[cgl_discussion] Hot Plug support for PICMG2.16 - Closure?

Peter Badovinatz tabmowzo at us.ibm.com
Mon Jan 20 06:42:05 PST 2003

John Grana wrote:
> Greetings,
>         In an attempt to get the PICMG 2.16 Hot Plug issue to some kind of closure,
> I wanted to summarize where I think things are and (might) go forward.
>         To recap. Hot Plug for PICMG 2.16 and future CompactPCI architectures
> require a change to the way the present 2.5 code handles Hot Plug. Since
> there is no PCI signaling on these new backplanes, the Identity portion
> needs to use a different method to identify what card is in a slot. I had
> proposed 4 ideas - 3 based on ethernet/TCP and 1 on IPMI. The few responses
> I received felt the network method was not ideal - took too long, etc. So,
> IPMI was the choice. IPMI is optional in PICMG 2.16 (but I believe a huge
> majority of manufacturers implement it) and is required in CompactTCA and
> AdvancedTCA. So, summary - IPMI is the method.

SAF HPI proposes a generic interface layer for exploiting hardware
support such as IPMI, Advanced TCA, etc.  The definitions are dependent
upon there being HW support "underneath" it for the features.  SAF has
worked to fix its licensing, the OSDL lawyers believe that it should now
be conducive to open source development.  It should be updated on their
web soon, and included in their specs.

>         Now, the next hurdle is to specify how to use IPMI. Andrew Cress proposed a
> library and driver from the SAForum that can be the basis. The one issue
> appears to be in licensing - see thread below. So, we can either
> beg/plead/pray they whatever the issues are get resolved or just
> architect/write our own. It would be a shame to duplicate the work, but what
> the heck. For Hot Identity, there is not that much to do really. I am
> willing to start the begging/pleading to see how to get this done.
>         Summary - IPMI is the underlying method to do Hot Identity in PICMG 2.16,
> CompactTCA and AdvancedTCA systems.
>         We can hope the SAforum opens it's IMPI work, or implement just what we
> need.

SAForum itself isn't implementing, but member companies are so something
should come available (see below).  In any case, the specification
licensing should at least be fixed by now.
> Thanks for the responses and suggestions.
> John Grana
> jjg at pt.com
> Performance Technologies, Inc.
> www.pt.com
> -----Original Message-----
> From: Cress, Andrew R [mailto:andrew.r.cress at intel.com]
> Sent: Tuesday, January 14, 2003 2:44 PM
> To: 'Mika Kukkonen'; Howell, David P
> Cc: John Grana; greg k-h; cgl_discussion at lists.osdl.org
> Subject: RE: [cgl_discussion] Re: Hot Plug support for PICMG2.16
> Mika,
> There is a development effort under way (in SOW stage now) to implement a
> library for the SAForum HPI interface on Linux IPMI platforms.  Since the
> OpenIPMI driver is already going to be in the Linux 2.5+ kernel, this should
> help.  The library will be in user-space.
> We anticipate that the licensing issues will be resolved, such that either
> the library will be open, or the non-open parts of the library will be
> isolated to a shared library.  There can also be other HPI libraries for
> other non-IPMI platforms.
> Andy
> -----Original Message-----
> From: Mika Kukkonen [mailto:mika at osdl.org]
> Sent: Monday, January 13, 2003 1:20 PM
> To: Howell, David P
> Cc: John Grana; greg k-h; cgl_discussion at lists.osdl.org
> Subject: RE: [cgl_discussion] Re: Hot Plug support for PICMG2.16
> On ma, 2003-01-13 at 06:38, Howell, David P wrote:
> > There have been many discussions already about IPMI being closed, mostly
> > coming from members providing non-IPMI hardware solutions. Seems like an
> > abstraction layer that could map IPMI or other implementations would be
> > in order to solve this for both.
> >
> > Wasn't there something in the SAF platform specs to address this?
> As our resident SAForum insider (Peter) is out of town, I'll take a shot
> on this.
> Yes, SAF platform spec is specified so that other implementations than
> IPMI can meet it and provide the required functionality. In other words,
> I am pretty sure it is possible to make non-IPMI (c)PCI iplementation of
> the SAF platform spec.
> On the other hand, some people seem to think that IPMI is best since
> sliced bread, so in CGL we have not seen any reason to venture outside
> IPMI, although as an overall philosophy we try to stay as HW agnostic
> as is reasonable (i.e. without sacrificing functionality).
> Now the real question (and the main reason for this email) is that who
> is going to implement the SAF platform spec on top of Linux? Assuming
> IPMI there should not be much impact on the kernel, but as a first step
> towards the implementation somebody should wade through the spec with
> those two things (IPMI & Linux kernel) in mind and find out what needs
> to be done.
> Any volunteers? :-)
> --MiKu

Peter R. Badovinatz aka 'Wombat' -- IBM Linux Technology Center
preferred: tabmowzo at us.ibm.com / alternate: wombat at us.ibm.com
These are my opinions and absolutely not official opinions of IBM, Corp.

More information about the cgl_discussion mailing list