Re: [PATCH v2 3/3] mm/swap: remove swap_cache_info statistics

From: Huang, Ying
Date: Mon Jun 20 2022 - 04:08:52 EST


Miaohe Lin <linmiaohe@xxxxxxxxxx> writes:

> swap_cache_info are not statistics that could be easily used to tune system
> performance because they are not easily accessile. Also they can't provide
> really useful info when OOM occurs. Remove these statistics can also help
> mitigate unneeded global swap_cache_info cacheline contention.
>
> Suggested-by: David Hildenbrand <david@xxxxxxxxxx>
> Signed-off-by: Miaohe Lin <linmiaohe@xxxxxxxxxx>
> ---
> mm/swap_state.c | 17 -----------------
> 1 file changed, 17 deletions(-)
>
> diff --git a/mm/swap_state.c b/mm/swap_state.c
> index 0a2021fc55ad..41c6a6053d5c 100644
> --- a/mm/swap_state.c
> +++ b/mm/swap_state.c
> @@ -59,24 +59,11 @@ static bool enable_vma_readahead __read_mostly = true;
> #define GET_SWAP_RA_VAL(vma) \
> (atomic_long_read(&(vma)->swap_readahead_info) ? : 4)
>
> -#define INC_CACHE_INFO(x) data_race(swap_cache_info.x++)
> -#define ADD_CACHE_INFO(x, nr) data_race(swap_cache_info.x += (nr))
> -
> -static struct {
> - unsigned long add_total;
> - unsigned long del_total;
> - unsigned long find_success;
> - unsigned long find_total;
> -} swap_cache_info;
> -
> static atomic_t swapin_readahead_hits = ATOMIC_INIT(4);
>
> void show_swap_cache_info(void)
> {
> printk("%lu pages in swap cache\n", total_swapcache_pages());
> - printk("Swap cache stats: add %lu, delete %lu, find %lu/%lu\n",
> - swap_cache_info.add_total, swap_cache_info.del_total,
> - swap_cache_info.find_success, swap_cache_info.find_total);
> printk("Free swap = %ldkB\n",
> get_nr_swap_pages() << (PAGE_SHIFT - 10));
> printk("Total swap = %lukB\n", total_swap_pages << (PAGE_SHIFT - 10));
> @@ -133,7 +120,6 @@ int add_to_swap_cache(struct page *page, swp_entry_t entry,
> address_space->nrpages += nr;
> __mod_node_page_state(page_pgdat(page), NR_FILE_PAGES, nr);
> __mod_lruvec_page_state(page, NR_SWAPCACHE, nr);
> - ADD_CACHE_INFO(add_total, nr);
> unlock:
> xas_unlock_irq(&xas);
> } while (xas_nomem(&xas, gfp));
> @@ -172,7 +158,6 @@ void __delete_from_swap_cache(struct page *page,
> address_space->nrpages -= nr;
> __mod_node_page_state(page_pgdat(page), NR_FILE_PAGES, -nr);
> __mod_lruvec_page_state(page, NR_SWAPCACHE, -nr);
> - ADD_CACHE_INFO(del_total, nr);
> }
>
> /**
> @@ -348,12 +333,10 @@ struct page *lookup_swap_cache(swp_entry_t entry, struct vm_area_struct *vma,
> page = find_get_page(swap_address_space(entry), swp_offset(entry));
> put_swap_device(si);
>
> - INC_CACHE_INFO(find_total);
> if (page) {
> bool vma_ra = swap_use_vma_readahead();
> bool readahead;
>
> - INC_CACHE_INFO(find_success);
> /*
> * At the moment, we don't support PG_readahead for anon THP
> * so let's bail out rather than confusing the readahead stat.

This looks reasonable. And if we want to do some statistics for swap
cache in the future, we can use BPF, that is even more convenient.

Acked-by: "Huang, Ying" <ying.huang@xxxxxxxxx>

Best Regards,
Huang, Ying