Re: [PATCH v2] drivers/base/node.c: Simplify unregister_memory_block_under_nodes()

[David Hildenbrand]

On 19.07.19 15:52, David Hildenbrand wrote:
> We don't allow to offline memory block devices that belong to multiple
> numa nodes. Therefore, such devices can never get removed. It is
> sufficient to process a single node when removing the memory block. No
> need to iterate over each and every PFN.
> 
> We already have the nid stored for each memory block. Make sure that
> the nid always has a sane value.
> 
> Please note that checking for node_online(nid) is not required. If we
> would have a memory block belonging to a node that is no longer offline,
> then we would have a BUG in the node offlining code.
> 
> Cc: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> Cc: "Rafael J. Wysocki" <rafael@xxxxxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> Cc: David Hildenbrand <david@xxxxxxxxxx>
> Cc: Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx>
> Cc: Pavel Tatashin <pasha.tatashin@xxxxxxxxxx>
> Cc: Michal Hocko <mhocko@xxxxxxxx>
> Cc: Oscar Salvador <osalvador@xxxxxxx>
> Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>
> ---
> 
> v1 -> v2:
> - Remove the "mixed nid" part, add a comment instead. Drop the warning.
> 
> ---
>  drivers/base/memory.c |  1 +
>  drivers/base/node.c   | 39 +++++++++++++++------------------------
>  2 files changed, 16 insertions(+), 24 deletions(-)
> 
> diff --git a/drivers/base/memory.c b/drivers/base/memory.c
> index 20c39d1bcef8..154d5d4a0779 100644
> --- a/drivers/base/memory.c
> +++ b/drivers/base/memory.c
> @@ -674,6 +674,7 @@ static int init_memory_block(struct memory_block **memory,
>  	mem->state = state;
>  	start_pfn = section_nr_to_pfn(mem->start_section_nr);
>  	mem->phys_device = arch_get_memory_phys_device(start_pfn);
> +	mem->nid = NUMA_NO_NODE;
>  
>  	ret = register_memory(mem);
>  
> diff --git a/drivers/base/node.c b/drivers/base/node.c
> index 75b7e6f6535b..840c95baa1d8 100644
> --- a/drivers/base/node.c
> +++ b/drivers/base/node.c
> @@ -759,8 +759,6 @@ static int register_mem_sect_under_node(struct memory_block *mem_blk,
>  	int ret, nid = *(int *)arg;
>  	unsigned long pfn, sect_start_pfn, sect_end_pfn;
>  
> -	mem_blk->nid = nid;
> -
>  	sect_start_pfn = section_nr_to_pfn(mem_blk->start_section_nr);
>  	sect_end_pfn = section_nr_to_pfn(mem_blk->end_section_nr);
>  	sect_end_pfn += PAGES_PER_SECTION - 1;
> @@ -789,6 +787,13 @@ static int register_mem_sect_under_node(struct memory_block *mem_blk,
>  			if (page_nid != nid)
>  				continue;
>  		}
> +
> +		/*
> +		 * If this memory block spans multiple nodes, we only indicate
> +		 * the last processed node.
> +		 */
> +		mem_blk->nid = nid;
> +
>  		ret = sysfs_create_link_nowarn(&node_devices[nid]->dev.kobj,
>  					&mem_blk->dev.kobj,
>  					kobject_name(&mem_blk->dev.kobj));
> @@ -804,32 +809,18 @@ static int register_mem_sect_under_node(struct memory_block *mem_blk,
>  }
>  
>  /*
> - * Unregister memory block device under all nodes that it spans.
> - * Has to be called with mem_sysfs_mutex held (due to unlinked_nodes).
> + * Unregister a memory block device under the node it spans. Memory blocks
> + * with multiple nodes cannot be offlined and therefore also never be removed.
>   */
>  void unregister_memory_block_under_nodes(struct memory_block *mem_blk)
>  {
> -	unsigned long pfn, sect_start_pfn, sect_end_pfn;
> -	static nodemask_t unlinked_nodes;
> -
> -	nodes_clear(unlinked_nodes);
> -	sect_start_pfn = section_nr_to_pfn(mem_blk->start_section_nr);
> -	sect_end_pfn = section_nr_to_pfn(mem_blk->end_section_nr);
> -	for (pfn = sect_start_pfn; pfn <= sect_end_pfn; pfn++) {
> -		int nid;
> +	if (mem_blk->nid == NUMA_NO_NODE)
> +		return;
>  
> -		nid = get_nid_for_pfn(pfn);
> -		if (nid < 0)
> -			continue;
> -		if (!node_online(nid))
> -			continue;
> -		if (node_test_and_set(nid, unlinked_nodes))
> -			continue;
> -		sysfs_remove_link(&node_devices[nid]->dev.kobj,
> -			 kobject_name(&mem_blk->dev.kobj));
> -		sysfs_remove_link(&mem_blk->dev.kobj,
> -			 kobject_name(&node_devices[nid]->dev.kobj));
> -	}
> +	sysfs_remove_link(&node_devices[mem_blk->nid]->dev.kobj,
> +			  kobject_name(&mem_blk->dev.kobj));
> +	sysfs_remove_link(&mem_blk->dev.kobj,
> +			  kobject_name(&node_devices[mem_blk->nid]->dev.kobj));
>  }
>  
>  int link_mem_sections(int nid, unsigned long start_pfn, unsigned long end_pfn)
> 


Just a note that this was actually also a bugfix as noted by Chris.

If the memory we are removing was never onlined,
get_nid_for_pfn()->pfn_to_nid() will return garbage. Removing will
succeed but links will remain in place.

Can be triggered by

1. hotplugging a DIMM to node 1
2. not onlining the memory blocks
3. unplugging it
4. re-plugging it to node 1

We will trigger the BUG_ON(ret) in add_memory_resource(), because
link_mem_sections() will return with -EEXIST.

-- 
Thanks,

David / dhildenb