Xen users - Jun 2010 - GPLPV on NetBSD dom0

Finally got a chance to look at this again.

Trying to install GPLPV on a Windows domU while running under a NetBSD/dom0
(Xen 3.3.2) causes a STOP 0x0000007B at boot.  Going through the debug
output a few months ago, James thought it had to do with the backend state.

Looking at the code, XenPci_XenConfigDeviceSpecifyBuffers()
calls XenPci_ChangeFrontendState() with a backend_state_response
of XenbusStateInitWait, however the current backend state
is XenbusStateConnected.  Because of this it eventually times out, fails,
and then Windows can''t find the drive and STOPs.

As a quick kludge, in XenPci_ChangeFrontendState() before the while() loop I
added:

  if ((backend_state_response ==
XenbusStateInitWait)&&(xppdd->backend_state
== XenbusStateConnected))
          backend_state_response = XenbusStateConnected;

After recompiling the drivers and loading them, everything appears to work
fine.

So my question is, is this a NetBSD bug (bringing the devices up to the the
wrong state too early), or it is a GPLPV bug?  I tried comparing the NetBSD
Xen kernel code against the code in the Linux kernel and I didn''t see
anything  that stood out.  If it''s a NetBSD problem, I''ll open
a PR in their
bug system.

Thanks!
-Dustin


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

> As a quick kludge, in XenPci_ChangeFrontendState() before the while()
loop I
> added:
> 
>   if ((backend_state_response
=XenbusStateInitWait)&&(xppdd->backend_state
> == XenbusStateConnected))
>           backend_state_response = XenbusStateConnected;
> 
> After recompiling the drivers and loading them, everything appears to
work
> fine.
> 
> So my question is, is this a NetBSD bug (bringing the devices up to
the the
> wrong state too early), or it is a GPLPV bug?  I tried comparing the
NetBSD
> Xen kernel code against the code in the Linux kernel and I didn''t
see
anything
> that stood out.  If it''s a NetBSD problem, I''ll open a PR
in their bug
system.
> 
Well if a Linux frontend works with a NetBSD backend and the other way
around then I''d be confident that the bugs are in GPLPV. The state
transitions are a bit different between net and vbd but I''ve never
quite
figured out exactly where the problem is.

James

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