This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


[Xen-devel] Re: [1/4] [NET] back: Fix maximum fragment check

On 1 Jul 2006, at 04:33, Herbert Xu wrote:

Good point.  I'll get rid of it.

Actually, we do need it for two reasons:

1. To indicate protocol for drivers that can cope with malformed packets.
   The header verification will be skipped for such drivers.
2. To carry extra flags such as ECN that cannot harm the host if set

Fair enough, that makes sense.

Given that Linux will cope with malformed headers or a bogus gso_type, I'd
really like to keep the type value uniform between Linux and Xen.

I'm uncomfortable with this, even though it makes things a little easier now. For sanity I want to see netfront/netback explicitly grok flags rather than dumbly pass them through. I'd prefer uint8_t protocol and uint8_t flags. Former is a protocol enumeration; latter is unused now but we can add ECN and so on later. By the way: will we need netback to advertise support for the ECN flag? I'm not sure exactly what it will mean, and whether it can just be ignored by netbacks that don't support it?

 -- Keir

Xen-devel mailing list