[patch 0/8] x86: UV - XPC fixes with related support functionality V2.
Robin Holt
The UV BIOS has been updated to implement some of our interface
functionality differently than originally expected. These patches
update us to that working bios and include a few minor bug fixes
which prevent us from doing significant testing on real hardware.
Changes from V1:
- Actually include the patch introducing the gru_read_gpa. This
was missed in the V1 submission.
- One additional BIOS change has the OS no longer passing blade
to the BIOS when registering a message queue watchlist.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majo...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Ingo Molnar
* Robin Holt <ho...@sgi.com> wrote:
>
> The UV BIOS has been updated to implement some of our interface
> functionality differently than originally expected. These patches
> update us to that working bios and include a few minor bug fixes
> which prevent us from doing significant testing on real hardware.
>
> Changes from V1:
>
> - Actually include the patch introducing the gru_read_gpa. This
> was missed in the V1 submission.
>
> - One additional BIOS change has the OS no longer passing blade
> to the BIOS when registering a message queue watchlist.
Note, the patches are a bit hard to review without diffstats - mind
generating them for future patches? Also, putting a 'summary' diffstat
(the total diffstat effect of all patches) into the 0/8 description mail
is common courtesy that is much welcome by maintainers and reviewers as
well.
Thanks,
Ingo
Ingo Molnar
* Ingo Molnar <mi...@elte.hu> wrote:
> * Robin Holt <ho...@sgi.com> wrote:
>
> >
> > The UV BIOS has been updated to implement some of our interface
> > functionality differently than originally expected. These patches
> > update us to that working bios and include a few minor bug fixes
> > which prevent us from doing significant testing on real hardware.
> >
> > Changes from V1:
> >
> > - Actually include the patch introducing the gru_read_gpa. This
> > was missed in the V1 submission.
> >
> > - One additional BIOS change has the OS no longer passing blade
> > to the BIOS when registering a message queue watchlist.
>
> Note, the patches are a bit hard to review without diffstats - mind
> generating them for future patches? Also, putting a 'summary' diffstat
> (the total diffstat effect of all patches) into the 0/8 description mail
> is common courtesy that is much welcome by maintainers and reviewers as
> well.
Ah, sorry - they were there, just merged together with the patch so i
missed them.
Regarding the patches - i still very much dislike how the code
interfaces to the 'BIOS'. The drivers/misc/sgi-xp/ code looks rather
messy.
Why isnt a pure hardware interface exposed, as both hpa and me suggested
in the past?
Anyway ... these bits got merged via -mm in the past so i guess it could
be done there in the future too - the series does not look horrible -
but the whole concept ought to be done in a cleaner way.
Robin Holt
> interfaces to the 'BIOS'. The drivers/misc/sgi-xp/ code looks rather
> messy.
>
> Why isnt a pure hardware interface exposed, as both hpa and me suggested
> in the past?
The information is stored on hardware (UV-Hub and RAM) outside this instance of hardware.
The only safe way to access that hardware is via the GRU.
No GRU resources are available to the BIOS.
Ergo, this interface method with multiple passes is needed.
If there is a better way, I am open to suggestions.
> Anyway ... these bits got merged via -mm in the past so i guess it could
> be done there in the future too - the series does not look horrible -
> but the whole concept ought to be done in a cleaner way.
I will ask Andrew to take them.
Thanks,
Robin
Ingo Molnar
* Robin Holt <ho...@sgi.com> wrote:
> > Regarding the patches - i still very much dislike how the code
> > interfaces to the 'BIOS'. The drivers/misc/sgi-xp/ code looks rather
> > messy.
> >
> > Why isnt a pure hardware interface exposed, as both hpa and me
> > suggested in the past?
>
> The information is stored on hardware (UV-Hub and RAM) outside this
> instance of hardware.
>
> The only safe way to access that hardware is via the GRU.
>
> No GRU resources are available to the BIOS.
>
> Ergo, this interface method with multiple passes is needed.
>
> If there is a better way, I am open to suggestions.
Erm, the better way is to not use the BIOS but to access those registers
that the BIOS accesses? (which is currently hidden by the BIOS)
In other words, we want to remove arch/x86/kernel/bios_uv.c and the
uv_bios_call*() APIs and replace them via clear implementations for
clear hardware abstractions. That 'bios message queue' thing is absolute
madness for example. But the type-opaque software API too.
Ingo