Linux PCI subsystem development
 help / color / mirror / Atom feed
* override B:D:F assignments for SRIOV VFs
@ 2013-10-02 19:10 Andre Richter
  2013-10-02 19:25 ` Alex Williamson
  0 siblings, 1 reply; 2+ messages in thread
From: Andre Richter @ 2013-10-02 19:10 UTC (permalink / raw)
  To: linux-pci

Hello all,

i have an SRIOV capable device which has its PF behind 01:00.0.
After enabling SRIOV, the VFs show up as 01:08.0, 01:08.1 and so on.

Due to several stupid reasons, I'd like to do a quick and dirty hack
that makes my VFs show up as 01:00.1, 01:00.2 and so on.

I'd be so happy for any hints where to start with this.

What I found already is virtfn_devfn() in iov.c, where the devfns for
the VFs are computed, but changing it here doesn't work. I guess this
has to be done way before?
However, I wonder where the rule for computing the devfn comes from?

Thanks, Andre

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: override B:D:F assignments for SRIOV VFs
  2013-10-02 19:10 override B:D:F assignments for SRIOV VFs Andre Richter
@ 2013-10-02 19:25 ` Alex Williamson
  0 siblings, 0 replies; 2+ messages in thread
From: Alex Williamson @ 2013-10-02 19:25 UTC (permalink / raw)
  To: Andre Richter; +Cc: linux-pci

On Wed, 2013-10-02 at 21:10 +0200, Andre Richter wrote:
> Hello all,
> 
> i have an SRIOV capable device which has its PF behind 01:00.0.
> After enabling SRIOV, the VFs show up as 01:08.0, 01:08.1 and so on.
> 
> Due to several stupid reasons, I'd like to do a quick and dirty hack
> that makes my VFs show up as 01:00.1, 01:00.2 and so on.
> 
> I'd be so happy for any hints where to start with this.
> 
> What I found already is virtfn_devfn() in iov.c, where the devfns for
> the VFs are computed, but changing it here doesn't work. I guess this
> has to be done way before?
> However, I wonder where the rule for computing the devfn comes from?

Can't be done, the VF address is determined by the stride and offset in
the SR-IOV capability on the PF which are read-only per the SR-IOV spec.
The kernel doesn't get to decide where they appear.  Thanks,

Alex


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2013-10-02 19:25 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-10-02 19:10 override B:D:F assignments for SRIOV VFs Andre Richter
2013-10-02 19:25 ` Alex Williamson

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox