Re: [net-next v3 2/9] ice: Create and register virtual bus for RDMA

[Greg KH <gregkh@linuxfoundation.org>]

On Thu, May 07, 2020 at 02:04:04PM +0000, Saleem, Shiraz wrote:
> > Subject: Re: [net-next v3 2/9] ice: Create and register virtual bus for RDMA
> > 
> > On Wed, May 06, 2020 at 02:04:58PM -0700, Jeff Kirsher wrote:
> > > From: Dave Ertman <david.m.ertman@intel.com>
> > >
> > > The RDMA block does not have its own PCI function, instead it must
> > > utilize the ice driver to gain access to the PCI device. Create a
> > > virtual bus device so the irdma driver can register a virtual bus
> > > driver to bind to it and receive device data. The device data contains
> > > all of the relevant information that the irdma peer will need to
> > > access this PF's IIDC API callbacks.
> > 
> > But there is no virtual bus driver in this patch!
> 
> Hi Greg - 
> 
> The irdma driver is the virtbus driver that would bind to the virtual devices created
> in this netdev driver.

Then why even have the virtbus code in this patch if there are no users?

And without any users, you are creating "virtbus devices" that live on
what bus?  How does that work at all?  Kind of defeats the purpose of a
virtual bus entirely if you do not use it, right?

> It is decoupled from this series as it was deemed in a prior discussion that irdma driver
> would go in a +1 cycle from net series to avoid conflicts. See discussion here --
> https://lore.kernel.org/netdev/46ed855e75f9eda89118bfad9c6f7b16dd372c71.camel@intel.com/
> 
> The irdma driver is currently posted as an RFC series with its most recent submission here --
> https://lore.kernel.org/linux-rdma/20200417171251.1533371-1-jeffrey.t.kirsher@intel.com/

I can't accept that this series is using a virtual bus properly without
actually using the virtual bus driver code, can you?

If this is the case, it better be REALLY REALLY REALLY well documented
when one would, and would not, want to do that, as it doesn't make sense
to me...

thanks,

greg k-h