watchdog/hardlockup: detect hard lockups using secondary (buddy) CPUs

Implement a hardlockup detector that doesn't doesn't need any extra
arch-specific support code to detect lockups.  Instead of using something
arch-specific we will use the buddy system, where each CPU watches out for
another one.  Specifically, each CPU will use its softlockup hrtimer to
check that the next CPU is processing hrtimer interrupts by verifying that
a counter is increasing.

NOTE: unlike the other hard lockup detectors, the buddy one can't easily
show what's happening on the CPU that locked up just by doing a simple
backtrace.  It relies on some other mechanism in the system to get
information about the locked up CPUs.  This could be support for NMI
backtraces like [1], it could be a mechanism for printing the PC of locked
CPUs at panic time like [2] / [3], or it could be something else.  Even
though that means we still rely on arch-specific code, this arch-specific
code seems to often be implemented even on architectures that don't have a
hardlockup detector.

This style of hardlockup detector originated in some downstream Android
trees and has been rebased on / carried in ChromeOS trees for quite a long
time for use on arm and arm64 boards.  Historically on these boards we've
leveraged mechanism [2] / [3] to get information about hung CPUs, but we
could move to [1].

Although the original motivation for the buddy system was for use on
systems without an arch-specific hardlockup detector, it can still be
useful to use even on systems that _do_ have an arch-specific hardlockup
detector.  On x86, for instance, there is a 24-part patch series [4] in
progress switching the arch-specific hard lockup detector from a scarce
perf counter to a less-scarce hardware resource.  Potentially the buddy
system could be a simpler alternative to free up the perf counter but
still get hard lockup detection.

Overall, pros (+) and cons (-) of the buddy system compared to an
arch-specific hardlockup detector (which might be implemented using
perf):
+ The buddy system is usable on systems that don't have an
  arch-specific hardlockup detector, like arm32 and arm64 (though it's
  being worked on for arm64 [5]).
+ The buddy system may free up scarce hardware resources.
+ If a CPU totally goes out to lunch (can't process NMIs) the buddy
  system could still detect the problem (though it would be unlikely
  to be able to get a stack trace).
+ The buddy system uses the same timer function to pet the hardlockup
  detector on the running CPU as it uses to detect hardlockups on
  other CPUs. Compared to other hardlockup detectors, this means it
  generates fewer interrupts and thus is likely better able to let
  CPUs stay idle longer.
- If all CPUs are hard locked up at the same time the buddy system
  can't detect it.
- If we don't have SMP we can't use the buddy system.
- The buddy system needs an arch-specific mechanism (possibly NMI
  backtrace) to get info about the locked up CPU.

[1] https://lore.kernel.org/r/20230419225604.21204-1-dianders@chromium.org
[2] https://issuetracker.google.com/172213129
[3] https://docs.kernel.org/trace/coresight/coresight-cpu-debug.html
[4] https://lore.kernel.org/lkml/20230301234753.28582-1-ricardo.neri-calderon@linux.intel.com/
[5] https://lore.kernel.org/linux-arm-kernel/20220903093415.15850-1-lecopzer.chen@mediatek.com/

Link: https://lkml.kernel.org/r/20230519101840.v5.14.I6bf789d21d0c3d75d382e7e51a804a7a51315f2c@changeid
Signed-off-by: Colin Cross <ccross@android.com>
Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Guenter Roeck <groeck@chromium.org>
Signed-off-by: Tzung-Bi Shih <tzungbi@chromium.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Chen-Yu Tsai <wens@csie.org>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Daniel Thompson <daniel.thompson@linaro.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Ian Rogers <irogers@google.com>
Cc: Marc Zyngier <maz@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Masayoshi Mizuma <msys.mizuma@gmail.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Pingfan Liu <kernelfans@gmail.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: "Ravi V. Shankar" <ravi.v.shankar@intel.com>
Cc: Ricardo Neri <ricardo.neri@intel.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Stephen Boyd <swboyd@chromium.org>
Cc: Sumit Garg <sumit.garg@linaro.org>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

3 files changed, 116 insertions, 7 deletions

diff --git a/kernel/Makefile b/kernel/Makefile
index 7eb72033143c..f9e3fd9195d9 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -91,6 +91,7 @@ obj-$(CONFIG_FAIL_FUNCTION) += fail_function.o
 obj-$(CONFIG_KGDB) += debug/
 obj-$(CONFIG_DETECT_HUNG_TASK) += hung_task.o
 obj-$(CONFIG_LOCKUP_DETECTOR) += watchdog.o
+obj-$(CONFIG_HARDLOCKUP_DETECTOR_BUDDY) += watchdog_buddy.o
 obj-$(CONFIG_HARDLOCKUP_DETECTOR_PERF) += watchdog_perf.o
 obj-$(CONFIG_SECCOMP) += seccomp.o
 obj-$(CONFIG_RELAY) += relay.o
diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index e67125f64719..10947c835079 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -85,7 +85,7 @@ __setup("nmi_watchdog=", hardlockup_panic_setup);
 
 #endif /* CONFIG_HARDLOCKUP_DETECTOR */
 
-#if defined(CONFIG_HARDLOCKUP_DETECTOR_PERF)
+#if defined(CONFIG_HARDLOCKUP_DETECTOR_COUNTS_HRTIMER)
 
 static DEFINE_PER_CPU(atomic_t, hrtimer_interrupts);
 static DEFINE_PER_CPU(int, hrtimer_interrupts_saved);
@@ -106,6 +106,14 @@ notrace void arch_touch_nmi_watchdog(void)
 }
 EXPORT_SYMBOL(arch_touch_nmi_watchdog);
 
+void watchdog_hardlockup_touch_cpu(unsigned int cpu)
+{
+	per_cpu(watchdog_hardlockup_touched, cpu) = true;
+
+	/* Match with smp_rmb() in watchdog_hardlockup_check() */
+	smp_wmb();
+}
+
 static bool is_hardlockup(unsigned int cpu)
 {
 	int hrint = atomic_read(&per_cpu(hrtimer_interrupts, cpu));
@@ -123,13 +131,16 @@ static bool is_hardlockup(unsigned int cpu)
 	return false;
 }
 
-static void watchdog_hardlockup_kick(void)
+static unsigned long watchdog_hardlockup_kick(void)
 {
-	atomic_inc(raw_cpu_ptr(&hrtimer_interrupts));
+	return atomic_inc_return(raw_cpu_ptr(&hrtimer_interrupts));
 }
 
 void watchdog_hardlockup_check(unsigned int cpu, struct pt_regs *regs)
 {
+	/* Match with smp_wmb() in watchdog_hardlockup_touch_cpu() */
+	smp_rmb();
+
 	if (per_cpu(watchdog_hardlockup_touched, cpu)) {
 		per_cpu(watchdog_hardlockup_touched, cpu) = false;
 		return;
@@ -180,11 +191,11 @@ void watchdog_hardlockup_check(unsigned int cpu, struct pt_regs *regs)
 	}
 }
 
-#else /* CONFIG_HARDLOCKUP_DETECTOR_PERF */
+#else /* CONFIG_HARDLOCKUP_DETECTOR_COUNTS_HRTIMER */
 
-static inline void watchdog_hardlockup_kick(void) { }
+static inline unsigned long watchdog_hardlockup_kick(void) { return 0; }
 
-#endif /* !CONFIG_HARDLOCKUP_DETECTOR_PERF */
+#endif /* !CONFIG_HARDLOCKUP_DETECTOR_COUNTS_HRTIMER */
 
 /*
  * These functions can be overridden based on the configured hardlockdup detector.
@@ -443,11 +454,15 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
 	struct pt_regs *regs = get_irq_regs();
 	int duration;
 	int softlockup_all_cpu_backtrace = sysctl_softlockup_all_cpu_backtrace;
+	unsigned long hrtimer_interrupts;
 
 	if (!watchdog_enabled)
 		return HRTIMER_NORESTART;
 
-	watchdog_hardlockup_kick();
+	hrtimer_interrupts = watchdog_hardlockup_kick();
+
+	/* test for hardlockups */
+	watchdog_buddy_check_hardlockup(hrtimer_interrupts);
 
 	/* kick the softlockup detector */
 	if (completion_done(this_cpu_ptr(&softlockup_completion))) {
diff --git a/kernel/watchdog_buddy.c b/kernel/watchdog_buddy.c
new file mode 100644
index 000000000000..fee45af2e5bd
--- /dev/null
+++ b/kernel/watchdog_buddy.c
@@ -0,0 +1,93 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/cpu.h>
+#include <linux/cpumask.h>
+#include <linux/kernel.h>
+#include <linux/nmi.h>
+#include <linux/percpu-defs.h>
+
+static cpumask_t __read_mostly watchdog_cpus;
+
+static unsigned int watchdog_next_cpu(unsigned int cpu)
+{
+	cpumask_t cpus = watchdog_cpus;
+	unsigned int next_cpu;
+
+	next_cpu = cpumask_next(cpu, &cpus);
+	if (next_cpu >= nr_cpu_ids)
+		next_cpu = cpumask_first(&cpus);
+
+	if (next_cpu == cpu)
+		return nr_cpu_ids;
+
+	return next_cpu;
+}
+
+int __init watchdog_hardlockup_probe(void)
+{
+	return 0;
+}
+
+void watchdog_hardlockup_enable(unsigned int cpu)
+{
+	unsigned int next_cpu;
+
+	/*
+	 * The new CPU will be marked online before the hrtimer interrupt
+	 * gets a chance to run on it. If another CPU tests for a
+	 * hardlockup on the new CPU before it has run its the hrtimer
+	 * interrupt, it will get a false positive. Touch the watchdog on
+	 * the new CPU to delay the check for at least 3 sampling periods
+	 * to guarantee one hrtimer has run on the new CPU.
+	 */
+	watchdog_hardlockup_touch_cpu(cpu);
+
+	/*
+	 * We are going to check the next CPU. Our watchdog_hrtimer
+	 * need not be zero if the CPU has already been online earlier.
+	 * Touch the watchdog on the next CPU to avoid false positive
+	 * if we try to check it in less then 3 interrupts.
+	 */
+	next_cpu = watchdog_next_cpu(cpu);
+	if (next_cpu < nr_cpu_ids)
+		watchdog_hardlockup_touch_cpu(next_cpu);
+
+	cpumask_set_cpu(cpu, &watchdog_cpus);
+}
+
+void watchdog_hardlockup_disable(unsigned int cpu)
+{
+	unsigned int next_cpu = watchdog_next_cpu(cpu);
+
+	/*
+	 * Offlining this CPU will cause the CPU before this one to start
+	 * checking the one after this one. If this CPU just finished checking
+	 * the next CPU and updating hrtimer_interrupts_saved, and then the
+	 * previous CPU checks it within one sample period, it will trigger a
+	 * false positive. Touch the watchdog on the next CPU to prevent it.
+	 */
+	if (next_cpu < nr_cpu_ids)
+		watchdog_hardlockup_touch_cpu(next_cpu);
+
+	cpumask_clear_cpu(cpu, &watchdog_cpus);
+}
+
+void watchdog_buddy_check_hardlockup(unsigned long hrtimer_interrupts)
+{
+	unsigned int next_cpu;
+
+	/*
+	 * Test for hardlockups every 3 samples. The sample period is
+	 *  watchdog_thresh * 2 / 5, so 3 samples gets us back to slightly over
+	 *  watchdog_thresh (over by 20%).
+	 */
+	if (hrtimer_interrupts % 3 != 0)
+		return;
+
+	/* check for a hardlockup on the next CPU */
+	next_cpu = watchdog_next_cpu(smp_processor_id());
+	if (next_cpu >= nr_cpu_ids)
+		return;
+
+	watchdog_hardlockup_check(next_cpu, NULL);
+}