Re: [PATCH 1/2] mm,memory_hotplug: Implement numa node notifier

[Harry Yoo <harry.yoo@oracle.com>]

On Tue, Apr 01, 2025 at 11:27:15AM +0200, Oscar Salvador wrote:
> There are at least four consumers of hotplug_memory_notifier that what they
> really are interested in is whether any numa node changed its state, e.g: going
> from being memory aware to becoming memoryless.
> 
> Implement a specific notifier for numa nodes when their state gets changed,
> and have those consumers that only care about numa node state changes use it.
> 
> Signed-off-by: Oscar Salvador <osalvador@suse.de>
> ---

Hi Oscar,

Please Cc harry.yoo@oracle.com instead of 42.hyeyoo@gmail.com
in the future—I recently changed my email.

FWIW, I reviewed this patch and it looks good to me:

Reviewed-by: Harry Yoo <harry.yoo@oracle.com>

>  drivers/acpi/numa/hmat.c  |  6 +--
>  drivers/base/node.c       | 19 +++++++++
>  drivers/cxl/core/region.c | 14 +++----
>  drivers/cxl/cxl.h         |  4 +-
>  include/linux/memory.h    | 38 ++++++++++++++++++
>  kernel/cgroup/cpuset.c    |  2 +-
>  mm/memory-tiers.c         |  8 ++--
>  mm/memory_hotplug.c       | 84 +++++++++++++++++++++++++++++----------
>  mm/slub.c                 | 22 +++++-----
>  9 files changed, 148 insertions(+), 49 deletions(-)
>  
> diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
> index 75401866fb76..4bb9ff282ec9 100644
> --- a/mm/memory_hotplug.c
> +++ b/mm/memory_hotplug.c
> @@ -1963,7 +1987,9 @@ int offline_pages(unsigned long start_pfn, unsigned long nr_pages,
>  	unsigned long pfn, managed_pages, system_ram_pages = 0;
>  	const int node = zone_to_nid(zone);
>  	unsigned long flags;
> -	struct memory_notify arg;
> +	struct memory_notify mem_arg;
> +	struct node_notify node_arg;
> +	bool cancel_mem_notifier_on_err = false, cancel_node_notifier_on_err = false;
>  	char *reason;
>  	int ret;
>  

[...]

> +	if (node_arg.status_change_nid >= 0) {
>  		kcompactd_stop(node);
>  		kswapd_stop(node);
> +		/*Node went memoryless. Notifiy interested consumers */
> +		node_notify(NODE_BECAME_MEMORYLESS, &node_arg);
>  	}

nit: /* Node went memoryless. ... */
       ^ a whitespace omitted.

>  	writeback_set_ratelimit();
>  
> -	memory_notify(MEM_OFFLINE, &arg);
> +	memory_notify(MEM_OFFLINE, &mem_arg);
>  	remove_pfn_range_from_zone(zone, start_pfn, nr_pages);
>  	return 0;

> diff --git a/mm/slub.c b/mm/slub.c
> index 184fd2b14758..74350f6c8ddd 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -5928,10 +5928,10 @@ static int slab_mem_going_offline_callback(void *arg)
>  
>  static void slab_mem_offline_callback(void *arg)
>  {
> -	struct memory_notify *marg = arg;
> +	struct node_notify *narg = arg;
>  	int offline_node;
>  
> -	offline_node = marg->status_change_nid_normal;
> +	offline_node = narg->status_change_nid_normal;
>  
>  	/*
>  	 * If the node still has available memory. we need kmem_cache_node

I was wondering if this offline_node check is still necessary after
this patch, but as SLUB is still notified for N_HIGH_MEMORY-only nodes,
it is necessary.

-- 
Cheers,
Harry (formerly known as Hyeonggon)