Re: [RFC] Proposal to address hfi1 UI and EPROM devices

[Leon Romanovsky <leon-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>]

[-- Attachment #1: Type: text/plain, Size: 2239 bytes --]

On Tue, May 03, 2016 at 09:42:18PM +0300, Leon Romanovsky wrote:
> On Tue, May 03, 2016 at 12:54:05PM -0400, Dennis Dalessandro wrote:
> > On Tue, May 03, 2016 at 07:24:57PM +0300, Leon Romanovsky wrote:
> > >On Mon, May 02, 2016 at 03:55:02PM -0400, Dennis Dalessandro wrote:
> > >>We also should be able to use any of these schemes to handle our eprom
> > >>reading/writing. Adding eprom to IPoIB as Doug suggested is a fine plan, but
> > >>we technically don't need to do it right now when we could make do with the
> > >>"UI" functionality and hide the details in user space. In the future there
> > >>may well be value in having an eprom capability in the rdma sub-system, but
> > >>for now we believe we can avoid extending the kernel in this regard.
> > >
> > >Just to understand better your RFC. Are you asking from the community
> > >excuse do not implement core functionality just because you don't need
> > >it? Am I right?
> > 
> > The purpose of the RFC is to get the community's feedback on our plans to
> > solve our UI/EPROM issue. The paragraphs [1] not snipped in your response
> > focus on that. It comes down to:
> > 
> > Use resource0 which does not support locking vs use a file in
> > /sys/kernel/debug.
> 
> I didn't quote the whole email because I didn't see any question in
> these 3 options:
> * First option - doesn't meet the requirements and hard to extend.
> * Second option - doesn't meet the requirements and HW limitations
> * Third option - presented as one possible option and looks like the
>   correct one.
> 
> And the real question is "where" do you need to implement third option.

And to remove the tension, all what you was supposed to ask can be
summarized in one question: "Do we have other customers for EEPROM in
our RDMA stack?". If the answer is yes, you will need to implement it in
core, and if the answer is no, you will implement it in your driver.

Hope it helps.

> 
> > 
> > The point of the particular paragraph you are questioning is that I don't
> > think we should add more code and complexity to the kernel for something
> > that is not needed.
> > 
> > [1] http://marc.info/?l=linux-rdma&m=146221891012428&w=2
> > 
> > -Denny



[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]