linux-kernel - Re: [PATCH v2 10/28] gup: Turn hpage_pincount_sub() into page_pincount

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <5d29d866-2d5d-6a64-eb35-735468ff8649@nvidia.com>
Date:   Mon, 10 Jan 2022 22:40:46 -0800
From:   John Hubbard <jhubbard@...dia.com>
To:     "Matthew Wilcox (Oracle)" <willy@...radead.org>, linux-mm@...ck.org
Cc:     Christoph Hellwig <hch@...radead.org>,
        William Kucharski <william.kucharski@...cle.com>,
        linux-kernel@...r.kernel.org, Jason Gunthorpe <jgg@...pe.ca>
Subject: Re: [PATCH v2 10/28] gup: Turn hpage_pincount_sub() into
 page_pincount_sub()

On 1/9/22 20:23, Matthew Wilcox (Oracle) wrote:
> Remove an unnecessary VM_BUG_ON by handling pages both with and without
> a pincount field in page_pincount_sub().
> 
> Signed-off-by: Matthew Wilcox (Oracle) <willy@...radead.org>
> ---
>   mm/gup.c | 15 +++++++--------
>   1 file changed, 7 insertions(+), 8 deletions(-)
> 
> diff --git a/mm/gup.c b/mm/gup.c
> index 3ed9907f3c8d..aed48de3912e 100644
> --- a/mm/gup.c
> +++ b/mm/gup.c
> @@ -48,12 +48,15 @@ static void page_pincount_add(struct page *page, int refs)
>   		page_ref_add(page, refs * (GUP_PIN_COUNTING_BIAS - 1));
>   }
>   
> -static void hpage_pincount_sub(struct page *page, int refs)

OK, this is a correct transformation. But it does make the function, as
small as it is, rather hard to understand. It is now factored out at a
strange place, as evidenced in the difficulty in documenting it. Here's
my best attempt so far at documenting it here:

/*
  * Subtract @refs from the compound_pincount, if applicable. Otherwise, do
  * nothing. And in all cases, return the number of pages that should be passed
  * to put_page_refs().
  */

...which I think is worth adding, if we can't find a better factoring.

Either way, it's correct, so:


Reviewed-by: John Hubbard <jhubbard@...dia.com>


thanks,
-- 
John Hubbard
NVIDIA

> +static int page_pincount_sub(struct page *page, int refs)
>   {
> -	VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
>   	VM_BUG_ON_PAGE(page != compound_head(page), page);
>   
> -	atomic_sub(refs, compound_pincount_ptr(page));
> +	if (hpage_pincount_available(page))
> +		atomic_sub(refs, compound_pincount_ptr(page));
> +	else
> +		refs *= GUP_PIN_COUNTING_BIAS;
> +	return refs;
>   }
>   
>   /* Equivalent to calling put_page() @refs times. */
> @@ -177,11 +180,7 @@ static void put_compound_head(struct page *page, int refs, unsigned int flags)
>   	if (flags & FOLL_PIN) {
>   		mod_node_page_state(page_pgdat(page), NR_FOLL_PIN_RELEASED,
>   				    refs);
> -
> -		if (hpage_pincount_available(page))
> -			hpage_pincount_sub(page, refs);
> -		else
> -			refs *= GUP_PIN_COUNTING_BIAS;
> +		refs = page_pincount_sub(page, refs);
>   	}
>   
>   	put_page_refs(page, refs);