[Qemu-devel] Re: [PATCH] qemu/virtio-net: remove wrong s/g layout assumptions

[Anthony Liguori <aliguori@linux.vnet.ibm.com>]

Michael S. Tsirkin wrote:
> It's useful because this way I won't have to maintain the fix, and it
> will make it possible for guests to experiment with layouts, without
> hacking qemu. If someone wants to make it a product, that's a different
> thing.
>   

Advertising a new feature is not hard.  It's one line of code in qemu 
with Rusty's ACK.

> Also, it might be a valid thing for a guest to say that host needs to be
> fixed. Not everyone might care about running on any possible broken qemu
> version.
>
> Finally - where do we draw the line? Does any bugfix need a feature bit?
>   

This is a spec bug, not a qemu bug IMHO.

The kernel drivers have always behaved this way and when this was all 
written originally, the semantics were never defined.  All the userspace 
implementations relied on this.  The fact that the spec claims a 
different behavior is a result of deciding that it should be different 
after the fact.

Thinking about it more, your patch is broken.  If you run it against a 
really old guest, it will break badly.

You assume that the guest header is always a fixed size.  It's not, 
we've added fields as we've added feature bits.  You actually have to 
look at the set of Ack'd features bits to determine how large the header is.

Which is why I now remember why this has never changed.  It's a PITA 
:-)  I'm not even sure you can do it in a correct way.

-- 
Regards,

Anthony Liguori