Back to home page




0001 ===============================================================
0002 Softlockup detector and hardlockup detector (aka nmi_watchdog)
0003 ===============================================================
0005 The Linux kernel can act as a watchdog to detect both soft and hard
0006 lockups.
0008 A 'softlockup' is defined as a bug that causes the kernel to loop in
0009 kernel mode for more than 20 seconds (see "Implementation" below for
0010 details), without giving other tasks a chance to run. The current
0011 stack trace is displayed upon detection and, by default, the system
0012 will stay locked up. Alternatively, the kernel can be configured to
0013 panic; a sysctl, "kernel.softlockup_panic", a kernel parameter,
0014 "softlockup_panic" (see "Documentation/admin-guide/kernel-parameters.rst" for
0015 details), and a compile option, "BOOTPARAM_SOFTLOCKUP_PANIC", are
0016 provided for this.
0018 A 'hardlockup' is defined as a bug that causes the CPU to loop in
0019 kernel mode for more than 10 seconds (see "Implementation" below for
0020 details), without letting other interrupts have a chance to run.
0021 Similarly to the softlockup case, the current stack trace is displayed
0022 upon detection and the system will stay locked up unless the default
0023 behavior is changed, which can be done through a sysctl,
0024 'hardlockup_panic', a compile time knob, "BOOTPARAM_HARDLOCKUP_PANIC",
0025 and a kernel parameter, "nmi_watchdog"
0026 (see "Documentation/admin-guide/kernel-parameters.rst" for details).
0028 The panic option can be used in combination with panic_timeout (this
0029 timeout is set through the confusingly named "kernel.panic" sysctl),
0030 to cause the system to reboot automatically after a specified amount
0031 of time.
0033 === Implementation ===
0035 The soft and hard lockup detectors are built on top of the hrtimer and
0036 perf subsystems, respectively. A direct consequence of this is that,
0037 in principle, they should work in any architecture where these
0038 subsystems are present.
0040 A periodic hrtimer runs to generate interrupts and kick the watchdog
0041 task. An NMI perf event is generated every "watchdog_thresh"
0042 (compile-time initialized to 10 and configurable through sysctl of the
0043 same name) seconds to check for hardlockups. If any CPU in the system
0044 does not receive any hrtimer interrupt during that time the
0045 'hardlockup detector' (the handler for the NMI perf event) will
0046 generate a kernel warning or call panic, depending on the
0047 configuration.
0049 The watchdog task is a high priority kernel thread that updates a
0050 timestamp every time it is scheduled. If that timestamp is not updated
0051 for 2*watchdog_thresh seconds (the softlockup threshold) the
0052 'softlockup detector' (coded inside the hrtimer callback function)
0053 will dump useful debug information to the system log, after which it
0054 will call panic if it was instructed to do so or resume execution of
0055 other kernel code.
0057 The period of the hrtimer is 2*watchdog_thresh/5, which means it has
0058 two or three chances to generate an interrupt before the hardlockup
0059 detector kicks in.
0061 As explained above, a kernel knob is provided that allows
0062 administrators to configure the period of the hrtimer and the perf
0063 event. The right value for a particular environment is a trade-off
0064 between fast response to lockups and detection overhead.
0066 By default, the watchdog runs on all online cores.  However, on a
0067 kernel configured with NO_HZ_FULL, by default the watchdog runs only
0068 on the housekeeping cores, not the cores specified in the "nohz_full"
0069 boot argument.  If we allowed the watchdog to run by default on
0070 the "nohz_full" cores, we would have to run timer ticks to activate
0071 the scheduler, which would prevent the "nohz_full" functionality
0072 from protecting the user code on those cores from the kernel.
0073 Of course, disabling it by default on the nohz_full cores means that
0074 when those cores do enter the kernel, by default we will not be
0075 able to detect if they lock up.  However, allowing the watchdog
0076 to continue to run on the housekeeping (non-tickless) cores means
0077 that we will continue to detect lockups properly on those cores.
0079 In either case, the set of cores excluded from running the watchdog
0080 may be adjusted via the kernel.watchdog_cpumask sysctl.  For
0081 nohz_full cores, this may be useful for debugging a case where the
0082 kernel seems to be hanging on the nohz_full cores.