Commit | Line | Data |
---|---|---|
9919cba7 FLVC |
1 | =============================================================== |
2 | Softlockup detector and hardlockup detector (aka nmi_watchdog) | |
3 | =============================================================== | |
4 | ||
5 | The Linux kernel can act as a watchdog to detect both soft and hard | |
6 | lockups. | |
7 | ||
8 | A 'softlockup' is defined as a bug that causes the kernel to loop in | |
9 | kernel mode for more than 20 seconds (see "Implementation" below for | |
10 | details), without giving other tasks a chance to run. The current | |
11 | stack trace is displayed upon detection and, by default, the system | |
12 | will stay locked up. Alternatively, the kernel can be configured to | |
13 | panic; a sysctl, "kernel.softlockup_panic", a kernel parameter, | |
14 | "softlockup_panic" (see "Documentation/kernel-parameters.txt" for | |
8ae34ea7 | 15 | details), and a compile option, "BOOTPARAM_SOFTLOCKUP_PANIC", are |
9919cba7 FLVC |
16 | provided for this. |
17 | ||
18 | A 'hardlockup' is defined as a bug that causes the CPU to loop in | |
19 | kernel mode for more than 10 seconds (see "Implementation" below for | |
20 | details), without letting other interrupts have a chance to run. | |
21 | Similarly to the softlockup case, the current stack trace is displayed | |
22 | upon detection and the system will stay locked up unless the default | |
23 | behavior is changed, which can be done through a compile time knob, | |
24 | "BOOTPARAM_HARDLOCKUP_PANIC", and a kernel parameter, "nmi_watchdog" | |
25 | (see "Documentation/kernel-parameters.txt" for details). | |
26 | ||
27 | The panic option can be used in combination with panic_timeout (this | |
28 | timeout is set through the confusingly named "kernel.panic" sysctl), | |
29 | to cause the system to reboot automatically after a specified amount | |
30 | of time. | |
31 | ||
32 | === Implementation === | |
33 | ||
34 | The soft and hard lockup detectors are built on top of the hrtimer and | |
35 | perf subsystems, respectively. A direct consequence of this is that, | |
36 | in principle, they should work in any architecture where these | |
37 | subsystems are present. | |
38 | ||
39 | A periodic hrtimer runs to generate interrupts and kick the watchdog | |
40 | task. An NMI perf event is generated every "watchdog_thresh" | |
41 | (compile-time initialized to 10 and configurable through sysctl of the | |
42 | same name) seconds to check for hardlockups. If any CPU in the system | |
43 | does not receive any hrtimer interrupt during that time the | |
44 | 'hardlockup detector' (the handler for the NMI perf event) will | |
45 | generate a kernel warning or call panic, depending on the | |
46 | configuration. | |
47 | ||
48 | The watchdog task is a high priority kernel thread that updates a | |
49 | timestamp every time it is scheduled. If that timestamp is not updated | |
50 | for 2*watchdog_thresh seconds (the softlockup threshold) the | |
51 | 'softlockup detector' (coded inside the hrtimer callback function) | |
52 | will dump useful debug information to the system log, after which it | |
53 | will call panic if it was instructed to do so or resume execution of | |
54 | other kernel code. | |
55 | ||
56 | The period of the hrtimer is 2*watchdog_thresh/5, which means it has | |
57 | two or three chances to generate an interrupt before the hardlockup | |
58 | detector kicks in. | |
59 | ||
60 | As explained above, a kernel knob is provided that allows | |
61 | administrators to configure the period of the hrtimer and the perf | |
62 | event. The right value for a particular environment is a trade-off | |
63 | between fast response to lockups and detection overhead. | |
fe4ba3c3 CM |
64 | |
65 | By default, the watchdog runs on all online cores. However, on a | |
66 | kernel configured with NO_HZ_FULL, by default the watchdog runs only | |
67 | on the housekeeping cores, not the cores specified in the "nohz_full" | |
68 | boot argument. If we allowed the watchdog to run by default on | |
69 | the "nohz_full" cores, we would have to run timer ticks to activate | |
70 | the scheduler, which would prevent the "nohz_full" functionality | |
71 | from protecting the user code on those cores from the kernel. | |
72 | Of course, disabling it by default on the nohz_full cores means that | |
73 | when those cores do enter the kernel, by default we will not be | |
74 | able to detect if they lock up. However, allowing the watchdog | |
75 | to continue to run on the housekeeping (non-tickless) cores means | |
76 | that we will continue to detect lockups properly on those cores. | |
77 | ||
78 | In either case, the set of cores excluded from running the watchdog | |
79 | may be adjusted via the kernel.watchdog_cpumask sysctl. For | |
80 | nohz_full cores, this may be useful for debugging a case where the | |
81 | kernel seems to be hanging on the nohz_full cores. |