Re: [PATCH v1] mm, hugetlb: skip irrelevant nodes in hugetlb_show_meminfo()

[Muchun Song <songmuchun@bytedance.com>]

On Wed, Jun 22, 2022 at 12:00:42PM +0800, Gang Li wrote:
> show_free_areas() allows to filter out node specific data which is
> irrelevant to the allocation request. But hugetlb_show_meminfo() still
> show hugetlb on all nodes, which is redundant and unnecessary.
> 
> Use show_mem_node_skip() to skip irrelevant nodes in
> hugetlb_show_meminfo().
> 
> Signed-off-by: Gang Li <ligang.bdlg@bytedance.com>
> ---
>  include/linux/hugetlb.h | 4 ++--
>  include/linux/mm.h      | 1 +
>  mm/hugetlb.c            | 7 +++++--
>  mm/page_alloc.c         | 4 ++--
>  4 files changed, 10 insertions(+), 6 deletions(-)
> 
> diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
> index 642a39016f9a..1913ac6bf10a 100644
> --- a/include/linux/hugetlb.h
> +++ b/include/linux/hugetlb.h
> @@ -155,7 +155,7 @@ void __unmap_hugepage_range_final(struct mmu_gather *tlb,
>  			  struct page *ref_page, zap_flags_t zap_flags);
>  void hugetlb_report_meminfo(struct seq_file *);
>  int hugetlb_report_node_meminfo(char *buf, int len, int nid);
> -void hugetlb_show_meminfo(void);
> +void hugetlb_show_meminfo(unsigned int filter, nodemask_t *nodemask);
>  unsigned long hugetlb_total_pages(void);
>  vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
>  			unsigned long address, unsigned int flags);
> @@ -300,7 +300,7 @@ static inline int hugetlb_report_node_meminfo(char *buf, int len, int nid)
>  	return 0;
>  }
>  
> -static inline void hugetlb_show_meminfo(void)
> +static inline void hugetlb_show_meminfo(unsigned int filter, nodemask_t *nodemask)
>  {
>  }
>  
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 764dc0fdae5e..f72c1b21cde3 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -1873,6 +1873,7 @@ extern void pagefault_out_of_memory(void);
>   */
>  #define SHOW_MEM_FILTER_NODES		(0x0001u)	/* disallowed nodes */
>  
> +extern bool show_mem_node_skip(unsigned int flags, int nid, nodemask_t *nodemask);

Exporting this to the only user of HugeTLB is not a good idea.

>  extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
>  
>  #ifdef CONFIG_MMU
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index 98492733cc64..632826e6fea5 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -4486,7 +4486,7 @@ int hugetlb_report_node_meminfo(char *buf, int len, int nid)
>  			     nid, h->surplus_huge_pages_node[nid]);
>  }
>  
> -void hugetlb_show_meminfo(void)
> +void hugetlb_show_meminfo(unsigned int filter, nodemask_t *nodemask)
>  {

I suggest making this function show meminfo for a specific node.
like hugetlb_show_meminfo_node(int nid) which only show meminfo
for node @nid.

>  	struct hstate *h;
>  	int nid;
> @@ -4494,7 +4494,9 @@ void hugetlb_show_meminfo(void)
>  	if (!hugepages_supported())
>  		return;
>  
> -	for_each_node_state(nid, N_MEMORY)
> +	for_each_node_state(nid, N_MEMORY) {
> +		if (show_mem_node_skip(filter, nid, nodemask))
> +			continue;
>  		for_each_hstate(h)
>  			pr_info("Node %d hugepages_total=%u hugepages_free=%u hugepages_surp=%u hugepages_size=%lukB\n",
>  				nid,
> @@ -4502,6 +4504,7 @@ void hugetlb_show_meminfo(void)
>  				h->free_huge_pages_node[nid],
>  				h->surplus_huge_pages_node[nid],
>  				huge_page_size(h) / SZ_1K);
> +	}
>  }
>  
>  void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm)
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 84781094b478..5896b5a9101f 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -5951,7 +5951,7 @@ void si_meminfo_node(struct sysinfo *val, int nid)
>   * Determine whether the node should be displayed or not, depending on whether
>   * SHOW_MEM_FILTER_NODES was passed to show_free_areas().
>   */
> -static bool show_mem_node_skip(unsigned int flags, int nid, nodemask_t *nodemask)
> +bool show_mem_node_skip(unsigned int flags, int nid, nodemask_t *nodemask)
>  {
>  	if (!(flags & SHOW_MEM_FILTER_NODES))
>  		return false;
> @@ -6196,7 +6196,7 @@ void show_free_areas(unsigned int filter, nodemask_t *nodemask)
>  		printk(KERN_CONT "= %lukB\n", K(total));
>  	}
>

Just like the above case. We can check nid here instead of in hugetlb_show_meminfo().
Then you do not need to export show_mem_node_skip().

for_each_online_node(nid) {
	if (show_mem_node_skip(filter, nid, nodemask))
		continue;
	hugetlb_show_meminfo_node(nid);
}

Thanks.

> -	hugetlb_show_meminfo();
> +	hugetlb_show_meminfo(filter, nodemask);
>  
>  	printk("%ld total pagecache pages\n", global_node_page_state(NR_FILE_PAGES));
>  
> -- 
> 2.20.1
> 
>