]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * hangcheck-timer.c | |
3 | * | |
4 | * Driver for a little io fencing timer. | |
5 | * | |
6 | * Copyright (C) 2002 Oracle Corporation. All rights reserved. | |
7 | * | |
8 | * Author: Joel Becker <joel.becker@oracle.com> | |
9 | * | |
10 | * This program is free software; you can redistribute it and/or | |
11 | * modify it under the terms of the GNU General Public | |
12 | * License version 2 as published by the Free Software Foundation. | |
13 | * | |
14 | * This program is distributed in the hope that it will be useful, | |
15 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
17 | * General Public License for more details. | |
18 | * | |
19 | * You should have received a copy of the GNU General Public | |
20 | * License along with this program; if not, write to the | |
21 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
22 | * Boston, MA 021110-1307, USA. | |
23 | */ | |
24 | ||
25 | /* | |
26 | * The hangcheck-timer driver uses the TSC to catch delays that | |
27 | * jiffies does not notice. A timer is set. When the timer fires, it | |
28 | * checks whether it was delayed and if that delay exceeds a given | |
29 | * margin of error. The hangcheck_tick module paramter takes the timer | |
30 | * duration in seconds. The hangcheck_margin parameter defines the | |
31 | * margin of error, in seconds. The defaults are 60 seconds for the | |
32 | * timer and 180 seconds for the margin of error. IOW, a timer is set | |
33 | * for 60 seconds. When the timer fires, the callback checks the | |
34 | * actual duration that the timer waited. If the duration exceeds the | |
35 | * alloted time and margin (here 60 + 180, or 240 seconds), the machine | |
36 | * is restarted. A healthy machine will have the duration match the | |
37 | * expected timeout very closely. | |
38 | */ | |
39 | ||
40 | #include <linux/module.h> | |
41 | #include <linux/moduleparam.h> | |
42 | #include <linux/types.h> | |
43 | #include <linux/kernel.h> | |
44 | #include <linux/fs.h> | |
45 | #include <linux/mm.h> | |
46 | #include <linux/reboot.h> | |
47 | #include <linux/init.h> | |
48 | #include <asm/uaccess.h> | |
49 | ||
50 | ||
51 | #define VERSION_STR "0.5.0" | |
52 | ||
53 | #define DEFAULT_IOFENCE_MARGIN 60 /* Default fudge factor, in seconds */ | |
54 | #define DEFAULT_IOFENCE_TICK 180 /* Default timer timeout, in seconds */ | |
55 | ||
56 | static int hangcheck_tick = DEFAULT_IOFENCE_TICK; | |
57 | static int hangcheck_margin = DEFAULT_IOFENCE_MARGIN; | |
58 | static int hangcheck_reboot; /* Defaults to not reboot */ | |
59 | ||
60 | /* Driver options */ | |
61 | module_param(hangcheck_tick, int, 0); | |
62 | MODULE_PARM_DESC(hangcheck_tick, "Timer delay."); | |
63 | module_param(hangcheck_margin, int, 0); | |
64 | MODULE_PARM_DESC(hangcheck_margin, "If the hangcheck timer has been delayed more than hangcheck_margin seconds, the driver will fire."); | |
65 | module_param(hangcheck_reboot, int, 0); | |
66 | MODULE_PARM_DESC(hangcheck_reboot, "If nonzero, the machine will reboot when the timer margin is exceeded."); | |
67 | ||
68 | MODULE_AUTHOR("Joel Becker"); | |
69 | MODULE_DESCRIPTION("Hangcheck-timer detects when the system has gone out to lunch past a certain margin."); | |
70 | MODULE_LICENSE("GPL"); | |
71 | ||
72 | ||
73 | /* Last time scheduled */ | |
74 | static unsigned long long hangcheck_tsc, hangcheck_tsc_margin; | |
75 | ||
76 | static void hangcheck_fire(unsigned long); | |
77 | ||
78 | static struct timer_list hangcheck_ticktock = | |
79 | TIMER_INITIALIZER(hangcheck_fire, 0, 0); | |
80 | ||
81 | extern unsigned long long monotonic_clock(void); | |
82 | ||
83 | static void hangcheck_fire(unsigned long data) | |
84 | { | |
85 | unsigned long long cur_tsc, tsc_diff; | |
86 | ||
87 | cur_tsc = monotonic_clock(); | |
88 | ||
89 | if (cur_tsc > hangcheck_tsc) | |
90 | tsc_diff = cur_tsc - hangcheck_tsc; | |
91 | else | |
92 | tsc_diff = (cur_tsc + (~0ULL - hangcheck_tsc)); /* or something */ | |
93 | ||
94 | if (tsc_diff > hangcheck_tsc_margin) { | |
95 | if (hangcheck_reboot) { | |
96 | printk(KERN_CRIT "Hangcheck: hangcheck is restarting the machine.\n"); | |
97 | machine_restart(NULL); | |
98 | } else { | |
99 | printk(KERN_CRIT "Hangcheck: hangcheck value past margin!\n"); | |
100 | } | |
101 | } | |
102 | mod_timer(&hangcheck_ticktock, jiffies + (hangcheck_tick*HZ)); | |
103 | hangcheck_tsc = monotonic_clock(); | |
104 | } | |
105 | ||
106 | ||
107 | static int __init hangcheck_init(void) | |
108 | { | |
109 | printk("Hangcheck: starting hangcheck timer %s (tick is %d seconds, margin is %d seconds).\n", | |
110 | VERSION_STR, hangcheck_tick, hangcheck_margin); | |
111 | ||
112 | hangcheck_tsc_margin = hangcheck_margin + hangcheck_tick; | |
113 | hangcheck_tsc_margin *= 1000000000; | |
114 | ||
115 | ||
116 | hangcheck_tsc = monotonic_clock(); | |
117 | mod_timer(&hangcheck_ticktock, jiffies + (hangcheck_tick*HZ)); | |
118 | ||
119 | return 0; | |
120 | } | |
121 | ||
122 | ||
123 | static void __exit hangcheck_exit(void) | |
124 | { | |
125 | del_timer_sync(&hangcheck_ticktock); | |
126 | } | |
127 | ||
128 | module_init(hangcheck_init); | |
129 | module_exit(hangcheck_exit); |