Commit 633fe795b80693a8198e7d82f66538a72d2bbba2

Authored by Randy Dunlap
Committed by Ingo Molnar
1 parent 7c526e1fef

timers: add missing kernel-doc

Add missing kernel-doc parameter notation and change function
name to its new name:

  Warning(kernel/timer.c:543): No description found for parameter 'name'
  Warning(kernel/timer.c:543): No description found for parameter 'key'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: akpm <akpm@linux-foundation.org>
Cc: Johannes Berg <johannes@sipsolutions.net>
LKML-Reference: <20090401174723.f0bea0eb.randy.dunlap@oracle.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Showing 1 changed file with 5 additions and 2 deletions Inline Diff

1 /* 1 /*
2 * linux/kernel/timer.c 2 * linux/kernel/timer.c
3 * 3 *
4 * Kernel internal timers, basic process system calls 4 * Kernel internal timers, basic process system calls
5 * 5 *
6 * Copyright (C) 1991, 1992 Linus Torvalds 6 * Copyright (C) 1991, 1992 Linus Torvalds
7 * 7 *
8 * 1997-01-28 Modified by Finn Arne Gangstad to make timers scale better. 8 * 1997-01-28 Modified by Finn Arne Gangstad to make timers scale better.
9 * 9 *
10 * 1997-09-10 Updated NTP code according to technical memorandum Jan '96 10 * 1997-09-10 Updated NTP code according to technical memorandum Jan '96
11 * "A Kernel Model for Precision Timekeeping" by Dave Mills 11 * "A Kernel Model for Precision Timekeeping" by Dave Mills
12 * 1998-12-24 Fixed a xtime SMP race (we need the xtime_lock rw spinlock to 12 * 1998-12-24 Fixed a xtime SMP race (we need the xtime_lock rw spinlock to
13 * serialize accesses to xtime/lost_ticks). 13 * serialize accesses to xtime/lost_ticks).
14 * Copyright (C) 1998 Andrea Arcangeli 14 * Copyright (C) 1998 Andrea Arcangeli
15 * 1999-03-10 Improved NTP compatibility by Ulrich Windl 15 * 1999-03-10 Improved NTP compatibility by Ulrich Windl
16 * 2002-05-31 Move sys_sysinfo here and make its locking sane, Robert Love 16 * 2002-05-31 Move sys_sysinfo here and make its locking sane, Robert Love
17 * 2000-10-05 Implemented scalable SMP per-CPU timer handling. 17 * 2000-10-05 Implemented scalable SMP per-CPU timer handling.
18 * Copyright (C) 2000, 2001, 2002 Ingo Molnar 18 * Copyright (C) 2000, 2001, 2002 Ingo Molnar
19 * Designed by David S. Miller, Alexey Kuznetsov and Ingo Molnar 19 * Designed by David S. Miller, Alexey Kuznetsov and Ingo Molnar
20 */ 20 */
21 21
22 #include <linux/kernel_stat.h> 22 #include <linux/kernel_stat.h>
23 #include <linux/module.h> 23 #include <linux/module.h>
24 #include <linux/interrupt.h> 24 #include <linux/interrupt.h>
25 #include <linux/percpu.h> 25 #include <linux/percpu.h>
26 #include <linux/init.h> 26 #include <linux/init.h>
27 #include <linux/mm.h> 27 #include <linux/mm.h>
28 #include <linux/swap.h> 28 #include <linux/swap.h>
29 #include <linux/pid_namespace.h> 29 #include <linux/pid_namespace.h>
30 #include <linux/notifier.h> 30 #include <linux/notifier.h>
31 #include <linux/thread_info.h> 31 #include <linux/thread_info.h>
32 #include <linux/time.h> 32 #include <linux/time.h>
33 #include <linux/jiffies.h> 33 #include <linux/jiffies.h>
34 #include <linux/posix-timers.h> 34 #include <linux/posix-timers.h>
35 #include <linux/cpu.h> 35 #include <linux/cpu.h>
36 #include <linux/syscalls.h> 36 #include <linux/syscalls.h>
37 #include <linux/delay.h> 37 #include <linux/delay.h>
38 #include <linux/tick.h> 38 #include <linux/tick.h>
39 #include <linux/kallsyms.h> 39 #include <linux/kallsyms.h>
40 40
41 #include <asm/uaccess.h> 41 #include <asm/uaccess.h>
42 #include <asm/unistd.h> 42 #include <asm/unistd.h>
43 #include <asm/div64.h> 43 #include <asm/div64.h>
44 #include <asm/timex.h> 44 #include <asm/timex.h>
45 #include <asm/io.h> 45 #include <asm/io.h>
46 46
47 u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES; 47 u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES;
48 48
49 EXPORT_SYMBOL(jiffies_64); 49 EXPORT_SYMBOL(jiffies_64);
50 50
51 /* 51 /*
52 * per-CPU timer vector definitions: 52 * per-CPU timer vector definitions:
53 */ 53 */
54 #define TVN_BITS (CONFIG_BASE_SMALL ? 4 : 6) 54 #define TVN_BITS (CONFIG_BASE_SMALL ? 4 : 6)
55 #define TVR_BITS (CONFIG_BASE_SMALL ? 6 : 8) 55 #define TVR_BITS (CONFIG_BASE_SMALL ? 6 : 8)
56 #define TVN_SIZE (1 << TVN_BITS) 56 #define TVN_SIZE (1 << TVN_BITS)
57 #define TVR_SIZE (1 << TVR_BITS) 57 #define TVR_SIZE (1 << TVR_BITS)
58 #define TVN_MASK (TVN_SIZE - 1) 58 #define TVN_MASK (TVN_SIZE - 1)
59 #define TVR_MASK (TVR_SIZE - 1) 59 #define TVR_MASK (TVR_SIZE - 1)
60 60
61 struct tvec { 61 struct tvec {
62 struct list_head vec[TVN_SIZE]; 62 struct list_head vec[TVN_SIZE];
63 }; 63 };
64 64
65 struct tvec_root { 65 struct tvec_root {
66 struct list_head vec[TVR_SIZE]; 66 struct list_head vec[TVR_SIZE];
67 }; 67 };
68 68
69 struct tvec_base { 69 struct tvec_base {
70 spinlock_t lock; 70 spinlock_t lock;
71 struct timer_list *running_timer; 71 struct timer_list *running_timer;
72 unsigned long timer_jiffies; 72 unsigned long timer_jiffies;
73 struct tvec_root tv1; 73 struct tvec_root tv1;
74 struct tvec tv2; 74 struct tvec tv2;
75 struct tvec tv3; 75 struct tvec tv3;
76 struct tvec tv4; 76 struct tvec tv4;
77 struct tvec tv5; 77 struct tvec tv5;
78 } ____cacheline_aligned; 78 } ____cacheline_aligned;
79 79
80 struct tvec_base boot_tvec_bases; 80 struct tvec_base boot_tvec_bases;
81 EXPORT_SYMBOL(boot_tvec_bases); 81 EXPORT_SYMBOL(boot_tvec_bases);
82 static DEFINE_PER_CPU(struct tvec_base *, tvec_bases) = &boot_tvec_bases; 82 static DEFINE_PER_CPU(struct tvec_base *, tvec_bases) = &boot_tvec_bases;
83 83
84 /* 84 /*
85 * Note that all tvec_bases are 2 byte aligned and lower bit of 85 * Note that all tvec_bases are 2 byte aligned and lower bit of
86 * base in timer_list is guaranteed to be zero. Use the LSB for 86 * base in timer_list is guaranteed to be zero. Use the LSB for
87 * the new flag to indicate whether the timer is deferrable 87 * the new flag to indicate whether the timer is deferrable
88 */ 88 */
89 #define TBASE_DEFERRABLE_FLAG (0x1) 89 #define TBASE_DEFERRABLE_FLAG (0x1)
90 90
91 /* Functions below help us manage 'deferrable' flag */ 91 /* Functions below help us manage 'deferrable' flag */
92 static inline unsigned int tbase_get_deferrable(struct tvec_base *base) 92 static inline unsigned int tbase_get_deferrable(struct tvec_base *base)
93 { 93 {
94 return ((unsigned int)(unsigned long)base & TBASE_DEFERRABLE_FLAG); 94 return ((unsigned int)(unsigned long)base & TBASE_DEFERRABLE_FLAG);
95 } 95 }
96 96
97 static inline struct tvec_base *tbase_get_base(struct tvec_base *base) 97 static inline struct tvec_base *tbase_get_base(struct tvec_base *base)
98 { 98 {
99 return ((struct tvec_base *)((unsigned long)base & ~TBASE_DEFERRABLE_FLAG)); 99 return ((struct tvec_base *)((unsigned long)base & ~TBASE_DEFERRABLE_FLAG));
100 } 100 }
101 101
102 static inline void timer_set_deferrable(struct timer_list *timer) 102 static inline void timer_set_deferrable(struct timer_list *timer)
103 { 103 {
104 timer->base = ((struct tvec_base *)((unsigned long)(timer->base) | 104 timer->base = ((struct tvec_base *)((unsigned long)(timer->base) |
105 TBASE_DEFERRABLE_FLAG)); 105 TBASE_DEFERRABLE_FLAG));
106 } 106 }
107 107
108 static inline void 108 static inline void
109 timer_set_base(struct timer_list *timer, struct tvec_base *new_base) 109 timer_set_base(struct timer_list *timer, struct tvec_base *new_base)
110 { 110 {
111 timer->base = (struct tvec_base *)((unsigned long)(new_base) | 111 timer->base = (struct tvec_base *)((unsigned long)(new_base) |
112 tbase_get_deferrable(timer->base)); 112 tbase_get_deferrable(timer->base));
113 } 113 }
114 114
115 static unsigned long round_jiffies_common(unsigned long j, int cpu, 115 static unsigned long round_jiffies_common(unsigned long j, int cpu,
116 bool force_up) 116 bool force_up)
117 { 117 {
118 int rem; 118 int rem;
119 unsigned long original = j; 119 unsigned long original = j;
120 120
121 /* 121 /*
122 * We don't want all cpus firing their timers at once hitting the 122 * We don't want all cpus firing their timers at once hitting the
123 * same lock or cachelines, so we skew each extra cpu with an extra 123 * same lock or cachelines, so we skew each extra cpu with an extra
124 * 3 jiffies. This 3 jiffies came originally from the mm/ code which 124 * 3 jiffies. This 3 jiffies came originally from the mm/ code which
125 * already did this. 125 * already did this.
126 * The skew is done by adding 3*cpunr, then round, then subtract this 126 * The skew is done by adding 3*cpunr, then round, then subtract this
127 * extra offset again. 127 * extra offset again.
128 */ 128 */
129 j += cpu * 3; 129 j += cpu * 3;
130 130
131 rem = j % HZ; 131 rem = j % HZ;
132 132
133 /* 133 /*
134 * If the target jiffie is just after a whole second (which can happen 134 * If the target jiffie is just after a whole second (which can happen
135 * due to delays of the timer irq, long irq off times etc etc) then 135 * due to delays of the timer irq, long irq off times etc etc) then
136 * we should round down to the whole second, not up. Use 1/4th second 136 * we should round down to the whole second, not up. Use 1/4th second
137 * as cutoff for this rounding as an extreme upper bound for this. 137 * as cutoff for this rounding as an extreme upper bound for this.
138 * But never round down if @force_up is set. 138 * But never round down if @force_up is set.
139 */ 139 */
140 if (rem < HZ/4 && !force_up) /* round down */ 140 if (rem < HZ/4 && !force_up) /* round down */
141 j = j - rem; 141 j = j - rem;
142 else /* round up */ 142 else /* round up */
143 j = j - rem + HZ; 143 j = j - rem + HZ;
144 144
145 /* now that we have rounded, subtract the extra skew again */ 145 /* now that we have rounded, subtract the extra skew again */
146 j -= cpu * 3; 146 j -= cpu * 3;
147 147
148 if (j <= jiffies) /* rounding ate our timeout entirely; */ 148 if (j <= jiffies) /* rounding ate our timeout entirely; */
149 return original; 149 return original;
150 return j; 150 return j;
151 } 151 }
152 152
153 /** 153 /**
154 * __round_jiffies - function to round jiffies to a full second 154 * __round_jiffies - function to round jiffies to a full second
155 * @j: the time in (absolute) jiffies that should be rounded 155 * @j: the time in (absolute) jiffies that should be rounded
156 * @cpu: the processor number on which the timeout will happen 156 * @cpu: the processor number on which the timeout will happen
157 * 157 *
158 * __round_jiffies() rounds an absolute time in the future (in jiffies) 158 * __round_jiffies() rounds an absolute time in the future (in jiffies)
159 * up or down to (approximately) full seconds. This is useful for timers 159 * up or down to (approximately) full seconds. This is useful for timers
160 * for which the exact time they fire does not matter too much, as long as 160 * for which the exact time they fire does not matter too much, as long as
161 * they fire approximately every X seconds. 161 * they fire approximately every X seconds.
162 * 162 *
163 * By rounding these timers to whole seconds, all such timers will fire 163 * By rounding these timers to whole seconds, all such timers will fire
164 * at the same time, rather than at various times spread out. The goal 164 * at the same time, rather than at various times spread out. The goal
165 * of this is to have the CPU wake up less, which saves power. 165 * of this is to have the CPU wake up less, which saves power.
166 * 166 *
167 * The exact rounding is skewed for each processor to avoid all 167 * The exact rounding is skewed for each processor to avoid all
168 * processors firing at the exact same time, which could lead 168 * processors firing at the exact same time, which could lead
169 * to lock contention or spurious cache line bouncing. 169 * to lock contention or spurious cache line bouncing.
170 * 170 *
171 * The return value is the rounded version of the @j parameter. 171 * The return value is the rounded version of the @j parameter.
172 */ 172 */
173 unsigned long __round_jiffies(unsigned long j, int cpu) 173 unsigned long __round_jiffies(unsigned long j, int cpu)
174 { 174 {
175 return round_jiffies_common(j, cpu, false); 175 return round_jiffies_common(j, cpu, false);
176 } 176 }
177 EXPORT_SYMBOL_GPL(__round_jiffies); 177 EXPORT_SYMBOL_GPL(__round_jiffies);
178 178
179 /** 179 /**
180 * __round_jiffies_relative - function to round jiffies to a full second 180 * __round_jiffies_relative - function to round jiffies to a full second
181 * @j: the time in (relative) jiffies that should be rounded 181 * @j: the time in (relative) jiffies that should be rounded
182 * @cpu: the processor number on which the timeout will happen 182 * @cpu: the processor number on which the timeout will happen
183 * 183 *
184 * __round_jiffies_relative() rounds a time delta in the future (in jiffies) 184 * __round_jiffies_relative() rounds a time delta in the future (in jiffies)
185 * up or down to (approximately) full seconds. This is useful for timers 185 * up or down to (approximately) full seconds. This is useful for timers
186 * for which the exact time they fire does not matter too much, as long as 186 * for which the exact time they fire does not matter too much, as long as
187 * they fire approximately every X seconds. 187 * they fire approximately every X seconds.
188 * 188 *
189 * By rounding these timers to whole seconds, all such timers will fire 189 * By rounding these timers to whole seconds, all such timers will fire
190 * at the same time, rather than at various times spread out. The goal 190 * at the same time, rather than at various times spread out. The goal
191 * of this is to have the CPU wake up less, which saves power. 191 * of this is to have the CPU wake up less, which saves power.
192 * 192 *
193 * The exact rounding is skewed for each processor to avoid all 193 * The exact rounding is skewed for each processor to avoid all
194 * processors firing at the exact same time, which could lead 194 * processors firing at the exact same time, which could lead
195 * to lock contention or spurious cache line bouncing. 195 * to lock contention or spurious cache line bouncing.
196 * 196 *
197 * The return value is the rounded version of the @j parameter. 197 * The return value is the rounded version of the @j parameter.
198 */ 198 */
199 unsigned long __round_jiffies_relative(unsigned long j, int cpu) 199 unsigned long __round_jiffies_relative(unsigned long j, int cpu)
200 { 200 {
201 unsigned long j0 = jiffies; 201 unsigned long j0 = jiffies;
202 202
203 /* Use j0 because jiffies might change while we run */ 203 /* Use j0 because jiffies might change while we run */
204 return round_jiffies_common(j + j0, cpu, false) - j0; 204 return round_jiffies_common(j + j0, cpu, false) - j0;
205 } 205 }
206 EXPORT_SYMBOL_GPL(__round_jiffies_relative); 206 EXPORT_SYMBOL_GPL(__round_jiffies_relative);
207 207
208 /** 208 /**
209 * round_jiffies - function to round jiffies to a full second 209 * round_jiffies - function to round jiffies to a full second
210 * @j: the time in (absolute) jiffies that should be rounded 210 * @j: the time in (absolute) jiffies that should be rounded
211 * 211 *
212 * round_jiffies() rounds an absolute time in the future (in jiffies) 212 * round_jiffies() rounds an absolute time in the future (in jiffies)
213 * up or down to (approximately) full seconds. This is useful for timers 213 * up or down to (approximately) full seconds. This is useful for timers
214 * for which the exact time they fire does not matter too much, as long as 214 * for which the exact time they fire does not matter too much, as long as
215 * they fire approximately every X seconds. 215 * they fire approximately every X seconds.
216 * 216 *
217 * By rounding these timers to whole seconds, all such timers will fire 217 * By rounding these timers to whole seconds, all such timers will fire
218 * at the same time, rather than at various times spread out. The goal 218 * at the same time, rather than at various times spread out. The goal
219 * of this is to have the CPU wake up less, which saves power. 219 * of this is to have the CPU wake up less, which saves power.
220 * 220 *
221 * The return value is the rounded version of the @j parameter. 221 * The return value is the rounded version of the @j parameter.
222 */ 222 */
223 unsigned long round_jiffies(unsigned long j) 223 unsigned long round_jiffies(unsigned long j)
224 { 224 {
225 return round_jiffies_common(j, raw_smp_processor_id(), false); 225 return round_jiffies_common(j, raw_smp_processor_id(), false);
226 } 226 }
227 EXPORT_SYMBOL_GPL(round_jiffies); 227 EXPORT_SYMBOL_GPL(round_jiffies);
228 228
229 /** 229 /**
230 * round_jiffies_relative - function to round jiffies to a full second 230 * round_jiffies_relative - function to round jiffies to a full second
231 * @j: the time in (relative) jiffies that should be rounded 231 * @j: the time in (relative) jiffies that should be rounded
232 * 232 *
233 * round_jiffies_relative() rounds a time delta in the future (in jiffies) 233 * round_jiffies_relative() rounds a time delta in the future (in jiffies)
234 * up or down to (approximately) full seconds. This is useful for timers 234 * up or down to (approximately) full seconds. This is useful for timers
235 * for which the exact time they fire does not matter too much, as long as 235 * for which the exact time they fire does not matter too much, as long as
236 * they fire approximately every X seconds. 236 * they fire approximately every X seconds.
237 * 237 *
238 * By rounding these timers to whole seconds, all such timers will fire 238 * By rounding these timers to whole seconds, all such timers will fire
239 * at the same time, rather than at various times spread out. The goal 239 * at the same time, rather than at various times spread out. The goal
240 * of this is to have the CPU wake up less, which saves power. 240 * of this is to have the CPU wake up less, which saves power.
241 * 241 *
242 * The return value is the rounded version of the @j parameter. 242 * The return value is the rounded version of the @j parameter.
243 */ 243 */
244 unsigned long round_jiffies_relative(unsigned long j) 244 unsigned long round_jiffies_relative(unsigned long j)
245 { 245 {
246 return __round_jiffies_relative(j, raw_smp_processor_id()); 246 return __round_jiffies_relative(j, raw_smp_processor_id());
247 } 247 }
248 EXPORT_SYMBOL_GPL(round_jiffies_relative); 248 EXPORT_SYMBOL_GPL(round_jiffies_relative);
249 249
250 /** 250 /**
251 * __round_jiffies_up - function to round jiffies up to a full second 251 * __round_jiffies_up - function to round jiffies up to a full second
252 * @j: the time in (absolute) jiffies that should be rounded 252 * @j: the time in (absolute) jiffies that should be rounded
253 * @cpu: the processor number on which the timeout will happen 253 * @cpu: the processor number on which the timeout will happen
254 * 254 *
255 * This is the same as __round_jiffies() except that it will never 255 * This is the same as __round_jiffies() except that it will never
256 * round down. This is useful for timeouts for which the exact time 256 * round down. This is useful for timeouts for which the exact time
257 * of firing does not matter too much, as long as they don't fire too 257 * of firing does not matter too much, as long as they don't fire too
258 * early. 258 * early.
259 */ 259 */
260 unsigned long __round_jiffies_up(unsigned long j, int cpu) 260 unsigned long __round_jiffies_up(unsigned long j, int cpu)
261 { 261 {
262 return round_jiffies_common(j, cpu, true); 262 return round_jiffies_common(j, cpu, true);
263 } 263 }
264 EXPORT_SYMBOL_GPL(__round_jiffies_up); 264 EXPORT_SYMBOL_GPL(__round_jiffies_up);
265 265
266 /** 266 /**
267 * __round_jiffies_up_relative - function to round jiffies up to a full second 267 * __round_jiffies_up_relative - function to round jiffies up to a full second
268 * @j: the time in (relative) jiffies that should be rounded 268 * @j: the time in (relative) jiffies that should be rounded
269 * @cpu: the processor number on which the timeout will happen 269 * @cpu: the processor number on which the timeout will happen
270 * 270 *
271 * This is the same as __round_jiffies_relative() except that it will never 271 * This is the same as __round_jiffies_relative() except that it will never
272 * round down. This is useful for timeouts for which the exact time 272 * round down. This is useful for timeouts for which the exact time
273 * of firing does not matter too much, as long as they don't fire too 273 * of firing does not matter too much, as long as they don't fire too
274 * early. 274 * early.
275 */ 275 */
276 unsigned long __round_jiffies_up_relative(unsigned long j, int cpu) 276 unsigned long __round_jiffies_up_relative(unsigned long j, int cpu)
277 { 277 {
278 unsigned long j0 = jiffies; 278 unsigned long j0 = jiffies;
279 279
280 /* Use j0 because jiffies might change while we run */ 280 /* Use j0 because jiffies might change while we run */
281 return round_jiffies_common(j + j0, cpu, true) - j0; 281 return round_jiffies_common(j + j0, cpu, true) - j0;
282 } 282 }
283 EXPORT_SYMBOL_GPL(__round_jiffies_up_relative); 283 EXPORT_SYMBOL_GPL(__round_jiffies_up_relative);
284 284
285 /** 285 /**
286 * round_jiffies_up - function to round jiffies up to a full second 286 * round_jiffies_up - function to round jiffies up to a full second
287 * @j: the time in (absolute) jiffies that should be rounded 287 * @j: the time in (absolute) jiffies that should be rounded
288 * 288 *
289 * This is the same as round_jiffies() except that it will never 289 * This is the same as round_jiffies() except that it will never
290 * round down. This is useful for timeouts for which the exact time 290 * round down. This is useful for timeouts for which the exact time
291 * of firing does not matter too much, as long as they don't fire too 291 * of firing does not matter too much, as long as they don't fire too
292 * early. 292 * early.
293 */ 293 */
294 unsigned long round_jiffies_up(unsigned long j) 294 unsigned long round_jiffies_up(unsigned long j)
295 { 295 {
296 return round_jiffies_common(j, raw_smp_processor_id(), true); 296 return round_jiffies_common(j, raw_smp_processor_id(), true);
297 } 297 }
298 EXPORT_SYMBOL_GPL(round_jiffies_up); 298 EXPORT_SYMBOL_GPL(round_jiffies_up);
299 299
300 /** 300 /**
301 * round_jiffies_up_relative - function to round jiffies up to a full second 301 * round_jiffies_up_relative - function to round jiffies up to a full second
302 * @j: the time in (relative) jiffies that should be rounded 302 * @j: the time in (relative) jiffies that should be rounded
303 * 303 *
304 * This is the same as round_jiffies_relative() except that it will never 304 * This is the same as round_jiffies_relative() except that it will never
305 * round down. This is useful for timeouts for which the exact time 305 * round down. This is useful for timeouts for which the exact time
306 * of firing does not matter too much, as long as they don't fire too 306 * of firing does not matter too much, as long as they don't fire too
307 * early. 307 * early.
308 */ 308 */
309 unsigned long round_jiffies_up_relative(unsigned long j) 309 unsigned long round_jiffies_up_relative(unsigned long j)
310 { 310 {
311 return __round_jiffies_up_relative(j, raw_smp_processor_id()); 311 return __round_jiffies_up_relative(j, raw_smp_processor_id());
312 } 312 }
313 EXPORT_SYMBOL_GPL(round_jiffies_up_relative); 313 EXPORT_SYMBOL_GPL(round_jiffies_up_relative);
314 314
315 315
316 static inline void set_running_timer(struct tvec_base *base, 316 static inline void set_running_timer(struct tvec_base *base,
317 struct timer_list *timer) 317 struct timer_list *timer)
318 { 318 {
319 #ifdef CONFIG_SMP 319 #ifdef CONFIG_SMP
320 base->running_timer = timer; 320 base->running_timer = timer;
321 #endif 321 #endif
322 } 322 }
323 323
324 static void internal_add_timer(struct tvec_base *base, struct timer_list *timer) 324 static void internal_add_timer(struct tvec_base *base, struct timer_list *timer)
325 { 325 {
326 unsigned long expires = timer->expires; 326 unsigned long expires = timer->expires;
327 unsigned long idx = expires - base->timer_jiffies; 327 unsigned long idx = expires - base->timer_jiffies;
328 struct list_head *vec; 328 struct list_head *vec;
329 329
330 if (idx < TVR_SIZE) { 330 if (idx < TVR_SIZE) {
331 int i = expires & TVR_MASK; 331 int i = expires & TVR_MASK;
332 vec = base->tv1.vec + i; 332 vec = base->tv1.vec + i;
333 } else if (idx < 1 << (TVR_BITS + TVN_BITS)) { 333 } else if (idx < 1 << (TVR_BITS + TVN_BITS)) {
334 int i = (expires >> TVR_BITS) & TVN_MASK; 334 int i = (expires >> TVR_BITS) & TVN_MASK;
335 vec = base->tv2.vec + i; 335 vec = base->tv2.vec + i;
336 } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) { 336 } else if (idx < 1 << (TVR_BITS + 2 * TVN_BITS)) {
337 int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK; 337 int i = (expires >> (TVR_BITS + TVN_BITS)) & TVN_MASK;
338 vec = base->tv3.vec + i; 338 vec = base->tv3.vec + i;
339 } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) { 339 } else if (idx < 1 << (TVR_BITS + 3 * TVN_BITS)) {
340 int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK; 340 int i = (expires >> (TVR_BITS + 2 * TVN_BITS)) & TVN_MASK;
341 vec = base->tv4.vec + i; 341 vec = base->tv4.vec + i;
342 } else if ((signed long) idx < 0) { 342 } else if ((signed long) idx < 0) {
343 /* 343 /*
344 * Can happen if you add a timer with expires == jiffies, 344 * Can happen if you add a timer with expires == jiffies,
345 * or you set a timer to go off in the past 345 * or you set a timer to go off in the past
346 */ 346 */
347 vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK); 347 vec = base->tv1.vec + (base->timer_jiffies & TVR_MASK);
348 } else { 348 } else {
349 int i; 349 int i;
350 /* If the timeout is larger than 0xffffffff on 64-bit 350 /* If the timeout is larger than 0xffffffff on 64-bit
351 * architectures then we use the maximum timeout: 351 * architectures then we use the maximum timeout:
352 */ 352 */
353 if (idx > 0xffffffffUL) { 353 if (idx > 0xffffffffUL) {
354 idx = 0xffffffffUL; 354 idx = 0xffffffffUL;
355 expires = idx + base->timer_jiffies; 355 expires = idx + base->timer_jiffies;
356 } 356 }
357 i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK; 357 i = (expires >> (TVR_BITS + 3 * TVN_BITS)) & TVN_MASK;
358 vec = base->tv5.vec + i; 358 vec = base->tv5.vec + i;
359 } 359 }
360 /* 360 /*
361 * Timers are FIFO: 361 * Timers are FIFO:
362 */ 362 */
363 list_add_tail(&timer->entry, vec); 363 list_add_tail(&timer->entry, vec);
364 } 364 }
365 365
366 #ifdef CONFIG_TIMER_STATS 366 #ifdef CONFIG_TIMER_STATS
367 void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr) 367 void __timer_stats_timer_set_start_info(struct timer_list *timer, void *addr)
368 { 368 {
369 if (timer->start_site) 369 if (timer->start_site)
370 return; 370 return;
371 371
372 timer->start_site = addr; 372 timer->start_site = addr;
373 memcpy(timer->start_comm, current->comm, TASK_COMM_LEN); 373 memcpy(timer->start_comm, current->comm, TASK_COMM_LEN);
374 timer->start_pid = current->pid; 374 timer->start_pid = current->pid;
375 } 375 }
376 376
377 static void timer_stats_account_timer(struct timer_list *timer) 377 static void timer_stats_account_timer(struct timer_list *timer)
378 { 378 {
379 unsigned int flag = 0; 379 unsigned int flag = 0;
380 380
381 if (unlikely(tbase_get_deferrable(timer->base))) 381 if (unlikely(tbase_get_deferrable(timer->base)))
382 flag |= TIMER_STATS_FLAG_DEFERRABLE; 382 flag |= TIMER_STATS_FLAG_DEFERRABLE;
383 383
384 timer_stats_update_stats(timer, timer->start_pid, timer->start_site, 384 timer_stats_update_stats(timer, timer->start_pid, timer->start_site,
385 timer->function, timer->start_comm, flag); 385 timer->function, timer->start_comm, flag);
386 } 386 }
387 387
388 #else 388 #else
389 static void timer_stats_account_timer(struct timer_list *timer) {} 389 static void timer_stats_account_timer(struct timer_list *timer) {}
390 #endif 390 #endif
391 391
392 #ifdef CONFIG_DEBUG_OBJECTS_TIMERS 392 #ifdef CONFIG_DEBUG_OBJECTS_TIMERS
393 393
394 static struct debug_obj_descr timer_debug_descr; 394 static struct debug_obj_descr timer_debug_descr;
395 395
396 /* 396 /*
397 * fixup_init is called when: 397 * fixup_init is called when:
398 * - an active object is initialized 398 * - an active object is initialized
399 */ 399 */
400 static int timer_fixup_init(void *addr, enum debug_obj_state state) 400 static int timer_fixup_init(void *addr, enum debug_obj_state state)
401 { 401 {
402 struct timer_list *timer = addr; 402 struct timer_list *timer = addr;
403 403
404 switch (state) { 404 switch (state) {
405 case ODEBUG_STATE_ACTIVE: 405 case ODEBUG_STATE_ACTIVE:
406 del_timer_sync(timer); 406 del_timer_sync(timer);
407 debug_object_init(timer, &timer_debug_descr); 407 debug_object_init(timer, &timer_debug_descr);
408 return 1; 408 return 1;
409 default: 409 default:
410 return 0; 410 return 0;
411 } 411 }
412 } 412 }
413 413
414 /* 414 /*
415 * fixup_activate is called when: 415 * fixup_activate is called when:
416 * - an active object is activated 416 * - an active object is activated
417 * - an unknown object is activated (might be a statically initialized object) 417 * - an unknown object is activated (might be a statically initialized object)
418 */ 418 */
419 static int timer_fixup_activate(void *addr, enum debug_obj_state state) 419 static int timer_fixup_activate(void *addr, enum debug_obj_state state)
420 { 420 {
421 struct timer_list *timer = addr; 421 struct timer_list *timer = addr;
422 422
423 switch (state) { 423 switch (state) {
424 424
425 case ODEBUG_STATE_NOTAVAILABLE: 425 case ODEBUG_STATE_NOTAVAILABLE:
426 /* 426 /*
427 * This is not really a fixup. The timer was 427 * This is not really a fixup. The timer was
428 * statically initialized. We just make sure that it 428 * statically initialized. We just make sure that it
429 * is tracked in the object tracker. 429 * is tracked in the object tracker.
430 */ 430 */
431 if (timer->entry.next == NULL && 431 if (timer->entry.next == NULL &&
432 timer->entry.prev == TIMER_ENTRY_STATIC) { 432 timer->entry.prev == TIMER_ENTRY_STATIC) {
433 debug_object_init(timer, &timer_debug_descr); 433 debug_object_init(timer, &timer_debug_descr);
434 debug_object_activate(timer, &timer_debug_descr); 434 debug_object_activate(timer, &timer_debug_descr);
435 return 0; 435 return 0;
436 } else { 436 } else {
437 WARN_ON_ONCE(1); 437 WARN_ON_ONCE(1);
438 } 438 }
439 return 0; 439 return 0;
440 440
441 case ODEBUG_STATE_ACTIVE: 441 case ODEBUG_STATE_ACTIVE:
442 WARN_ON(1); 442 WARN_ON(1);
443 443
444 default: 444 default:
445 return 0; 445 return 0;
446 } 446 }
447 } 447 }
448 448
449 /* 449 /*
450 * fixup_free is called when: 450 * fixup_free is called when:
451 * - an active object is freed 451 * - an active object is freed
452 */ 452 */
453 static int timer_fixup_free(void *addr, enum debug_obj_state state) 453 static int timer_fixup_free(void *addr, enum debug_obj_state state)
454 { 454 {
455 struct timer_list *timer = addr; 455 struct timer_list *timer = addr;
456 456
457 switch (state) { 457 switch (state) {
458 case ODEBUG_STATE_ACTIVE: 458 case ODEBUG_STATE_ACTIVE:
459 del_timer_sync(timer); 459 del_timer_sync(timer);
460 debug_object_free(timer, &timer_debug_descr); 460 debug_object_free(timer, &timer_debug_descr);
461 return 1; 461 return 1;
462 default: 462 default:
463 return 0; 463 return 0;
464 } 464 }
465 } 465 }
466 466
467 static struct debug_obj_descr timer_debug_descr = { 467 static struct debug_obj_descr timer_debug_descr = {
468 .name = "timer_list", 468 .name = "timer_list",
469 .fixup_init = timer_fixup_init, 469 .fixup_init = timer_fixup_init,
470 .fixup_activate = timer_fixup_activate, 470 .fixup_activate = timer_fixup_activate,
471 .fixup_free = timer_fixup_free, 471 .fixup_free = timer_fixup_free,
472 }; 472 };
473 473
474 static inline void debug_timer_init(struct timer_list *timer) 474 static inline void debug_timer_init(struct timer_list *timer)
475 { 475 {
476 debug_object_init(timer, &timer_debug_descr); 476 debug_object_init(timer, &timer_debug_descr);
477 } 477 }
478 478
479 static inline void debug_timer_activate(struct timer_list *timer) 479 static inline void debug_timer_activate(struct timer_list *timer)
480 { 480 {
481 debug_object_activate(timer, &timer_debug_descr); 481 debug_object_activate(timer, &timer_debug_descr);
482 } 482 }
483 483
484 static inline void debug_timer_deactivate(struct timer_list *timer) 484 static inline void debug_timer_deactivate(struct timer_list *timer)
485 { 485 {
486 debug_object_deactivate(timer, &timer_debug_descr); 486 debug_object_deactivate(timer, &timer_debug_descr);
487 } 487 }
488 488
489 static inline void debug_timer_free(struct timer_list *timer) 489 static inline void debug_timer_free(struct timer_list *timer)
490 { 490 {
491 debug_object_free(timer, &timer_debug_descr); 491 debug_object_free(timer, &timer_debug_descr);
492 } 492 }
493 493
494 static void __init_timer(struct timer_list *timer); 494 static void __init_timer(struct timer_list *timer);
495 495
496 void init_timer_on_stack(struct timer_list *timer) 496 void init_timer_on_stack(struct timer_list *timer)
497 { 497 {
498 debug_object_init_on_stack(timer, &timer_debug_descr); 498 debug_object_init_on_stack(timer, &timer_debug_descr);
499 __init_timer(timer); 499 __init_timer(timer);
500 } 500 }
501 EXPORT_SYMBOL_GPL(init_timer_on_stack); 501 EXPORT_SYMBOL_GPL(init_timer_on_stack);
502 502
503 void destroy_timer_on_stack(struct timer_list *timer) 503 void destroy_timer_on_stack(struct timer_list *timer)
504 { 504 {
505 debug_object_free(timer, &timer_debug_descr); 505 debug_object_free(timer, &timer_debug_descr);
506 } 506 }
507 EXPORT_SYMBOL_GPL(destroy_timer_on_stack); 507 EXPORT_SYMBOL_GPL(destroy_timer_on_stack);
508 508
509 #else 509 #else
510 static inline void debug_timer_init(struct timer_list *timer) { } 510 static inline void debug_timer_init(struct timer_list *timer) { }
511 static inline void debug_timer_activate(struct timer_list *timer) { } 511 static inline void debug_timer_activate(struct timer_list *timer) { }
512 static inline void debug_timer_deactivate(struct timer_list *timer) { } 512 static inline void debug_timer_deactivate(struct timer_list *timer) { }
513 #endif 513 #endif
514 514
515 static void __init_timer(struct timer_list *timer) 515 static void __init_timer(struct timer_list *timer)
516 { 516 {
517 timer->entry.next = NULL; 517 timer->entry.next = NULL;
518 timer->base = __raw_get_cpu_var(tvec_bases); 518 timer->base = __raw_get_cpu_var(tvec_bases);
519 #ifdef CONFIG_TIMER_STATS 519 #ifdef CONFIG_TIMER_STATS
520 timer->start_site = NULL; 520 timer->start_site = NULL;
521 timer->start_pid = -1; 521 timer->start_pid = -1;
522 memset(timer->start_comm, 0, TASK_COMM_LEN); 522 memset(timer->start_comm, 0, TASK_COMM_LEN);
523 #endif 523 #endif
524 } 524 }
525 525
526 /** 526 /**
527 * init_timer - initialize a timer. 527 * init_timer_key - initialize a timer
528 * @timer: the timer to be initialized 528 * @timer: the timer to be initialized
529 * @name: name of the timer
530 * @key: lockdep class key of the fake lock used for tracking timer
531 * sync lock dependencies
529 * 532 *
530 * init_timer() must be done to a timer prior calling *any* of the 533 * init_timer_key() must be done to a timer prior calling *any* of the
531 * other timer functions. 534 * other timer functions.
532 */ 535 */
533 void init_timer(struct timer_list *timer) 536 void init_timer(struct timer_list *timer)
534 { 537 {
535 debug_timer_init(timer); 538 debug_timer_init(timer);
536 __init_timer(timer); 539 __init_timer(timer);
537 } 540 }
538 EXPORT_SYMBOL(init_timer); 541 EXPORT_SYMBOL(init_timer);
539 542
540 void init_timer_deferrable(struct timer_list *timer) 543 void init_timer_deferrable(struct timer_list *timer)
541 { 544 {
542 init_timer(timer); 545 init_timer(timer);
543 timer_set_deferrable(timer); 546 timer_set_deferrable(timer);
544 } 547 }
545 EXPORT_SYMBOL(init_timer_deferrable); 548 EXPORT_SYMBOL(init_timer_deferrable);
546 549
547 static inline void detach_timer(struct timer_list *timer, 550 static inline void detach_timer(struct timer_list *timer,
548 int clear_pending) 551 int clear_pending)
549 { 552 {
550 struct list_head *entry = &timer->entry; 553 struct list_head *entry = &timer->entry;
551 554
552 debug_timer_deactivate(timer); 555 debug_timer_deactivate(timer);
553 556
554 __list_del(entry->prev, entry->next); 557 __list_del(entry->prev, entry->next);
555 if (clear_pending) 558 if (clear_pending)
556 entry->next = NULL; 559 entry->next = NULL;
557 entry->prev = LIST_POISON2; 560 entry->prev = LIST_POISON2;
558 } 561 }
559 562
560 /* 563 /*
561 * We are using hashed locking: holding per_cpu(tvec_bases).lock 564 * We are using hashed locking: holding per_cpu(tvec_bases).lock
562 * means that all timers which are tied to this base via timer->base are 565 * means that all timers which are tied to this base via timer->base are
563 * locked, and the base itself is locked too. 566 * locked, and the base itself is locked too.
564 * 567 *
565 * So __run_timers/migrate_timers can safely modify all timers which could 568 * So __run_timers/migrate_timers can safely modify all timers which could
566 * be found on ->tvX lists. 569 * be found on ->tvX lists.
567 * 570 *
568 * When the timer's base is locked, and the timer removed from list, it is 571 * When the timer's base is locked, and the timer removed from list, it is
569 * possible to set timer->base = NULL and drop the lock: the timer remains 572 * possible to set timer->base = NULL and drop the lock: the timer remains
570 * locked. 573 * locked.
571 */ 574 */
572 static struct tvec_base *lock_timer_base(struct timer_list *timer, 575 static struct tvec_base *lock_timer_base(struct timer_list *timer,
573 unsigned long *flags) 576 unsigned long *flags)
574 __acquires(timer->base->lock) 577 __acquires(timer->base->lock)
575 { 578 {
576 struct tvec_base *base; 579 struct tvec_base *base;
577 580
578 for (;;) { 581 for (;;) {
579 struct tvec_base *prelock_base = timer->base; 582 struct tvec_base *prelock_base = timer->base;
580 base = tbase_get_base(prelock_base); 583 base = tbase_get_base(prelock_base);
581 if (likely(base != NULL)) { 584 if (likely(base != NULL)) {
582 spin_lock_irqsave(&base->lock, *flags); 585 spin_lock_irqsave(&base->lock, *flags);
583 if (likely(prelock_base == timer->base)) 586 if (likely(prelock_base == timer->base))
584 return base; 587 return base;
585 /* The timer has migrated to another CPU */ 588 /* The timer has migrated to another CPU */
586 spin_unlock_irqrestore(&base->lock, *flags); 589 spin_unlock_irqrestore(&base->lock, *flags);
587 } 590 }
588 cpu_relax(); 591 cpu_relax();
589 } 592 }
590 } 593 }
591 594
592 static inline int 595 static inline int
593 __mod_timer(struct timer_list *timer, unsigned long expires, bool pending_only) 596 __mod_timer(struct timer_list *timer, unsigned long expires, bool pending_only)
594 { 597 {
595 struct tvec_base *base, *new_base; 598 struct tvec_base *base, *new_base;
596 unsigned long flags; 599 unsigned long flags;
597 int ret; 600 int ret;
598 601
599 ret = 0; 602 ret = 0;
600 603
601 timer_stats_timer_set_start_info(timer); 604 timer_stats_timer_set_start_info(timer);
602 BUG_ON(!timer->function); 605 BUG_ON(!timer->function);
603 606
604 base = lock_timer_base(timer, &flags); 607 base = lock_timer_base(timer, &flags);
605 608
606 if (timer_pending(timer)) { 609 if (timer_pending(timer)) {
607 detach_timer(timer, 0); 610 detach_timer(timer, 0);
608 ret = 1; 611 ret = 1;
609 } else { 612 } else {
610 if (pending_only) 613 if (pending_only)
611 goto out_unlock; 614 goto out_unlock;
612 } 615 }
613 616
614 debug_timer_activate(timer); 617 debug_timer_activate(timer);
615 618
616 new_base = __get_cpu_var(tvec_bases); 619 new_base = __get_cpu_var(tvec_bases);
617 620
618 if (base != new_base) { 621 if (base != new_base) {
619 /* 622 /*
620 * We are trying to schedule the timer on the local CPU. 623 * We are trying to schedule the timer on the local CPU.
621 * However we can't change timer's base while it is running, 624 * However we can't change timer's base while it is running,
622 * otherwise del_timer_sync() can't detect that the timer's 625 * otherwise del_timer_sync() can't detect that the timer's
623 * handler yet has not finished. This also guarantees that 626 * handler yet has not finished. This also guarantees that
624 * the timer is serialized wrt itself. 627 * the timer is serialized wrt itself.
625 */ 628 */
626 if (likely(base->running_timer != timer)) { 629 if (likely(base->running_timer != timer)) {
627 /* See the comment in lock_timer_base() */ 630 /* See the comment in lock_timer_base() */
628 timer_set_base(timer, NULL); 631 timer_set_base(timer, NULL);
629 spin_unlock(&base->lock); 632 spin_unlock(&base->lock);
630 base = new_base; 633 base = new_base;
631 spin_lock(&base->lock); 634 spin_lock(&base->lock);
632 timer_set_base(timer, base); 635 timer_set_base(timer, base);
633 } 636 }
634 } 637 }
635 638
636 timer->expires = expires; 639 timer->expires = expires;
637 internal_add_timer(base, timer); 640 internal_add_timer(base, timer);
638 641
639 out_unlock: 642 out_unlock:
640 spin_unlock_irqrestore(&base->lock, flags); 643 spin_unlock_irqrestore(&base->lock, flags);
641 644
642 return ret; 645 return ret;
643 } 646 }
644 647
645 /** 648 /**
646 * mod_timer_pending - modify a pending timer's timeout 649 * mod_timer_pending - modify a pending timer's timeout
647 * @timer: the pending timer to be modified 650 * @timer: the pending timer to be modified
648 * @expires: new timeout in jiffies 651 * @expires: new timeout in jiffies
649 * 652 *
650 * mod_timer_pending() is the same for pending timers as mod_timer(), 653 * mod_timer_pending() is the same for pending timers as mod_timer(),
651 * but will not re-activate and modify already deleted timers. 654 * but will not re-activate and modify already deleted timers.
652 * 655 *
653 * It is useful for unserialized use of timers. 656 * It is useful for unserialized use of timers.
654 */ 657 */
655 int mod_timer_pending(struct timer_list *timer, unsigned long expires) 658 int mod_timer_pending(struct timer_list *timer, unsigned long expires)
656 { 659 {
657 return __mod_timer(timer, expires, true); 660 return __mod_timer(timer, expires, true);
658 } 661 }
659 EXPORT_SYMBOL(mod_timer_pending); 662 EXPORT_SYMBOL(mod_timer_pending);
660 663
661 /** 664 /**
662 * mod_timer - modify a timer's timeout 665 * mod_timer - modify a timer's timeout
663 * @timer: the timer to be modified 666 * @timer: the timer to be modified
664 * @expires: new timeout in jiffies 667 * @expires: new timeout in jiffies
665 * 668 *
666 * mod_timer() is a more efficient way to update the expire field of an 669 * mod_timer() is a more efficient way to update the expire field of an
667 * active timer (if the timer is inactive it will be activated) 670 * active timer (if the timer is inactive it will be activated)
668 * 671 *
669 * mod_timer(timer, expires) is equivalent to: 672 * mod_timer(timer, expires) is equivalent to:
670 * 673 *
671 * del_timer(timer); timer->expires = expires; add_timer(timer); 674 * del_timer(timer); timer->expires = expires; add_timer(timer);
672 * 675 *
673 * Note that if there are multiple unserialized concurrent users of the 676 * Note that if there are multiple unserialized concurrent users of the
674 * same timer, then mod_timer() is the only safe way to modify the timeout, 677 * same timer, then mod_timer() is the only safe way to modify the timeout,
675 * since add_timer() cannot modify an already running timer. 678 * since add_timer() cannot modify an already running timer.
676 * 679 *
677 * The function returns whether it has modified a pending timer or not. 680 * The function returns whether it has modified a pending timer or not.
678 * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an 681 * (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an
679 * active timer returns 1.) 682 * active timer returns 1.)
680 */ 683 */
681 int mod_timer(struct timer_list *timer, unsigned long expires) 684 int mod_timer(struct timer_list *timer, unsigned long expires)
682 { 685 {
683 /* 686 /*
684 * This is a common optimization triggered by the 687 * This is a common optimization triggered by the
685 * networking code - if the timer is re-modified 688 * networking code - if the timer is re-modified
686 * to be the same thing then just return: 689 * to be the same thing then just return:
687 */ 690 */
688 if (timer->expires == expires && timer_pending(timer)) 691 if (timer->expires == expires && timer_pending(timer))
689 return 1; 692 return 1;
690 693
691 return __mod_timer(timer, expires, false); 694 return __mod_timer(timer, expires, false);
692 } 695 }
693 EXPORT_SYMBOL(mod_timer); 696 EXPORT_SYMBOL(mod_timer);
694 697
695 /** 698 /**
696 * add_timer - start a timer 699 * add_timer - start a timer
697 * @timer: the timer to be added 700 * @timer: the timer to be added
698 * 701 *
699 * The kernel will do a ->function(->data) callback from the 702 * The kernel will do a ->function(->data) callback from the
700 * timer interrupt at the ->expires point in the future. The 703 * timer interrupt at the ->expires point in the future. The
701 * current time is 'jiffies'. 704 * current time is 'jiffies'.
702 * 705 *
703 * The timer's ->expires, ->function (and if the handler uses it, ->data) 706 * The timer's ->expires, ->function (and if the handler uses it, ->data)
704 * fields must be set prior calling this function. 707 * fields must be set prior calling this function.
705 * 708 *
706 * Timers with an ->expires field in the past will be executed in the next 709 * Timers with an ->expires field in the past will be executed in the next
707 * timer tick. 710 * timer tick.
708 */ 711 */
709 void add_timer(struct timer_list *timer) 712 void add_timer(struct timer_list *timer)
710 { 713 {
711 BUG_ON(timer_pending(timer)); 714 BUG_ON(timer_pending(timer));
712 mod_timer(timer, timer->expires); 715 mod_timer(timer, timer->expires);
713 } 716 }
714 EXPORT_SYMBOL(add_timer); 717 EXPORT_SYMBOL(add_timer);
715 718
716 /** 719 /**
717 * add_timer_on - start a timer on a particular CPU 720 * add_timer_on - start a timer on a particular CPU
718 * @timer: the timer to be added 721 * @timer: the timer to be added
719 * @cpu: the CPU to start it on 722 * @cpu: the CPU to start it on
720 * 723 *
721 * This is not very scalable on SMP. Double adds are not possible. 724 * This is not very scalable on SMP. Double adds are not possible.
722 */ 725 */
723 void add_timer_on(struct timer_list *timer, int cpu) 726 void add_timer_on(struct timer_list *timer, int cpu)
724 { 727 {
725 struct tvec_base *base = per_cpu(tvec_bases, cpu); 728 struct tvec_base *base = per_cpu(tvec_bases, cpu);
726 unsigned long flags; 729 unsigned long flags;
727 730
728 timer_stats_timer_set_start_info(timer); 731 timer_stats_timer_set_start_info(timer);
729 BUG_ON(timer_pending(timer) || !timer->function); 732 BUG_ON(timer_pending(timer) || !timer->function);
730 spin_lock_irqsave(&base->lock, flags); 733 spin_lock_irqsave(&base->lock, flags);
731 timer_set_base(timer, base); 734 timer_set_base(timer, base);
732 debug_timer_activate(timer); 735 debug_timer_activate(timer);
733 internal_add_timer(base, timer); 736 internal_add_timer(base, timer);
734 /* 737 /*
735 * Check whether the other CPU is idle and needs to be 738 * Check whether the other CPU is idle and needs to be
736 * triggered to reevaluate the timer wheel when nohz is 739 * triggered to reevaluate the timer wheel when nohz is
737 * active. We are protected against the other CPU fiddling 740 * active. We are protected against the other CPU fiddling
738 * with the timer by holding the timer base lock. This also 741 * with the timer by holding the timer base lock. This also
739 * makes sure that a CPU on the way to idle can not evaluate 742 * makes sure that a CPU on the way to idle can not evaluate
740 * the timer wheel. 743 * the timer wheel.
741 */ 744 */
742 wake_up_idle_cpu(cpu); 745 wake_up_idle_cpu(cpu);
743 spin_unlock_irqrestore(&base->lock, flags); 746 spin_unlock_irqrestore(&base->lock, flags);
744 } 747 }
745 748
746 /** 749 /**
747 * del_timer - deactive a timer. 750 * del_timer - deactive a timer.
748 * @timer: the timer to be deactivated 751 * @timer: the timer to be deactivated
749 * 752 *
750 * del_timer() deactivates a timer - this works on both active and inactive 753 * del_timer() deactivates a timer - this works on both active and inactive
751 * timers. 754 * timers.
752 * 755 *
753 * The function returns whether it has deactivated a pending timer or not. 756 * The function returns whether it has deactivated a pending timer or not.
754 * (ie. del_timer() of an inactive timer returns 0, del_timer() of an 757 * (ie. del_timer() of an inactive timer returns 0, del_timer() of an
755 * active timer returns 1.) 758 * active timer returns 1.)
756 */ 759 */
757 int del_timer(struct timer_list *timer) 760 int del_timer(struct timer_list *timer)
758 { 761 {
759 struct tvec_base *base; 762 struct tvec_base *base;
760 unsigned long flags; 763 unsigned long flags;
761 int ret = 0; 764 int ret = 0;
762 765
763 timer_stats_timer_clear_start_info(timer); 766 timer_stats_timer_clear_start_info(timer);
764 if (timer_pending(timer)) { 767 if (timer_pending(timer)) {
765 base = lock_timer_base(timer, &flags); 768 base = lock_timer_base(timer, &flags);
766 if (timer_pending(timer)) { 769 if (timer_pending(timer)) {
767 detach_timer(timer, 1); 770 detach_timer(timer, 1);
768 ret = 1; 771 ret = 1;
769 } 772 }
770 spin_unlock_irqrestore(&base->lock, flags); 773 spin_unlock_irqrestore(&base->lock, flags);
771 } 774 }
772 775
773 return ret; 776 return ret;
774 } 777 }
775 EXPORT_SYMBOL(del_timer); 778 EXPORT_SYMBOL(del_timer);
776 779
777 #ifdef CONFIG_SMP 780 #ifdef CONFIG_SMP
778 /** 781 /**
779 * try_to_del_timer_sync - Try to deactivate a timer 782 * try_to_del_timer_sync - Try to deactivate a timer
780 * @timer: timer do del 783 * @timer: timer do del
781 * 784 *
782 * This function tries to deactivate a timer. Upon successful (ret >= 0) 785 * This function tries to deactivate a timer. Upon successful (ret >= 0)
783 * exit the timer is not queued and the handler is not running on any CPU. 786 * exit the timer is not queued and the handler is not running on any CPU.
784 * 787 *
785 * It must not be called from interrupt contexts. 788 * It must not be called from interrupt contexts.
786 */ 789 */
787 int try_to_del_timer_sync(struct timer_list *timer) 790 int try_to_del_timer_sync(struct timer_list *timer)
788 { 791 {
789 struct tvec_base *base; 792 struct tvec_base *base;
790 unsigned long flags; 793 unsigned long flags;
791 int ret = -1; 794 int ret = -1;
792 795
793 base = lock_timer_base(timer, &flags); 796 base = lock_timer_base(timer, &flags);
794 797
795 if (base->running_timer == timer) 798 if (base->running_timer == timer)
796 goto out; 799 goto out;
797 800
798 ret = 0; 801 ret = 0;
799 if (timer_pending(timer)) { 802 if (timer_pending(timer)) {
800 detach_timer(timer, 1); 803 detach_timer(timer, 1);
801 ret = 1; 804 ret = 1;
802 } 805 }
803 out: 806 out:
804 spin_unlock_irqrestore(&base->lock, flags); 807 spin_unlock_irqrestore(&base->lock, flags);
805 808
806 return ret; 809 return ret;
807 } 810 }
808 EXPORT_SYMBOL(try_to_del_timer_sync); 811 EXPORT_SYMBOL(try_to_del_timer_sync);
809 812
810 /** 813 /**
811 * del_timer_sync - deactivate a timer and wait for the handler to finish. 814 * del_timer_sync - deactivate a timer and wait for the handler to finish.
812 * @timer: the timer to be deactivated 815 * @timer: the timer to be deactivated
813 * 816 *
814 * This function only differs from del_timer() on SMP: besides deactivating 817 * This function only differs from del_timer() on SMP: besides deactivating
815 * the timer it also makes sure the handler has finished executing on other 818 * the timer it also makes sure the handler has finished executing on other
816 * CPUs. 819 * CPUs.
817 * 820 *
818 * Synchronization rules: Callers must prevent restarting of the timer, 821 * Synchronization rules: Callers must prevent restarting of the timer,
819 * otherwise this function is meaningless. It must not be called from 822 * otherwise this function is meaningless. It must not be called from
820 * interrupt contexts. The caller must not hold locks which would prevent 823 * interrupt contexts. The caller must not hold locks which would prevent
821 * completion of the timer's handler. The timer's handler must not call 824 * completion of the timer's handler. The timer's handler must not call
822 * add_timer_on(). Upon exit the timer is not queued and the handler is 825 * add_timer_on(). Upon exit the timer is not queued and the handler is
823 * not running on any CPU. 826 * not running on any CPU.
824 * 827 *
825 * The function returns whether it has deactivated a pending timer or not. 828 * The function returns whether it has deactivated a pending timer or not.
826 */ 829 */
827 int del_timer_sync(struct timer_list *timer) 830 int del_timer_sync(struct timer_list *timer)
828 { 831 {
829 for (;;) { 832 for (;;) {
830 int ret = try_to_del_timer_sync(timer); 833 int ret = try_to_del_timer_sync(timer);
831 if (ret >= 0) 834 if (ret >= 0)
832 return ret; 835 return ret;
833 cpu_relax(); 836 cpu_relax();
834 } 837 }
835 } 838 }
836 EXPORT_SYMBOL(del_timer_sync); 839 EXPORT_SYMBOL(del_timer_sync);
837 #endif 840 #endif
838 841
839 static int cascade(struct tvec_base *base, struct tvec *tv, int index) 842 static int cascade(struct tvec_base *base, struct tvec *tv, int index)
840 { 843 {
841 /* cascade all the timers from tv up one level */ 844 /* cascade all the timers from tv up one level */
842 struct timer_list *timer, *tmp; 845 struct timer_list *timer, *tmp;
843 struct list_head tv_list; 846 struct list_head tv_list;
844 847
845 list_replace_init(tv->vec + index, &tv_list); 848 list_replace_init(tv->vec + index, &tv_list);
846 849
847 /* 850 /*
848 * We are removing _all_ timers from the list, so we 851 * We are removing _all_ timers from the list, so we
849 * don't have to detach them individually. 852 * don't have to detach them individually.
850 */ 853 */
851 list_for_each_entry_safe(timer, tmp, &tv_list, entry) { 854 list_for_each_entry_safe(timer, tmp, &tv_list, entry) {
852 BUG_ON(tbase_get_base(timer->base) != base); 855 BUG_ON(tbase_get_base(timer->base) != base);
853 internal_add_timer(base, timer); 856 internal_add_timer(base, timer);
854 } 857 }
855 858
856 return index; 859 return index;
857 } 860 }
858 861
859 #define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK) 862 #define INDEX(N) ((base->timer_jiffies >> (TVR_BITS + (N) * TVN_BITS)) & TVN_MASK)
860 863
861 /** 864 /**
862 * __run_timers - run all expired timers (if any) on this CPU. 865 * __run_timers - run all expired timers (if any) on this CPU.
863 * @base: the timer vector to be processed. 866 * @base: the timer vector to be processed.
864 * 867 *
865 * This function cascades all vectors and executes all expired timer 868 * This function cascades all vectors and executes all expired timer
866 * vectors. 869 * vectors.
867 */ 870 */
868 static inline void __run_timers(struct tvec_base *base) 871 static inline void __run_timers(struct tvec_base *base)
869 { 872 {
870 struct timer_list *timer; 873 struct timer_list *timer;
871 874
872 spin_lock_irq(&base->lock); 875 spin_lock_irq(&base->lock);
873 while (time_after_eq(jiffies, base->timer_jiffies)) { 876 while (time_after_eq(jiffies, base->timer_jiffies)) {
874 struct list_head work_list; 877 struct list_head work_list;
875 struct list_head *head = &work_list; 878 struct list_head *head = &work_list;
876 int index = base->timer_jiffies & TVR_MASK; 879 int index = base->timer_jiffies & TVR_MASK;
877 880
878 /* 881 /*
879 * Cascade timers: 882 * Cascade timers:
880 */ 883 */
881 if (!index && 884 if (!index &&
882 (!cascade(base, &base->tv2, INDEX(0))) && 885 (!cascade(base, &base->tv2, INDEX(0))) &&
883 (!cascade(base, &base->tv3, INDEX(1))) && 886 (!cascade(base, &base->tv3, INDEX(1))) &&
884 !cascade(base, &base->tv4, INDEX(2))) 887 !cascade(base, &base->tv4, INDEX(2)))
885 cascade(base, &base->tv5, INDEX(3)); 888 cascade(base, &base->tv5, INDEX(3));
886 ++base->timer_jiffies; 889 ++base->timer_jiffies;
887 list_replace_init(base->tv1.vec + index, &work_list); 890 list_replace_init(base->tv1.vec + index, &work_list);
888 while (!list_empty(head)) { 891 while (!list_empty(head)) {
889 void (*fn)(unsigned long); 892 void (*fn)(unsigned long);
890 unsigned long data; 893 unsigned long data;
891 894
892 timer = list_first_entry(head, struct timer_list,entry); 895 timer = list_first_entry(head, struct timer_list,entry);
893 fn = timer->function; 896 fn = timer->function;
894 data = timer->data; 897 data = timer->data;
895 898
896 timer_stats_account_timer(timer); 899 timer_stats_account_timer(timer);
897 900
898 set_running_timer(base, timer); 901 set_running_timer(base, timer);
899 detach_timer(timer, 1); 902 detach_timer(timer, 1);
900 spin_unlock_irq(&base->lock); 903 spin_unlock_irq(&base->lock);
901 { 904 {
902 int preempt_count = preempt_count(); 905 int preempt_count = preempt_count();
903 fn(data); 906 fn(data);
904 if (preempt_count != preempt_count()) { 907 if (preempt_count != preempt_count()) {
905 printk(KERN_ERR "huh, entered %p " 908 printk(KERN_ERR "huh, entered %p "
906 "with preempt_count %08x, exited" 909 "with preempt_count %08x, exited"
907 " with %08x?\n", 910 " with %08x?\n",
908 fn, preempt_count, 911 fn, preempt_count,
909 preempt_count()); 912 preempt_count());
910 BUG(); 913 BUG();
911 } 914 }
912 } 915 }
913 spin_lock_irq(&base->lock); 916 spin_lock_irq(&base->lock);
914 } 917 }
915 } 918 }
916 set_running_timer(base, NULL); 919 set_running_timer(base, NULL);
917 spin_unlock_irq(&base->lock); 920 spin_unlock_irq(&base->lock);
918 } 921 }
919 922
920 #ifdef CONFIG_NO_HZ 923 #ifdef CONFIG_NO_HZ
921 /* 924 /*
922 * Find out when the next timer event is due to happen. This 925 * Find out when the next timer event is due to happen. This
923 * is used on S/390 to stop all activity when a cpus is idle. 926 * is used on S/390 to stop all activity when a cpus is idle.
924 * This functions needs to be called disabled. 927 * This functions needs to be called disabled.
925 */ 928 */
926 static unsigned long __next_timer_interrupt(struct tvec_base *base) 929 static unsigned long __next_timer_interrupt(struct tvec_base *base)
927 { 930 {
928 unsigned long timer_jiffies = base->timer_jiffies; 931 unsigned long timer_jiffies = base->timer_jiffies;
929 unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA; 932 unsigned long expires = timer_jiffies + NEXT_TIMER_MAX_DELTA;
930 int index, slot, array, found = 0; 933 int index, slot, array, found = 0;
931 struct timer_list *nte; 934 struct timer_list *nte;
932 struct tvec *varray[4]; 935 struct tvec *varray[4];
933 936
934 /* Look for timer events in tv1. */ 937 /* Look for timer events in tv1. */
935 index = slot = timer_jiffies & TVR_MASK; 938 index = slot = timer_jiffies & TVR_MASK;
936 do { 939 do {
937 list_for_each_entry(nte, base->tv1.vec + slot, entry) { 940 list_for_each_entry(nte, base->tv1.vec + slot, entry) {
938 if (tbase_get_deferrable(nte->base)) 941 if (tbase_get_deferrable(nte->base))
939 continue; 942 continue;
940 943
941 found = 1; 944 found = 1;
942 expires = nte->expires; 945 expires = nte->expires;
943 /* Look at the cascade bucket(s)? */ 946 /* Look at the cascade bucket(s)? */
944 if (!index || slot < index) 947 if (!index || slot < index)
945 goto cascade; 948 goto cascade;
946 return expires; 949 return expires;
947 } 950 }
948 slot = (slot + 1) & TVR_MASK; 951 slot = (slot + 1) & TVR_MASK;
949 } while (slot != index); 952 } while (slot != index);
950 953
951 cascade: 954 cascade:
952 /* Calculate the next cascade event */ 955 /* Calculate the next cascade event */
953 if (index) 956 if (index)
954 timer_jiffies += TVR_SIZE - index; 957 timer_jiffies += TVR_SIZE - index;
955 timer_jiffies >>= TVR_BITS; 958 timer_jiffies >>= TVR_BITS;
956 959
957 /* Check tv2-tv5. */ 960 /* Check tv2-tv5. */
958 varray[0] = &base->tv2; 961 varray[0] = &base->tv2;
959 varray[1] = &base->tv3; 962 varray[1] = &base->tv3;
960 varray[2] = &base->tv4; 963 varray[2] = &base->tv4;
961 varray[3] = &base->tv5; 964 varray[3] = &base->tv5;
962 965
963 for (array = 0; array < 4; array++) { 966 for (array = 0; array < 4; array++) {
964 struct tvec *varp = varray[array]; 967 struct tvec *varp = varray[array];
965 968
966 index = slot = timer_jiffies & TVN_MASK; 969 index = slot = timer_jiffies & TVN_MASK;
967 do { 970 do {
968 list_for_each_entry(nte, varp->vec + slot, entry) { 971 list_for_each_entry(nte, varp->vec + slot, entry) {
969 found = 1; 972 found = 1;
970 if (time_before(nte->expires, expires)) 973 if (time_before(nte->expires, expires))
971 expires = nte->expires; 974 expires = nte->expires;
972 } 975 }
973 /* 976 /*
974 * Do we still search for the first timer or are 977 * Do we still search for the first timer or are
975 * we looking up the cascade buckets ? 978 * we looking up the cascade buckets ?
976 */ 979 */
977 if (found) { 980 if (found) {
978 /* Look at the cascade bucket(s)? */ 981 /* Look at the cascade bucket(s)? */
979 if (!index || slot < index) 982 if (!index || slot < index)
980 break; 983 break;
981 return expires; 984 return expires;
982 } 985 }
983 slot = (slot + 1) & TVN_MASK; 986 slot = (slot + 1) & TVN_MASK;
984 } while (slot != index); 987 } while (slot != index);
985 988
986 if (index) 989 if (index)
987 timer_jiffies += TVN_SIZE - index; 990 timer_jiffies += TVN_SIZE - index;
988 timer_jiffies >>= TVN_BITS; 991 timer_jiffies >>= TVN_BITS;
989 } 992 }
990 return expires; 993 return expires;
991 } 994 }
992 995
993 /* 996 /*
994 * Check, if the next hrtimer event is before the next timer wheel 997 * Check, if the next hrtimer event is before the next timer wheel
995 * event: 998 * event:
996 */ 999 */
997 static unsigned long cmp_next_hrtimer_event(unsigned long now, 1000 static unsigned long cmp_next_hrtimer_event(unsigned long now,
998 unsigned long expires) 1001 unsigned long expires)
999 { 1002 {
1000 ktime_t hr_delta = hrtimer_get_next_event(); 1003 ktime_t hr_delta = hrtimer_get_next_event();
1001 struct timespec tsdelta; 1004 struct timespec tsdelta;
1002 unsigned long delta; 1005 unsigned long delta;
1003 1006
1004 if (hr_delta.tv64 == KTIME_MAX) 1007 if (hr_delta.tv64 == KTIME_MAX)
1005 return expires; 1008 return expires;
1006 1009
1007 /* 1010 /*
1008 * Expired timer available, let it expire in the next tick 1011 * Expired timer available, let it expire in the next tick
1009 */ 1012 */
1010 if (hr_delta.tv64 <= 0) 1013 if (hr_delta.tv64 <= 0)
1011 return now + 1; 1014 return now + 1;
1012 1015
1013 tsdelta = ktime_to_timespec(hr_delta); 1016 tsdelta = ktime_to_timespec(hr_delta);
1014 delta = timespec_to_jiffies(&tsdelta); 1017 delta = timespec_to_jiffies(&tsdelta);
1015 1018
1016 /* 1019 /*
1017 * Limit the delta to the max value, which is checked in 1020 * Limit the delta to the max value, which is checked in
1018 * tick_nohz_stop_sched_tick(): 1021 * tick_nohz_stop_sched_tick():
1019 */ 1022 */
1020 if (delta > NEXT_TIMER_MAX_DELTA) 1023 if (delta > NEXT_TIMER_MAX_DELTA)
1021 delta = NEXT_TIMER_MAX_DELTA; 1024 delta = NEXT_TIMER_MAX_DELTA;
1022 1025
1023 /* 1026 /*
1024 * Take rounding errors in to account and make sure, that it 1027 * Take rounding errors in to account and make sure, that it
1025 * expires in the next tick. Otherwise we go into an endless 1028 * expires in the next tick. Otherwise we go into an endless
1026 * ping pong due to tick_nohz_stop_sched_tick() retriggering 1029 * ping pong due to tick_nohz_stop_sched_tick() retriggering
1027 * the timer softirq 1030 * the timer softirq
1028 */ 1031 */
1029 if (delta < 1) 1032 if (delta < 1)
1030 delta = 1; 1033 delta = 1;
1031 now += delta; 1034 now += delta;
1032 if (time_before(now, expires)) 1035 if (time_before(now, expires))
1033 return now; 1036 return now;
1034 return expires; 1037 return expires;
1035 } 1038 }
1036 1039
1037 /** 1040 /**
1038 * get_next_timer_interrupt - return the jiffy of the next pending timer 1041 * get_next_timer_interrupt - return the jiffy of the next pending timer
1039 * @now: current time (in jiffies) 1042 * @now: current time (in jiffies)
1040 */ 1043 */
1041 unsigned long get_next_timer_interrupt(unsigned long now) 1044 unsigned long get_next_timer_interrupt(unsigned long now)
1042 { 1045 {
1043 struct tvec_base *base = __get_cpu_var(tvec_bases); 1046 struct tvec_base *base = __get_cpu_var(tvec_bases);
1044 unsigned long expires; 1047 unsigned long expires;
1045 1048
1046 spin_lock(&base->lock); 1049 spin_lock(&base->lock);
1047 expires = __next_timer_interrupt(base); 1050 expires = __next_timer_interrupt(base);
1048 spin_unlock(&base->lock); 1051 spin_unlock(&base->lock);
1049 1052
1050 if (time_before_eq(expires, now)) 1053 if (time_before_eq(expires, now))
1051 return now; 1054 return now;
1052 1055
1053 return cmp_next_hrtimer_event(now, expires); 1056 return cmp_next_hrtimer_event(now, expires);
1054 } 1057 }
1055 #endif 1058 #endif
1056 1059
1057 /* 1060 /*
1058 * Called from the timer interrupt handler to charge one tick to the current 1061 * Called from the timer interrupt handler to charge one tick to the current
1059 * process. user_tick is 1 if the tick is user time, 0 for system. 1062 * process. user_tick is 1 if the tick is user time, 0 for system.
1060 */ 1063 */
1061 void update_process_times(int user_tick) 1064 void update_process_times(int user_tick)
1062 { 1065 {
1063 struct task_struct *p = current; 1066 struct task_struct *p = current;
1064 int cpu = smp_processor_id(); 1067 int cpu = smp_processor_id();
1065 1068
1066 /* Note: this timer irq context must be accounted for as well. */ 1069 /* Note: this timer irq context must be accounted for as well. */
1067 account_process_tick(p, user_tick); 1070 account_process_tick(p, user_tick);
1068 run_local_timers(); 1071 run_local_timers();
1069 if (rcu_pending(cpu)) 1072 if (rcu_pending(cpu))
1070 rcu_check_callbacks(cpu, user_tick); 1073 rcu_check_callbacks(cpu, user_tick);
1071 printk_tick(); 1074 printk_tick();
1072 scheduler_tick(); 1075 scheduler_tick();
1073 run_posix_cpu_timers(p); 1076 run_posix_cpu_timers(p);
1074 } 1077 }
1075 1078
1076 /* 1079 /*
1077 * Nr of active tasks - counted in fixed-point numbers 1080 * Nr of active tasks - counted in fixed-point numbers
1078 */ 1081 */
1079 static unsigned long count_active_tasks(void) 1082 static unsigned long count_active_tasks(void)
1080 { 1083 {
1081 return nr_active() * FIXED_1; 1084 return nr_active() * FIXED_1;
1082 } 1085 }
1083 1086
1084 /* 1087 /*
1085 * Hmm.. Changed this, as the GNU make sources (load.c) seems to 1088 * Hmm.. Changed this, as the GNU make sources (load.c) seems to
1086 * imply that avenrun[] is the standard name for this kind of thing. 1089 * imply that avenrun[] is the standard name for this kind of thing.
1087 * Nothing else seems to be standardized: the fractional size etc 1090 * Nothing else seems to be standardized: the fractional size etc
1088 * all seem to differ on different machines. 1091 * all seem to differ on different machines.
1089 * 1092 *
1090 * Requires xtime_lock to access. 1093 * Requires xtime_lock to access.
1091 */ 1094 */
1092 unsigned long avenrun[3]; 1095 unsigned long avenrun[3];
1093 1096
1094 EXPORT_SYMBOL(avenrun); 1097 EXPORT_SYMBOL(avenrun);
1095 1098
1096 /* 1099 /*
1097 * calc_load - given tick count, update the avenrun load estimates. 1100 * calc_load - given tick count, update the avenrun load estimates.
1098 * This is called while holding a write_lock on xtime_lock. 1101 * This is called while holding a write_lock on xtime_lock.
1099 */ 1102 */
1100 static inline void calc_load(unsigned long ticks) 1103 static inline void calc_load(unsigned long ticks)
1101 { 1104 {
1102 unsigned long active_tasks; /* fixed-point */ 1105 unsigned long active_tasks; /* fixed-point */
1103 static int count = LOAD_FREQ; 1106 static int count = LOAD_FREQ;
1104 1107
1105 count -= ticks; 1108 count -= ticks;
1106 if (unlikely(count < 0)) { 1109 if (unlikely(count < 0)) {
1107 active_tasks = count_active_tasks(); 1110 active_tasks = count_active_tasks();
1108 do { 1111 do {
1109 CALC_LOAD(avenrun[0], EXP_1, active_tasks); 1112 CALC_LOAD(avenrun[0], EXP_1, active_tasks);
1110 CALC_LOAD(avenrun[1], EXP_5, active_tasks); 1113 CALC_LOAD(avenrun[1], EXP_5, active_tasks);
1111 CALC_LOAD(avenrun[2], EXP_15, active_tasks); 1114 CALC_LOAD(avenrun[2], EXP_15, active_tasks);
1112 count += LOAD_FREQ; 1115 count += LOAD_FREQ;
1113 } while (count < 0); 1116 } while (count < 0);
1114 } 1117 }
1115 } 1118 }
1116 1119
1117 /* 1120 /*
1118 * This function runs timers and the timer-tq in bottom half context. 1121 * This function runs timers and the timer-tq in bottom half context.
1119 */ 1122 */
1120 static void run_timer_softirq(struct softirq_action *h) 1123 static void run_timer_softirq(struct softirq_action *h)
1121 { 1124 {
1122 struct tvec_base *base = __get_cpu_var(tvec_bases); 1125 struct tvec_base *base = __get_cpu_var(tvec_bases);
1123 1126
1124 hrtimer_run_pending(); 1127 hrtimer_run_pending();
1125 1128
1126 if (time_after_eq(jiffies, base->timer_jiffies)) 1129 if (time_after_eq(jiffies, base->timer_jiffies))
1127 __run_timers(base); 1130 __run_timers(base);
1128 } 1131 }
1129 1132
1130 /* 1133 /*
1131 * Called by the local, per-CPU timer interrupt on SMP. 1134 * Called by the local, per-CPU timer interrupt on SMP.
1132 */ 1135 */
1133 void run_local_timers(void) 1136 void run_local_timers(void)
1134 { 1137 {
1135 hrtimer_run_queues(); 1138 hrtimer_run_queues();
1136 raise_softirq(TIMER_SOFTIRQ); 1139 raise_softirq(TIMER_SOFTIRQ);
1137 softlockup_tick(); 1140 softlockup_tick();
1138 } 1141 }
1139 1142
1140 /* 1143 /*
1141 * Called by the timer interrupt. xtime_lock must already be taken 1144 * Called by the timer interrupt. xtime_lock must already be taken
1142 * by the timer IRQ! 1145 * by the timer IRQ!
1143 */ 1146 */
1144 static inline void update_times(unsigned long ticks) 1147 static inline void update_times(unsigned long ticks)
1145 { 1148 {
1146 update_wall_time(); 1149 update_wall_time();
1147 calc_load(ticks); 1150 calc_load(ticks);
1148 } 1151 }
1149 1152
1150 /* 1153 /*
1151 * The 64-bit jiffies value is not atomic - you MUST NOT read it 1154 * The 64-bit jiffies value is not atomic - you MUST NOT read it
1152 * without sampling the sequence number in xtime_lock. 1155 * without sampling the sequence number in xtime_lock.
1153 * jiffies is defined in the linker script... 1156 * jiffies is defined in the linker script...
1154 */ 1157 */
1155 1158
1156 void do_timer(unsigned long ticks) 1159 void do_timer(unsigned long ticks)
1157 { 1160 {
1158 jiffies_64 += ticks; 1161 jiffies_64 += ticks;
1159 update_times(ticks); 1162 update_times(ticks);
1160 } 1163 }
1161 1164
1162 #ifdef __ARCH_WANT_SYS_ALARM 1165 #ifdef __ARCH_WANT_SYS_ALARM
1163 1166
1164 /* 1167 /*
1165 * For backwards compatibility? This can be done in libc so Alpha 1168 * For backwards compatibility? This can be done in libc so Alpha
1166 * and all newer ports shouldn't need it. 1169 * and all newer ports shouldn't need it.
1167 */ 1170 */
1168 SYSCALL_DEFINE1(alarm, unsigned int, seconds) 1171 SYSCALL_DEFINE1(alarm, unsigned int, seconds)
1169 { 1172 {
1170 return alarm_setitimer(seconds); 1173 return alarm_setitimer(seconds);
1171 } 1174 }
1172 1175
1173 #endif 1176 #endif
1174 1177
1175 #ifndef __alpha__ 1178 #ifndef __alpha__
1176 1179
1177 /* 1180 /*
1178 * The Alpha uses getxpid, getxuid, and getxgid instead. Maybe this 1181 * The Alpha uses getxpid, getxuid, and getxgid instead. Maybe this
1179 * should be moved into arch/i386 instead? 1182 * should be moved into arch/i386 instead?
1180 */ 1183 */
1181 1184
1182 /** 1185 /**
1183 * sys_getpid - return the thread group id of the current process 1186 * sys_getpid - return the thread group id of the current process
1184 * 1187 *
1185 * Note, despite the name, this returns the tgid not the pid. The tgid and 1188 * Note, despite the name, this returns the tgid not the pid. The tgid and
1186 * the pid are identical unless CLONE_THREAD was specified on clone() in 1189 * the pid are identical unless CLONE_THREAD was specified on clone() in
1187 * which case the tgid is the same in all threads of the same group. 1190 * which case the tgid is the same in all threads of the same group.
1188 * 1191 *
1189 * This is SMP safe as current->tgid does not change. 1192 * This is SMP safe as current->tgid does not change.
1190 */ 1193 */
1191 SYSCALL_DEFINE0(getpid) 1194 SYSCALL_DEFINE0(getpid)
1192 { 1195 {
1193 return task_tgid_vnr(current); 1196 return task_tgid_vnr(current);
1194 } 1197 }
1195 1198
1196 /* 1199 /*
1197 * Accessing ->real_parent is not SMP-safe, it could 1200 * Accessing ->real_parent is not SMP-safe, it could
1198 * change from under us. However, we can use a stale 1201 * change from under us. However, we can use a stale
1199 * value of ->real_parent under rcu_read_lock(), see 1202 * value of ->real_parent under rcu_read_lock(), see
1200 * release_task()->call_rcu(delayed_put_task_struct). 1203 * release_task()->call_rcu(delayed_put_task_struct).
1201 */ 1204 */
1202 SYSCALL_DEFINE0(getppid) 1205 SYSCALL_DEFINE0(getppid)
1203 { 1206 {
1204 int pid; 1207 int pid;
1205 1208
1206 rcu_read_lock(); 1209 rcu_read_lock();
1207 pid = task_tgid_vnr(current->real_parent); 1210 pid = task_tgid_vnr(current->real_parent);
1208 rcu_read_unlock(); 1211 rcu_read_unlock();
1209 1212
1210 return pid; 1213 return pid;
1211 } 1214 }
1212 1215
1213 SYSCALL_DEFINE0(getuid) 1216 SYSCALL_DEFINE0(getuid)
1214 { 1217 {
1215 /* Only we change this so SMP safe */ 1218 /* Only we change this so SMP safe */
1216 return current_uid(); 1219 return current_uid();
1217 } 1220 }
1218 1221
1219 SYSCALL_DEFINE0(geteuid) 1222 SYSCALL_DEFINE0(geteuid)
1220 { 1223 {
1221 /* Only we change this so SMP safe */ 1224 /* Only we change this so SMP safe */
1222 return current_euid(); 1225 return current_euid();
1223 } 1226 }
1224 1227
1225 SYSCALL_DEFINE0(getgid) 1228 SYSCALL_DEFINE0(getgid)
1226 { 1229 {
1227 /* Only we change this so SMP safe */ 1230 /* Only we change this so SMP safe */
1228 return current_gid(); 1231 return current_gid();
1229 } 1232 }
1230 1233
1231 SYSCALL_DEFINE0(getegid) 1234 SYSCALL_DEFINE0(getegid)
1232 { 1235 {
1233 /* Only we change this so SMP safe */ 1236 /* Only we change this so SMP safe */
1234 return current_egid(); 1237 return current_egid();
1235 } 1238 }
1236 1239
1237 #endif 1240 #endif
1238 1241
1239 static void process_timeout(unsigned long __data) 1242 static void process_timeout(unsigned long __data)
1240 { 1243 {
1241 wake_up_process((struct task_struct *)__data); 1244 wake_up_process((struct task_struct *)__data);
1242 } 1245 }
1243 1246
1244 /** 1247 /**
1245 * schedule_timeout - sleep until timeout 1248 * schedule_timeout - sleep until timeout
1246 * @timeout: timeout value in jiffies 1249 * @timeout: timeout value in jiffies
1247 * 1250 *
1248 * Make the current task sleep until @timeout jiffies have 1251 * Make the current task sleep until @timeout jiffies have
1249 * elapsed. The routine will return immediately unless 1252 * elapsed. The routine will return immediately unless
1250 * the current task state has been set (see set_current_state()). 1253 * the current task state has been set (see set_current_state()).
1251 * 1254 *
1252 * You can set the task state as follows - 1255 * You can set the task state as follows -
1253 * 1256 *
1254 * %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to 1257 * %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to
1255 * pass before the routine returns. The routine will return 0 1258 * pass before the routine returns. The routine will return 0
1256 * 1259 *
1257 * %TASK_INTERRUPTIBLE - the routine may return early if a signal is 1260 * %TASK_INTERRUPTIBLE - the routine may return early if a signal is
1258 * delivered to the current task. In this case the remaining time 1261 * delivered to the current task. In this case the remaining time
1259 * in jiffies will be returned, or 0 if the timer expired in time 1262 * in jiffies will be returned, or 0 if the timer expired in time
1260 * 1263 *
1261 * The current task state is guaranteed to be TASK_RUNNING when this 1264 * The current task state is guaranteed to be TASK_RUNNING when this
1262 * routine returns. 1265 * routine returns.
1263 * 1266 *
1264 * Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule 1267 * Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule
1265 * the CPU away without a bound on the timeout. In this case the return 1268 * the CPU away without a bound on the timeout. In this case the return
1266 * value will be %MAX_SCHEDULE_TIMEOUT. 1269 * value will be %MAX_SCHEDULE_TIMEOUT.
1267 * 1270 *
1268 * In all cases the return value is guaranteed to be non-negative. 1271 * In all cases the return value is guaranteed to be non-negative.
1269 */ 1272 */
1270 signed long __sched schedule_timeout(signed long timeout) 1273 signed long __sched schedule_timeout(signed long timeout)
1271 { 1274 {
1272 struct timer_list timer; 1275 struct timer_list timer;
1273 unsigned long expire; 1276 unsigned long expire;
1274 1277
1275 switch (timeout) 1278 switch (timeout)
1276 { 1279 {
1277 case MAX_SCHEDULE_TIMEOUT: 1280 case MAX_SCHEDULE_TIMEOUT:
1278 /* 1281 /*
1279 * These two special cases are useful to be comfortable 1282 * These two special cases are useful to be comfortable
1280 * in the caller. Nothing more. We could take 1283 * in the caller. Nothing more. We could take
1281 * MAX_SCHEDULE_TIMEOUT from one of the negative value 1284 * MAX_SCHEDULE_TIMEOUT from one of the negative value
1282 * but I' d like to return a valid offset (>=0) to allow 1285 * but I' d like to return a valid offset (>=0) to allow
1283 * the caller to do everything it want with the retval. 1286 * the caller to do everything it want with the retval.
1284 */ 1287 */
1285 schedule(); 1288 schedule();
1286 goto out; 1289 goto out;
1287 default: 1290 default:
1288 /* 1291 /*
1289 * Another bit of PARANOID. Note that the retval will be 1292 * Another bit of PARANOID. Note that the retval will be
1290 * 0 since no piece of kernel is supposed to do a check 1293 * 0 since no piece of kernel is supposed to do a check
1291 * for a negative retval of schedule_timeout() (since it 1294 * for a negative retval of schedule_timeout() (since it
1292 * should never happens anyway). You just have the printk() 1295 * should never happens anyway). You just have the printk()
1293 * that will tell you if something is gone wrong and where. 1296 * that will tell you if something is gone wrong and where.
1294 */ 1297 */
1295 if (timeout < 0) { 1298 if (timeout < 0) {
1296 printk(KERN_ERR "schedule_timeout: wrong timeout " 1299 printk(KERN_ERR "schedule_timeout: wrong timeout "
1297 "value %lx\n", timeout); 1300 "value %lx\n", timeout);
1298 dump_stack(); 1301 dump_stack();
1299 current->state = TASK_RUNNING; 1302 current->state = TASK_RUNNING;
1300 goto out; 1303 goto out;
1301 } 1304 }
1302 } 1305 }
1303 1306
1304 expire = timeout + jiffies; 1307 expire = timeout + jiffies;
1305 1308
1306 setup_timer_on_stack(&timer, process_timeout, (unsigned long)current); 1309 setup_timer_on_stack(&timer, process_timeout, (unsigned long)current);
1307 __mod_timer(&timer, expire, false); 1310 __mod_timer(&timer, expire, false);
1308 schedule(); 1311 schedule();
1309 del_singleshot_timer_sync(&timer); 1312 del_singleshot_timer_sync(&timer);
1310 1313
1311 /* Remove the timer from the object tracker */ 1314 /* Remove the timer from the object tracker */
1312 destroy_timer_on_stack(&timer); 1315 destroy_timer_on_stack(&timer);
1313 1316
1314 timeout = expire - jiffies; 1317 timeout = expire - jiffies;
1315 1318
1316 out: 1319 out:
1317 return timeout < 0 ? 0 : timeout; 1320 return timeout < 0 ? 0 : timeout;
1318 } 1321 }
1319 EXPORT_SYMBOL(schedule_timeout); 1322 EXPORT_SYMBOL(schedule_timeout);
1320 1323
1321 /* 1324 /*
1322 * We can use __set_current_state() here because schedule_timeout() calls 1325 * We can use __set_current_state() here because schedule_timeout() calls
1323 * schedule() unconditionally. 1326 * schedule() unconditionally.
1324 */ 1327 */
1325 signed long __sched schedule_timeout_interruptible(signed long timeout) 1328 signed long __sched schedule_timeout_interruptible(signed long timeout)
1326 { 1329 {
1327 __set_current_state(TASK_INTERRUPTIBLE); 1330 __set_current_state(TASK_INTERRUPTIBLE);
1328 return schedule_timeout(timeout); 1331 return schedule_timeout(timeout);
1329 } 1332 }
1330 EXPORT_SYMBOL(schedule_timeout_interruptible); 1333 EXPORT_SYMBOL(schedule_timeout_interruptible);
1331 1334
1332 signed long __sched schedule_timeout_killable(signed long timeout) 1335 signed long __sched schedule_timeout_killable(signed long timeout)
1333 { 1336 {
1334 __set_current_state(TASK_KILLABLE); 1337 __set_current_state(TASK_KILLABLE);
1335 return schedule_timeout(timeout); 1338 return schedule_timeout(timeout);
1336 } 1339 }
1337 EXPORT_SYMBOL(schedule_timeout_killable); 1340 EXPORT_SYMBOL(schedule_timeout_killable);
1338 1341
1339 signed long __sched schedule_timeout_uninterruptible(signed long timeout) 1342 signed long __sched schedule_timeout_uninterruptible(signed long timeout)
1340 { 1343 {
1341 __set_current_state(TASK_UNINTERRUPTIBLE); 1344 __set_current_state(TASK_UNINTERRUPTIBLE);
1342 return schedule_timeout(timeout); 1345 return schedule_timeout(timeout);
1343 } 1346 }
1344 EXPORT_SYMBOL(schedule_timeout_uninterruptible); 1347 EXPORT_SYMBOL(schedule_timeout_uninterruptible);
1345 1348
1346 /* Thread ID - the internal kernel "pid" */ 1349 /* Thread ID - the internal kernel "pid" */
1347 SYSCALL_DEFINE0(gettid) 1350 SYSCALL_DEFINE0(gettid)
1348 { 1351 {
1349 return task_pid_vnr(current); 1352 return task_pid_vnr(current);
1350 } 1353 }
1351 1354
1352 /** 1355 /**
1353 * do_sysinfo - fill in sysinfo struct 1356 * do_sysinfo - fill in sysinfo struct
1354 * @info: pointer to buffer to fill 1357 * @info: pointer to buffer to fill
1355 */ 1358 */
1356 int do_sysinfo(struct sysinfo *info) 1359 int do_sysinfo(struct sysinfo *info)
1357 { 1360 {
1358 unsigned long mem_total, sav_total; 1361 unsigned long mem_total, sav_total;
1359 unsigned int mem_unit, bitcount; 1362 unsigned int mem_unit, bitcount;
1360 unsigned long seq; 1363 unsigned long seq;
1361 1364
1362 memset(info, 0, sizeof(struct sysinfo)); 1365 memset(info, 0, sizeof(struct sysinfo));
1363 1366
1364 do { 1367 do {
1365 struct timespec tp; 1368 struct timespec tp;
1366 seq = read_seqbegin(&xtime_lock); 1369 seq = read_seqbegin(&xtime_lock);
1367 1370
1368 /* 1371 /*
1369 * This is annoying. The below is the same thing 1372 * This is annoying. The below is the same thing
1370 * posix_get_clock_monotonic() does, but it wants to 1373 * posix_get_clock_monotonic() does, but it wants to
1371 * take the lock which we want to cover the loads stuff 1374 * take the lock which we want to cover the loads stuff
1372 * too. 1375 * too.
1373 */ 1376 */
1374 1377
1375 getnstimeofday(&tp); 1378 getnstimeofday(&tp);
1376 tp.tv_sec += wall_to_monotonic.tv_sec; 1379 tp.tv_sec += wall_to_monotonic.tv_sec;
1377 tp.tv_nsec += wall_to_monotonic.tv_nsec; 1380 tp.tv_nsec += wall_to_monotonic.tv_nsec;
1378 monotonic_to_bootbased(&tp); 1381 monotonic_to_bootbased(&tp);
1379 if (tp.tv_nsec - NSEC_PER_SEC >= 0) { 1382 if (tp.tv_nsec - NSEC_PER_SEC >= 0) {
1380 tp.tv_nsec = tp.tv_nsec - NSEC_PER_SEC; 1383 tp.tv_nsec = tp.tv_nsec - NSEC_PER_SEC;
1381 tp.tv_sec++; 1384 tp.tv_sec++;
1382 } 1385 }
1383 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); 1386 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
1384 1387
1385 info->loads[0] = avenrun[0] << (SI_LOAD_SHIFT - FSHIFT); 1388 info->loads[0] = avenrun[0] << (SI_LOAD_SHIFT - FSHIFT);
1386 info->loads[1] = avenrun[1] << (SI_LOAD_SHIFT - FSHIFT); 1389 info->loads[1] = avenrun[1] << (SI_LOAD_SHIFT - FSHIFT);
1387 info->loads[2] = avenrun[2] << (SI_LOAD_SHIFT - FSHIFT); 1390 info->loads[2] = avenrun[2] << (SI_LOAD_SHIFT - FSHIFT);
1388 1391
1389 info->procs = nr_threads; 1392 info->procs = nr_threads;
1390 } while (read_seqretry(&xtime_lock, seq)); 1393 } while (read_seqretry(&xtime_lock, seq));
1391 1394
1392 si_meminfo(info); 1395 si_meminfo(info);
1393 si_swapinfo(info); 1396 si_swapinfo(info);
1394 1397
1395 /* 1398 /*
1396 * If the sum of all the available memory (i.e. ram + swap) 1399 * If the sum of all the available memory (i.e. ram + swap)
1397 * is less than can be stored in a 32 bit unsigned long then 1400 * is less than can be stored in a 32 bit unsigned long then
1398 * we can be binary compatible with 2.2.x kernels. If not, 1401 * we can be binary compatible with 2.2.x kernels. If not,
1399 * well, in that case 2.2.x was broken anyways... 1402 * well, in that case 2.2.x was broken anyways...
1400 * 1403 *
1401 * -Erik Andersen <andersee@debian.org> 1404 * -Erik Andersen <andersee@debian.org>
1402 */ 1405 */
1403 1406
1404 mem_total = info->totalram + info->totalswap; 1407 mem_total = info->totalram + info->totalswap;
1405 if (mem_total < info->totalram || mem_total < info->totalswap) 1408 if (mem_total < info->totalram || mem_total < info->totalswap)
1406 goto out; 1409 goto out;
1407 bitcount = 0; 1410 bitcount = 0;
1408 mem_unit = info->mem_unit; 1411 mem_unit = info->mem_unit;
1409 while (mem_unit > 1) { 1412 while (mem_unit > 1) {
1410 bitcount++; 1413 bitcount++;
1411 mem_unit >>= 1; 1414 mem_unit >>= 1;
1412 sav_total = mem_total; 1415 sav_total = mem_total;
1413 mem_total <<= 1; 1416 mem_total <<= 1;
1414 if (mem_total < sav_total) 1417 if (mem_total < sav_total)
1415 goto out; 1418 goto out;
1416 } 1419 }
1417 1420
1418 /* 1421 /*
1419 * If mem_total did not overflow, multiply all memory values by 1422 * If mem_total did not overflow, multiply all memory values by
1420 * info->mem_unit and set it to 1. This leaves things compatible 1423 * info->mem_unit and set it to 1. This leaves things compatible
1421 * with 2.2.x, and also retains compatibility with earlier 2.4.x 1424 * with 2.2.x, and also retains compatibility with earlier 2.4.x
1422 * kernels... 1425 * kernels...
1423 */ 1426 */
1424 1427
1425 info->mem_unit = 1; 1428 info->mem_unit = 1;
1426 info->totalram <<= bitcount; 1429 info->totalram <<= bitcount;
1427 info->freeram <<= bitcount; 1430 info->freeram <<= bitcount;
1428 info->sharedram <<= bitcount; 1431 info->sharedram <<= bitcount;
1429 info->bufferram <<= bitcount; 1432 info->bufferram <<= bitcount;
1430 info->totalswap <<= bitcount; 1433 info->totalswap <<= bitcount;
1431 info->freeswap <<= bitcount; 1434 info->freeswap <<= bitcount;
1432 info->totalhigh <<= bitcount; 1435 info->totalhigh <<= bitcount;
1433 info->freehigh <<= bitcount; 1436 info->freehigh <<= bitcount;
1434 1437
1435 out: 1438 out:
1436 return 0; 1439 return 0;
1437 } 1440 }
1438 1441
1439 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info) 1442 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
1440 { 1443 {
1441 struct sysinfo val; 1444 struct sysinfo val;
1442 1445
1443 do_sysinfo(&val); 1446 do_sysinfo(&val);
1444 1447
1445 if (copy_to_user(info, &val, sizeof(struct sysinfo))) 1448 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
1446 return -EFAULT; 1449 return -EFAULT;
1447 1450
1448 return 0; 1451 return 0;
1449 } 1452 }
1450 1453
1451 static int __cpuinit init_timers_cpu(int cpu) 1454 static int __cpuinit init_timers_cpu(int cpu)
1452 { 1455 {
1453 int j; 1456 int j;
1454 struct tvec_base *base; 1457 struct tvec_base *base;
1455 static char __cpuinitdata tvec_base_done[NR_CPUS]; 1458 static char __cpuinitdata tvec_base_done[NR_CPUS];
1456 1459
1457 if (!tvec_base_done[cpu]) { 1460 if (!tvec_base_done[cpu]) {
1458 static char boot_done; 1461 static char boot_done;
1459 1462
1460 if (boot_done) { 1463 if (boot_done) {
1461 /* 1464 /*
1462 * The APs use this path later in boot 1465 * The APs use this path later in boot
1463 */ 1466 */
1464 base = kmalloc_node(sizeof(*base), 1467 base = kmalloc_node(sizeof(*base),
1465 GFP_KERNEL | __GFP_ZERO, 1468 GFP_KERNEL | __GFP_ZERO,
1466 cpu_to_node(cpu)); 1469 cpu_to_node(cpu));
1467 if (!base) 1470 if (!base)
1468 return -ENOMEM; 1471 return -ENOMEM;
1469 1472
1470 /* Make sure that tvec_base is 2 byte aligned */ 1473 /* Make sure that tvec_base is 2 byte aligned */
1471 if (tbase_get_deferrable(base)) { 1474 if (tbase_get_deferrable(base)) {
1472 WARN_ON(1); 1475 WARN_ON(1);
1473 kfree(base); 1476 kfree(base);
1474 return -ENOMEM; 1477 return -ENOMEM;
1475 } 1478 }
1476 per_cpu(tvec_bases, cpu) = base; 1479 per_cpu(tvec_bases, cpu) = base;
1477 } else { 1480 } else {
1478 /* 1481 /*
1479 * This is for the boot CPU - we use compile-time 1482 * This is for the boot CPU - we use compile-time
1480 * static initialisation because per-cpu memory isn't 1483 * static initialisation because per-cpu memory isn't
1481 * ready yet and because the memory allocators are not 1484 * ready yet and because the memory allocators are not
1482 * initialised either. 1485 * initialised either.
1483 */ 1486 */
1484 boot_done = 1; 1487 boot_done = 1;
1485 base = &boot_tvec_bases; 1488 base = &boot_tvec_bases;
1486 } 1489 }
1487 tvec_base_done[cpu] = 1; 1490 tvec_base_done[cpu] = 1;
1488 } else { 1491 } else {
1489 base = per_cpu(tvec_bases, cpu); 1492 base = per_cpu(tvec_bases, cpu);
1490 } 1493 }
1491 1494
1492 spin_lock_init(&base->lock); 1495 spin_lock_init(&base->lock);
1493 1496
1494 for (j = 0; j < TVN_SIZE; j++) { 1497 for (j = 0; j < TVN_SIZE; j++) {
1495 INIT_LIST_HEAD(base->tv5.vec + j); 1498 INIT_LIST_HEAD(base->tv5.vec + j);
1496 INIT_LIST_HEAD(base->tv4.vec + j); 1499 INIT_LIST_HEAD(base->tv4.vec + j);
1497 INIT_LIST_HEAD(base->tv3.vec + j); 1500 INIT_LIST_HEAD(base->tv3.vec + j);
1498 INIT_LIST_HEAD(base->tv2.vec + j); 1501 INIT_LIST_HEAD(base->tv2.vec + j);
1499 } 1502 }
1500 for (j = 0; j < TVR_SIZE; j++) 1503 for (j = 0; j < TVR_SIZE; j++)
1501 INIT_LIST_HEAD(base->tv1.vec + j); 1504 INIT_LIST_HEAD(base->tv1.vec + j);
1502 1505
1503 base->timer_jiffies = jiffies; 1506 base->timer_jiffies = jiffies;
1504 return 0; 1507 return 0;
1505 } 1508 }
1506 1509
1507 #ifdef CONFIG_HOTPLUG_CPU 1510 #ifdef CONFIG_HOTPLUG_CPU
1508 static void migrate_timer_list(struct tvec_base *new_base, struct list_head *head) 1511 static void migrate_timer_list(struct tvec_base *new_base, struct list_head *head)
1509 { 1512 {
1510 struct timer_list *timer; 1513 struct timer_list *timer;
1511 1514
1512 while (!list_empty(head)) { 1515 while (!list_empty(head)) {
1513 timer = list_first_entry(head, struct timer_list, entry); 1516 timer = list_first_entry(head, struct timer_list, entry);
1514 detach_timer(timer, 0); 1517 detach_timer(timer, 0);
1515 timer_set_base(timer, new_base); 1518 timer_set_base(timer, new_base);
1516 internal_add_timer(new_base, timer); 1519 internal_add_timer(new_base, timer);
1517 } 1520 }
1518 } 1521 }
1519 1522
1520 static void __cpuinit migrate_timers(int cpu) 1523 static void __cpuinit migrate_timers(int cpu)
1521 { 1524 {
1522 struct tvec_base *old_base; 1525 struct tvec_base *old_base;
1523 struct tvec_base *new_base; 1526 struct tvec_base *new_base;
1524 int i; 1527 int i;
1525 1528
1526 BUG_ON(cpu_online(cpu)); 1529 BUG_ON(cpu_online(cpu));
1527 old_base = per_cpu(tvec_bases, cpu); 1530 old_base = per_cpu(tvec_bases, cpu);
1528 new_base = get_cpu_var(tvec_bases); 1531 new_base = get_cpu_var(tvec_bases);
1529 /* 1532 /*
1530 * The caller is globally serialized and nobody else 1533 * The caller is globally serialized and nobody else
1531 * takes two locks at once, deadlock is not possible. 1534 * takes two locks at once, deadlock is not possible.
1532 */ 1535 */
1533 spin_lock_irq(&new_base->lock); 1536 spin_lock_irq(&new_base->lock);
1534 spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING); 1537 spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING);
1535 1538
1536 BUG_ON(old_base->running_timer); 1539 BUG_ON(old_base->running_timer);
1537 1540
1538 for (i = 0; i < TVR_SIZE; i++) 1541 for (i = 0; i < TVR_SIZE; i++)
1539 migrate_timer_list(new_base, old_base->tv1.vec + i); 1542 migrate_timer_list(new_base, old_base->tv1.vec + i);
1540 for (i = 0; i < TVN_SIZE; i++) { 1543 for (i = 0; i < TVN_SIZE; i++) {
1541 migrate_timer_list(new_base, old_base->tv2.vec + i); 1544 migrate_timer_list(new_base, old_base->tv2.vec + i);
1542 migrate_timer_list(new_base, old_base->tv3.vec + i); 1545 migrate_timer_list(new_base, old_base->tv3.vec + i);
1543 migrate_timer_list(new_base, old_base->tv4.vec + i); 1546 migrate_timer_list(new_base, old_base->tv4.vec + i);
1544 migrate_timer_list(new_base, old_base->tv5.vec + i); 1547 migrate_timer_list(new_base, old_base->tv5.vec + i);
1545 } 1548 }
1546 1549
1547 spin_unlock(&old_base->lock); 1550 spin_unlock(&old_base->lock);
1548 spin_unlock_irq(&new_base->lock); 1551 spin_unlock_irq(&new_base->lock);
1549 put_cpu_var(tvec_bases); 1552 put_cpu_var(tvec_bases);
1550 } 1553 }
1551 #endif /* CONFIG_HOTPLUG_CPU */ 1554 #endif /* CONFIG_HOTPLUG_CPU */
1552 1555
1553 static int __cpuinit timer_cpu_notify(struct notifier_block *self, 1556 static int __cpuinit timer_cpu_notify(struct notifier_block *self,
1554 unsigned long action, void *hcpu) 1557 unsigned long action, void *hcpu)
1555 { 1558 {
1556 long cpu = (long)hcpu; 1559 long cpu = (long)hcpu;
1557 switch(action) { 1560 switch(action) {
1558 case CPU_UP_PREPARE: 1561 case CPU_UP_PREPARE:
1559 case CPU_UP_PREPARE_FROZEN: 1562 case CPU_UP_PREPARE_FROZEN:
1560 if (init_timers_cpu(cpu) < 0) 1563 if (init_timers_cpu(cpu) < 0)
1561 return NOTIFY_BAD; 1564 return NOTIFY_BAD;
1562 break; 1565 break;
1563 #ifdef CONFIG_HOTPLUG_CPU 1566 #ifdef CONFIG_HOTPLUG_CPU
1564 case CPU_DEAD: 1567 case CPU_DEAD:
1565 case CPU_DEAD_FROZEN: 1568 case CPU_DEAD_FROZEN:
1566 migrate_timers(cpu); 1569 migrate_timers(cpu);
1567 break; 1570 break;
1568 #endif 1571 #endif
1569 default: 1572 default:
1570 break; 1573 break;
1571 } 1574 }
1572 return NOTIFY_OK; 1575 return NOTIFY_OK;
1573 } 1576 }
1574 1577
1575 static struct notifier_block __cpuinitdata timers_nb = { 1578 static struct notifier_block __cpuinitdata timers_nb = {
1576 .notifier_call = timer_cpu_notify, 1579 .notifier_call = timer_cpu_notify,
1577 }; 1580 };
1578 1581
1579 1582
1580 void __init init_timers(void) 1583 void __init init_timers(void)
1581 { 1584 {
1582 int err = timer_cpu_notify(&timers_nb, (unsigned long)CPU_UP_PREPARE, 1585 int err = timer_cpu_notify(&timers_nb, (unsigned long)CPU_UP_PREPARE,
1583 (void *)(long)smp_processor_id()); 1586 (void *)(long)smp_processor_id());
1584 1587
1585 init_timer_stats(); 1588 init_timer_stats();
1586 1589
1587 BUG_ON(err == NOTIFY_BAD); 1590 BUG_ON(err == NOTIFY_BAD);
1588 register_cpu_notifier(&timers_nb); 1591 register_cpu_notifier(&timers_nb);
1589 open_softirq(TIMER_SOFTIRQ, run_timer_softirq); 1592 open_softirq(TIMER_SOFTIRQ, run_timer_softirq);
1590 } 1593 }
1591 1594
1592 /** 1595 /**
1593 * msleep - sleep safely even with waitqueue interruptions 1596 * msleep - sleep safely even with waitqueue interruptions
1594 * @msecs: Time in milliseconds to sleep for 1597 * @msecs: Time in milliseconds to sleep for
1595 */ 1598 */
1596 void msleep(unsigned int msecs) 1599 void msleep(unsigned int msecs)
1597 { 1600 {
1598 unsigned long timeout = msecs_to_jiffies(msecs) + 1; 1601 unsigned long timeout = msecs_to_jiffies(msecs) + 1;
1599 1602
1600 while (timeout) 1603 while (timeout)
1601 timeout = schedule_timeout_uninterruptible(timeout); 1604 timeout = schedule_timeout_uninterruptible(timeout);
1602 } 1605 }
1603 1606
1604 EXPORT_SYMBOL(msleep); 1607 EXPORT_SYMBOL(msleep);
1605 1608
1606 /** 1609 /**
1607 * msleep_interruptible - sleep waiting for signals 1610 * msleep_interruptible - sleep waiting for signals
1608 * @msecs: Time in milliseconds to sleep for 1611 * @msecs: Time in milliseconds to sleep for
1609 */ 1612 */
1610 unsigned long msleep_interruptible(unsigned int msecs) 1613 unsigned long msleep_interruptible(unsigned int msecs)
1611 { 1614 {
1612 unsigned long timeout = msecs_to_jiffies(msecs) + 1; 1615 unsigned long timeout = msecs_to_jiffies(msecs) + 1;
1613 1616
1614 while (timeout && !signal_pending(current)) 1617 while (timeout && !signal_pending(current))
1615 timeout = schedule_timeout_interruptible(timeout); 1618 timeout = schedule_timeout_interruptible(timeout);
1616 return jiffies_to_msecs(timeout); 1619 return jiffies_to_msecs(timeout);
1617 } 1620 }
1618 1621
1619 EXPORT_SYMBOL(msleep_interruptible); 1622 EXPORT_SYMBOL(msleep_interruptible);
1620 1623