blob: 7f21c6a57178239e29f337432f8325938264faf0 [file] [log] [blame]
rjw1f884582022-01-06 17:20:42 +08001// SPDX-License-Identifier: GPL-2.0
2/*
3 * cpuidle-pseries - idle state cpuidle driver.
4 * Adapted from drivers/idle/intel_idle.c and
5 * drivers/acpi/processor_idle.c
6 *
7 */
8
9#include <linux/kernel.h>
10#include <linux/module.h>
11#include <linux/init.h>
12#include <linux/moduleparam.h>
13#include <linux/cpuidle.h>
14#include <linux/cpu.h>
15#include <linux/notifier.h>
16
17#include <asm/paca.h>
18#include <asm/reg.h>
19#include <asm/machdep.h>
20#include <asm/firmware.h>
21#include <asm/runlatch.h>
22#include <asm/plpar_wrappers.h>
23
24struct cpuidle_driver pseries_idle_driver = {
25 .name = "pseries_idle",
26 .owner = THIS_MODULE,
27};
28
29static int max_idle_state __read_mostly;
30static struct cpuidle_state *cpuidle_state_table __read_mostly;
31static u64 snooze_timeout __read_mostly;
32static bool snooze_timeout_en __read_mostly;
33
34static inline void idle_loop_prolog(unsigned long *in_purr)
35{
36 ppc64_runlatch_off();
37 *in_purr = mfspr(SPRN_PURR);
38 /*
39 * Indicate to the HV that we are idle. Now would be
40 * a good time to find other work to dispatch.
41 */
42 get_lppaca()->idle = 1;
43}
44
45static inline void idle_loop_epilog(unsigned long in_purr)
46{
47 u64 wait_cycles;
48
49 wait_cycles = be64_to_cpu(get_lppaca()->wait_state_cycles);
50 wait_cycles += mfspr(SPRN_PURR) - in_purr;
51 get_lppaca()->wait_state_cycles = cpu_to_be64(wait_cycles);
52 get_lppaca()->idle = 0;
53
54 if (irqs_disabled())
55 local_irq_enable();
56 ppc64_runlatch_on();
57}
58
59static int snooze_loop(struct cpuidle_device *dev,
60 struct cpuidle_driver *drv,
61 int index)
62{
63 unsigned long in_purr;
64 u64 snooze_exit_time;
65
66 set_thread_flag(TIF_POLLING_NRFLAG);
67
68 idle_loop_prolog(&in_purr);
69 local_irq_enable();
70 snooze_exit_time = get_tb() + snooze_timeout;
71
72 while (!need_resched()) {
73 HMT_low();
74 HMT_very_low();
75 if (likely(snooze_timeout_en) && get_tb() > snooze_exit_time) {
76 /*
77 * Task has not woken up but we are exiting the polling
78 * loop anyway. Require a barrier after polling is
79 * cleared to order subsequent test of need_resched().
80 */
81 clear_thread_flag(TIF_POLLING_NRFLAG);
82 smp_mb();
83 break;
84 }
85 }
86
87 HMT_medium();
88 clear_thread_flag(TIF_POLLING_NRFLAG);
89
90 idle_loop_epilog(in_purr);
91
92 return index;
93}
94
95static void check_and_cede_processor(void)
96{
97 /*
98 * Ensure our interrupt state is properly tracked,
99 * also checks if no interrupt has occurred while we
100 * were soft-disabled
101 */
102 if (prep_irq_for_idle()) {
103 cede_processor();
104#ifdef CONFIG_TRACE_IRQFLAGS
105 /* Ensure that H_CEDE returns with IRQs on */
106 if (WARN_ON(!(mfmsr() & MSR_EE)))
107 __hard_irq_enable();
108#endif
109 }
110}
111
112static int dedicated_cede_loop(struct cpuidle_device *dev,
113 struct cpuidle_driver *drv,
114 int index)
115{
116 unsigned long in_purr;
117
118 idle_loop_prolog(&in_purr);
119 get_lppaca()->donate_dedicated_cpu = 1;
120
121 HMT_medium();
122 check_and_cede_processor();
123
124 get_lppaca()->donate_dedicated_cpu = 0;
125
126 idle_loop_epilog(in_purr);
127
128 return index;
129}
130
131static int shared_cede_loop(struct cpuidle_device *dev,
132 struct cpuidle_driver *drv,
133 int index)
134{
135 unsigned long in_purr;
136
137 idle_loop_prolog(&in_purr);
138
139 /*
140 * Yield the processor to the hypervisor. We return if
141 * an external interrupt occurs (which are driven prior
142 * to returning here) or if a prod occurs from another
143 * processor. When returning here, external interrupts
144 * are enabled.
145 */
146 check_and_cede_processor();
147
148 idle_loop_epilog(in_purr);
149
150 return index;
151}
152
153/*
154 * States for dedicated partition case.
155 */
156static struct cpuidle_state dedicated_states[] = {
157 { /* Snooze */
158 .name = "snooze",
159 .desc = "snooze",
160 .exit_latency = 0,
161 .target_residency = 0,
162 .enter = &snooze_loop },
163 { /* CEDE */
164 .name = "CEDE",
165 .desc = "CEDE",
166 .exit_latency = 10,
167 .target_residency = 100,
168 .enter = &dedicated_cede_loop },
169};
170
171/*
172 * States for shared partition case.
173 */
174static struct cpuidle_state shared_states[] = {
175 { /* Shared Cede */
176 .name = "Shared Cede",
177 .desc = "Shared Cede",
178 .exit_latency = 0,
179 .target_residency = 0,
180 .enter = &shared_cede_loop },
181};
182
183static int pseries_cpuidle_cpu_online(unsigned int cpu)
184{
185 struct cpuidle_device *dev = per_cpu(cpuidle_devices, cpu);
186
187 if (dev && cpuidle_get_driver()) {
188 cpuidle_pause_and_lock();
189 cpuidle_enable_device(dev);
190 cpuidle_resume_and_unlock();
191 }
192 return 0;
193}
194
195static int pseries_cpuidle_cpu_dead(unsigned int cpu)
196{
197 struct cpuidle_device *dev = per_cpu(cpuidle_devices, cpu);
198
199 if (dev && cpuidle_get_driver()) {
200 cpuidle_pause_and_lock();
201 cpuidle_disable_device(dev);
202 cpuidle_resume_and_unlock();
203 }
204 return 0;
205}
206
207/*
208 * pseries_cpuidle_driver_init()
209 */
210static int pseries_cpuidle_driver_init(void)
211{
212 int idle_state;
213 struct cpuidle_driver *drv = &pseries_idle_driver;
214
215 drv->state_count = 0;
216
217 for (idle_state = 0; idle_state < max_idle_state; ++idle_state) {
218 /* Is the state not enabled? */
219 if (cpuidle_state_table[idle_state].enter == NULL)
220 continue;
221
222 drv->states[drv->state_count] = /* structure copy */
223 cpuidle_state_table[idle_state];
224
225 drv->state_count += 1;
226 }
227
228 return 0;
229}
230
231/*
232 * pseries_idle_probe()
233 * Choose state table for shared versus dedicated partition
234 */
235static int pseries_idle_probe(void)
236{
237
238 if (cpuidle_disable != IDLE_NO_OVERRIDE)
239 return -ENODEV;
240
241 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
242 /*
243 * Use local_paca instead of get_lppaca() since
244 * preemption is not disabled, and it is not required in
245 * fact, since lppaca_ptr does not need to be the value
246 * associated to the current CPU, it can be from any CPU.
247 */
248 if (lppaca_shared_proc(local_paca->lppaca_ptr)) {
249 cpuidle_state_table = shared_states;
250 max_idle_state = ARRAY_SIZE(shared_states);
251 } else {
252 cpuidle_state_table = dedicated_states;
253 max_idle_state = ARRAY_SIZE(dedicated_states);
254 }
255 } else
256 return -ENODEV;
257
258 if (max_idle_state > 1) {
259 snooze_timeout_en = true;
260 snooze_timeout = cpuidle_state_table[1].target_residency *
261 tb_ticks_per_usec;
262 }
263 return 0;
264}
265
266static int __init pseries_processor_idle_init(void)
267{
268 int retval;
269
270 retval = pseries_idle_probe();
271 if (retval)
272 return retval;
273
274 pseries_cpuidle_driver_init();
275 retval = cpuidle_register(&pseries_idle_driver, NULL);
276 if (retval) {
277 printk(KERN_DEBUG "Registration of pseries driver failed.\n");
278 return retval;
279 }
280
281 retval = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN,
282 "cpuidle/pseries:online",
283 pseries_cpuidle_cpu_online, NULL);
284 WARN_ON(retval < 0);
285 retval = cpuhp_setup_state_nocalls(CPUHP_CPUIDLE_DEAD,
286 "cpuidle/pseries:DEAD", NULL,
287 pseries_cpuidle_cpu_dead);
288 WARN_ON(retval < 0);
289 printk(KERN_DEBUG "pseries_idle_driver registered\n");
290 return 0;
291}
292
293device_initcall(pseries_processor_idle_init);