[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20210915035103.15586-4-kernelfans@gmail.com>
Date: Wed, 15 Sep 2021 11:51:01 +0800
From: Pingfan Liu <kernelfans@...il.com>
To: linux-kernel@...r.kernel.org
Cc: Pingfan Liu <kernelfans@...il.com>,
Sumit Garg <sumit.garg@...aro.org>,
Catalin Marinas <catalin.marinas@....com>,
Will Deacon <will@...nel.org>, Ingo Molnar <mingo@...hat.com>,
Arnaldo Carvalho de Melo <acme@...nel.org>,
Mark Rutland <mark.rutland@....com>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
Jiri Olsa <jolsa@...hat.com>,
Namhyung Kim <namhyung@...nel.org>,
Marc Zyngier <maz@...nel.org>,
Julien Thierry <jthierry@...hat.com>,
Kees Cook <keescook@...omium.org>,
Masahiro Yamada <masahiroy@...nel.org>,
Sami Tolvanen <samitolvanen@...gle.com>,
Petr Mladek <pmladek@...e.com>,
Andrew Morton <akpm@...ux-foundation.org>,
Wang Qing <wangqing@...o.com>,
"Peter Zijlstra (Intel)" <peterz@...radead.org>,
Santosh Sivaraj <santosh@...six.org>
Subject: [PATCH 3/5] kernel/watchdog: adapt the watchdog_hld interface for async model
When lockup_detector_init()->watchdog_nmi_probe(), PMU may be not ready
yet. E.g. on arm64, PMU is not ready until
device_initcall(armv8_pmu_driver_init). And it is deeply integrated
with the driver model and cpuhp. Hence it is hard to push this
initialization before smp_init().
But it is easy to take an opposite approach by enabling watchdog_hld to
get the capability of PMU async.
The async model is achieved by introducing an extra parameter notifier
of watchdog_nmi_probe().
Note after this patch, the async model, which is utilized by the next
patch, does not take effect yet.
Signed-off-by: Pingfan Liu <kernelfans@...il.com>
Cc: Sumit Garg <sumit.garg@...aro.org>
Cc: Catalin Marinas <catalin.marinas@....com>
Cc: Will Deacon <will@...nel.org>
Cc: Ingo Molnar <mingo@...hat.com>
Cc: Arnaldo Carvalho de Melo <acme@...nel.org>
Cc: Mark Rutland <mark.rutland@....com>
Cc: Alexander Shishkin <alexander.shishkin@...ux.intel.com>
Cc: Jiri Olsa <jolsa@...hat.com>
Cc: Namhyung Kim <namhyung@...nel.org>
Cc: Marc Zyngier <maz@...nel.org>
Cc: Julien Thierry <jthierry@...hat.com>
Cc: Kees Cook <keescook@...omium.org>
Cc: Masahiro Yamada <masahiroy@...nel.org>
Cc: Sami Tolvanen <samitolvanen@...gle.com>
Cc: Petr Mladek <pmladek@...e.com>
Cc: Andrew Morton <akpm@...ux-foundation.org>
Cc: Wang Qing <wangqing@...o.com>
Cc: "Peter Zijlstra (Intel)" <peterz@...radead.org>
Cc: Santosh Sivaraj <santosh@...six.org>
To: linux-kernel@...r.kernel.org
---
include/linux/nmi.h | 12 +++++++++--
kernel/watchdog.c | 49 +++++++++++++++++++++++++++++++++++----------
2 files changed, 48 insertions(+), 13 deletions(-)
diff --git a/include/linux/nmi.h b/include/linux/nmi.h
index 750c7f395ca9..70665fa6e0a9 100644
--- a/include/linux/nmi.h
+++ b/include/linux/nmi.h
@@ -78,8 +78,10 @@ static inline void reset_hung_task_detector(void) { }
*/
#define NMI_WATCHDOG_ENABLED_BIT 0
#define SOFT_WATCHDOG_ENABLED_BIT 1
+#define NMI_WATCHDOG_UNDETERMINED_BIT 2
#define NMI_WATCHDOG_ENABLED (1 << NMI_WATCHDOG_ENABLED_BIT)
#define SOFT_WATCHDOG_ENABLED (1 << SOFT_WATCHDOG_ENABLED_BIT)
+#define NMI_WATCHDOG_UNDETERMINED (1 << NMI_WATCHDOG_UNDETERMINED_BIT)
#if defined(CONFIG_HARDLOCKUP_DETECTOR)
extern void hardlockup_detector_disable(void);
@@ -116,10 +118,16 @@ static inline int hardlockup_detector_perf_init(void) { return 0; }
# endif
#endif
+struct watchdog_nmi_status {
+ unsigned int cpu;
+ int status;
+};
+
+typedef void (*watchdog_nmi_status_reporter)(struct watchdog_nmi_status *);
void watchdog_nmi_stop(void);
void watchdog_nmi_start(void);
-int watchdog_nmi_probe(void);
-int watchdog_nmi_enable(unsigned int cpu);
+int watchdog_nmi_probe(watchdog_nmi_status_reporter notifier);
+void watchdog_nmi_enable(unsigned int cpu);
void watchdog_nmi_disable(unsigned int cpu);
/**
diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index e2a9e3331416..4ab71943d65f 100644
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -40,7 +40,7 @@ int __read_mostly watchdog_user_enabled = 1;
int __read_mostly nmi_watchdog_user_enabled = NMI_WATCHDOG_DEFAULT;
int __read_mostly soft_watchdog_user_enabled = 1;
int __read_mostly watchdog_thresh = 10;
-static int __read_mostly nmi_watchdog_available;
+static int __read_mostly nmi_watchdog_status;
struct cpumask watchdog_cpumask __read_mostly;
unsigned long *watchdog_cpumask_bits = cpumask_bits(&watchdog_cpumask);
@@ -85,6 +85,10 @@ __setup("nmi_watchdog=", hardlockup_panic_setup);
#endif /* CONFIG_HARDLOCKUP_DETECTOR */
+static void lockup_detector_update_enable(void);
+
+static watchdog_nmi_status_reporter status_reporter;
+
/*
* These functions can be overridden if an architecture implements its
* own hardlockup detector.
@@ -93,10 +97,9 @@ __setup("nmi_watchdog=", hardlockup_panic_setup);
* softlockup watchdog start and stop. The arch must select the
* SOFTLOCKUP_DETECTOR Kconfig.
*/
-int __weak watchdog_nmi_enable(unsigned int cpu)
+void __weak watchdog_nmi_enable(unsigned int cpu)
{
hardlockup_detector_perf_enable();
- return 0;
}
void __weak watchdog_nmi_disable(unsigned int cpu)
@@ -104,8 +107,28 @@ 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 watchdog_nmi_probe(void)
+static void watchdog_nmi_report_capability(struct watchdog_nmi_status *data)
+{
+ /* Set status to 1 temporary to block any further access */
+ if (atomic_cmpxchg((atomic_t *)&nmi_watchdog_status, -EBUSY, 1)
+ == -EBUSY) {
+ if (!data->status) {
+ nmi_watchdog_status = 0;
+ lockup_detector_update_enable();
+ } else {
+ nmi_watchdog_status = -ENODEV;
+ /* turn offf watchdog_enabled forever */
+ lockup_detector_update_enable();
+ pr_info("Perf NMI watchdog permanently disabled\n");
+ }
+ }
+}
+
+/*
+ * Return 0, if a NMI watchdog is available. -ENODEV if unavailable. -EBUSY if
+ * undetermined at this stage, and async notifier will update later.
+ */
+int __weak __init watchdog_nmi_probe(watchdog_nmi_status_reporter notifier)
{
return hardlockup_detector_perf_init();
}
@@ -144,8 +167,12 @@ static void lockup_detector_update_enable(void)
watchdog_enabled = 0;
if (!watchdog_user_enabled)
return;
- if (nmi_watchdog_available && nmi_watchdog_user_enabled)
- watchdog_enabled |= NMI_WATCHDOG_ENABLED;
+ if (nmi_watchdog_user_enabled) {
+ if (nmi_watchdog_status == 0)
+ watchdog_enabled |= NMI_WATCHDOG_ENABLED;
+ else if (nmi_watchdog_status == -EBUSY)
+ watchdog_enabled |= NMI_WATCHDOG_UNDETERMINED;
+ }
if (soft_watchdog_user_enabled)
watchdog_enabled |= SOFT_WATCHDOG_ENABLED;
}
@@ -467,7 +494,8 @@ static void watchdog_enable(unsigned int cpu)
/* Initialize timestamp */
update_touch_ts();
/* Enable the perf event */
- if (watchdog_enabled & NMI_WATCHDOG_ENABLED)
+ if (watchdog_enabled &
+ (NMI_WATCHDOG_ENABLED | NMI_WATCHDOG_UNDETERMINED))
watchdog_nmi_enable(cpu);
}
@@ -682,7 +710,7 @@ int proc_watchdog(struct ctl_table *table, int write,
int proc_nmi_watchdog(struct ctl_table *table, int write,
void *buffer, size_t *lenp, loff_t *ppos)
{
- if (!nmi_watchdog_available && write)
+ if (!nmi_watchdog_status && write)
return -ENOTSUPP;
return proc_watchdog_common(NMI_WATCHDOG_ENABLED,
table, write, buffer, lenp, ppos);
@@ -748,7 +776,6 @@ void __init lockup_detector_init(void)
cpumask_copy(&watchdog_cpumask,
housekeeping_cpumask(HK_FLAG_TIMER));
- if (!watchdog_nmi_probe())
- nmi_watchdog_available = true;
+ nmi_watchdog_status = watchdog_nmi_probe(watchdog_nmi_report_capability);
lockup_detector_setup();
}
--
2.31.1
Powered by blists - more mailing lists