[openfabrics-ewg] [PATCH] ehca: ehca_main.c for OFED-1.1-rc5

[Michael S. Tsirkin]

Quoting r. Hoang-Nam Nguyen <HNGUYEN at de.ibm.com>:
> Subject: Re: [PATCH] ehca: ehca_main.c for OFED-1.1-rc5
> 
> Michael,
> > OK, I'm taking this in.
> Thanks
> > Unfortunately I can't integrate the rest of the backport patches - the
> > changes are risky for the rest of the stack - so only 2.6.18 will be supported
> > with ehca in OFED 1.1.
> > OK?
> It would be great help if we can accomplish this for 2.6.9-RH and 2.6.16 as

Is 2.6.16 the yellowdog?

BTW, please note RH is now 3 platforms (U2,U3,U4). These are significantly
different kernel dialects, we test them all and you really must at least build test
on all of these before you ask us to enable ehca there.

> two "major" distros want to see ehca in OFED-1.1.

Well, RC6 is Wed and final should be next week Tue.
If I see the small patches as we discussed by tomorrow and they look safe,
we could integrate conditionally for when ehca is enabled, I guess, although I
still have qualms - people tend to do build/all, just in case. I'm looking for
something to mitigate the risk of touching DMA macros used all over the code.

Would it OK with you if build-all has ehca disabled, and user must
enable it separately?

> At least 2.6.16 should be feasible because we only have a hvcall patch,
> which
> remaps some defines to existing ones and affects only ppc64. And I'm pretty
> sure
> that it's really harmless.
> Thx
> Nam Nguyen

I agree that hvcall patch is harmless, I did not realize this was
the only patch for 2.6.16. I'll put it in, and ask Vlad to enable ehca
on ppc on 2.6.16.

-- 
MST