netdev - Re: [PATCH net-next] atl1c: add 4 RX/TX queue support for Mikrotik 10/25G NIC

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]

Date:   Tue, 25 May 2021 10:27:52 +0300
From:   Gatis Peisenieks <gatis@...rotik.com>
To:     Chris Snook <chris.snook@...il.com>
Cc:     "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        Heiner Kallweit <hkallweit1@...il.com>,
        jesse.brandeburg@...el.com, dchickles@...vell.com,
        tully@...rotik.com, Eric Dumazet <eric.dumazet@...il.com>,
        netdev <netdev@...r.kernel.org>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH net-next] atl1c: add 4 RX/TX queue support for Mikrotik
 10/25G NIC

Chris, thank you for taking a look at this!

In my experience L4 hashing (adding TCP/UDP ports to the hash to
determine rx queue) can cause problems with fragmented packets when
packet parser ignores the "More Fragments" and/or "Fragment Offset"
fields of the IPv4 header. Only the first fragment contains the ports,
so if parser blindly assumes ports to be at start of L4 offset, then
packets belonging to same connection get scattered among the rx queues
which is not good for performance.

Mikrotik 10/25G NIC RX parser stops at L3 if it sees any of those set.
Essentially it falls back to L2/L3 hashing for fragmented packets.
So it is ok in that regard.

On 2021-05-24 22:52, Chris Snook wrote:
> Is the L4 part of that hash configurable? That sort of thing tends to
> cause performance problems for fragmenting workloads, such as NFS over
> UDP.
> 
> - Chris
>