netdev - Driver i40e have XDP-redirect bug

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-next>] [day] [month] [year] [list]

Message-ID: <20210301131832.0d765179@carbon>
Date:   Mon, 1 Mar 2021 13:18:32 +0100
From:   Jesper Dangaard Brouer <brouer@...hat.com>
To:     netdev <netdev@...r.kernel.org>,
        Björn Töpel <bjorn.topel@...il.com>,
        Jesse Brandeburg <jesse.brandeburg@...el.com>,
        Tony Nguyen <anthony.l.nguyen@...el.com>,
        Marek Majtyka <alardam@...il.com>
Cc:     brouer@...hat.com,
        intel-wired-lan <intel-wired-lan@...ts.osuosl.org>,
        Jakub Kicinski <kuba@...nel.org>,
        Alexei Starovoitov <alexei.starovoitov@...il.com>,
        Maciej Fijalkowski <maciej.fijalkowski@...el.com>,
        Tony Brelinski <tonyx.brelinski@...el.com>,
        Toke Høiland-Jørgensen <toke@...hat.com>
Subject: Driver i40e have XDP-redirect bug

Hi i40e-people + XDP-feature-people, 

The driver i40e have a XDP-redirect bug, where is it partly broken. It can
transmit a redirected xdp_frame (from another driver). But broken when
redirecting a xdp_frame that is received by the driver itself.

This reminds me about lacking XDP-features, as this "state" is actually
"supported" (for Intel drivers), when running in 'legacy-rx' mode.  This can
be configured (via: 'ethtool --set-priv-flags i40e2 legacy-rx on').  When
running in 'legacy-rx' mode the headroom is zero, which means that xdp_frame
cannot be created as it is stored in this headroom, but an XDP-prog can
still run a (DDoS) filter.  (Hint grepping after xdp_redirect stats is not enough).

The BUG I experience *is* that headroom is zero, but 'legacy-rx' mode is off:

  $ ethtool --show-priv-flags i40e2 | grep legacy-rx
  legacy-rx             : off

This is clearly a driver initialization bug as the headroom should not
be zero in this configuration. Further indication that this is related
to init order: If while xdp_redirect is running, I change RX-ring size
(e.g. 'ethtool -G i40e2 rx 1024') then redirect starts working again.

I will continue to find the offending commit... (to-be-continued)
(p.s. testing on net-next on top of commit d310ec03a34e92).

-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer