[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <Y35AAV2H4xiPZYed@casper.infradead.org>
Date: Wed, 23 Nov 2022 15:45:05 +0000
From: Matthew Wilcox <willy@...radead.org>
To: wonder_rock@....com
Cc: akpm@...ux-foundation.org, linux-mm@...ck.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/1] mm/page_alloc: merge two get_pfnblock_migratetype()
into one
On Wed, Nov 23, 2022 at 09:40:43PM +0800, wonder_rock@....com wrote:
> From: Deyan Wang <wonder_rock@....com>
>
> It is safe to get_pfnblock_migratetype() in critical section,
> so just call get_pfnblock_migratetype() once.
No, you don't understand. zone->lock is HOT and we want to avoid
calling get_pfnblock_migratetype() under the lock, if we can avoid it.
Which we usually can.
> Signed-off-by: Deyan Wang <wonder_rock@....com>
> ---
> mm/page_alloc.c | 7 +------
> 1 file changed, 1 insertion(+), 6 deletions(-)
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 6652df1f5cc8..c7ba9d65d1c0 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -1704,13 +1704,8 @@ static void __free_pages_ok(struct page *page, unsigned int order,
> if (!free_pages_prepare(page, order, true, fpi_flags))
> return;
>
> - migratetype = get_pfnblock_migratetype(page, pfn);
> -
> spin_lock_irqsave(&zone->lock, flags);
> - if (unlikely(has_isolate_pageblock(zone) ||
> - is_migrate_isolate(migratetype))) {
> - migratetype = get_pfnblock_migratetype(page, pfn);
> - }
> + migratetype = get_pfnblock_migratetype(page, pfn);
> __free_one_page(page, pfn, zone, order, migratetype, fpi_flags);
> spin_unlock_irqrestore(&zone->lock, flags);
>
> --
> 2.25.1
>
>
Powered by blists - more mailing lists