[ofa-general] Re: [PATCH] for-2.6.23 ib/umad: add partition support

[Michael S. Tsirkin]

> Quoting Roland Dreier <rdreier at cisco.com>:
> Subject: Re: [PATCH] for-2.6.23 ib/umad: add partition support
> 
>  > Did you have something in mind?  (new ioctl?  re-using existing fields?)
>  > 
>  > Not all fields are used for both reads and writes.  E.g. status is
>  > unused on a write, and retries is unused on a read.  Storing the
>  > pkey_index on a read seems doable.  I think if we do anything on a
>  > write, we need to make an assumption that the data is currently set to
>  > 0 by the app.
> 
> I hadn't really thought about it.
> 
> One other thing is that the top 8 bits of flow_label aren't used.  I
> guess we could steal that, although it's a little ugly.  I doubt it
> would break existing userspace.
> 
> There is the problem of old kernels silently ignoring the pkey index
> though.  I'm not sure I see a good way around that.
> 
> I'm beginning to think that just updating the ABI might be the right
> answer.

Ugh. OFED 1.2 (with the old ABI) just went out.
I wonder - is it time to start making the kernel backwards-compatible?
It would be trivial to have userspace supply its own ABI
version and have kernel support both new and old ABI if we want to.
What do you think?

> But let's try to make this be the last ABI break.  Are we
> pretty sure there's *nothing* else we might ever want to add to the
> structure?  I can't think of anything right now...

It'd be easy to add some extra padding just in case ...

-- 
MST