xen-pciback: redo VF placement in the virtual topology
authorJan Beulich <jbeulich@suse.com>
Tue, 18 May 2021 16:13:42 +0000 (18:13 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 10 Jun 2021 11:24:09 +0000 (13:24 +0200)
commited6a024f4888df6fa2b8a3e92dea5f6a63649b84
treeade9237b0e5635bd0fcad93c42538b01f402ee2d
parentb5cd7f229609cf328664aedb056a6fd744e0068b
xen-pciback: redo VF placement in the virtual topology

The commit referenced below was incomplete: It merely affected what
would get written to the vdev-<N> xenstore node. The guest would still
find the function at the original function number as long as
__xen_pcibk_get_pci_dev() wouldn't be in sync. The same goes for AER wrt
__xen_pcibk_get_pcifront_dev().

Undo overriding the function to zero and instead make sure that VFs at
function zero remain alone in their slot. This has the added benefit of
improving overall capacity, considering that there's only a total of 32
slots available right now (PCI segment and bus can both only ever be
zero at present).

This is upstream commit 4ba50e7c423c29639878c00573288869aa627068.

Fixes: 8a5248fe10b1 ("xen PV passthru: assign SR-IOV virtual functions to
separate virtual slots")
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Link: https://lore.kernel.org/r/8def783b-404c-3452-196d-3f3fd4d72c9e@suse.com
Signed-off-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/xen/xen-pciback/vpci.c