WARNING - OLD ARCHIVES

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/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-devel

Re: [Xen-devel] Re: USB virt

On Tue, 2005-10-25 at 19:21 -0400, Sean Dague wrote:
> Wow, 18k lines of code so far.

Actually, the USB driver is only 8k raw lines, most of which is
copyright comments.  It's only 900 ";" of code which I think is pretty
small.

Some of the rest of the code is duplicated because I'm in the middle of
refactoring it.

The remainder of the rest of the code is the common code that is
generally required to turn Xenbus and grant-tables into a useful
inter-domain communication mechanism.  This is the code that's currently
replicated in all the other drivers.  The reason that this portion of
the code is bigger in my driver is because I'm coding a general solution
which a) will work for device driver domains with driver restarts and
loadable driver modules and b) is a service which can be used by new
drivers in the future to save them from having yet another copy of the
code.

> 
> Just out of curiosity, a friend pointed me to
> http://heanet.dl.sourceforge.net/sourceforge/usbip/usbip-0.1-2.6.13.tar.gz
> which he was looking at for trying to find a way to do usb on Xen 2.0.
> 
> Would an approach like this be worth looking at?

The API of the inter-domain communication mechanism that I wrote the USB
driver to is currently implemented using shared memory (grant-tables)
for efficiency but is compatible with a network transparent
implementation so could be extended to work through domain migration in
the future.  Naturally, if you extended the implementation of my
interdomain communication API then any other drivers that had been
written to use it would also benefit from the new functionality.

The difference between what I have done and the above IP based solution
is that I have explicitly created a solution which can have an efficient
shared memory implementation in the normal case when domains are
resident on the same physical machine and my solution is intentionally
available to any other drivers that might choose to use it.

Thanks for your input.  I think I'll continue with what I was doing for
the time being.

Harry.

> 
>       -Sean
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel

<Prev in Thread] Current Thread [Next in Thread>