[PATCH v7 2/8] x86/crash: Introduce new options to support cpu and memory hotplug

Eric DeVolder eric.devolder at oracle.com
Tue Apr 26 13:08:26 PDT 2022


Boris,
Many thanks for the time taken to illustrate for me!
I've one question below.
Eric

On 4/25/22 14:25, Borislav Petkov wrote:
> On Tue, Apr 19, 2022 at 04:58:47PM -0500, Eric DeVolder wrote:
>> So in taking this concept and looking at, in particular, patch 4/8
>> "crash: add generic infrastructure for crash hotplug support", I'm not
>> exactly sure how to apply this technique.
> 
> So I took your patch 4 and maimed into what I think it should look like,
> see below.
> 
> Now there's a single ifdef there and the registration routines are
> wrapped in IS_ENABLED() so that you register a callback only when the
> respective stuff - HOTPLUG_CPU or MEMORY_HOTPLUG - is enabled. Otherwise
> the couple of functions are unused but that's not that big of a deal.
> 
> I've also fixed up some other issues I've encountered along the way.
> 
> Holler if there are questions.
> 
> HTH.
> 
> ---
> diff --git a/include/linux/crash_core.h b/include/linux/crash_core.h
> index de62a722431e..b3c32e04d3f0 100644
> --- a/include/linux/crash_core.h
> +++ b/include/linux/crash_core.h
> @@ -84,4 +84,12 @@ int parse_crashkernel_high(char *cmdline, unsigned long long system_ram,
>   int parse_crashkernel_low(char *cmdline, unsigned long long system_ram,
>   		unsigned long long *crash_size, unsigned long long *crash_base);
>   
> +#define KEXEC_CRASH_HP_REMOVE_CPU	0
> +#define KEXEC_CRASH_HP_ADD_CPU		1
> +#define KEXEC_CRASH_HP_REMOVE_MEMORY	2
> +#define KEXEC_CRASH_HP_ADD_MEMORY	3
> +
> +struct kimage;
> +void arch_crash_handle_hotplug_event(struct kimage *image, unsigned int hp_action,
> +				     unsigned int cpu);
>   #endif /* LINUX_CRASH_CORE_H */
> diff --git a/include/linux/kexec.h b/include/linux/kexec.h
> index 02daff1f47dd..d907a1f0d3da 100644
> --- a/include/linux/kexec.h
> +++ b/include/linux/kexec.h
> @@ -307,13 +307,11 @@ struct kimage {
>   	/* Information for loading purgatory */
>   	struct purgatory_info purgatory_info;
>   
> -#ifdef CONFIG_CRASH_HOTPLUG
>   	bool hotplug_event;
>   	unsigned int offlinecpu;
>   	bool elfcorehdr_index_valid;
>   	int elfcorehdr_index;
>   #endif
> -#endif
>   
>   #ifdef CONFIG_IMA_KEXEC
>   	/* Virtual address of IMA measurement buffer for kexec syscall */
> @@ -329,15 +327,6 @@ struct kimage {
>   	unsigned long elf_load_addr;
>   };
>   
> -#ifdef CONFIG_CRASH_HOTPLUG
> -void arch_crash_hotplug_handler(struct kimage *image,
> -	unsigned int hp_action, unsigned int cpu);
> -#define KEXEC_CRASH_HP_REMOVE_CPU   0
> -#define KEXEC_CRASH_HP_ADD_CPU      1
> -#define KEXEC_CRASH_HP_REMOVE_MEMORY 2
> -#define KEXEC_CRASH_HP_ADD_MEMORY   3
> -#endif /* CONFIG_CRASH_HOTPLUG */
> -
>   /* kexec interface functions */
>   extern void machine_kexec(struct kimage *image);
>   extern int machine_kexec_prepare(struct kimage *image);
> diff --git a/kernel/crash_core.c b/kernel/crash_core.c
> index ecf746243ab2..a6c4ee1d5c86 100644
> --- a/kernel/crash_core.c
> +++ b/kernel/crash_core.c
> @@ -497,57 +497,50 @@ static int __init crash_save_vmcoreinfo_init(void)
>   
>   subsys_initcall(crash_save_vmcoreinfo_init);
>   
> -#ifdef CONFIG_CRASH_HOTPLUG
> -void __weak arch_crash_hotplug_handler(struct kimage *image,
> -	unsigned int hp_action, unsigned int cpu)
> +void __weak arch_crash_handle_hotplug_event(struct kimage *image, unsigned int hp_action,
> +					    unsigned int cpu)
>   {
> -	pr_warn("crash hp: %s not implemented", __func__);
> +	WARN(1, "crash hotplug handler not implemented");
>   }
>   
> -static void crash_hotplug_handler(unsigned int hp_action,
> -	unsigned int cpu)
> +#if defined(CONFIG_HOTPLUG_CPU) || defined(CONFIG_MEMORY_HOTPLUG)
> +static void handle_hotplug_event(unsigned int hp_action, unsigned int cpu)
>   {
> -	/* Obtain lock while changing crash information */
> -	if (!mutex_trylock(&kexec_mutex))
> +	if (!kexec_crash_image)
>   		return;
Why is it safe to examine kexec_crash_image outside the mutex? As I understand it, there is still 
the (very rare) opportunity for a kdump load/unload initiated via userland and this check to 
collide. (Similarly, I believe the mutex entry is almost always assured/likely.)

>   
> -	/* Check kdump is loaded */
> -	if (kexec_crash_image) {
> -		pr_debug("crash hp: hp_action %u, cpu %u", hp_action, cpu);
> +	mutex_lock(&kexec_mutex);
>   
> -		/* Needed in order for the segments to be updated */
> -		arch_kexec_unprotect_crashkres();
> +	pr_debug("crash hotplug: hp_action %u, cpu %u", hp_action, cpu);
>   
> -		/* Flag to differentiate between normal load and hotplug */
> -		kexec_crash_image->hotplug_event = true;
> +	/* Needed in order for the segments to be updated */
> +	arch_kexec_unprotect_crashkres();
>   
> -		/* Now invoke arch-specific update handler */
> -		arch_crash_hotplug_handler(kexec_crash_image, hp_action, cpu);
> +	/* Flag to differentiate between normal load and hotplug */
> +	kexec_crash_image->hotplug_event = true;
>   
> -		/* No longer handling a hotplug event */
> -		kexec_crash_image->hotplug_event = false;
> +	/* Now invoke arch-specific update handler */
> +	arch_crash_handle_hotplug_event(kexec_crash_image, hp_action, cpu);
>   
> -		/* Change back to read-only */
> -		arch_kexec_protect_crashkres();
> -	}
> +	/* No longer handling a hotplug event */
> +	kexec_crash_image->hotplug_event = false;
> +
> +	/* Change back to read-only */
> +	arch_kexec_protect_crashkres();
>   
>   	/* Release lock now that update complete */
>   	mutex_unlock(&kexec_mutex);
>   }
>   
> -#if defined(CONFIG_MEMORY_HOTPLUG)
> -static int crash_memhp_notifier(struct notifier_block *nb,
> -	unsigned long val, void *v)
> +static int crash_memhp_notifier(struct notifier_block *nb, unsigned long val, void *v)
>   {
> -	struct memory_notify *mhp = v;
> -
>   	switch (val) {
>   	case MEM_ONLINE:
> -		crash_hotplug_handler(KEXEC_CRASH_HP_ADD_MEMORY, -1U);
> +		handle_hotplug_event(KEXEC_CRASH_HP_ADD_MEMORY, -1U);
>   		break;
>   
>   	case MEM_OFFLINE:
> -		crash_hotplug_handler(KEXEC_CRASH_HP_REMOVE_MEMORY, -1U);
> +		handle_hotplug_event(KEXEC_CRASH_HP_REMOVE_MEMORY, -1U);
>   		break;
>   	}
>   	return NOTIFY_OK;
> @@ -557,38 +550,33 @@ static struct notifier_block crash_memhp_nb = {
>   	.notifier_call = crash_memhp_notifier,
>   	.priority = 0
>   };
> -#endif
>   
> -#if defined(CONFIG_HOTPLUG_CPU)
>   static int crash_cpuhp_online(unsigned int cpu)
>   {
> -	crash_hotplug_handler(KEXEC_CRASH_HP_ADD_CPU, cpu);
> +	handle_hotplug_event(KEXEC_CRASH_HP_ADD_CPU, cpu);
>   	return 0;
>   }
>   
>   static int crash_cpuhp_offline(unsigned int cpu)
>   {
> -	crash_hotplug_handler(KEXEC_CRASH_HP_REMOVE_CPU, cpu);
> +	handle_hotplug_event(KEXEC_CRASH_HP_REMOVE_CPU, cpu);
>   	return 0;
>   }
> -#endif
>   
>   static int __init crash_hotplug_init(void)
>   {
>   	int result = 0;
>   
> -#if defined(CONFIG_MEMORY_HOTPLUG)
> -	register_memory_notifier(&crash_memhp_nb);
> -#endif
> +	if (IS_ENABLED(CONFIG_MEMORY_HOTPLUG))
> +		register_memory_notifier(&crash_memhp_nb);
>   
> -#if defined(CONFIG_HOTPLUG_CPU)
> -	result = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN,
> -				"crash/cpuhp",
> -				crash_cpuhp_online, crash_cpuhp_offline);
> -#endif
> +	if (IS_ENABLED(CONFIG_HOTPLUG_CPU))
> +		result = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN,
> +						   "crash/cpuhp",
> +						   crash_cpuhp_online,
> +						   crash_cpuhp_offline);
>   
>   	return result;
>   }
> -
>   subsys_initcall(crash_hotplug_init);
> -#endif /* CONFIG_CRASH_HOTPLUG */
> +#endif
> 



More information about the kexec mailing list