netdev - Re: [PATCH net-next] skbuff: Optimize SKB coalescing for page pool case

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]

Date: Fri, 30 Jun 2023 16:07:09 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: Liang Chen <liangchen.linux@...il.com>
Cc: ilias.apalodimas@...aro.org, hawk@...nel.org, davem@...emloft.net,
 edumazet@...gle.com, pabeni@...hat.com, linyunsheng@...wei.com,
 netdev@...r.kernel.org
Subject: Re: [PATCH net-next] skbuff: Optimize SKB coalescing for page pool
 case

On Wed, 28 Jun 2023 20:11:50 +0800 Liang Chen wrote:
> +static inline void page_pool_page_ref(struct page *page)
> +{
> +	struct page *head_page = compound_head(page);
> +
> +	if (page_pool_is_pp_page(head_page) &&
> +			page_pool_is_pp_page_frag(head_page))
> +		atomic_long_inc(&head_page->pp_frag_count);
> +	else
> +		get_page(head_page);

AFAICT this is not correct, you cannot take an extra *pp* reference 
on a pp page, unless it is a frag. If the @to skb is a pp skb (and it
must be, if we get here) we will have two skbs with pp_recycle = 1.
Which means if they both get freed at the same time they will both
try to return / release the page.

I haven't looked at Yunsheng's v5 yet, but that's why I was asking 
to rename the pp_frag_count to pp_ref_count. pp_frag_count is a true
refcount (rather than skb->pp_recycle acting as a poor man's single
shot refcount), so in case of frag'd pages / after Yunsheng's work
we will be able to take new refs.

Long story short please wait until Yunsheng's changes are finalized.