Doug / smarc-fsl-linux-kernel

Download as
Browse Code ยป

Commit 5f1601261050251a5ca293378b492a69d590dacb

Authored by Stuart Hayes
Committed by Len Brown
1 parent 76e10d158e
Exists in smarc-l5.0.0_1.0.0-ga and in 5 other branches imx_3.10.17_1.0.1_ga, imx_3.10.53_1.1.0_ga, imx_3.14.28_1.0.0_ga, smarc-imx_3.10.53_1.1.0_ga, smarc-imx_3.14.28_1.0.0_ga

acpi_pad: fix power_saving thread deadlock 

The acpi_pad driver can get stuck in destroy_power_saving_task()
waiting for kthread_stop() to stop a power_saving thread.  The problem
is that the isolated_cpus_lock mutex is owned when
destroy_power_saving_task() calls kthread_stop(), which waits for a
power_saving thread to end, and the power_saving thread tries to
acquire the isolated_cpus_lock when it calls round_robin_cpu().  This
patch fixes the issue by making round_robin_cpu() use its own mutex.

https://bugzilla.kernel.org/show_bug.cgi?id=42981

Cc: stable@vger.kernel.org
Signed-off-by: Stuart Hayes <Stuart_Hayes@Dell.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Showing 1 changed file with 4 additions and 3 deletions Inline Diff

drivers/acpi/acpi_pad.c
Diff comments   View file @ 5f16012
1 /* 1 /*
2 * acpi_pad.c ACPI Processor Aggregator Driver 2 * acpi_pad.c ACPI Processor Aggregator Driver
3 * 3 *
4 * Copyright (c) 2009, Intel Corporation. 4 * Copyright (c) 2009, Intel Corporation.
5 * 5 *
6 * This program is free software; you can redistribute it and/or modify it 6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms and conditions of the GNU General Public License, 7 * under the terms and conditions of the GNU General Public License,
8 * version 2, as published by the Free Software Foundation. 8 * version 2, as published by the Free Software Foundation.
9 * 9 *
10 * This program is distributed in the hope it will be useful, but WITHOUT 10 * This program is distributed in the hope it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for 12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * more details. 13 * more details.
14 * 14 *
15 * You should have received a copy of the GNU General Public License along with 15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc., 16 * this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. 17 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 * 18 *
19 */ 19 */
20 20
21 #include <linux/kernel.h> 21 #include <linux/kernel.h>
22 #include <linux/cpumask.h> 22 #include <linux/cpumask.h>
23 #include <linux/module.h> 23 #include <linux/module.h>
24 #include <linux/init.h> 24 #include <linux/init.h>
25 #include <linux/types.h> 25 #include <linux/types.h>
26 #include <linux/kthread.h> 26 #include <linux/kthread.h>
27 #include <linux/freezer.h> 27 #include <linux/freezer.h>
28 #include <linux/cpu.h> 28 #include <linux/cpu.h>
29 #include <linux/clockchips.h> 29 #include <linux/clockchips.h>
30 #include <linux/slab.h> 30 #include <linux/slab.h>
31 #include <acpi/acpi_bus.h> 31 #include <acpi/acpi_bus.h>
32 #include <acpi/acpi_drivers.h> 32 #include <acpi/acpi_drivers.h>
33 #include <asm/mwait.h> 33 #include <asm/mwait.h>
34 34
35 #define ACPI_PROCESSOR_AGGREGATOR_CLASS "acpi_pad" 35 #define ACPI_PROCESSOR_AGGREGATOR_CLASS "acpi_pad"
36 #define ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME "Processor Aggregator" 36 #define ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME "Processor Aggregator"
37 #define ACPI_PROCESSOR_AGGREGATOR_NOTIFY 0x80 37 #define ACPI_PROCESSOR_AGGREGATOR_NOTIFY 0x80
38 static DEFINE_MUTEX(isolated_cpus_lock); 38 static DEFINE_MUTEX(isolated_cpus_lock);
39 static DEFINE_MUTEX(round_robin_lock);
39 40
40 static unsigned long power_saving_mwait_eax; 41 static unsigned long power_saving_mwait_eax;
41 42
42 static unsigned char tsc_detected_unstable; 43 static unsigned char tsc_detected_unstable;
43 static unsigned char tsc_marked_unstable; 44 static unsigned char tsc_marked_unstable;
44 static unsigned char lapic_detected_unstable; 45 static unsigned char lapic_detected_unstable;
45 static unsigned char lapic_marked_unstable; 46 static unsigned char lapic_marked_unstable;
46 47
47 static void power_saving_mwait_init(void) 48 static void power_saving_mwait_init(void)
48 { 49 {
49 unsigned int eax, ebx, ecx, edx; 50 unsigned int eax, ebx, ecx, edx;
50 unsigned int highest_cstate = 0; 51 unsigned int highest_cstate = 0;
51 unsigned int highest_subcstate = 0; 52 unsigned int highest_subcstate = 0;
52 int i; 53 int i;
53 54
54 if (!boot_cpu_has(X86_FEATURE_MWAIT)) 55 if (!boot_cpu_has(X86_FEATURE_MWAIT))
55 return; 56 return;
56 if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF) 57 if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
57 return; 58 return;
58 59
59 cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx); 60 cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx);
60 61
61 if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) || 62 if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) ||
62 !(ecx & CPUID5_ECX_INTERRUPT_BREAK)) 63 !(ecx & CPUID5_ECX_INTERRUPT_BREAK))
63 return; 64 return;
64 65
65 edx >>= MWAIT_SUBSTATE_SIZE; 66 edx >>= MWAIT_SUBSTATE_SIZE;
66 for (i = 0; i < 7 && edx; i++, edx >>= MWAIT_SUBSTATE_SIZE) { 67 for (i = 0; i < 7 && edx; i++, edx >>= MWAIT_SUBSTATE_SIZE) {
67 if (edx & MWAIT_SUBSTATE_MASK) { 68 if (edx & MWAIT_SUBSTATE_MASK) {
68 highest_cstate = i; 69 highest_cstate = i;
69 highest_subcstate = edx & MWAIT_SUBSTATE_MASK; 70 highest_subcstate = edx & MWAIT_SUBSTATE_MASK;
70 } 71 }
71 } 72 }
72 power_saving_mwait_eax = (highest_cstate << MWAIT_SUBSTATE_SIZE) | 73 power_saving_mwait_eax = (highest_cstate << MWAIT_SUBSTATE_SIZE) |
73 (highest_subcstate - 1); 74 (highest_subcstate - 1);
74 75
75 #if defined(CONFIG_X86) 76 #if defined(CONFIG_X86)
76 switch (boot_cpu_data.x86_vendor) { 77 switch (boot_cpu_data.x86_vendor) {
77 case X86_VENDOR_AMD: 78 case X86_VENDOR_AMD:
78 case X86_VENDOR_INTEL: 79 case X86_VENDOR_INTEL:
79 /* 80 /*
80 * AMD Fam10h TSC will tick in all 81 * AMD Fam10h TSC will tick in all
81 * C/P/S0/S1 states when this bit is set. 82 * C/P/S0/S1 states when this bit is set.
82 */ 83 */
83 if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC)) 84 if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
84 tsc_detected_unstable = 1; 85 tsc_detected_unstable = 1;
85 if (!boot_cpu_has(X86_FEATURE_ARAT)) 86 if (!boot_cpu_has(X86_FEATURE_ARAT))
86 lapic_detected_unstable = 1; 87 lapic_detected_unstable = 1;
87 break; 88 break;
88 default: 89 default:
89 /* TSC & LAPIC could halt in idle */ 90 /* TSC & LAPIC could halt in idle */
90 tsc_detected_unstable = 1; 91 tsc_detected_unstable = 1;
91 lapic_detected_unstable = 1; 92 lapic_detected_unstable = 1;
92 } 93 }
93 #endif 94 #endif
94 } 95 }
95 96
96 static unsigned long cpu_weight[NR_CPUS]; 97 static unsigned long cpu_weight[NR_CPUS];
97 static int tsk_in_cpu[NR_CPUS] = {[0 ... NR_CPUS-1] = -1}; 98 static int tsk_in_cpu[NR_CPUS] = {[0 ... NR_CPUS-1] = -1};
98 static DECLARE_BITMAP(pad_busy_cpus_bits, NR_CPUS); 99 static DECLARE_BITMAP(pad_busy_cpus_bits, NR_CPUS);
99 static void round_robin_cpu(unsigned int tsk_index) 100 static void round_robin_cpu(unsigned int tsk_index)
100 { 101 {
101 struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits); 102 struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
102 cpumask_var_t tmp; 103 cpumask_var_t tmp;
103 int cpu; 104 int cpu;
104 unsigned long min_weight = -1; 105 unsigned long min_weight = -1;
105 unsigned long uninitialized_var(preferred_cpu); 106 unsigned long uninitialized_var(preferred_cpu);
106 107
107 if (!alloc_cpumask_var(&tmp, GFP_KERNEL)) 108 if (!alloc_cpumask_var(&tmp, GFP_KERNEL))
108 return; 109 return;
109 110
110 mutex_lock(&isolated_cpus_lock); 111 mutex_lock(&round_robin_lock);
111 cpumask_clear(tmp); 112 cpumask_clear(tmp);
112 for_each_cpu(cpu, pad_busy_cpus) 113 for_each_cpu(cpu, pad_busy_cpus)
113 cpumask_or(tmp, tmp, topology_thread_cpumask(cpu)); 114 cpumask_or(tmp, tmp, topology_thread_cpumask(cpu));
114 cpumask_andnot(tmp, cpu_online_mask, tmp); 115 cpumask_andnot(tmp, cpu_online_mask, tmp);
115 /* avoid HT sibilings if possible */ 116 /* avoid HT sibilings if possible */
116 if (cpumask_empty(tmp)) 117 if (cpumask_empty(tmp))
117 cpumask_andnot(tmp, cpu_online_mask, pad_busy_cpus); 118 cpumask_andnot(tmp, cpu_online_mask, pad_busy_cpus);
118 if (cpumask_empty(tmp)) { 119 if (cpumask_empty(tmp)) {
119 mutex_unlock(&isolated_cpus_lock); 120 mutex_unlock(&round_robin_lock);
120 return; 121 return;
121 } 122 }
122 for_each_cpu(cpu, tmp) { 123 for_each_cpu(cpu, tmp) {
123 if (cpu_weight[cpu] < min_weight) { 124 if (cpu_weight[cpu] < min_weight) {
124 min_weight = cpu_weight[cpu]; 125 min_weight = cpu_weight[cpu];
125 preferred_cpu = cpu; 126 preferred_cpu = cpu;
126 } 127 }
127 } 128 }
128 129
129 if (tsk_in_cpu[tsk_index] != -1) 130 if (tsk_in_cpu[tsk_index] != -1)
130 cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus); 131 cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
131 tsk_in_cpu[tsk_index] = preferred_cpu; 132 tsk_in_cpu[tsk_index] = preferred_cpu;
132 cpumask_set_cpu(preferred_cpu, pad_busy_cpus); 133 cpumask_set_cpu(preferred_cpu, pad_busy_cpus);
133 cpu_weight[preferred_cpu]++; 134 cpu_weight[preferred_cpu]++;
134 mutex_unlock(&isolated_cpus_lock); 135 mutex_unlock(&round_robin_lock);
135 136
136 set_cpus_allowed_ptr(current, cpumask_of(preferred_cpu)); 137 set_cpus_allowed_ptr(current, cpumask_of(preferred_cpu));
137 } 138 }
138 139
139 static void exit_round_robin(unsigned int tsk_index) 140 static void exit_round_robin(unsigned int tsk_index)
140 { 141 {
141 struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits); 142 struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
142 cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus); 143 cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
143 tsk_in_cpu[tsk_index] = -1; 144 tsk_in_cpu[tsk_index] = -1;
144 } 145 }
145 146
146 static unsigned int idle_pct = 5; /* percentage */ 147 static unsigned int idle_pct = 5; /* percentage */
147 static unsigned int round_robin_time = 10; /* second */ 148 static unsigned int round_robin_time = 10; /* second */
148 static int power_saving_thread(void *data) 149 static int power_saving_thread(void *data)
149 { 150 {
150 struct sched_param param = {.sched_priority = 1}; 151 struct sched_param param = {.sched_priority = 1};
151 int do_sleep; 152 int do_sleep;
152 unsigned int tsk_index = (unsigned long)data; 153 unsigned int tsk_index = (unsigned long)data;
153 u64 last_jiffies = 0; 154 u64 last_jiffies = 0;
154 155
155 sched_setscheduler(current, SCHED_RR, &param); 156 sched_setscheduler(current, SCHED_RR, &param);
156 157
157 while (!kthread_should_stop()) { 158 while (!kthread_should_stop()) {
158 int cpu; 159 int cpu;
159 u64 expire_time; 160 u64 expire_time;
160 161
161 try_to_freeze(); 162 try_to_freeze();
162 163
163 /* round robin to cpus */ 164 /* round robin to cpus */
164 if (last_jiffies + round_robin_time * HZ < jiffies) { 165 if (last_jiffies + round_robin_time * HZ < jiffies) {
165 last_jiffies = jiffies; 166 last_jiffies = jiffies;
166 round_robin_cpu(tsk_index); 167 round_robin_cpu(tsk_index);
167 } 168 }
168 169
169 do_sleep = 0; 170 do_sleep = 0;
170 171
171 expire_time = jiffies + HZ * (100 - idle_pct) / 100; 172 expire_time = jiffies + HZ * (100 - idle_pct) / 100;
172 173
173 while (!need_resched()) { 174 while (!need_resched()) {
174 if (tsc_detected_unstable && !tsc_marked_unstable) { 175 if (tsc_detected_unstable && !tsc_marked_unstable) {
175 /* TSC could halt in idle, so notify users */ 176 /* TSC could halt in idle, so notify users */
176 mark_tsc_unstable("TSC halts in idle"); 177 mark_tsc_unstable("TSC halts in idle");
177 tsc_marked_unstable = 1; 178 tsc_marked_unstable = 1;
178 } 179 }
179 if (lapic_detected_unstable && !lapic_marked_unstable) { 180 if (lapic_detected_unstable && !lapic_marked_unstable) {
180 int i; 181 int i;
181 /* LAPIC could halt in idle, so notify users */ 182 /* LAPIC could halt in idle, so notify users */
182 for_each_online_cpu(i) 183 for_each_online_cpu(i)
183 clockevents_notify( 184 clockevents_notify(
184 CLOCK_EVT_NOTIFY_BROADCAST_ON, 185 CLOCK_EVT_NOTIFY_BROADCAST_ON,
185 &i); 186 &i);
186 lapic_marked_unstable = 1; 187 lapic_marked_unstable = 1;
187 } 188 }
188 local_irq_disable(); 189 local_irq_disable();
189 cpu = smp_processor_id(); 190 cpu = smp_processor_id();
190 if (lapic_marked_unstable) 191 if (lapic_marked_unstable)
191 clockevents_notify( 192 clockevents_notify(
192 CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu); 193 CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu);
193 stop_critical_timings(); 194 stop_critical_timings();
194 195
195 __monitor((void *)&current_thread_info()->flags, 0, 0); 196 __monitor((void *)&current_thread_info()->flags, 0, 0);
196 smp_mb(); 197 smp_mb();
197 if (!need_resched()) 198 if (!need_resched())
198 __mwait(power_saving_mwait_eax, 1); 199 __mwait(power_saving_mwait_eax, 1);
199 200
200 start_critical_timings(); 201 start_critical_timings();
201 if (lapic_marked_unstable) 202 if (lapic_marked_unstable)
202 clockevents_notify( 203 clockevents_notify(
203 CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu); 204 CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu);
204 local_irq_enable(); 205 local_irq_enable();
205 206
206 if (jiffies > expire_time) { 207 if (jiffies > expire_time) {
207 do_sleep = 1; 208 do_sleep = 1;
208 break; 209 break;
209 } 210 }
210 } 211 }
211 212
212 /* 213 /*
213 * current sched_rt has threshold for rt task running time. 214 * current sched_rt has threshold for rt task running time.
214 * When a rt task uses 95% CPU time, the rt thread will be 215 * When a rt task uses 95% CPU time, the rt thread will be
215 * scheduled out for 5% CPU time to not starve other tasks. But 216 * scheduled out for 5% CPU time to not starve other tasks. But
216 * the mechanism only works when all CPUs have RT task running, 217 * the mechanism only works when all CPUs have RT task running,
217 * as if one CPU hasn't RT task, RT task from other CPUs will 218 * as if one CPU hasn't RT task, RT task from other CPUs will
218 * borrow CPU time from this CPU and cause RT task use > 95% 219 * borrow CPU time from this CPU and cause RT task use > 95%
219 * CPU time. To make 'avoid starvation' work, takes a nap here. 220 * CPU time. To make 'avoid starvation' work, takes a nap here.
220 */ 221 */
221 if (do_sleep) 222 if (do_sleep)
222 schedule_timeout_killable(HZ * idle_pct / 100); 223 schedule_timeout_killable(HZ * idle_pct / 100);
223 } 224 }
224 225
225 exit_round_robin(tsk_index); 226 exit_round_robin(tsk_index);
226 return 0; 227 return 0;
227 } 228 }
228 229
229 static struct task_struct *ps_tsks[NR_CPUS]; 230 static struct task_struct *ps_tsks[NR_CPUS];
230 static unsigned int ps_tsk_num; 231 static unsigned int ps_tsk_num;
231 static int create_power_saving_task(void) 232 static int create_power_saving_task(void)
232 { 233 {
233 int rc = -ENOMEM; 234 int rc = -ENOMEM;
234 235
235 ps_tsks[ps_tsk_num] = kthread_run(power_saving_thread, 236 ps_tsks[ps_tsk_num] = kthread_run(power_saving_thread,
236 (void *)(unsigned long)ps_tsk_num, 237 (void *)(unsigned long)ps_tsk_num,
237 "power_saving/%d", ps_tsk_num); 238 "power_saving/%d", ps_tsk_num);
238 rc = IS_ERR(ps_tsks[ps_tsk_num]) ? PTR_ERR(ps_tsks[ps_tsk_num]) : 0; 239 rc = IS_ERR(ps_tsks[ps_tsk_num]) ? PTR_ERR(ps_tsks[ps_tsk_num]) : 0;
239 if (!rc) 240 if (!rc)
240 ps_tsk_num++; 241 ps_tsk_num++;
241 else 242 else
242 ps_tsks[ps_tsk_num] = NULL; 243 ps_tsks[ps_tsk_num] = NULL;
243 244
244 return rc; 245 return rc;
245 } 246 }
246 247
247 static void destroy_power_saving_task(void) 248 static void destroy_power_saving_task(void)
248 { 249 {
249 if (ps_tsk_num > 0) { 250 if (ps_tsk_num > 0) {
250 ps_tsk_num--; 251 ps_tsk_num--;
251 kthread_stop(ps_tsks[ps_tsk_num]); 252 kthread_stop(ps_tsks[ps_tsk_num]);
252 ps_tsks[ps_tsk_num] = NULL; 253 ps_tsks[ps_tsk_num] = NULL;
253 } 254 }
254 } 255 }
255 256
256 static void set_power_saving_task_num(unsigned int num) 257 static void set_power_saving_task_num(unsigned int num)
257 { 258 {
258 if (num > ps_tsk_num) { 259 if (num > ps_tsk_num) {
259 while (ps_tsk_num < num) { 260 while (ps_tsk_num < num) {
260 if (create_power_saving_task()) 261 if (create_power_saving_task())
261 return; 262 return;
262 } 263 }
263 } else if (num < ps_tsk_num) { 264 } else if (num < ps_tsk_num) {
264 while (ps_tsk_num > num) 265 while (ps_tsk_num > num)
265 destroy_power_saving_task(); 266 destroy_power_saving_task();
266 } 267 }
267 } 268 }
268 269
269 static void acpi_pad_idle_cpus(unsigned int num_cpus) 270 static void acpi_pad_idle_cpus(unsigned int num_cpus)
270 { 271 {
271 get_online_cpus(); 272 get_online_cpus();
272 273
273 num_cpus = min_t(unsigned int, num_cpus, num_online_cpus()); 274 num_cpus = min_t(unsigned int, num_cpus, num_online_cpus());
274 set_power_saving_task_num(num_cpus); 275 set_power_saving_task_num(num_cpus);
275 276
276 put_online_cpus(); 277 put_online_cpus();
277 } 278 }
278 279
279 static uint32_t acpi_pad_idle_cpus_num(void) 280 static uint32_t acpi_pad_idle_cpus_num(void)
280 { 281 {
281 return ps_tsk_num; 282 return ps_tsk_num;
282 } 283 }
283 284
284 static ssize_t acpi_pad_rrtime_store(struct device *dev, 285 static ssize_t acpi_pad_rrtime_store(struct device *dev,
285 struct device_attribute *attr, const char *buf, size_t count) 286 struct device_attribute *attr, const char *buf, size_t count)
286 { 287 {
287 unsigned long num; 288 unsigned long num;
288 if (strict_strtoul(buf, 0, &num)) 289 if (strict_strtoul(buf, 0, &num))
289 return -EINVAL; 290 return -EINVAL;
290 if (num < 1 || num >= 100) 291 if (num < 1 || num >= 100)
291 return -EINVAL; 292 return -EINVAL;
292 mutex_lock(&isolated_cpus_lock); 293 mutex_lock(&isolated_cpus_lock);
293 round_robin_time = num; 294 round_robin_time = num;
294 mutex_unlock(&isolated_cpus_lock); 295 mutex_unlock(&isolated_cpus_lock);
295 return count; 296 return count;
296 } 297 }
297 298
298 static ssize_t acpi_pad_rrtime_show(struct device *dev, 299 static ssize_t acpi_pad_rrtime_show(struct device *dev,
299 struct device_attribute *attr, char *buf) 300 struct device_attribute *attr, char *buf)
300 { 301 {
301 return scnprintf(buf, PAGE_SIZE, "%d\n", round_robin_time); 302 return scnprintf(buf, PAGE_SIZE, "%d\n", round_robin_time);
302 } 303 }
303 static DEVICE_ATTR(rrtime, S_IRUGO|S_IWUSR, 304 static DEVICE_ATTR(rrtime, S_IRUGO|S_IWUSR,
304 acpi_pad_rrtime_show, 305 acpi_pad_rrtime_show,
305 acpi_pad_rrtime_store); 306 acpi_pad_rrtime_store);
306 307
307 static ssize_t acpi_pad_idlepct_store(struct device *dev, 308 static ssize_t acpi_pad_idlepct_store(struct device *dev,
308 struct device_attribute *attr, const char *buf, size_t count) 309 struct device_attribute *attr, const char *buf, size_t count)
309 { 310 {
310 unsigned long num; 311 unsigned long num;
311 if (strict_strtoul(buf, 0, &num)) 312 if (strict_strtoul(buf, 0, &num))
312 return -EINVAL; 313 return -EINVAL;
313 if (num < 1 || num >= 100) 314 if (num < 1 || num >= 100)
314 return -EINVAL; 315 return -EINVAL;
315 mutex_lock(&isolated_cpus_lock); 316 mutex_lock(&isolated_cpus_lock);
316 idle_pct = num; 317 idle_pct = num;
317 mutex_unlock(&isolated_cpus_lock); 318 mutex_unlock(&isolated_cpus_lock);
318 return count; 319 return count;
319 } 320 }
320 321
321 static ssize_t acpi_pad_idlepct_show(struct device *dev, 322 static ssize_t acpi_pad_idlepct_show(struct device *dev,
322 struct device_attribute *attr, char *buf) 323 struct device_attribute *attr, char *buf)
323 { 324 {
324 return scnprintf(buf, PAGE_SIZE, "%d\n", idle_pct); 325 return scnprintf(buf, PAGE_SIZE, "%d\n", idle_pct);
325 } 326 }
326 static DEVICE_ATTR(idlepct, S_IRUGO|S_IWUSR, 327 static DEVICE_ATTR(idlepct, S_IRUGO|S_IWUSR,
327 acpi_pad_idlepct_show, 328 acpi_pad_idlepct_show,
328 acpi_pad_idlepct_store); 329 acpi_pad_idlepct_store);
329 330
330 static ssize_t acpi_pad_idlecpus_store(struct device *dev, 331 static ssize_t acpi_pad_idlecpus_store(struct device *dev,
331 struct device_attribute *attr, const char *buf, size_t count) 332 struct device_attribute *attr, const char *buf, size_t count)
332 { 333 {
333 unsigned long num; 334 unsigned long num;
334 if (strict_strtoul(buf, 0, &num)) 335 if (strict_strtoul(buf, 0, &num))
335 return -EINVAL; 336 return -EINVAL;
336 mutex_lock(&isolated_cpus_lock); 337 mutex_lock(&isolated_cpus_lock);
337 acpi_pad_idle_cpus(num); 338 acpi_pad_idle_cpus(num);
338 mutex_unlock(&isolated_cpus_lock); 339 mutex_unlock(&isolated_cpus_lock);
339 return count; 340 return count;
340 } 341 }
341 342
342 static ssize_t acpi_pad_idlecpus_show(struct device *dev, 343 static ssize_t acpi_pad_idlecpus_show(struct device *dev,
343 struct device_attribute *attr, char *buf) 344 struct device_attribute *attr, char *buf)
344 { 345 {
345 int n = 0; 346 int n = 0;
346 n = cpumask_scnprintf(buf, PAGE_SIZE-2, to_cpumask(pad_busy_cpus_bits)); 347 n = cpumask_scnprintf(buf, PAGE_SIZE-2, to_cpumask(pad_busy_cpus_bits));
347 buf[n++] = '\n'; 348 buf[n++] = '\n';
348 buf[n] = '\0'; 349 buf[n] = '\0';
349 return n; 350 return n;
350 } 351 }
351 static DEVICE_ATTR(idlecpus, S_IRUGO|S_IWUSR, 352 static DEVICE_ATTR(idlecpus, S_IRUGO|S_IWUSR,
352 acpi_pad_idlecpus_show, 353 acpi_pad_idlecpus_show,
353 acpi_pad_idlecpus_store); 354 acpi_pad_idlecpus_store);
354 355
355 static int acpi_pad_add_sysfs(struct acpi_device *device) 356 static int acpi_pad_add_sysfs(struct acpi_device *device)
356 { 357 {
357 int result; 358 int result;
358 359
359 result = device_create_file(&device->dev, &dev_attr_idlecpus); 360 result = device_create_file(&device->dev, &dev_attr_idlecpus);
360 if (result) 361 if (result)
361 return -ENODEV; 362 return -ENODEV;
362 result = device_create_file(&device->dev, &dev_attr_idlepct); 363 result = device_create_file(&device->dev, &dev_attr_idlepct);
363 if (result) { 364 if (result) {
364 device_remove_file(&device->dev, &dev_attr_idlecpus); 365 device_remove_file(&device->dev, &dev_attr_idlecpus);
365 return -ENODEV; 366 return -ENODEV;
366 } 367 }
367 result = device_create_file(&device->dev, &dev_attr_rrtime); 368 result = device_create_file(&device->dev, &dev_attr_rrtime);
368 if (result) { 369 if (result) {
369 device_remove_file(&device->dev, &dev_attr_idlecpus); 370 device_remove_file(&device->dev, &dev_attr_idlecpus);
370 device_remove_file(&device->dev, &dev_attr_idlepct); 371 device_remove_file(&device->dev, &dev_attr_idlepct);
371 return -ENODEV; 372 return -ENODEV;
372 } 373 }
373 return 0; 374 return 0;
374 } 375 }
375 376
376 static void acpi_pad_remove_sysfs(struct acpi_device *device) 377 static void acpi_pad_remove_sysfs(struct acpi_device *device)
377 { 378 {
378 device_remove_file(&device->dev, &dev_attr_idlecpus); 379 device_remove_file(&device->dev, &dev_attr_idlecpus);
379 device_remove_file(&device->dev, &dev_attr_idlepct); 380 device_remove_file(&device->dev, &dev_attr_idlepct);
380 device_remove_file(&device->dev, &dev_attr_rrtime); 381 device_remove_file(&device->dev, &dev_attr_rrtime);
381 } 382 }
382 383
383 /* 384 /*
384 * Query firmware how many CPUs should be idle 385 * Query firmware how many CPUs should be idle
385 * return -1 on failure 386 * return -1 on failure
386 */ 387 */
387 static int acpi_pad_pur(acpi_handle handle) 388 static int acpi_pad_pur(acpi_handle handle)
388 { 389 {
389 struct acpi_buffer buffer = {ACPI_ALLOCATE_BUFFER, NULL}; 390 struct acpi_buffer buffer = {ACPI_ALLOCATE_BUFFER, NULL};
390 union acpi_object *package; 391 union acpi_object *package;
391 int num = -1; 392 int num = -1;
392 393
393 if (ACPI_FAILURE(acpi_evaluate_object(handle, "_PUR", NULL, &buffer))) 394 if (ACPI_FAILURE(acpi_evaluate_object(handle, "_PUR", NULL, &buffer)))
394 return num; 395 return num;
395 396
396 if (!buffer.length || !buffer.pointer) 397 if (!buffer.length || !buffer.pointer)
397 return num; 398 return num;
398 399
399 package = buffer.pointer; 400 package = buffer.pointer;
400 401
401 if (package->type == ACPI_TYPE_PACKAGE && 402 if (package->type == ACPI_TYPE_PACKAGE &&
402 package->package.count == 2 && 403 package->package.count == 2 &&
403 package->package.elements[0].integer.value == 1) /* rev 1 */ 404 package->package.elements[0].integer.value == 1) /* rev 1 */
404 405
405 num = package->package.elements[1].integer.value; 406 num = package->package.elements[1].integer.value;
406 407
407 kfree(buffer.pointer); 408 kfree(buffer.pointer);
408 return num; 409 return num;
409 } 410 }
410 411
411 /* Notify firmware how many CPUs are idle */ 412 /* Notify firmware how many CPUs are idle */
412 static void acpi_pad_ost(acpi_handle handle, int stat, 413 static void acpi_pad_ost(acpi_handle handle, int stat,
413 uint32_t idle_cpus) 414 uint32_t idle_cpus)
414 { 415 {
415 union acpi_object params[3] = { 416 union acpi_object params[3] = {
416 {.type = ACPI_TYPE_INTEGER,}, 417 {.type = ACPI_TYPE_INTEGER,},
417 {.type = ACPI_TYPE_INTEGER,}, 418 {.type = ACPI_TYPE_INTEGER,},
418 {.type = ACPI_TYPE_BUFFER,}, 419 {.type = ACPI_TYPE_BUFFER,},
419 }; 420 };
420 struct acpi_object_list arg_list = {3, params}; 421 struct acpi_object_list arg_list = {3, params};
421 422
422 params[0].integer.value = ACPI_PROCESSOR_AGGREGATOR_NOTIFY; 423 params[0].integer.value = ACPI_PROCESSOR_AGGREGATOR_NOTIFY;
423 params[1].integer.value = stat; 424 params[1].integer.value = stat;
424 params[2].buffer.length = 4; 425 params[2].buffer.length = 4;
425 params[2].buffer.pointer = (void *)&idle_cpus; 426 params[2].buffer.pointer = (void *)&idle_cpus;
426 acpi_evaluate_object(handle, "_OST", &arg_list, NULL); 427 acpi_evaluate_object(handle, "_OST", &arg_list, NULL);
427 } 428 }
428 429
429 static void acpi_pad_handle_notify(acpi_handle handle) 430 static void acpi_pad_handle_notify(acpi_handle handle)
430 { 431 {
431 int num_cpus; 432 int num_cpus;
432 uint32_t idle_cpus; 433 uint32_t idle_cpus;
433 434
434 mutex_lock(&isolated_cpus_lock); 435 mutex_lock(&isolated_cpus_lock);
435 num_cpus = acpi_pad_pur(handle); 436 num_cpus = acpi_pad_pur(handle);
436 if (num_cpus < 0) { 437 if (num_cpus < 0) {
437 mutex_unlock(&isolated_cpus_lock); 438 mutex_unlock(&isolated_cpus_lock);
438 return; 439 return;
439 } 440 }
440 acpi_pad_idle_cpus(num_cpus); 441 acpi_pad_idle_cpus(num_cpus);
441 idle_cpus = acpi_pad_idle_cpus_num(); 442 idle_cpus = acpi_pad_idle_cpus_num();
442 acpi_pad_ost(handle, 0, idle_cpus); 443 acpi_pad_ost(handle, 0, idle_cpus);
443 mutex_unlock(&isolated_cpus_lock); 444 mutex_unlock(&isolated_cpus_lock);
444 } 445 }
445 446
446 static void acpi_pad_notify(acpi_handle handle, u32 event, 447 static void acpi_pad_notify(acpi_handle handle, u32 event,
447 void *data) 448 void *data)
448 { 449 {
449 struct acpi_device *device = data; 450 struct acpi_device *device = data;
450 451
451 switch (event) { 452 switch (event) {
452 case ACPI_PROCESSOR_AGGREGATOR_NOTIFY: 453 case ACPI_PROCESSOR_AGGREGATOR_NOTIFY:
453 acpi_pad_handle_notify(handle); 454 acpi_pad_handle_notify(handle);
454 acpi_bus_generate_proc_event(device, event, 0); 455 acpi_bus_generate_proc_event(device, event, 0);
455 acpi_bus_generate_netlink_event(device->pnp.device_class, 456 acpi_bus_generate_netlink_event(device->pnp.device_class,
456 dev_name(&device->dev), event, 0); 457 dev_name(&device->dev), event, 0);
457 break; 458 break;
458 default: 459 default:
459 printk(KERN_WARNING "Unsupported event [0x%x]\n", event); 460 printk(KERN_WARNING "Unsupported event [0x%x]\n", event);
460 break; 461 break;
461 } 462 }
462 } 463 }
463 464
464 static int acpi_pad_add(struct acpi_device *device) 465 static int acpi_pad_add(struct acpi_device *device)
465 { 466 {
466 acpi_status status; 467 acpi_status status;
467 468
468 strcpy(acpi_device_name(device), ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME); 469 strcpy(acpi_device_name(device), ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME);
469 strcpy(acpi_device_class(device), ACPI_PROCESSOR_AGGREGATOR_CLASS); 470 strcpy(acpi_device_class(device), ACPI_PROCESSOR_AGGREGATOR_CLASS);
470 471
471 if (acpi_pad_add_sysfs(device)) 472 if (acpi_pad_add_sysfs(device))
472 return -ENODEV; 473 return -ENODEV;
473 474
474 status = acpi_install_notify_handler(device->handle, 475 status = acpi_install_notify_handler(device->handle,
475 ACPI_DEVICE_NOTIFY, acpi_pad_notify, device); 476 ACPI_DEVICE_NOTIFY, acpi_pad_notify, device);
476 if (ACPI_FAILURE(status)) { 477 if (ACPI_FAILURE(status)) {
477 acpi_pad_remove_sysfs(device); 478 acpi_pad_remove_sysfs(device);
478 return -ENODEV; 479 return -ENODEV;
479 } 480 }
480 481
481 return 0; 482 return 0;
482 } 483 }
483 484
484 static int acpi_pad_remove(struct acpi_device *device, 485 static int acpi_pad_remove(struct acpi_device *device,
485 int type) 486 int type)
486 { 487 {
487 mutex_lock(&isolated_cpus_lock); 488 mutex_lock(&isolated_cpus_lock);
488 acpi_pad_idle_cpus(0); 489 acpi_pad_idle_cpus(0);
489 mutex_unlock(&isolated_cpus_lock); 490 mutex_unlock(&isolated_cpus_lock);
490 491
491 acpi_remove_notify_handler(device->handle, 492 acpi_remove_notify_handler(device->handle,
492 ACPI_DEVICE_NOTIFY, acpi_pad_notify); 493 ACPI_DEVICE_NOTIFY, acpi_pad_notify);
493 acpi_pad_remove_sysfs(device); 494 acpi_pad_remove_sysfs(device);
494 return 0; 495 return 0;
495 } 496 }
496 497
497 static const struct acpi_device_id pad_device_ids[] = { 498 static const struct acpi_device_id pad_device_ids[] = {
498 {"ACPI000C", 0}, 499 {"ACPI000C", 0},
499 {"", 0}, 500 {"", 0},
500 }; 501 };
501 MODULE_DEVICE_TABLE(acpi, pad_device_ids); 502 MODULE_DEVICE_TABLE(acpi, pad_device_ids);
502 503
503 static struct acpi_driver acpi_pad_driver = { 504 static struct acpi_driver acpi_pad_driver = {
504 .name = "processor_aggregator", 505 .name = "processor_aggregator",
505 .class = ACPI_PROCESSOR_AGGREGATOR_CLASS, 506 .class = ACPI_PROCESSOR_AGGREGATOR_CLASS,
506 .ids = pad_device_ids, 507 .ids = pad_device_ids,
507 .ops = { 508 .ops = {
508 .add = acpi_pad_add, 509 .add = acpi_pad_add,
509 .remove = acpi_pad_remove, 510 .remove = acpi_pad_remove,
510 }, 511 },
511 }; 512 };
512 513
513 static int __init acpi_pad_init(void) 514 static int __init acpi_pad_init(void)
514 { 515 {
515 power_saving_mwait_init(); 516 power_saving_mwait_init();
516 if (power_saving_mwait_eax == 0) 517 if (power_saving_mwait_eax == 0)
517 return -EINVAL; 518 return -EINVAL;
518 519
519 return acpi_bus_register_driver(&acpi_pad_driver); 520 return acpi_bus_register_driver(&acpi_pad_driver);
520 } 521 }
521 522
522 static void __exit acpi_pad_exit(void) 523 static void __exit acpi_pad_exit(void)
523 { 524 {
524 acpi_bus_unregister_driver(&acpi_pad_driver); 525 acpi_bus_unregister_driver(&acpi_pad_driver);
525 } 526 }
526 527
527 module_init(acpi_pad_init); 528 module_init(acpi_pad_init);
528 module_exit(acpi_pad_exit); 529 module_exit(acpi_pad_exit);
529 MODULE_AUTHOR("Shaohua Li<shaohua.li@intel.com>"); 530 MODULE_AUTHOR("Shaohua Li<shaohua.li@intel.com>");
530 MODULE_DESCRIPTION("ACPI Processor Aggregator Driver"); 531 MODULE_DESCRIPTION("ACPI Processor Aggregator Driver");
531 MODULE_LICENSE("GPL"); 532 MODULE_LICENSE("GPL");
532 533