netdev - Re: [PATCH 1/3] net: mvpp2: classifier flow remove tagged

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: <20230315075638.x7osulon5zqn7spu@SvensMacbookPro.hq.voleatech.com>
Date:   Wed, 15 Mar 2023 08:56:38 +0100
From:   Sven Auhagen <sven.auhagen@...eatech.de>
To:     Jakub Kicinski <kuba@...nel.org>
Cc:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "mw@...ihalf.com" <mw@...ihalf.com>,
        "linux@...linux.org.uk" <linux@...linux.org.uk>,
        "davem@...emloft.net" <davem@...emloft.net>,
        "maxime.chevallier@...tlin.com" <maxime.chevallier@...tlin.com>
Subject: Re: [PATCH 1/3] net: mvpp2: classifier flow remove tagged

On Wed, Mar 15, 2023 at 07:58:53AM +0100, Jakub Kicinski wrote:
> On Sat, 11 Mar 2023 08:09:48 +0100 Sven Auhagen wrote:
> > The classifier attribute MVPP22_CLS_HEK_TAGGED
> > has no effect in the definition and is filtered out by default.
> 
> filtered out in the code somewhere? 

It is filtered out in mvpp2_port_rss_hash_opts_set in this line:

	hash_opts = flow->supported_hash_opts & requested_opts;

Please see my email to Maxime which explains why the HEK can not use it
in case of a 5 Tuple.

> 
> > Even if it is applied to the classifier, it would discard double
> > or tripple tagged vlans.
> 
> So the MVPP22_CLS_HEK_TAGGED change is a nop?
> 

Yes, it is currently not applied to the classifier.

> > Also add missing IP Fragmentation Flag.
> 
> What's the impact of this change? Seems like it should be a separate
> patch.

I will seperate it out in a new patch.
The classifier should distinguish between fragmented and unfragmented packets.
At the moment this is not correct and all matches are only for unfragmented packets.