Roadmap for AST2500?

[Rick Altherr]

On Fri, Jun 3, 2016 at 1:12 PM, Joel Stanley <joel at jms.id.au> wrote:

> Hello Rick,
>
> On Thu, Jun 2, 2016 at 1:05 PM, Rick Altherr <raltherr at google.com> wrote:
> > AFAIK, Zaius will have an option for NC-SI in addition to an on-board
> PHY.
> > I suspect there will be some work required for NC-SI with the NICs Google
> > will be using.
>
> We've got a proposed NC-SI stack in our kernels we're using for
> Barreleye and on a ast2500 based system. That's been worked on by
> Gavin; he's keen to find someone who works in the Linux network stack
> to provide meaningful review.
>

Let me ask if we have anyone here at Google who can help.


>
> > Are you planning to stick with 4.7-rc1 as a base for AST2500 support?  I
> was
> > expecting to start from a normal release such as 4.6.
>
> We tend to keep moving forward so any work we do is relevant for
> upstreaming. I have some basic aspeed support upstream in 4.7, and I
> moved our drivers up so I could test this. As I didn't run in to any
> trouble on the evb when testing we're running with 4.7.
>
> I intend to continue cleaning up our drivers and submitting them
> before the 4.8 merge window opens. Assistance on this front is
> welcome.
>
>
As we discussed in IRC, I've backported the upstream aspeed support from
4.7-rc1 to 4.6 and then rebased your dev-4.7 branch on top.  I've pushed
the result to https://github.com/kc8apf/linux/tree/v4.6-ast2500.  This
seems to be a good starting point for linux-obmc so I'll send a pull
request soon.  While you move your dev branch along with upstream, I'll
backport changes into the 4.6-based linux-obmc branch until 4.7 is
released.  Then we'll move linux-obmc to 4.7 and repeat.  That way, poky
can be aimed at the 4.6-based linux-obmc and let work on the BSP and
machine layers continue.


> > I'll ask one of my people to start looking into u-boot.  I know we need
> to
> > do some work for Google-specific changes but we should also be able to
> help
> > with getting the Aspeed changes cleaned up for upstream.
>
> We've got two people working on that front. Teddy has been focusing on
> the ast2400 and sent some mail to the u-boot list relating to that
> today.
>
> Cedric has been focusing on ast2500 support. He has that booting on
> Qemu and will try that on hardware next week.
>
>
I look forward to seeing the changes.  Full obmc-phosphor-images will be
blocked until we have a u-boot that can at least build for ast2500.


> Finally, Brad has added a ast2500-evb board to our poky tree. Expect
> to see patches for that on the list in the next day or two.
>
>
As we discussed on IRC, Xo and I have also put together a BSP and machine
layer that seems to build everything except u-boot.  So far when using the
SDK u-boot, the image seems to boot up to starting login consoles.  There's
something not right with the ttys so we never get a login prompt.  Xo is
poking at that now.


> Cheers,
>
> Joel
>
> >
> >
> > On Wed, Jun 1, 2016 at 7:26 PM, Patrick Williams <patrick at stwcx.xyz>
> wrote:
> >>
> >> Rick,
> >>
> >> Joel has been working on the AST2500 support for the kernel. I believe
> >> his current working tree is here:
> >>     https://github.com/shenki/linux/tree/dev-4.7
> >>
> >> We are currently using the u-boot from the Aspeed SDK without much
> >> (any?) change at all. I think this is his tree for that:
> >>     https://github.com/shenki/u-boot/tree/v2013.01-aspeed-ast2500
> >>
> >> We are also in the process of creating the AST2500 BSP recipes to pull
> >> this work all together. You should see some commits in the next few days
> >> for this. In the worst case, it will be up by the end of next week.
> >>
> >> One area of need we have is taking the u-boot changes from Aspeed and
> >> getting them cleaned up, applicable to upstream master, and sent
> >> upstream. Joel and Teddy from Facebook have both done a little work in
> >> this area, but I think both of them have been busy with other tasks.
> >>
> >> Have you looked at support for Zaius' NIC for the BMC? Some of the
> >> systems use an NCSI shared interface and some have an independent PHY.
> >> This has tended to be a "problematic" area for both u-boot and kernel.
> >> If you're not using an already-supported interface there will be code
> >> you need to write there. U-boot isn't required but is handy for netboot
> >> in development.
> >>
> >> I don't have the roadmap ready for the userspace work we are planning to
> >> do. There is a significant amount of work we have planned over the next
> >> N months for a machine we are developing in a parallel time-frame with
> >> Zaius. I am really striving to get this all out on Github so we can
> >> coordinate this work with your team, but had some snags that I am
> >> working out.
> >>
> >> I am going to be in town next Wed 6/8. We can discuss some more on the
> >> userspace plans face-to-face. Hopefully, I can get our plans for the
> >> next few months published in Github by the end of this month.
> >>
> >> On Wed, Jun 01, 2016 at 12:54:51PM -0700, Rick Altherr wrote:
> >> > For Zaius, Google and Rackspace's Power9 server, we'll be including an
> >> > AST2500 and plan to run OpenBMC.  I have a few people on my team
> >> > available
> >> > to contribute to AST2500 bringup and Zaius-specific OpenBMC
> development.
> >> > I'm trying to figure out what all needs to be done, what AST2500 work
> is
> >> > already being done, and where my team's contributions can best be
> >> > utilized.
> >> >
> >> > I know Xo Wang (xow at google.com) has already been talking with some of
> >> > you
> >> > about the kernel work needed for AST2500.  What is needed in Yocto and
> >> > the
> >> > userspace apps?  I can easily put together a AST2500 bsp layer but
> >> > without
> >> > a corresponding kernel, that's not particularly useful.
> >> >
> >> > Is there a published roadmap somewhere?  Where do you see the biggest
> >> > need
> >> > for help?
> >> >
> >> > Rick
> >>
> >> > _______________________________________________
> >> > openbmc mailing list
> >> > openbmc at lists.ozlabs.org
> >> > https://lists.ozlabs.org/listinfo/openbmc
> >>
> >>
> >> --
> >> Patrick Williams
> >
> >
> >
> > _______________________________________________
> > openbmc mailing list
> > openbmc at lists.ozlabs.org
> > https://lists.ozlabs.org/listinfo/openbmc
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ozlabs.org/pipermail/openbmc/attachments/20160603/d6d90cb1/attachment-0001.html>