netdev - Re: netdev stats for AF

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: Thu, 30 Nov 2023 17:10:20 +0530
From: Srivats P <pstavirs@...il.com>
To: Jesper Dangaard Brouer <jbrouer@...hat.com>
Cc: Xdp <xdp-newbies@...r.kernel.org>, 
	Network Development <netdev@...r.kernel.org>, brouer@...hat.com, 
	Lorenzo Bianconi <lorenzo.bianconi@...hat.com>, David Ahern <dsahern@...nel.org>, 
	Jakub Kicinski <kuba@...nel.org>
Subject: Re: netdev stats for AF_XDP traffic

Hi,

> > Is there any update on the "consistency for XDP statistics" [1]
> > high-prio work item from the xdp-project.net project management page?
>
> Lorenzo (cc) have done cleanups in several drivers, both mvneta and
> veth, but not mlx5. Ahern (cc) also had an interest earlier.
>
>
> > More specifically the fact that the mlx driver (are there others too?)
> > don't update netdev/ifconfig stats for XDP/AF_XDP traffic?
>
> IMHO the drivers MUST update the netdev/ifconfig RX stats, even when an
> XDP action is taken.  XDP-progs are user installed software, and not
> updating netdev RX-stats (for some XDP actions) is very confusing for
> all existing stats collector tools.

Does anyone know if the mlx4/mlx5 driver in recent kernel versions
update the netdev stats for AF_XDP tx/rx traffic?

> The mlx drivers are (sadly) the most inconsistent if I remember correctly.
>
> > The email thread started for that item dates from 2018 and hasn't
> > seen any updates after that.
> I think nobody have really had the time to fix up this inconsistent
> stats mess.
>
> Are you interested in working on this?
>
>
> > I see some recent patch/commit activity about XDP specific stats via
> > netlink but I believe this is different from the standard
> > netdev/ifconfig stats.
>
> Yes, if taking this "task", I recommend focusing on standardizing
> netdev/ifconfig stats behavior for XDP.  Solving the other XDP specific
> stats is a much larger task with much upstream discussions required.
>
> > Srivats
> >
> > [1] https://xdp-project.net/#Consistency-for-statistics-with-XDP
> >
>
> --Jesper
>