cpuidle-pseries.c 6.4 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2
/*
3
 *  cpuidle-pseries - idle state cpuidle driver.
4 5 6 7 8 9 10 11 12 13 14
 *  Adapted from drivers/idle/intel_idle.c and
 *  drivers/acpi/processor_idle.c
 *
 */

#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/moduleparam.h>
#include <linux/cpuidle.h>
#include <linux/cpu.h>
15
#include <linux/notifier.h>
16 17 18 19 20

#include <asm/paca.h>
#include <asm/reg.h>
#include <asm/machdep.h>
#include <asm/firmware.h>
21
#include <asm/runlatch.h>
22
#include <asm/plpar_wrappers.h>
23 24

struct cpuidle_driver pseries_idle_driver = {
25 26
	.name             = "pseries_idle",
	.owner            = THIS_MODULE,
27 28
};

29 30 31 32
static int max_idle_state __read_mostly;
static struct cpuidle_state *cpuidle_state_table __read_mostly;
static u64 snooze_timeout __read_mostly;
static bool snooze_timeout_en __read_mostly;
33

34
static inline void idle_loop_prolog(unsigned long *in_purr)
35
{
36
	ppc64_runlatch_off();
37 38 39 40 41 42 43 44
	*in_purr = mfspr(SPRN_PURR);
	/*
	 * Indicate to the HV that we are idle. Now would be
	 * a good time to find other work to dispatch.
	 */
	get_lppaca()->idle = 1;
}

45
static inline void idle_loop_epilog(unsigned long in_purr)
46
{
47 48 49 50 51
	u64 wait_cycles;

	wait_cycles = be64_to_cpu(get_lppaca()->wait_state_cycles);
	wait_cycles += mfspr(SPRN_PURR) - in_purr;
	get_lppaca()->wait_state_cycles = cpu_to_be64(wait_cycles);
52
	get_lppaca()->idle = 0;
53 54

	ppc64_runlatch_on();
55 56 57 58 59 60 61
}

static int snooze_loop(struct cpuidle_device *dev,
			struct cpuidle_driver *drv,
			int index)
{
	unsigned long in_purr;
62
	u64 snooze_exit_time;
63

64 65
	set_thread_flag(TIF_POLLING_NRFLAG);

66
	idle_loop_prolog(&in_purr);
67
	local_irq_enable();
68
	snooze_exit_time = get_tb() + snooze_timeout;
69

70
	while (!need_resched()) {
71 72
		HMT_low();
		HMT_very_low();
73 74 75 76 77 78 79 80
		if (likely(snooze_timeout_en) && get_tb() > snooze_exit_time) {
			/*
			 * Task has not woken up but we are exiting the polling
			 * loop anyway. Require a barrier after polling is
			 * cleared to order subsequent test of need_resched().
			 */
			clear_thread_flag(TIF_POLLING_NRFLAG);
			smp_mb();
81
			break;
82
		}
83 84 85
	}

	HMT_medium();
86 87
	clear_thread_flag(TIF_POLLING_NRFLAG);

88 89
	local_irq_disable();

90 91
	idle_loop_epilog(in_purr);

92 93 94
	return index;
}

95 96 97
static void check_and_cede_processor(void)
{
	/*
98 99 100
	 * Ensure our interrupt state is properly tracked,
	 * also checks if no interrupt has occurred while we
	 * were soft-disabled
101
	 */
102
	if (prep_irq_for_idle()) {
103
		cede_processor();
104 105 106 107 108 109
#ifdef CONFIG_TRACE_IRQFLAGS
		/* Ensure that H_CEDE returns with IRQs on */
		if (WARN_ON(!(mfmsr() & MSR_EE)))
			__hard_irq_enable();
#endif
	}
110 111
}

112 113 114 115 116 117
static int dedicated_cede_loop(struct cpuidle_device *dev,
				struct cpuidle_driver *drv,
				int index)
{
	unsigned long in_purr;

118
	idle_loop_prolog(&in_purr);
119 120 121
	get_lppaca()->donate_dedicated_cpu = 1;

	HMT_medium();
122
	check_and_cede_processor();
123

124
	local_irq_disable();
125
	get_lppaca()->donate_dedicated_cpu = 0;
126 127 128

	idle_loop_epilog(in_purr);

129 130 131 132 133 134 135 136 137
	return index;
}

static int shared_cede_loop(struct cpuidle_device *dev,
			struct cpuidle_driver *drv,
			int index)
{
	unsigned long in_purr;

138
	idle_loop_prolog(&in_purr);
139 140 141 142 143 144 145 146

	/*
	 * Yield the processor to the hypervisor.  We return if
	 * an external interrupt occurs (which are driven prior
	 * to returning here) or if a prod occurs from another
	 * processor. When returning here, external interrupts
	 * are enabled.
	 */
147
	check_and_cede_processor();
148

149
	local_irq_disable();
150 151
	idle_loop_epilog(in_purr);

152 153 154 155 156 157
	return index;
}

/*
 * States for dedicated partition case.
 */
158
static struct cpuidle_state dedicated_states[] = {
159 160 161 162 163 164 165 166 167
	{ /* Snooze */
		.name = "snooze",
		.desc = "snooze",
		.exit_latency = 0,
		.target_residency = 0,
		.enter = &snooze_loop },
	{ /* CEDE */
		.name = "CEDE",
		.desc = "CEDE",
168 169
		.exit_latency = 10,
		.target_residency = 100,
170 171 172 173 174 175
		.enter = &dedicated_cede_loop },
};

/*
 * States for shared partition case.
 */
176
static struct cpuidle_state shared_states[] = {
177 178 179 180 181 182
	{ /* Snooze */
		.name = "snooze",
		.desc = "snooze",
		.exit_latency = 0,
		.target_residency = 0,
		.enter = &snooze_loop },
183 184 185
	{ /* Shared Cede */
		.name = "Shared Cede",
		.desc = "Shared Cede",
186 187
		.exit_latency = 10,
		.target_residency = 100,
188 189 190
		.enter = &shared_cede_loop },
};

191
static int pseries_cpuidle_cpu_online(unsigned int cpu)
192
{
193
	struct cpuidle_device *dev = per_cpu(cpuidle_devices, cpu);
194

195
	if (dev && cpuidle_get_driver()) {
196 197 198 199 200 201
		cpuidle_pause_and_lock();
		cpuidle_enable_device(dev);
		cpuidle_resume_and_unlock();
	}
	return 0;
}
202

203 204 205
static int pseries_cpuidle_cpu_dead(unsigned int cpu)
{
	struct cpuidle_device *dev = per_cpu(cpuidle_devices, cpu);
206

207 208 209 210
	if (dev && cpuidle_get_driver()) {
		cpuidle_pause_and_lock();
		cpuidle_disable_device(dev);
		cpuidle_resume_and_unlock();
211
	}
212
	return 0;
213 214 215 216 217 218 219 220 221 222 223 224
}

/*
 * pseries_cpuidle_driver_init()
 */
static int pseries_cpuidle_driver_init(void)
{
	int idle_state;
	struct cpuidle_driver *drv = &pseries_idle_driver;

	drv->state_count = 0;

225 226
	for (idle_state = 0; idle_state < max_idle_state; ++idle_state) {
		/* Is the state not enabled? */
227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
		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;
	}

	return 0;
}

/*
 * pseries_idle_probe()
 * Choose state table for shared versus dedicated partition
 */
static int pseries_idle_probe(void)
{

246 247 248
	if (cpuidle_disable != IDLE_NO_OVERRIDE)
		return -ENODEV;

249
	if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
250
		if (lppaca_shared_proc(get_lppaca())) {
251
			cpuidle_state_table = shared_states;
252 253
			max_idle_state = ARRAY_SIZE(shared_states);
		} else {
254
			cpuidle_state_table = dedicated_states;
255 256
			max_idle_state = ARRAY_SIZE(dedicated_states);
		}
257 258
	} else
		return -ENODEV;
259

260 261 262 263 264
	if (max_idle_state > 1) {
		snooze_timeout_en = true;
		snooze_timeout = cpuidle_state_table[1].target_residency *
				 tb_ticks_per_usec;
	}
265 266 267 268 269 270 271 272 273 274 275 276
	return 0;
}

static int __init pseries_processor_idle_init(void)
{
	int retval;

	retval = pseries_idle_probe();
	if (retval)
		return retval;

	pseries_cpuidle_driver_init();
277
	retval = cpuidle_register(&pseries_idle_driver, NULL);
278 279 280 281 282
	if (retval) {
		printk(KERN_DEBUG "Registration of pseries driver failed.\n");
		return retval;
	}

283 284 285 286 287 288 289 290
	retval = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN,
					   "cpuidle/pseries:online",
					   pseries_cpuidle_cpu_online, NULL);
	WARN_ON(retval < 0);
	retval = cpuhp_setup_state_nocalls(CPUHP_CPUIDLE_DEAD,
					   "cpuidle/pseries:DEAD", NULL,
					   pseries_cpuidle_cpu_dead);
	WARN_ON(retval < 0);
291 292 293 294
	printk(KERN_DEBUG "pseries_idle_driver registered\n");
	return 0;
}

295
device_initcall(pseries_processor_idle_init);