Experimental Physics and
Industrial Control System

Gedare Bloom via Core-talk <core-talk at aps.anl.gov> · Fri, 12 Feb 2021 12:09:47 -0700

On Thu, Feb 11, 2021 at 2:56 PM Michael Davidsaver
<mdavidsaver at gmail.com> wrote:
>
> On 2/11/21 12:32 PM, Gedare Bloom via Core-talk wrote:
> > Hello again EPICS Core,
> >
> > I'm looking to collect some feedback from you all related to
> > EPICS+RTEMS. I'm trying to also keep in touch with monitoring the
> > progress on EPICS7/RTEMS5 (And, I'm planning to participate with
> > students in the upcoming hackathon! if you have tasks related, I might
> > be able to help knock them out). My ask right now is:
> > * What feature improvements would you like to see in future RTEMS releases?
> > * What are your pain points with current RTEMS versions and/or workflows?
>
> * BSP Status
>
> I've been confused about the status of the various architectures/BSPs
> in the RTEMS >=5 era.  My questions for each are:
>
> 1. Is libbsd (new network stack) supported?
> 2. Does it build?
> 3. Does it boot in emulation?  (assuming there is an emulator)
> 4. Has it been tested on metal?
>
> I think I can answer "yes" to the first 3 for i386/pc686 and powerpc/mvme3100.
>
This is indeed really important information. We have tried to capture
some of this in the notion of a "Tiering" system.
https://docs.rtems.org/branches/master/user/hardware/tiers.html

Tier 1 BSP meets #1-4.

I'm hoping we can make the tiers more apparent to our user base, and
maybe do some kind of scoreboard (red/yellow/green status matrix) on
some of these points.

> Another surprising (to me) question which came up recently is whether
> the VME bridge driver is usable on targets without libbsd support.
> So maybe my question 1. should be which features depend on libbsd?
>
>
> The current set of BSPs with configuration in EPICS 7.0 series
> repository are:
This is really great information for us. I was actually going to put
together a survey to try to collect some of this data regarding BSPs
in use.

>
> arm
>   at91rm9200ek
The dev kit for this one is obsolete, so we would have a hard time to
do hw testing. We always welcome others to help us out with hw test
runs and reporting results back to us. Chris Johns might help guide
the setup. This is the only way to get BSPs to Tier 1.

We test some arm boards (Beaglebone Black hardware is pretty well
used, and Zynq in production) and under qemu simulation. We would be
curious to know if any sites contemplate other ARM BSPs for future
installations.

> powerpc
>   beatnik
>   mcp750
>   mvme2100
>   mvme2700
mcp750, mvme2100, mvme2700 are variants of the motorola_shared BSP, I believe

>   mvme3100
>   mvme5500
Supposedly the beatnik BSP supports both mvme5500 and mvme6100. It
would be good if we can identify if there is any gap between the two,
or if beatnik can handle these.

>   psim
> i386
>   pc386   (we're switching to pc686 w/ RTEMS5)
> m68k
>   gen68360
>   mvme167
>   uC5282
>
As I mentioned, I'm trying to get some uC5282 boards, and Joel dug up
an old patch for an old qemu version that Till sent out (over a decade
ago), that we might try to revive.
FYI: https://lists.rtems.org/pipermail/users/2009-September/021089.html
The m68k qemu target is maintained for emulating some old MAC machines, btw.

>
> * Toolchain builds
>
> It seems like everyone I know who is working with RTEMS5 is using a different
> process to build RTEMS.  Some people are using RSB in part or in full.  Others
> build some piece individually with autoconf or waf (I guess).  The result has
> been repeatability and "works for me" issues.  I'd love to have a recommended
> recipe to focus everyone on.

Of course we encourage and provide our effort toward the RSB at least
for toolchain builds, for exactly the reasons you mention. We try to
discourage people from rolling their own toolchains anymore. We don't
want to prescribe any particular methods for building RTEMS or
applications. The autoconf version is going to disappear before
RTEMS-6 releases, so building RTEMS itself will be waf-based moving
forward. There are numerous advantages including configuration control
and much faster build times.

-Gedare

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System