netdev - Re: [RFC PATCH net-next 2/5] net: split skb_checksum

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

Date:   Tue, 28 Feb 2017 11:31:11 +0100
From:   Davide Caratti <dcaratti@...hat.com>
To:     Tom Herbert <tom@...bertland.com>
Cc:     David Laight <David.Laight@...lab.com>,
        "David S. Miller" <davem@...emloft.net>,
        Linux Kernel Network Developers <netdev@...r.kernel.org>,
        "linux-sctp@...r.kernel.org" <linux-sctp@...r.kernel.org>,
        Marcelo Ricardo Leitner <marcelo.leitner@...il.com>
Subject: Re: [RFC PATCH net-next 2/5] net: split skb_checksum_help

On Mon, 2017-02-27 at 07:11 -0800, Tom Herbert wrote:
> CHECKSUM_PARTIAL is the preferred mechanism on the transmit path this
> defers defers the checksum computation as long as possible.
> Unfortunately, if SCTP is encapsulated in UDP we will probably need to
> run the SCTP CRC on the host which will be done with your changes to
> skb_checksum_help.

right. Tunnel devices have NETIF_F_SCTP_CRC bit cleared and
NETIF_F_HW_CSUM bit set: so, in this case csum_not_inet can help
recovering non-GSO SCTP packets having ip_summed equal to
CHECKSUM_PARTIAL.

> > I'm not sure if setting CHECKSUM_UNNECESSARY fits my case, because this would
> > implicitly skip RX validation when using devices like veth or loopback.
> >
> CHECKSUM_UNNECESSARY can be used in the transmit path (really the
> forwarding path), however this I think this must imply that the
> checksum in the packet must be correct. Please see my post about
> drivers that are mistakingly using CHECKSUM_UNNECESSARY with LRO since
> the checksum in the packet sent into the stack is not correct.

Ok, now I'm more convinced to use CHECKSUM_NONE :-)

thank you for the attention!
regards
--
davide