[patch V2 00/29] lockup_detector: Cure hotplug deadlocks and replace duct tape

From: Thomas Gleixner
Date: Tue Sep 12 2017 - 15:49:09 EST


The lockup detector is broken is several ways:

- It's deadlock prone vs. CPU hotplug in various ways. Some of these
are due to recursive cpus_read_lock() others are due to
cpus_read_lock() from CPU hotplug callbacks which immediately lock
the machine because cpus are write locked.

- The handling of the cpu hotplug threads happens sideways to the
smpboot thread infrastructure, which is racy and pointless

- The handling of the user space sysctl interface is a complete
trainwreck as it fiddles directly with variables which can be
modified or evaluated by the running watchdogs.

- The perf event initialization is a trainwreck as it tries to create
perf events over and over even if perf is not functional (no
hardware, ....). To avoid excessive dmesg spam it contains magic
printk ratelimiting along with either wrong or useless messages.

- The code structure is horrible as ifdef sections are scattered all
over the place which makes it unreadable

- There is more wreckage, but see the changelogs for the ugly details.

The following series sanitizes the facility and addresses the problems.

Changes since V1:

- Wrapped the perf specific calls into the weak watchdog_nmi_*
functions

- Fixed the compile error pointed out by Don

- Fixed the reconfiguration parameter inconsistency which broke
powerpc

- Picked up the updated version of patch 11/29

Delta patch below.

Thanks,

tglx
---
Diffstat for the series:

arch/parisc/kernel/process.c | 2
arch/powerpc/kernel/watchdog.c | 22 -
arch/x86/events/intel/core.c | 11
include/linux/nmi.h | 121 +++----
include/linux/smpboot.h | 4
kernel/cpu.c | 6
kernel/smpboot.c | 22 -
kernel/sysctl.c | 22 -
kernel/watchdog.c | 633 ++++++++++++++---------------------------
kernel/watchdog_hld.c | 193 ++++++------
10 files changed, 434 insertions(+), 602 deletions(-)

Delta patch vs. V1
8<------------------------
--- a/arch/powerpc/kernel/watchdog.c
+++ b/arch/powerpc/kernel/watchdog.c
@@ -355,12 +355,12 @@ static void watchdog_calc_timeouts(void)
wd_timer_period_ms = watchdog_thresh * 1000 * 2 / 5;
}

-void watchdog_nmi_reconfigure(bool stop)
+void watchdog_nmi_reconfigure(bool run)
{
int cpu;

cpus_read_lock();
- if (stop) {
+ if (!run) {
for_each_cpu(cpu, &wd_cpus_enabled)
stop_wd_on_cpu(cpu);
} else {
--- a/include/linux/nmi.h
+++ b/include/linux/nmi.h
@@ -102,7 +102,7 @@ static inline void hardlockup_detector_p
static inline void hardlockup_detector_perf_enable(void) { }
static inline void hardlockup_detector_perf_cleanup(void) { }
# if !defined(CONFIG_HAVE_NMI_WATCHDOG)
-static int hardlockup_detector_perf_init(void) { return -ENODEV; }
+static inline int hardlockup_detector_perf_init(void) { return -ENODEV; }
static inline void arch_touch_nmi_watchdog(void) {}
# else
static int hardlockup_detector_perf_init(void) { return 0; }
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -105,18 +105,32 @@ static int __init hardlockup_all_cpu_bac
* softlockup watchdog threads start and stop. The arch must select the
* SOFTLOCKUP_DETECTOR Kconfig.
*/
-int __weak watchdog_nmi_enable(unsigned int cpu) { return 0; }
-void __weak watchdog_nmi_disable(unsigned int cpu) { }
+int __weak watchdog_nmi_enable(unsigned int cpu)
+{
+ hardlockup_detector_perf_enable();
+ return 0;
+}
+
+void __weak watchdog_nmi_disable(unsigned int cpu)
+{
+ hardlockup_detector_perf_disable();
+}
+
+/* Return 0, if a NMI watchdog is available. Error code otherwise */
+int __weak __init void watchdog_nmi_probe(void)
+{
+ return hardlockup_detector_perf_init();
+}

/**
* watchdog_nmi_reconfigure - Optional function to reconfigure NMI watchdogs
- * @stop: If true stop the watchdogs on all enabled CPUs
- * If false start the watchdogs on all enabled CPUs
+ * @run: If false stop the watchdogs on all enabled CPUs
+ * If true start the watchdogs on all enabled CPUs
*
* The core call order is:
- * watchdog_nmi_reconfigure(true);
- * update_variables();
* watchdog_nmi_reconfigure(false);
+ * update_variables();
+ * watchdog_nmi_reconfigure(true);
*
* The second call which starts the watchdogs again guarantees that the
* following variables are stable across the call.
@@ -126,13 +140,13 @@ void __weak watchdog_nmi_disable(unsigne
*
* After the call the variables can be changed again.
*/
-void __weak watchdog_nmi_reconfigure(bool stop) { }
+void __weak watchdog_nmi_reconfigure(bool run) { }

/**
* lockup_detector_update_enable - Update the sysctl enable bit
*
* Caller needs to make sure that the NMI/perf watchdogs are off, so this
- * can't race with hardlockup_detector_disable().
+ * can't race with watchdog_nmi_disable().
*/
static void lockup_detector_update_enable(void)
{
@@ -453,8 +467,7 @@ static void watchdog_enable(unsigned int
__touch_watchdog();
/* Enable the perf event */
if (watchdog_enabled & NMI_WATCHDOG_ENABLED)
- hardlockup_detector_perf_enable();
- watchdog_nmi_enable(cpu);
+ watchdog_nmi_enable(cpu);

watchdog_set_prio(SCHED_FIFO, MAX_RT_PRIO - 1);
}
@@ -469,7 +482,6 @@ static void watchdog_disable(unsigned in
* between disabling the timer and disabling the perf event causes
* the perf NMI to detect a false positive.
*/
- hardlockup_detector_perf_disable();
watchdog_nmi_disable(cpu);
hrtimer_cancel(hrtimer);
}
@@ -745,12 +757,6 @@ int proc_watchdog_cpumask(struct ctl_tab
}
#endif /* CONFIG_SYSCTL */

-static __init void detect_nmi_watchdog(void)
-{
- if (!hardlockup_detector_perf_init())
- nmi_watchdog_available = true;
-}
-
void __init lockup_detector_init(void)
{
#ifdef CONFIG_NO_HZ_FULL
@@ -763,6 +769,7 @@ void __init lockup_detector_init(void)
cpumask_copy(&watchdog_cpumask, cpu_possible_mask);
#endif

- detect_nmi_watchdog();
+ if (!watchdog_nmi_probe())
+ nmi_watchdog_available = true;
softlockup_init_threads();
}