Re: [PATCH net-next] sunvnet: don't change gso data on clones

[David L Stevens <david.stevens@oracle.com>]

Yes, I've seen that patch. The bug I'm fixing has the same symptoms,
but the mangling is being done by the sunvnet driver and the fix is
the same-- don't mangle packets that are cloned (and unclone them if you
need to mangle them).

The original sunvnet code changed gso_size temporarily, but that still
had a race where TCP could see the driver-modified gso_size with low probability
and end up with a negative packets-in-flight.

The unclone removes that bug, created (by me) after your fix, but in the sunvnet driver
with the addition of TSO support.

						+-DLS

On 02/11/2015 09:52 PM, Eric Dumazet wrote:

> 
> Well, we had a very hard to find bug in TCP stack, I want to make sure
> we fixed all relevant points.
> 
> commit c52e2421f7368fd36cbe330d2cf41b10452e39a9
> Author: Eric Dumazet <edumazet@google.com>
> Date:   Tue Oct 15 11:54:30 2013 -0700
> 
>     tcp: must unclone packets before mangling them
>     
>     TCP stack should make sure it owns skbs before mangling them.
...