netdev - Re: [PATCH 1/6] net: page_pool: Add alloc stats and fast path stat

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] [day] [month] [year] [list]

Message-ID: <1302c302-1a4b-ff0d-c989-cbe851c92e06@redhat.com>
Date:   Wed, 2 Feb 2022 15:14:34 +0100
From:   Jesper Dangaard Brouer <jbrouer@...hat.com>
To:     Joe Damato <jdamato@...tly.com>, netdev@...r.kernel.org
Cc:     brouer@...hat.com, kuba@...nel.org, davem@...emloft.net,
        ilias.apalodimas@...aro.org, hawk@...nel.org,
        Tariq Toukan <ttoukan.linux@...il.com>,
        Saeed Mahameed <saeed@...nel.org>
Subject: Re: [PATCH 1/6] net: page_pool: Add alloc stats and fast path stat

On 26/01/2022 23.48, Joe Damato wrote:
 > @@ -86,6 +100,7 @@ struct page_pool_params {
 >
 >   struct page_pool {
 >   	struct page_pool_params p;
 > +	struct page_pool_stats ps;
 >
 >   	struct delayed_work release_dw;
The placement of page_pool_stats is problematic, due to cache-line 
placement.

It will be sharing a cache-line with page_pool_params, which is 
read-mostly.  As I say on benchmark email, this is likely the 
performance regression you saw.

I *do* know you have changed location (to percpu) in newer patch 
versions, but I think it is important to point out, that we have to be 
very careful about cache-line placements, and which part of 'struct 
page_pool' is accessed by which part of the code RX or 
DMA-TX-completion "return" code.

--Jesper