netdev - Re: RFC: Checksum offload and XDP

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, 11 Apr 2017 11:43:24 -0700
From:   Jakub Kicinski <kubakici@...pl>
To:     Edward Cree <ecree@...arflare.com>
Cc:     Tom Herbert <tom@...bertland.com>,
        Linux Kernel Network Developers <netdev@...r.kernel.org>,
        "Alexei Starovoitov" <ast@...nel.org>
Subject: Re: RFC: Checksum offload and XDP

On Tue, 11 Apr 2017 18:13:37 +0100, Edward Cree wrote:
> > Note that this only applies to
> > checksum_complete, if we were to allow XDP program to return
> > checksum_unnecessary for instance then it's more a leap of faith that
> > things are always correct.  
> Speaking of checksum_unnecessary, it might still be useful for the
>  verifier to tell us whether the program contains any writes, since if
>  so any drivers using checksum_unnecessary will have to clear it when
>  calling XDP or else do conversion to checksum_complete beforehand.

If we exposed csum_complete as a value in xdp buffer, we could make the
unnecessary -> complete conversion dependent on accesses to that field
rather than writes and header adjustments.  If program doesn't read or
write the csum it doesn't care about the csum.

And make invalidation of unnecessary dependent on writes and header
adjustments.