[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <3c6100a1-78a3-4211-a62d-eea07db91b62@suse.cz>
Date: Tue, 5 Aug 2025 11:20:48 +0200
From: Vlastimil Babka <vbabka@...e.cz>
To: Adrian Huang <adrianhuang0701@...il.com>,
Andrew Morton <akpm@...ux-foundation.org>
Cc: David Hildenbrand <david@...hat.com>,
Lorenzo Stoakes <lorenzo.stoakes@...cle.com>, Liam.Howlett@...cle.com,
Mike Rapoport <rppt@...nel.org>, Suren Baghdasaryan <surenb@...gle.com>,
Michal Hocko <mhocko@...e.com>, Feng Tang <feng.79.tang@...il.com>,
linux-mm@...ck.org, linux-kernel@...r.kernel.org,
Adrian Huang <ahuang12@...ovo.com>
Subject: Re: [PATCH 1/1] mm: Correct misleading comment on mmap_lock field in
mm_struct
On 8/5/25 10:47 AM, Adrian Huang wrote:
> From: Adrian Huang <ahuang12@...ovo.com>
>
> The comment previously described the offset of mmap_lock as 0x120 (hex),
> which is misleading. The correct value is 120 (decimal), and using '0x120'
> could confuse readers trying to understand why the count and owner fields
> reside in separate cachelines.
>
> This change also removes an unnecessary space for improved formatting.
>
> Fixes: 2e3025434a6b ("mm: relocate 'write_protect_seq' in struct mm_struct")
> Signed-off-by: Adrian Huang <ahuang12@...ovo.com>
That seems all true so
Acked-by: Vlastimil Babka <vbabka@...e.cz>
But I wonder why we just hope this remains true "with some kernel
config" and don't employ some explicit alignment to make sure it's true
(except perhaps with some debug options like lockdep bloating the
structures, but we don't care about perfmance in such configs).
> ---
> include/linux/mm_types.h | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
> index 1ec273b06691..ec90bbf22e2b 100644
> --- a/include/linux/mm_types.h
> +++ b/include/linux/mm_types.h
> @@ -1027,9 +1027,9 @@ struct mm_struct {
> */
> /*
> * With some kernel config, the current mmap_lock's offset
> - * inside 'mm_struct' is at 0x120, which is very optimal, as
> + * inside 'mm_struct' is at 120, which is very optimal, as
> * its two hot fields 'count' and 'owner' sit in 2 different
> - * cachelines, and when mmap_lock is highly contended, both
> + * cachelines, and when mmap_lock is highly contended, both
> * of the 2 fields will be accessed frequently, current layout
> * will help to reduce cache bouncing.
> *
Powered by blists - more mailing lists