On Wed, May 02, 2018 at 09:41:45AM -0500, Steve Wise wrote:
>
>
> On 5/1/2018 3:31 PM, Jason Gunthorpe wrote:
> > On Tue, May 01, 2018 at 01:45:03PM -0600, Jason Gunthorpe wrote:
> >> On Tue, May 01, 2018 at 12:30:02PM -0400, Doug Ledford wrote:
> >>> On Sat, 2018-04-28 at 08:53 -0700, Steve Wise wrote:
> >>>> This series adds new resource tracking attributes and APIs for
> >>>> rdma providers to add their own implementation specific attributes.
> >>>> Included in this series is the core changes, and a cxgb4-specific QP
> >>>> implementation.
> >>> I'm pretty happy with this version Steve. It allows the kernel to tell
> >>> what each type should look like (hex, string, whatever) but doesn't
> >>> attempt to tell it how to format anything beyond single elements, which
> >>> seems the right balance to me. We don't know things like screen
> >>> attributes or anything else that would play into those larger formatting
> >>> concerns. But knowing that a given element is best displayed as hex
> >>> instead of decimal makes lots of sense. I've applied this to my wip/dl-
> >>> for-next, so unless Jason has objections within the next 24 hours or so,
> >>> it'll be in the next official for-next.
> >> Looks OK to me
> > Oh, I have one minor quibble actually..
> >
> > + [RDMA_NLDEV_ATTR_PROVIDER] = { .type = NLA_NESTED },
> > + [RDMA_NLDEV_ATTR_PROVIDER_ENTRY] = { .type = NLA_NESTED },
> > + [RDMA_NLDEV_ATTR_PROVIDER_STRING] = { .type = NLA_NUL_STRING,
> >
> > I have been trying really hard to consistently call the kernel driver
> > the 'driver' and not use the word 'provider' or 'vendor'.
> >
> > eg in the kabi we are consistently calling all this stuff 'driver' now.
> >
> > The provider is what we call the libibverbs thing...
> >
> > Jason
>
> If Leon and Doug agree this is a good idea, I'll respin and change
> provider->driver and PROVIDER->DRIVER.
I don't have any opinion about that.
Thanks
>
> Steve.
Attachment:
signature.asc
Description: PGP signature
