Loading...
1/*
2 * linux/kernel/irq/spurious.c
3 *
4 * Copyright (C) 1992, 1998-2004 Linus Torvalds, Ingo Molnar
5 *
6 * This file contains spurious interrupt handling.
7 */
8
9#include <linux/jiffies.h>
10#include <linux/irq.h>
11#include <linux/module.h>
12#include <linux/kallsyms.h>
13#include <linux/interrupt.h>
14#include <linux/moduleparam.h>
15#include <linux/timer.h>
16
17#include "internals.h"
18
19static int irqfixup __read_mostly;
20
21#define POLL_SPURIOUS_IRQ_INTERVAL (HZ/10)
22static void poll_spurious_irqs(unsigned long dummy);
23static DEFINE_TIMER(poll_spurious_irq_timer, poll_spurious_irqs, 0, 0);
24static int irq_poll_cpu;
25static atomic_t irq_poll_active;
26
27/*
28 * We wait here for a poller to finish.
29 *
30 * If the poll runs on this CPU, then we yell loudly and return
31 * false. That will leave the interrupt line disabled in the worst
32 * case, but it should never happen.
33 *
34 * We wait until the poller is done and then recheck disabled and
35 * action (about to be disabled). Only if it's still active, we return
36 * true and let the handler run.
37 */
38bool irq_wait_for_poll(struct irq_desc *desc)
39{
40 if (WARN_ONCE(irq_poll_cpu == smp_processor_id(),
41 "irq poll in progress on cpu %d for irq %d\n",
42 smp_processor_id(), desc->irq_data.irq))
43 return false;
44
45#ifdef CONFIG_SMP
46 do {
47 raw_spin_unlock(&desc->lock);
48 while (irqd_irq_inprogress(&desc->irq_data))
49 cpu_relax();
50 raw_spin_lock(&desc->lock);
51 } while (irqd_irq_inprogress(&desc->irq_data));
52 /* Might have been disabled in meantime */
53 return !irqd_irq_disabled(&desc->irq_data) && desc->action;
54#else
55 return false;
56#endif
57}
58
59
60/*
61 * Recovery handler for misrouted interrupts.
62 */
63static int try_one_irq(int irq, struct irq_desc *desc, bool force)
64{
65 irqreturn_t ret = IRQ_NONE;
66 struct irqaction *action;
67
68 raw_spin_lock(&desc->lock);
69
70 /* PER_CPU and nested thread interrupts are never polled */
71 if (irq_settings_is_per_cpu(desc) || irq_settings_is_nested_thread(desc))
72 goto out;
73
74 /*
75 * Do not poll disabled interrupts unless the spurious
76 * disabled poller asks explicitely.
77 */
78 if (irqd_irq_disabled(&desc->irq_data) && !force)
79 goto out;
80
81 /*
82 * All handlers must agree on IRQF_SHARED, so we test just the
83 * first. Check for action->next as well.
84 */
85 action = desc->action;
86 if (!action || !(action->flags & IRQF_SHARED) ||
87 (action->flags & __IRQF_TIMER) ||
88 (action->handler(irq, action->dev_id) == IRQ_HANDLED) ||
89 !action->next)
90 goto out;
91
92 /* Already running on another processor */
93 if (irqd_irq_inprogress(&desc->irq_data)) {
94 /*
95 * Already running: If it is shared get the other
96 * CPU to go looking for our mystery interrupt too
97 */
98 desc->istate |= IRQS_PENDING;
99 goto out;
100 }
101
102 /* Mark it poll in progress */
103 desc->istate |= IRQS_POLL_INPROGRESS;
104 do {
105 if (handle_irq_event(desc) == IRQ_HANDLED)
106 ret = IRQ_HANDLED;
107 action = desc->action;
108 } while ((desc->istate & IRQS_PENDING) && action);
109 desc->istate &= ~IRQS_POLL_INPROGRESS;
110out:
111 raw_spin_unlock(&desc->lock);
112 return ret == IRQ_HANDLED;
113}
114
115static int misrouted_irq(int irq)
116{
117 struct irq_desc *desc;
118 int i, ok = 0;
119
120 if (atomic_inc_return(&irq_poll_active) != 1)
121 goto out;
122
123 irq_poll_cpu = smp_processor_id();
124
125 for_each_irq_desc(i, desc) {
126 if (!i)
127 continue;
128
129 if (i == irq) /* Already tried */
130 continue;
131
132 if (try_one_irq(i, desc, false))
133 ok = 1;
134 }
135out:
136 atomic_dec(&irq_poll_active);
137 /* So the caller can adjust the irq error counts */
138 return ok;
139}
140
141static void poll_spurious_irqs(unsigned long dummy)
142{
143 struct irq_desc *desc;
144 int i;
145
146 if (atomic_inc_return(&irq_poll_active) != 1)
147 goto out;
148 irq_poll_cpu = smp_processor_id();
149
150 for_each_irq_desc(i, desc) {
151 unsigned int state;
152
153 if (!i)
154 continue;
155
156 /* Racy but it doesn't matter */
157 state = desc->istate;
158 barrier();
159 if (!(state & IRQS_SPURIOUS_DISABLED))
160 continue;
161
162 local_irq_disable();
163 try_one_irq(i, desc, true);
164 local_irq_enable();
165 }
166out:
167 atomic_dec(&irq_poll_active);
168 mod_timer(&poll_spurious_irq_timer,
169 jiffies + POLL_SPURIOUS_IRQ_INTERVAL);
170}
171
172static inline int bad_action_ret(irqreturn_t action_ret)
173{
174 if (likely(action_ret <= (IRQ_HANDLED | IRQ_WAKE_THREAD)))
175 return 0;
176 return 1;
177}
178
179/*
180 * If 99,900 of the previous 100,000 interrupts have not been handled
181 * then assume that the IRQ is stuck in some manner. Drop a diagnostic
182 * and try to turn the IRQ off.
183 *
184 * (The other 100-of-100,000 interrupts may have been a correctly
185 * functioning device sharing an IRQ with the failing one)
186 */
187static void
188__report_bad_irq(unsigned int irq, struct irq_desc *desc,
189 irqreturn_t action_ret)
190{
191 struct irqaction *action;
192 unsigned long flags;
193
194 if (bad_action_ret(action_ret)) {
195 printk(KERN_ERR "irq event %d: bogus return value %x\n",
196 irq, action_ret);
197 } else {
198 printk(KERN_ERR "irq %d: nobody cared (try booting with "
199 "the \"irqpoll\" option)\n", irq);
200 }
201 dump_stack();
202 printk(KERN_ERR "handlers:\n");
203
204 /*
205 * We need to take desc->lock here. note_interrupt() is called
206 * w/o desc->lock held, but IRQ_PROGRESS set. We might race
207 * with something else removing an action. It's ok to take
208 * desc->lock here. See synchronize_irq().
209 */
210 raw_spin_lock_irqsave(&desc->lock, flags);
211 action = desc->action;
212 while (action) {
213 printk(KERN_ERR "[<%p>] %pf", action->handler, action->handler);
214 if (action->thread_fn)
215 printk(KERN_CONT " threaded [<%p>] %pf",
216 action->thread_fn, action->thread_fn);
217 printk(KERN_CONT "\n");
218 action = action->next;
219 }
220 raw_spin_unlock_irqrestore(&desc->lock, flags);
221}
222
223static void
224report_bad_irq(unsigned int irq, struct irq_desc *desc, irqreturn_t action_ret)
225{
226 static int count = 100;
227
228 if (count > 0) {
229 count--;
230 __report_bad_irq(irq, desc, action_ret);
231 }
232}
233
234static inline int
235try_misrouted_irq(unsigned int irq, struct irq_desc *desc,
236 irqreturn_t action_ret)
237{
238 struct irqaction *action;
239
240 if (!irqfixup)
241 return 0;
242
243 /* We didn't actually handle the IRQ - see if it was misrouted? */
244 if (action_ret == IRQ_NONE)
245 return 1;
246
247 /*
248 * But for 'irqfixup == 2' we also do it for handled interrupts if
249 * they are marked as IRQF_IRQPOLL (or for irq zero, which is the
250 * traditional PC timer interrupt.. Legacy)
251 */
252 if (irqfixup < 2)
253 return 0;
254
255 if (!irq)
256 return 1;
257
258 /*
259 * Since we don't get the descriptor lock, "action" can
260 * change under us. We don't really care, but we don't
261 * want to follow a NULL pointer. So tell the compiler to
262 * just load it once by using a barrier.
263 */
264 action = desc->action;
265 barrier();
266 return action && (action->flags & IRQF_IRQPOLL);
267}
268
269void note_interrupt(unsigned int irq, struct irq_desc *desc,
270 irqreturn_t action_ret)
271{
272 if (desc->istate & IRQS_POLL_INPROGRESS)
273 return;
274
275 /* we get here again via the threaded handler */
276 if (action_ret == IRQ_WAKE_THREAD)
277 return;
278
279 if (bad_action_ret(action_ret)) {
280 report_bad_irq(irq, desc, action_ret);
281 return;
282 }
283
284 if (unlikely(action_ret == IRQ_NONE)) {
285 /*
286 * If we are seeing only the odd spurious IRQ caused by
287 * bus asynchronicity then don't eventually trigger an error,
288 * otherwise the counter becomes a doomsday timer for otherwise
289 * working systems
290 */
291 if (time_after(jiffies, desc->last_unhandled + HZ/10))
292 desc->irqs_unhandled = 1;
293 else
294 desc->irqs_unhandled++;
295 desc->last_unhandled = jiffies;
296 }
297
298 if (unlikely(try_misrouted_irq(irq, desc, action_ret))) {
299 int ok = misrouted_irq(irq);
300 if (action_ret == IRQ_NONE)
301 desc->irqs_unhandled -= ok;
302 }
303
304 desc->irq_count++;
305 if (likely(desc->irq_count < 100000))
306 return;
307
308 desc->irq_count = 0;
309 if (unlikely(desc->irqs_unhandled > 99900)) {
310 /*
311 * The interrupt is stuck
312 */
313 __report_bad_irq(irq, desc, action_ret);
314 /*
315 * Now kill the IRQ
316 */
317 printk(KERN_EMERG "Disabling IRQ #%d\n", irq);
318 desc->istate |= IRQS_SPURIOUS_DISABLED;
319 desc->depth++;
320 irq_disable(desc);
321
322 mod_timer(&poll_spurious_irq_timer,
323 jiffies + POLL_SPURIOUS_IRQ_INTERVAL);
324 }
325 desc->irqs_unhandled = 0;
326}
327
328bool noirqdebug __read_mostly;
329
330int noirqdebug_setup(char *str)
331{
332 noirqdebug = 1;
333 printk(KERN_INFO "IRQ lockup detection disabled\n");
334
335 return 1;
336}
337
338__setup("noirqdebug", noirqdebug_setup);
339module_param(noirqdebug, bool, 0644);
340MODULE_PARM_DESC(noirqdebug, "Disable irq lockup detection when true");
341
342static int __init irqfixup_setup(char *str)
343{
344 irqfixup = 1;
345 printk(KERN_WARNING "Misrouted IRQ fixup support enabled.\n");
346 printk(KERN_WARNING "This may impact system performance.\n");
347
348 return 1;
349}
350
351__setup("irqfixup", irqfixup_setup);
352module_param(irqfixup, int, 0644);
353
354static int __init irqpoll_setup(char *str)
355{
356 irqfixup = 2;
357 printk(KERN_WARNING "Misrouted IRQ fixup and polling support "
358 "enabled\n");
359 printk(KERN_WARNING "This may significantly impact system "
360 "performance\n");
361 return 1;
362}
363
364__setup("irqpoll", irqpoll_setup);
1/*
2 * linux/kernel/irq/spurious.c
3 *
4 * Copyright (C) 1992, 1998-2004 Linus Torvalds, Ingo Molnar
5 *
6 * This file contains spurious interrupt handling.
7 */
8
9#include <linux/jiffies.h>
10#include <linux/irq.h>
11#include <linux/module.h>
12#include <linux/kallsyms.h>
13#include <linux/interrupt.h>
14#include <linux/moduleparam.h>
15#include <linux/timer.h>
16
17#include "internals.h"
18
19static int irqfixup __read_mostly;
20
21#define POLL_SPURIOUS_IRQ_INTERVAL (HZ/10)
22static void poll_spurious_irqs(unsigned long dummy);
23static DEFINE_TIMER(poll_spurious_irq_timer, poll_spurious_irqs, 0, 0);
24static int irq_poll_cpu;
25static atomic_t irq_poll_active;
26
27/*
28 * We wait here for a poller to finish.
29 *
30 * If the poll runs on this CPU, then we yell loudly and return
31 * false. That will leave the interrupt line disabled in the worst
32 * case, but it should never happen.
33 *
34 * We wait until the poller is done and then recheck disabled and
35 * action (about to be disabled). Only if it's still active, we return
36 * true and let the handler run.
37 */
38bool irq_wait_for_poll(struct irq_desc *desc)
39{
40 if (WARN_ONCE(irq_poll_cpu == smp_processor_id(),
41 "irq poll in progress on cpu %d for irq %d\n",
42 smp_processor_id(), desc->irq_data.irq))
43 return false;
44
45#ifdef CONFIG_SMP
46 do {
47 raw_spin_unlock(&desc->lock);
48 while (irqd_irq_inprogress(&desc->irq_data))
49 cpu_relax();
50 raw_spin_lock(&desc->lock);
51 } while (irqd_irq_inprogress(&desc->irq_data));
52 /* Might have been disabled in meantime */
53 return !irqd_irq_disabled(&desc->irq_data) && desc->action;
54#else
55 return false;
56#endif
57}
58
59
60/*
61 * Recovery handler for misrouted interrupts.
62 */
63static int try_one_irq(int irq, struct irq_desc *desc, bool force)
64{
65 irqreturn_t ret = IRQ_NONE;
66 struct irqaction *action;
67
68 raw_spin_lock(&desc->lock);
69
70 /* PER_CPU and nested thread interrupts are never polled */
71 if (irq_settings_is_per_cpu(desc) || irq_settings_is_nested_thread(desc))
72 goto out;
73
74 /*
75 * Do not poll disabled interrupts unless the spurious
76 * disabled poller asks explicitely.
77 */
78 if (irqd_irq_disabled(&desc->irq_data) && !force)
79 goto out;
80
81 /*
82 * All handlers must agree on IRQF_SHARED, so we test just the
83 * first. Check for action->next as well.
84 */
85 action = desc->action;
86 if (!action || !(action->flags & IRQF_SHARED) ||
87 (action->flags & __IRQF_TIMER) || !action->next)
88 goto out;
89
90 /* Already running on another processor */
91 if (irqd_irq_inprogress(&desc->irq_data)) {
92 /*
93 * Already running: If it is shared get the other
94 * CPU to go looking for our mystery interrupt too
95 */
96 desc->istate |= IRQS_PENDING;
97 goto out;
98 }
99
100 /* Mark it poll in progress */
101 desc->istate |= IRQS_POLL_INPROGRESS;
102 do {
103 if (handle_irq_event(desc) == IRQ_HANDLED)
104 ret = IRQ_HANDLED;
105 action = desc->action;
106 } while ((desc->istate & IRQS_PENDING) && action);
107 desc->istate &= ~IRQS_POLL_INPROGRESS;
108out:
109 raw_spin_unlock(&desc->lock);
110 return ret == IRQ_HANDLED;
111}
112
113static int misrouted_irq(int irq)
114{
115 struct irq_desc *desc;
116 int i, ok = 0;
117
118 if (atomic_inc_return(&irq_poll_active) == 1)
119 goto out;
120
121 irq_poll_cpu = smp_processor_id();
122
123 for_each_irq_desc(i, desc) {
124 if (!i)
125 continue;
126
127 if (i == irq) /* Already tried */
128 continue;
129
130 if (try_one_irq(i, desc, false))
131 ok = 1;
132 }
133out:
134 atomic_dec(&irq_poll_active);
135 /* So the caller can adjust the irq error counts */
136 return ok;
137}
138
139static void poll_spurious_irqs(unsigned long dummy)
140{
141 struct irq_desc *desc;
142 int i;
143
144 if (atomic_inc_return(&irq_poll_active) != 1)
145 goto out;
146 irq_poll_cpu = smp_processor_id();
147
148 for_each_irq_desc(i, desc) {
149 unsigned int state;
150
151 if (!i)
152 continue;
153
154 /* Racy but it doesn't matter */
155 state = desc->istate;
156 barrier();
157 if (!(state & IRQS_SPURIOUS_DISABLED))
158 continue;
159
160 local_irq_disable();
161 try_one_irq(i, desc, true);
162 local_irq_enable();
163 }
164out:
165 atomic_dec(&irq_poll_active);
166 mod_timer(&poll_spurious_irq_timer,
167 jiffies + POLL_SPURIOUS_IRQ_INTERVAL);
168}
169
170static inline int bad_action_ret(irqreturn_t action_ret)
171{
172 if (likely(action_ret <= (IRQ_HANDLED | IRQ_WAKE_THREAD)))
173 return 0;
174 return 1;
175}
176
177/*
178 * If 99,900 of the previous 100,000 interrupts have not been handled
179 * then assume that the IRQ is stuck in some manner. Drop a diagnostic
180 * and try to turn the IRQ off.
181 *
182 * (The other 100-of-100,000 interrupts may have been a correctly
183 * functioning device sharing an IRQ with the failing one)
184 */
185static void
186__report_bad_irq(unsigned int irq, struct irq_desc *desc,
187 irqreturn_t action_ret)
188{
189 struct irqaction *action;
190 unsigned long flags;
191
192 if (bad_action_ret(action_ret)) {
193 printk(KERN_ERR "irq event %d: bogus return value %x\n",
194 irq, action_ret);
195 } else {
196 printk(KERN_ERR "irq %d: nobody cared (try booting with "
197 "the \"irqpoll\" option)\n", irq);
198 }
199 dump_stack();
200 printk(KERN_ERR "handlers:\n");
201
202 /*
203 * We need to take desc->lock here. note_interrupt() is called
204 * w/o desc->lock held, but IRQ_PROGRESS set. We might race
205 * with something else removing an action. It's ok to take
206 * desc->lock here. See synchronize_irq().
207 */
208 raw_spin_lock_irqsave(&desc->lock, flags);
209 action = desc->action;
210 while (action) {
211 printk(KERN_ERR "[<%p>] %pf", action->handler, action->handler);
212 if (action->thread_fn)
213 printk(KERN_CONT " threaded [<%p>] %pf",
214 action->thread_fn, action->thread_fn);
215 printk(KERN_CONT "\n");
216 action = action->next;
217 }
218 raw_spin_unlock_irqrestore(&desc->lock, flags);
219}
220
221static void
222report_bad_irq(unsigned int irq, struct irq_desc *desc, irqreturn_t action_ret)
223{
224 static int count = 100;
225
226 if (count > 0) {
227 count--;
228 __report_bad_irq(irq, desc, action_ret);
229 }
230}
231
232static inline int
233try_misrouted_irq(unsigned int irq, struct irq_desc *desc,
234 irqreturn_t action_ret)
235{
236 struct irqaction *action;
237
238 if (!irqfixup)
239 return 0;
240
241 /* We didn't actually handle the IRQ - see if it was misrouted? */
242 if (action_ret == IRQ_NONE)
243 return 1;
244
245 /*
246 * But for 'irqfixup == 2' we also do it for handled interrupts if
247 * they are marked as IRQF_IRQPOLL (or for irq zero, which is the
248 * traditional PC timer interrupt.. Legacy)
249 */
250 if (irqfixup < 2)
251 return 0;
252
253 if (!irq)
254 return 1;
255
256 /*
257 * Since we don't get the descriptor lock, "action" can
258 * change under us. We don't really care, but we don't
259 * want to follow a NULL pointer. So tell the compiler to
260 * just load it once by using a barrier.
261 */
262 action = desc->action;
263 barrier();
264 return action && (action->flags & IRQF_IRQPOLL);
265}
266
267void note_interrupt(unsigned int irq, struct irq_desc *desc,
268 irqreturn_t action_ret)
269{
270 if (desc->istate & IRQS_POLL_INPROGRESS)
271 return;
272
273 /* we get here again via the threaded handler */
274 if (action_ret == IRQ_WAKE_THREAD)
275 return;
276
277 if (bad_action_ret(action_ret)) {
278 report_bad_irq(irq, desc, action_ret);
279 return;
280 }
281
282 if (unlikely(action_ret == IRQ_NONE)) {
283 /*
284 * If we are seeing only the odd spurious IRQ caused by
285 * bus asynchronicity then don't eventually trigger an error,
286 * otherwise the counter becomes a doomsday timer for otherwise
287 * working systems
288 */
289 if (time_after(jiffies, desc->last_unhandled + HZ/10))
290 desc->irqs_unhandled = 1;
291 else
292 desc->irqs_unhandled++;
293 desc->last_unhandled = jiffies;
294 }
295
296 if (unlikely(try_misrouted_irq(irq, desc, action_ret))) {
297 int ok = misrouted_irq(irq);
298 if (action_ret == IRQ_NONE)
299 desc->irqs_unhandled -= ok;
300 }
301
302 desc->irq_count++;
303 if (likely(desc->irq_count < 100000))
304 return;
305
306 desc->irq_count = 0;
307 if (unlikely(desc->irqs_unhandled > 99900)) {
308 /*
309 * The interrupt is stuck
310 */
311 __report_bad_irq(irq, desc, action_ret);
312 /*
313 * Now kill the IRQ
314 */
315 printk(KERN_EMERG "Disabling IRQ #%d\n", irq);
316 desc->istate |= IRQS_SPURIOUS_DISABLED;
317 desc->depth++;
318 irq_disable(desc);
319
320 mod_timer(&poll_spurious_irq_timer,
321 jiffies + POLL_SPURIOUS_IRQ_INTERVAL);
322 }
323 desc->irqs_unhandled = 0;
324}
325
326int noirqdebug __read_mostly;
327
328int noirqdebug_setup(char *str)
329{
330 noirqdebug = 1;
331 printk(KERN_INFO "IRQ lockup detection disabled\n");
332
333 return 1;
334}
335
336__setup("noirqdebug", noirqdebug_setup);
337module_param(noirqdebug, bool, 0644);
338MODULE_PARM_DESC(noirqdebug, "Disable irq lockup detection when true");
339
340static int __init irqfixup_setup(char *str)
341{
342 irqfixup = 1;
343 printk(KERN_WARNING "Misrouted IRQ fixup support enabled.\n");
344 printk(KERN_WARNING "This may impact system performance.\n");
345
346 return 1;
347}
348
349__setup("irqfixup", irqfixup_setup);
350module_param(irqfixup, int, 0644);
351
352static int __init irqpoll_setup(char *str)
353{
354 irqfixup = 2;
355 printk(KERN_WARNING "Misrouted IRQ fixup and polling support "
356 "enabled\n");
357 printk(KERN_WARNING "This may significantly impact system "
358 "performance\n");
359 return 1;
360}
361
362__setup("irqpoll", irqpoll_setup);