linux-kernel - Re: [PATCH net-next v2 3/3] net: stmmac: Set CIC bit only for TX queues with COE

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: <20250821071739.2e05316a@kernel.org>
Date: Thu, 21 Aug 2025 07:17:39 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: "G Thomas, Rohan" <rohan.g.thomas@...era.com>
Cc: Rohan G Thomas via B4 Relay
 <devnull+rohan.g.thomas.altera.com@...nel.org>, Andrew Lunn
 <andrew+netdev@...n.ch>, "David S. Miller" <davem@...emloft.net>, Eric
 Dumazet <edumazet@...gle.com>, Paolo Abeni <pabeni@...hat.com>, Maxime
 Coquelin <mcoquelin.stm32@...il.com>, Alexandre Torgue
 <alexandre.torgue@...s.st.com>, Serge Semin <fancer.lancer@...il.com>,
 Romain Gantois <romain.gantois@...tlin.com>, Jose Abreu
 <Jose.Abreu@...opsys.com>, Ong Boon Leong <boon.leong.ong@...el.com>,
 netdev@...r.kernel.org, linux-stm32@...md-mailman.stormreply.com,
 linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org, Matthew
 Gerlach <matthew.gerlach@...era.com>
Subject: Re: [PATCH net-next v2 3/3] net: stmmac: Set CIC bit only for TX
 queues with COE

On Thu, 21 Aug 2025 19:20:02 +0530 G Thomas, Rohan wrote:
> > To be clear -- this is just for context. I don't understand the details
> > of what the CIC bit controls, so the final decision is up to you.  
> 
> Currently, in the stmmac driver, even though tmo_request_checksum is not
> implemented, checksum offloading is still effectively enabled for AF_XDP
> frames, as CIC bit for tx desc are set, which implies checksum
> calculation and insertion by hardware for IP packets. So, I'm thinking
> it is better to keep it as false only for queues that do not support
> COE.

Oh, so the device parses the packet and inserts the checksum whether
user asked for it or not? Damn, I guess it may indeed be too late
to fix, but that certainly _not_ how AF_XDP is supposed to work.
The frame should not be modified without user asking for it..