186:(Advanced Programmable Interrupt Controller). Most computer peripherals generate interrupts through an APIC as the number of interrupts is most always less (typically 15 for the modern PC) than the number of devices. The OS must then query each driver registered to that interrupt to ask if the interrupt originated from its hardware. Faulty drivers may always claim "yes", causing the OS to not query other drivers registered to that interrupt (only one interrupt can be processed at a time). The device which originally requested the interrupt therefore does not get its interrupt serviced, so a new interrupt is generated (or is not cleared) and the processor becomes swamped with continuous interrupt signals. Any operating system can live lock under an interrupt storm caused by such a fault. A
217:
of interrupts or the resource use caused by an interrupt, passes certain thresholds. When these thresholds are no longer exceeded, an OS may then change the interrupting driver, interrupt, or interrupt handling globally, from an interrupt mode to a polling mode. Interrupt rate limiting in hardware usually negates the use of a polling mode, but can still happen during normal operation during intense I/O if the processor is unable switch contexts quickly enough to keep pace.
22:
277:
FIFO entry is marked as occupied. If at that point entry (tail−1) is free (cleared), an interrupt will be generated (level interrupt) and the tail pointer will be incremented. If the hardware requires the interrupt be acknowledged, the CPU (interrupt handler) will do that, handle the valid DMA descriptors at the head, and return from the interrupt.
216:
mode that queries for pending interrupts at fixed intervals or in a round-robin fashion. This mode can be set globally, on a per-driver, per-interrupt basis, or dynamically if the OS detects a fault condition or excessive interrupt generation. A polling mode may be enabled dynamically when the number
276:
Another interesting approach using hardware support is one where the device generates interrupt when the event queue state changes from "empty" to "not empty". Then, if there are no free DMA descriptors at the RX FIFO tail, the device drops the event. The event is then added to the tail and the
178:
controllers implement interrupt "rate limiting", which causes the controller to wait a programmable amount of time between each interrupt it generates. When not present within the device, similar functionality is usually written into the device driver, and/or the operating system itself.
163:. In such a state, the system is spending most of its resources processing interrupts instead of completing other work. To the end-user, it does not appear to be processing anything at all as there is often no output. An interrupt storm is sometimes mistaken for
208:
compatibility mode could not properly interact with the ISA interrupt routing. This would either cause interrupts to never be detected by the operating system, or the operating system would never be able to clear them, resulting in an interrupt storm.
245:
it receives from the network in between each interrupt. If the rate is set too low, the controller's buffer will overflow, and packets will be dropped. The rate must take into account how fast the buffer may fill between interrupts, and the
170:
Common causes include: misconfigured or faulty hardware, faulty device drivers, flaws in the operating system, or metastability in one or more components. The latter condition rarely occurs outside of prototype or amateur-built hardware.
193:
can usually break the storm by unloading the faulty driver, allowing the driver "underneath" the faulty one to clear the interrupt, if user input is still possible.
273:
then disables the interrupt and lets a thread/task handle the event(s) and then task polls the device, processing some number of events and enabling the interrupt.
130:
that consume the majority of the processor's time. Interrupt storms are typically caused by hardware devices that do not support interrupt rate limiting.
183:
303:
201:
105:
39:
86:
205:
43:
58:
174:
Most modern hardware and operating systems have methods for mitigating the effect of an interrupt storm. For example, most
164:
65:
374:
159:
213:
32:
291:
187:
72:
167:, since they both have similar symptoms (unresponsive or sluggish response to user input, little or no output).
226:
269:
is an example of the hardware-based approach: the system (driver) starts in interrupt enabled state, and the
54:
324:
297:
143:
369:
270:
247:
79:
154:
150:
119:
286:
237:
Interrupt rate limiting must be carefully configured for optimum results. For example, an
212:
As drivers are most often implemented by a 3rd party, most operating systems also have a
242:
363:
262:
detects interrupt storms and masks problematic interrupts for some time in response.
258:
There are hardware-based and software-based approaches to the problem. For example,
182:
The most common cause is when a device "behind" another signals an interrupt to an
153:, an interrupt storm will cause sluggish response to user input, or even appear to
147:
21:
139:
127:
238:
190:
175:
259:
197:
126:
is an event during which a processor receives an inordinate number of
250:
between the interrupt and the transfer of the buffer to the system.
266:
15:
325:"Problems updating FreeBSD's card system from ISA to PCI"
241:
controller with interrupt rate limiting will buffer the
157:
the system completely. This state is commonly known as
225:
Perhaps the first interrupt storm occurred during the
46:. Unsourced material may be challenged and removed.
8:
204:cards that were configured to operate in
106:Learn how and when to remove this message
352:. Simon and Schuster. pp. 345–355.
316:
196:This occurred in an older version of
7:
44:adding citations to reliable sources
14:
304:Programmable Interrupt Controller
20:
31:needs additional citations for
142:processing is typically a non-
1:
350:Apollo: The Race to the Moon
391:
292:Inter-processor interrupt
227:Apollo 11's lunar descent
348:Murray, Charles (1989).
298:Non-maskable interrupt
254:Interrupt mitigating
40:improve this article
265:The system used by
375:Software anomalies
271:Interrupt handler
248:interrupt latency
151:operating systems
120:operating systems
116:
115:
108:
90:
55:"Interrupt storm"
382:
354:
353:
345:
339:
338:
336:
335:
321:
111:
104:
100:
97:
91:
89:
48:
24:
16:
390:
389:
385:
384:
383:
381:
380:
379:
360:
359:
358:
357:
347:
346:
342:
333:
331:
323:
322:
318:
313:
287:Broadcast storm
283:
256:
235:
223:
136:
124:interrupt storm
112:
101:
95:
92:
49:
47:
37:
25:
12:
11:
5:
388:
386:
378:
377:
372:
362:
361:
356:
355:
340:
329:www.usenix.org
315:
314:
312:
309:
308:
307:
301:
295:
289:
282:
279:
255:
252:
234:
233:Considerations
231:
222:
219:
135:
132:
114:
113:
28:
26:
19:
13:
10:
9:
6:
4:
3:
2:
387:
376:
373:
371:
368:
367:
365:
351:
344:
341:
330:
326:
320:
317:
310:
305:
302:
299:
296:
293:
290:
288:
285:
284:
280:
278:
274:
272:
268:
263:
261:
253:
251:
249:
244:
240:
232:
230:
228:
220:
218:
215:
210:
207:
203:
199:
194:
192:
189:
185:
180:
177:
172:
168:
166:
162:
161:
156:
152:
149:
145:
141:
133:
131:
129:
125:
121:
110:
107:
99:
96:February 2013
88:
85:
81:
78:
74:
71:
67:
64:
60:
57: –
56:
52:
51:Find sources:
45:
41:
35:
34:
29:This article
27:
23:
18:
17:
349:
343:
332:. Retrieved
328:
319:
275:
264:
257:
236:
224:
211:
195:
181:
173:
169:
158:
148:time-sharing
137:
123:
117:
102:
93:
83:
76:
69:
62:
50:
38:Please help
33:verification
30:
144:preemptible
370:Interrupts
364:Categories
334:2024-05-07
311:References
134:Background
128:interrupts
66:newspapers
229:in 1969.
165:thrashing
160:live lock
140:interrupt
281:See also
239:Ethernet
200:, where
191:debugger
176:Ethernet
146:task in
138:Because
260:FreeBSD
243:packets
221:History
214:polling
198:FreeBSD
80:scholar
188:kernel
155:freeze
82:
75:
68:
61:
53:
306:(PIC)
300:(NMI)
294:(IPI)
122:, an
87:JSTOR
73:books
267:NAPI
184:APIC
59:news
206:ISA
202:PCI
118:In
42:by
366::
327:.
337:.
109:)
103:(
98:)
94:(
84:·
77:·
70:·
63:·
36:.
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.