Patchset to establish sanity in Xilinx Virtex support

Patchset to establish sanity in Xilinx Virtex support

[Grant Likely]

The Virtex support in Linus' tree is a mess.  When I ported the common
devices over to using the platform bus I tried the use the ppc_sys
infrastructure.  ppc_sys is intended for SoCs that have a fixed set of
devices.  It is the *wrong* approach for FPGA support, and I made a
royal mess of things.  This patch set fixes that problem and makes
a number of other changes that make supporting Virtex boards easier.

I do know that changes to arch/ppc are frowned upon at the moment.  However,
4xx support in arch/powerpc is not merged yet, and Virtex support needs
additional work beyond that.  I suspect that arch/ppc support for the
4xx parts is going to need to stay alive for the next 2-3 kernel releases
anyway while arch/powerpc support stablizes.  Getting these patches in
will make it easier to support the Virtex ports while I hack on getting
arch/powerpc stuff figured out.

 arch/ppc/boot/common/misc-common.c    |   15 ++-
 arch/ppc/boot/simple/Makefile         |    1 +
 arch/ppc/boot/simple/uartlite_tty.c   |   37 ++++++
 arch/ppc/platforms/4xx/Kconfig        |   13 +-
 arch/ppc/platforms/4xx/Makefile       |    1 -
 arch/ppc/platforms/4xx/virtex.c       |   56 ---------
 arch/ppc/platforms/4xx/virtex.h       |   47 +++++---
 arch/ppc/platforms/4xx/xilinx_ml300.c |   65 +----------
 arch/ppc/platforms/4xx/xilinx_ml300.h |   45 -------
 arch/ppc/platforms/4xx/xilinx_ml403.c |   66 +----------
 arch/ppc/platforms/4xx/xilinx_ml403.h |   49 --------
 arch/ppc/syslib/Makefile              |    3 +-
 arch/ppc/syslib/virtex_devices.c      |  209 +++++++++++++++++++++++++++++++++
 arch/ppc/syslib/virtex_devices.h      |   27 +++++
 include/asm-ppc/ibm4xx.h              |    8 +-
 include/asm-ppc/ppc_sys.h             |    2 -
 16 files changed, 335 insertions(+), 309 deletions(-)

Re: Patchset to establish sanity in Xilinx Virtex support

[Peter Mendham]

Grant Likely wrote:
> The Virtex support in Linus' tree is a mess.  When I ported the common
> devices over to using the platform bus I tried the use the ppc_sys
> infrastructure.  ppc_sys is intended for SoCs that have a fixed set of
> devices.  It is the *wrong* approach for FPGA support, and I made a
> royal mess of things.  This patch set fixes that problem and makes
> a number of other changes that make supporting Virtex boards easier.
>
> I do know that changes to arch/ppc are frowned upon at the moment.  However,
> 4xx support in arch/powerpc is not merged yet, and Virtex support needs
> additional work beyond that.  I suspect that arch/ppc support for the
> 4xx parts is going to need to stay alive for the next 2-3 kernel releases
> anyway while arch/powerpc support stablizes.  Getting these patches in
> will make it easier to support the Virtex ports while I hack on getting
> arch/powerpc stuff figured out.
>   
I have applied your patches to a 2.6.20.4 kernel, along with your 
SystemACE driver. I am fully up and running on a simple ML405-based 
system setup with a Uartlite and a SystemACE. (I made the Kconfig 
changes suggested by Dale Farnsworth to my copy). So, thank you very 
much. I appreciate the neatness of your patches, the fact that I now 
have early boot text support on my uartlite port and the fact that 
(somehow) you've shaved a good couple of seconds of my boot time against 
the Xilinx SystemACE driver. Thanks!

In your reply to Mohamed Sadegh Sadri you mentioned that you will be 
publishing your GIT tree with TEMAC support.  Two very quick questions: 
where do I find this, and does it support auto-negotiation properly?  I 
am having serious problems with auto-negotiation and the various 
flavours of driver out there.  It seems like the best long term plan 
would be to use the PHYLIB support to handle the PHY side of things but 
I have only just started looking into this.

Thanks again,
-- Peter



-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
MailScanner thanks transtec Computers for their support.

Re: Patchset to establish sanity in Xilinx Virtex support

[Grant Likely]

On 4/16/07, Peter Mendham <petermendham@computing.dundee.ac.uk> wrote:
> I have applied your patches to a 2.6.20.4 kernel, along with your
> SystemACE driver. I am fully up and running on a simple ML405-based
> system setup with a Uartlite and a SystemACE. (I made the Kconfig
> changes suggested by Dale Farnsworth to my copy). So, thank you very
> much. I appreciate the neatness of your patches, the fact that I now
> have early boot text support on my uartlite port and the fact that
> (somehow) you've shaved a good couple of seconds of my boot time against
> the Xilinx SystemACE driver. Thanks!

Thanks for your feedback.

I'm glad to hear that the driver is faster; I was hoping that was the
case, but I haven't benchmarked it against the Xilinx driver yet.
Please,
please, PLEASE let me know if you have any problems with it.  Any
information you can provide me on your usage would be fantastic.

So, I assume that your rootfs is mounted from the CF card, correct?
If so, very cool.  I hadn't gotten to that point in my testing, so I
didn't know if that would work.  :-)

> In your reply to Mohamed Sadegh Sadri you mentioned that you will be
> publishing your GIT tree with TEMAC support.  Two very quick questions:
> where do I find this, and does it support auto-negotiation properly?  I
> am having serious problems with auto-negotiation and the various
> flavours of driver out there.  It seems like the best long term plan
> would be to use the PHYLIB support to handle the PHY side of things but
> I have only just started looking into this.

I'll be posting it on my git server at http://git.secretlab.ca, and
yes it does support autonegotiation, at least when using the Phy that
is on the ML403 board.  I'm going to take a close look a the TEMAC
patch that Andrei recently posted on the list which makes use of phylib.  It
looks interesting.

Cheers and thanks,
g.

-- 
Grant Likely, B.Sc. P.Eng.
Secret Lab Technologies Ltd.
grant.likely@secretlab.ca
(403) 399-0195

Re: Patchset to establish sanity in Xilinx Virtex support

[Peter Mendham]

Grant Likely wrote:
> So, I assume that your rootfs is mounted from the CF card, correct?
It is indeed - no problems yet :)
-- Peter

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
MailScanner thanks transtec Computers for their support.