netdev - gso_type field for LRO packets

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>] [day] [month] [year] [list]

Date:   Tue, 12 Dec 2017 17:53:28 +0200
From:   Gal Pressman <galp@...lanox.com>
To:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        Eric Dumazet <edumazet@...gle.com>,
        "David S. Miller" <davem@...emloft.net>,
        Ben Hutchings <bhutchings@...arflare.com>,
        "Michael S. Tsirkin" <mst@...hat.com>
Cc:     Tariq Toukan <tariqt@...lanox.com>,
        Saeed Mahameed <saeedm@...lanox.com>
Subject: gso_type field for LRO packets

Hi all,
I've been exploring hardware LRO related code in the stack and came across an inconsistency I don't quite understand.
When receiving a hardware LRO packet, most of the drivers set skb_shinfo(skb)->gso_type (For example [1]) in order to prevent macvtap from breaking.
According to include/linux/skbuff.h +/skb_warn_if_lro, LRO should set gso_size but not gso_type, which contradicts the drivers implementation and makes the check redundant.

Can anyone advise what is the correct driver implementation? should we change skb_warn_if_lro() implementation?

Regards,
Gal

[1] commit bd69ba798e21 ("qlcnic: set gso_type")