On Tue, Nov 22, 2016 at 05:49:32PM -0700, Jason Gunthorpe wrote:
> > We could add a custom Interface between HFI1 driver and hfi_vnic drivers
> > without involving a bus.
>
> hfi is already registering on the infiniband class, just use that.
I don't understand what you mean here?
Get the struct ib_device for the hfi and then do something to get hfi
specific function calls.
Or work it backwards with a _register function..
OK, thanks for your feedback.
We can make the hfi_vnic module as an ib client (which it is) like other ULPs,
and do not have an in-built or custom bus for binding.
Then the hfi_vnic ULP by some mechanism will identify the device as hfi1 device
and will only serve that device.
In order to pass the hfi function pointers to the hfi_vnic ULP, I can,
a) Have hfi_vnic ULP define an interface API for hfi1 driver to call to
register its callback (as you pointed). Unfortunately there will be a module
dependency here.
Or,
b) Add a new member ‘struct vnic_ops’ either to the ib_device structure or
ib_port_immutable structure. As it is hfi1 specific, only hfi1 driver will set
it. No module dependency here.
And will move the hfi_vnic module under ‘drivers/infiniband/ulp/hfi_vnic’.
All these will remove undue complexity and fit the driver in current design
framework as per your suggestion.
Let me know your comments.
Niranjana
Jason
