When lockup_detector_init()->watchdog_hardlockup_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 and try to initialize the watchdog once again later. The delayed probe is called using workqueues. It need to allocate memory and must be proceed in a normal context. The delayed probe is able to use if watchdog_hardlockup_probe() returns non-zero which means the return code returned when PMU is not ready yet. Provide an API - lockup_detector_retry_init() for anyone who needs to delayed init lockup detector if they had ever failed at lockup_detector_init(). The original assumption is: nobody should use delayed probe after lockup_detector_check() which has __init attribute. That is, anyone uses this API must call between lockup_detector_init() and lockup_detector_check(), and the caller must have __init attribute
Unified · Split
| @@ -13,6 +13,7 @@ · 12 unmodified lines | ||
| 13 | 13 | |
| 14 | 14 | #ifdef CONFIG_LOCKUP_DETECTOR |
| 15 | 15 | void lockup_detector_init(void); |
| 16 | +void lockup_detector_retry_init(void); | |
| 16 | 17 | void lockup_detector_soft_poweroff(void); |
| 17 | 18 | void lockup_detector_cleanup(void); |
| 18 | 19 | |
| @@ -32,6 +33,7 @@ extern int sysctl_hardlockup_all_cpu_backtrace; · 13 unmodified lines | ||
| 32 | 33 | |
| 33 | 34 | #else /* CONFIG_LOCKUP_DETECTOR */ |
| 34 | 35 | static inline void lockup_detector_init(void) { } |
| 36 | +static inline void lockup_detector_retry_init(void) { } | |
| 35 | 37 | static inline void lockup_detector_soft_poweroff(void) { } |
| 36 | 38 | static inline void lockup_detector_cleanup(void) { } |
| 37 | 39 | #endif /* !CONFIG_LOCKUP_DETECTOR */ |
| @@ -208,7 +208,13 @@ void __weak watchdog_hardlockup_enable(unsigned int cpu) { } · 207 unmodified lines | ||
| 208 | 208 | |
| 209 | 209 | void __weak watchdog_hardlockup_disable(unsigned int cpu) { } |
| 210 | 210 | |
| 211 | −/* Return 0, if a hardlockup watchdog is available. Error code otherwise */ | |
| 211 | +/* | |
| 212 | + * Watchdog-detector specific API. | |
| 213 | + * | |
| 214 | + * Return 0 when hardlockup watchdog is available, negative value otherwise. | |
| 215 | + * Note that the negative value means that a delayed probe might | |
| 216 | + * succeed later. | |
| 217 | + */ | |
| 212 | 218 | int __weak __init watchdog_hardlockup_probe(void) |
| 213 | 219 | { |
| 214 | 220 | /* |
| @@ -954,6 +960,62 @@ static void __init watchdog_sysctl_init(void) · 739 unmodified lines | ||
| 954 | 960 | #define watchdog_sysctl_init() do { } while (0) |
| 955 | 961 | #endif /* CONFIG_SYSCTL */ |
| 956 | 962 | |
| 963 | +static void __init lockup_detector_delay_init(struct work_struct *work); | |
| 964 | +static bool allow_lockup_detector_init_retry __initdata; | |
| 965 | + | |
| 966 | +static struct work_struct detector_work __initdata = | |
| 967 | + __WORK_INITIALIZER(detector_work, lockup_detector_delay_init); | |
| 968 | + | |
| 969 | +static void __init lockup_detector_delay_init(struct work_struct *work) | |
| 970 | +{ | |
| 971 | + int ret; | |
| 972 | + | |
| 973 | + ret = watchdog_hardlockup_probe(); | |
| 974 | + if (ret) { | |
| 975 | + pr_info("Delayed init of the lockup detector failed: %d\n", ret); | |
| 976 | + pr_info("Hard watchdog permanently disabled\n"); | |
| 977 | + return; | |
| 978 | + } | |
| 979 | + | |
| 980 | + allow_lockup_detector_init_retry = false; | |
| 981 | + | |
| 982 | + watchdog_hardlockup_available = true; | |
| 983 | + lockup_detector_setup(); | |
| 984 | +} | |
| 985 | + | |
| 986 | +/* | |
| 987 | + * lockup_detector_retry_init - retry init lockup detector if possible. | |
| 988 | + * | |
| 989 | + * Retry hardlockup detector init. It is useful when it requires some | |
| 990 | + * functionality that has to be initialized later on a particular | |
| 991 | + * platform. | |
| 992 | + */ | |
| 993 | +void __init lockup_detector_retry_init(void) | |
| 994 | +{ | |
| 995 | + /* Must be called before late init calls */ | |
| 996 | + if (!allow_lockup_detector_init_retry) | |
| 997 | + return; | |
| 998 | + | |
| 999 | + schedule_work(&detector_work); | |
| 1000 | +} | |
| 1001 | + | |
| 1002 | +/* | |
| 1003 | + * Ensure that optional delayed hardlockup init is proceed before | |
| 1004 | + * the init code and memory is freed. | |
| 1005 | + */ | |
| 1006 | +static int __init lockup_detector_check(void) | |
| 1007 | +{ | |
| 1008 | + /* Prevent any later retry. */ | |
| 1009 | + allow_lockup_detector_init_retry = false; | |
| 1010 | + | |
| 1011 | + /* Make sure no work is pending. */ | |
| 1012 | + flush_work(&detector_work); | |
| 1013 | + | |
| 1014 | + return 0; | |
| 1015 | + | |
| 1016 | +} | |
| 1017 | +late_initcall_sync(lockup_detector_check); | |
| 1018 | + | |
| 957 | 1019 | void __init lockup_detector_init(void) |
| 958 | 1020 | { |
| 959 | 1021 | if (tick_nohz_full_enabled()) |
| @@ -964,6 +1026,9 @@ void __init lockup_detector_init(void) · 4 unmodified lines | ||
| 964 | 1026 | |
| 965 | 1027 | if (!watchdog_hardlockup_probe()) |
| 966 | 1028 | watchdog_hardlockup_available = true; |
| 1029 | + else | |
| 1030 | + allow_lockup_detector_init_retry = true; | |
| 1031 | + | |
| 967 | 1032 | lockup_detector_setup(); |
| 968 | 1033 | watchdog_sysctl_init(); |
| 969 | 1034 | } |