arm64: watchdog: add switch to select sdei_watchdog/pmu_watchdog

hulk inclusion
category: feature
bugzilla: NA
CVE: NA

-------------------------------------------------

On aarch64, we can compile both SDEI_WATCHODG and PMU_WATCHDOG code
instead of choosing one.  SDEI_WATCHDOG is used by default, and if
SDEI_WATCHDOG is disabled by kernel parameter "disable_sdei_nmi_watchdog",
PMU_WATCHDOG is used instead.
Signed-off-by: NJingyi Wang <wangjingyi11@huawei.com>
Signed-off-by: NWei Li <liwei391@huawei.com>
Reviewed-by: NXiongfeng Wang <wangxiongfeng2@huawei.com>
Signed-off-by: NYang Yingliang <yangyingliang@huawei.com>

arch/arm64/kernel/watchdog_sdei.c

@@ -24,7 +24,7 @@ static int sdei_watchdog_event_num;
 static bool disable_sdei_nmi_watchdog;
 static bool sdei_watchdog_registered;
 
-int watchdog_nmi_enable(unsigned int cpu)
+int watchdog_sdei_enable(unsigned int cpu)
 {
 	int ret;
 
@@ -47,7 +47,7 @@ int watchdog_nmi_enable(unsigned int cpu)
 	return 0;
 }
 
-void watchdog_nmi_disable(unsigned int cpu)
+void watchdog_sdei_disable(unsigned int cpu)
 {
 	int ret;
 
@@ -92,13 +92,10 @@ void sdei_watchdog_clear_eoi(void)
 		sdei_api_clear_eoi(SDEI_NMI_WATCHDOG_HWIRQ);
 }
 
-int __init watchdog_nmi_probe(void)
+int __init watchdog_sdei_probe(void)
 {
 	int ret;
 
-	if (disable_sdei_nmi_watchdog)
-		return -EINVAL;
-
 	if (!is_hyp_mode_available()) {
 		pr_err("Disable SDEI NMI Watchdog in VM\n");
 		return -EINVAL;
@@ -135,3 +132,17 @@ int __init watchdog_nmi_probe(void)
 
 	return 0;
 }
+
+static struct watchdog_operations arch_watchdog_ops = {
+	.watchdog_nmi_stop = &watchdog_nmi_stop,
+	.watchdog_nmi_start = &watchdog_nmi_start,
+	.watchdog_nmi_probe = &watchdog_sdei_probe,
+	.watchdog_nmi_enable = &watchdog_sdei_enable,
+	.watchdog_nmi_disable = &watchdog_sdei_disable,
+};
+
+void watchdog_ops_init(void)
+{
+	if (!disable_sdei_nmi_watchdog)
+		nmi_watchdog_ops = arch_watchdog_ops;
+}

include/linux/nmi.h

@@ -130,6 +130,17 @@ int watchdog_nmi_probe(void);
 int watchdog_nmi_enable(unsigned int cpu);
 void watchdog_nmi_disable(unsigned int cpu);
 
+struct watchdog_operations {
+	void (*watchdog_nmi_stop)(void);
+	void (*watchdog_nmi_start)(void);
+	int (*watchdog_nmi_probe)(void);
+	int (*watchdog_nmi_enable)(unsigned int cpu);
+	void (*watchdog_nmi_disable)(unsigned int cpu);
+};
+
+extern struct watchdog_operations nmi_watchdog_ops;
+void watchdog_ops_init(void);
+
 /**
  * touch_nmi_watchdog - restart NMI watchdog timeout.
  *

kernel/watchdog.c

@@ -50,7 +50,16 @@ struct cpumask watchdog_allowed_mask __read_mostly;
 struct cpumask watchdog_cpumask __read_mostly;
 unsigned long *watchdog_cpumask_bits = cpumask_bits(&watchdog_cpumask);
 
+struct watchdog_operations nmi_watchdog_ops = {
+	.watchdog_nmi_stop = &watchdog_nmi_stop,
+	.watchdog_nmi_start = &watchdog_nmi_start,
+	.watchdog_nmi_probe = &watchdog_nmi_probe,
+	.watchdog_nmi_enable = &watchdog_nmi_enable,
+	.watchdog_nmi_disable = &watchdog_nmi_disable,
+};
+
 #ifdef CONFIG_HARDLOCKUP_DETECTOR
+
 /*
  * Should we panic when a soft-lockup or hard-lockup occurs:
  */
@@ -496,7 +505,7 @@ static void watchdog_enable(unsigned int cpu)
 	__touch_watchdog();
 	/* Enable the perf event */
 	if (watchdog_enabled & NMI_WATCHDOG_ENABLED)
-		watchdog_nmi_enable(cpu);
+		nmi_watchdog_ops.watchdog_nmi_enable(cpu);
 }
 
 static void watchdog_disable(unsigned int cpu)
@@ -510,7 +519,7 @@ static void watchdog_disable(unsigned int cpu)
 	 * between disabling the timer and disabling the perf event causes
 	 * the perf NMI to detect a false positive.
 	 */
-	watchdog_nmi_disable(cpu);
+	nmi_watchdog_ops.watchdog_nmi_disable(cpu);
 	hrtimer_cancel(hrtimer);
 	wait_for_completion(this_cpu_ptr(&softlockup_completion));
 }
@@ -566,7 +575,7 @@ int lockup_detector_offline_cpu(unsigned int cpu)
 static void lockup_detector_reconfigure(void)
 {
 	cpus_read_lock();
-	watchdog_nmi_stop();
+	nmi_watchdog_ops.watchdog_nmi_stop();
 
 	softlockup_stop_all();
 	set_sample_period();
@@ -574,7 +583,7 @@ static void lockup_detector_reconfigure(void)
 	if (watchdog_enabled && watchdog_thresh)
 		softlockup_start_all();
 
-	watchdog_nmi_start();
+	nmi_watchdog_ops.watchdog_nmi_start();
 	cpus_read_unlock();
 	/*
 	 * Must be called outside the cpus locked section to prevent
@@ -612,9 +621,9 @@ static __init void lockup_detector_setup(void)
 static void lockup_detector_reconfigure(void)
 {
 	cpus_read_lock();
-	watchdog_nmi_stop();
+	nmi_watchdog_ops.watchdog_nmi_stop();
 	lockup_detector_update_enable();
-	watchdog_nmi_start();
+	nmi_watchdog_ops.watchdog_nmi_start();
 	cpus_read_unlock();
 }
 static inline void lockup_detector_setup(void)
@@ -772,15 +781,21 @@ int proc_watchdog_cpumask(struct ctl_table *table, int write,
 }
 #endif /* CONFIG_SYSCTL */
 
+void __weak watchdog_ops_init(void)
+{
+}
+
 void __init lockup_detector_init(void)
 {
+	watchdog_ops_init();
+
 	if (tick_nohz_full_enabled())
 		pr_info("Disabling watchdog on nohz_full cores by default\n");
 
 	cpumask_copy(&watchdog_cpumask,
 		     housekeeping_cpumask(HK_FLAG_TIMER));
 
-	if (!watchdog_nmi_probe())
+	if (!nmi_watchdog_ops.watchdog_nmi_probe())
 		nmi_watchdog_available = true;
 	lockup_detector_setup();
 }

lib/Kconfig.debug

@@ -835,11 +835,8 @@ config HARDLOCKUP_DETECTOR_PERF
 	bool
 	select SOFTLOCKUP_DETECTOR
 
-choice
-	prompt "aarch64 NMI watchdog method"
+menu "ARM64 NMI watchdog configuration"
 	depends on ARM64
-	help
-	  Watchdog implementation method configuration.
 
 config SDEI_WATCHDOG
 	bool "SDEI NMI Watchdog support"
@@ -852,7 +849,7 @@ config PMU_WATCHDOG
 	depends on PERF_EVENTS && HAVE_PERF_EVENTS_NMI
 	select HAVE_HARDLOCKUP_DETECTOR_PERF
 
-endchoice
+endmenu # "ARM64 NMI watchdog configuration"
 
 #
 # Enables a timestamp based low pass filter to compensate for perf based