On Tue, Nov 29, 2016 at 09:21:13AM -0700, Jason Gunthorpe wrote:
On Mon, Nov 28, 2016 at 10:29:38PM -0800, Vishwanathapura, Niranjana wrote:
On Thu, Nov 24, 2016 at 09:15:45AM -0700, Jason Gunthorpe wrote:
>>And will move the hfi_vnic module under
>>???drivers/infiniband/ulp/hfi_vnic???.
>
>I would prefer drivers/net/ethernet
>
>This is clearly not a ULP since it doesn't use verbs.
>
I understand it is not using verbs, but the control path (ib_device client)
is using verbs (IB MAD).
Our prefernce is to keep it somewhere under drivers/infiniband. Summarizing
reasons again here,
- VNIC control driver (ib_device client) is an IB MAD agent.
- It is purly a software construct, encapsualtes ethernet packets in
Omni-path packet and depends on hfi1 driver here for HW access.
Is the majority of the code MAD focused or net stack focused?
I'm not sure it matters, it isn't like we can review Intel's
proprietary mad stuff anyhow. :\
Jason
That is an intersting measure. In hfi_vnic driver, I would say, >60% of the
code is MAD focused, mainly interfacing with the IB MAD agent.
It also includes populating/parsing those MAD packets. At the least it is not
supporting the driver to be put under net folder.
Even in the remaining <40%, half of it is involved with encapsulating ethernet
frames with Omni-path header (does this makes it belong under
drivers/infiniband/hw?).
The net stack interface part is pretty standard, hence is not much of code.
I do see the reason to put it under net folder, but I am seeing more reason for
it to be somewhere under drivers/infiniband.
Niranjana