
}
static int powernv_cpuidle_cpu_dead(unsigned int cpu)
{
struct cpuidle_device *dev = per_cpu(cpuidle_devices, cpu);
if (dev && cpuidle_get_driver()) {
cpuidle_pause_and_lock();
cpuidle_disable_device(dev);
cpuidle_resume_and_unlock();
}
return 0;
}
/*
* 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;
}
/*
* 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[CPUIDLE_STATE_MAX];
u32 residency_ns[CPUIDLE_STATE_MAX];
u32 flags[CPUIDLE_STATE_MAX];
u64 psscr_val[CPUIDLE_STATE_MAX];
const char *names[CPUIDLE_STATE_MAX];
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;
}
/*
* Since snooze is used as first idle state, max idle states allowed is
* CPUIDLE_STATE_MAX -1
*/
if (dt_idle_states > CPUIDLE_STATE_MAX - 1) {
pr_warn("cpuidle-powernv: discovered idle states more than allowed");
dt_idle_states = CPUIDLE_STATE_MAX - 1;
}
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;
}
if (of_property_read_u32_array(power_mgt,
"ibm,cpu-idle-state-latencies-ns", latency_ns,
dt_idle_states)) {
pr_warn("cpuidle-powernv: missing ibm,cpu-idle-state-latencies-ns in DT\n");
goto out;
}
if (of_property_read_string_array(power_mgt,
"ibm,cpu-idle-state-names", names, dt_idle_states) < 0) {
pr_warn("cpuidle-powernv: missing ibm,cpu-idle-state-names in DT\n");
goto out;
}
/*
* If the idle states use stop instruction, probe for psscr values
* which are necessary to specify required stop level.
*/
if (flags[0] & (OPAL_PM_STOP_INST_FAST | OPAL_PM_STOP_INST_DEEP))
if (of_property_read_u64_array(power_mgt,
"ibm,cpu-idle-state-psscr", psscr_val, dt_idle_states)) {
pr_warn("cpuidle-powernv: missing ibm,cpu-idle-states-psscr in DT\n");
goto out;
}
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++) {
/*
* If an idle state has exit latency beyond
* POWERNV_THRESHOLD_LATENCY_NS then don't use it
* in cpu-idle.
*/
if (latency_ns[i] > POWERNV_THRESHOLD_LATENCY_NS)
continue;
/*
* 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;
} else if ((flags[i] & OPAL_PM_STOP_INST_FAST) &&
!(flags[i] & OPAL_PM_TIMEBASE_STOP)) {
strncpy(powernv_states[nr_idle_states].name,
names[i], CPUIDLE_NAME_LEN);
strncpy(powernv_states[nr_idle_states].desc,
names[i], CPUIDLE_NAME_LEN);
powernv_states[nr_idle_states].flags = 0;
powernv_states[nr_idle_states].enter = stop_loop;
stop_psscr_table[nr_idle_states] = psscr_val[i];
}
/*
* 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;
} else if ((flags[i] & OPAL_PM_STOP_INST_DEEP) &&
(flags[i] & OPAL_PM_TIMEBASE_STOP)) {
strncpy(powernv_states[nr_idle_states].name,
names[i], CPUIDLE_NAME_LEN);
strncpy(powernv_states[nr_idle_states].desc,
names[i], CPUIDLE_NAME_LEN);
powernv_states[nr_idle_states].flags = CPUIDLE_FLAG_TIMER_STOP;
powernv_states[nr_idle_states].enter = stop_loop;
stop_psscr_table[nr_idle_states] = psscr_val[i];
}
#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++;
}
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();
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;
}
retval = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN,
"cpuidle/powernv:online",
powernv_cpuidle_cpu_online, NULL);
WARN_ON(retval < 0);
retval = cpuhp_setup_state_nocalls(CPUHP_CPUIDLE_DEAD,
"cpuidle/powernv:dead", NULL,
powernv_cpuidle_cpu_dead);
WARN_ON(retval < 0);
printk(KERN_DEBUG "powernv_idle_driver registered\n");
return 0;
}
device_initcall(powernv_processor_idle_init);