blob: f4e72af9d7b9500af7b33bcd45feb7ce9b05e7d6 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
|
// SPDX-License-Identifier: GPL-2.0-or-later
/*
* Timer Wheel
* Copyright (C) 2016 Cumulus Networks, Inc.
* Donald Sharp
*/
#ifndef __WHEEL_H__
#define __WHEEL_H__
#ifdef __cplusplus
extern "C" {
#endif
struct timer_wheel {
char *name;
struct thread_master *master;
int slots;
long long curr_slot;
unsigned int period;
unsigned int nexttime;
unsigned int slots_to_skip;
struct list **wheel_slot_lists;
struct event *timer;
/*
* Key to determine what slot the item belongs in
*/
unsigned int (*slot_key)(const void *);
void (*slot_run)(void *);
};
/*
* Creates a timer wheel
*
* master - Thread master structure for the process
* period - The Time in seconds that the timer wheel will
* take before it starts issuing commands again
* for items in each slot
* slots - The number of slots to have in this particular
* timer wheel
* slot_key - A hashing function of some sort that will allow
* the timer wheel to put items into individual slots
* slot_run - The function to run over each item in a particular slot
*
* Creates a timer wheel that will wake up 'slots' times over the entire
* wheel. Each time the timer wheel wakes up it will iterate through
* and run the slot_run function for each item stored in that particular
* slot.
*
* The timer code is 'intelligent' in that it notices if anything is
* in a particular slot and can schedule the next timer to skip
* the empty slot.
*
* The general purpose of a timer wheel is to reduce events in a system.
* A perfect example of usage for this is say hello packets that need
* to be sent out to all your neighbors. Suppose a large routing protocol
* has to send keepalive packets every Y seconds to each of it's peers.
* At scale we can have a very large number of peers, X.
* This means that we will have X timing events every Y seconds.
* If you replace these events with a timer wheel that has Z slots
* you will have at most Y/Z timer events if each slot has a work item
* in it.
*
* When X is large the number of events in a system can quickly escalate
* and cause significant amount of time handling thread events instead
* of running your code.
*/
struct timer_wheel *wheel_init(struct thread_master *master, int period,
size_t slots,
unsigned int (*slot_key)(const void *),
void (*slot_run)(void *), const char *run_name);
/*
* Delete the specified timer wheel created
*/
void wheel_delete(struct timer_wheel *);
/*
* wheel - The Timer wheel being modified
* item - The generic data structure that will be handed
* to the slot_run function.
*
* Add item to a slot setup by the slot_key,
* possibly change next time pop.
*/
int wheel_add_item(struct timer_wheel *wheel, void *item);
/*
* wheel - The Timer wheel being modified.
* item - The item to remove from one of the slots in
* the timer wheel.
*
* Remove a item to a slot setup by the slot_key,
* possibly change next time pop.
*/
int wheel_remove_item(struct timer_wheel *wheel, void *item);
#ifdef __cplusplus
}
#endif
#endif
|
