Re: [Xen-users] xm->xl migration in Xen 4.2.2

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Mon, 2013-10-28 at 13:09 -0700, Saurabh Mishra wrote:
> Hi,
> 
> 
> One of the I noticed during migration from XM toolchain to XL in PCI
> Pass-through was that we used to specify target bdf in HVM guest as 
> 
> 
> pci = [ '0000:07:11.6=0@1a' ]

What was the effect of this? I guess it passes through host device
0000:07:11.6 which appears as some device derived from "0@1a" in the
guest?

> However with XL, it now seems like this is the correct syntax:
> 
> 
> pci = [ '0000:07:11.6@1a' ]
> 
> 
> Is that correct? I've verified that it works but what about zero in
> '0@' that we used to specify in XM. Was it bus-id?

I've no idea -- this was part of the general problem with Xend ;-)

Where did you find this =0@1a syntax from? Perhaps that might give a
clue.

Trying to understand
tools/python/xen/util/pci.py:parse_pci_name_extended (which I think
might be parsing the above strings for xend) suggests that the syntax
is:

<domain>:<bus>:<slot>.<fn>=<fn'>@<vdevfn>

IOW the =0@ specifies that the device should have function 0 inside the
guest (where in your example it has fn 6 in the host). I think with the
xl syntax you have the device will end up with function 6 instead of 0
within the guest. Is that right?

If this is function remapping functionality which you need please can
you post to xen-devel@ explaining the usecase and the syntax you are
used to using the xend and the affect which it has on the PCI devices
which are seen by the guest (i.e. the lspci with xend and with xl).

Thanks,
Ian.



_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users