[U-Boot-Users] IXP4xx Silicon version

[U-Boot-Users] IXP4xx Silicon version

[Jonathan Pratt]

Hi all

I'm not sure if anyone else has found it to be a problem but some boards
we recently had built have an IXP425 on them which reports a Product
Revision value 0x2 (lowest four bits of CP15, reg 0). This is only a
problem because the file

cpu/ixp/npe/npe.c

uses this field in the npe_initialize function to decide how to
determine how many ethernet ports there are. Unfortunately the case
statement only accounts for revision values 0x0
(IX_FEATURE_CTRL_SILICON_TYPE_A0) and 0x01
(IX_FEATURE_CTRL_SILICON_TYPE_B0) so that the existing code doesn't
think that these processors have any ethernet ports available. I found
that by adding a 'default:' statement before line #639 remedied the
problem. This file is part of u-boot, not the Intel npe library.

Since it seems that someone is immanently about to release a patch on
the npe stuff anyway, would it be too much trouble to add the 'default:'
statement or otherwise handle this product revision?

Thanks

Jonathan Pratt
Senior Development Engineer
ELPRO Technologies P/L 

[U-Boot-Users] IXP4xx Silicon version

[Stefan Roese]

Hi Jonathan,

On Thursday 07 June 2007, Jonathan Pratt wrote:
> I'm not sure if anyone else has found it to be a problem but some boards
> we recently had built have an IXP425 on them which reports a Product
> Revision value 0x2 (lowest four bits of CP15, reg 0).

Seems that Intel has released a new revision of the CPU.

> This is only a 
> problem because the file
>
> cpu/ixp/npe/npe.c
>
> uses this field in the npe_initialize function to decide how to
> determine how many ethernet ports there are. Unfortunately the case
> statement only accounts for revision values 0x0
> (IX_FEATURE_CTRL_SILICON_TYPE_A0) and 0x01
> (IX_FEATURE_CTRL_SILICON_TYPE_B0) so that the existing code doesn't
> think that these processors have any ethernet ports available. I found
> that by adding a 'default:' statement before line #639 remedied the
> problem. This file is part of u-boot, not the Intel npe library.

Would be interesting to know, what the differences of this new IXP425 revision 
are before "blindly" selecting default for Rev. A or Rev. B.

> Since it seems that someone is immanently about to release a patch on
> the npe stuff anyway, would it be too much trouble to add the 'default:'
> statement or otherwise handle this product revision?

Yes, "someone" is waiting for another revised patch. I could of course add 
this statement, but I can't test since I don't have IXP hardware right now. 
Did you test this change successfully on both ethernet ports?

Best regards,
Stefan

=====================================================================
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-0 Fax: +49-8142-66989-80  Email: office at denx.de
=====================================================================

[U-Boot-Users] IXP4xx Silicon version

[Jonathan Pratt]

> -----Original Message-----
> From: Stefan Roese [mailto:sr at denx.de] 
> Sent: Thursday, 7 June 2007 3:18 PM
> To: u-boot-users at lists.sourceforge.net
> Cc: Jonathan Pratt
> Subject: Re: [U-Boot-Users] IXP4xx Silicon version
> 
> Hi Jonathan,
> 
> On Thursday 07 June 2007, Jonathan Pratt wrote:
> > I'm not sure if anyone else has found it to be a problem but some 
> > boards we recently had built have an IXP425 on them which reports a 
> > Product Revision value 0x2 (lowest four bits of CP15, reg 0).
> 
> Seems that Intel has released a new revision of the CPU.

I couldn't find any reference to it in the Intel documentation (but the
web site is not friendly to navigate so I didn't try all that hard).

> > This is only a
> > problem because the file
> >
> > cpu/ixp/npe/npe.c
> >
> > uses this field in the npe_initialize function to decide how to 
> > determine how many ethernet ports there are. Unfortunately the case 
> > statement only accounts for revision values 0x0
> > (IX_FEATURE_CTRL_SILICON_TYPE_A0) and 0x01
> > (IX_FEATURE_CTRL_SILICON_TYPE_B0) so that the existing code doesn't 
> > think that these processors have any ethernet ports 
> available. I found 
> > that by adding a 'default:' statement before line #639 remedied the 
> > problem. This file is part of u-boot, not the Intel npe library.
> 
> Would be interesting to know, what the differences of this 
> new IXP425 revision are before "blindly" selecting default 
> for Rev. A or Rev. B.

The 'B' option uses the reported capabilities of the device (through the
ixFeatureCtrlComponentCheck function in the npe library) which reads the
feature control register to get the infomration. I would have thought
that (for determining whether ethernet MAC's are present) this would be
a pretty safe bet.

> > Since it seems that someone is immanently about to release 
> a patch on 
> > the npe stuff anyway, would it be too much trouble to add 
> the 'default:'
> > statement or otherwise handle this product revision?
> 
> Yes, "someone" is waiting for another revised patch. I could 
> of course add this statement, but I can't test since I don't 
> have IXP hardware right now. 
> Did you test this change successfully on both ethernet ports?

Sorry - this hardware only utilises one of the MACs (npe B = MAC 0).
 
> Best regards,
> Stefan
> 
> =====================================================================
> DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
> HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
> Phone: +49-8142-66989-0 Fax: +49-8142-66989-80  Email: 
> office at denx.de 
> =====================================================================

Thanks
Jonathan Pratt

[U-Boot-Users] IXP4xx Silicon version

[Christian Hohnstaedt]

On Thu, Jun 07, 2007 at 08:38:16AM +1000, Jonathan Pratt wrote:
> Hi all
> 
> I'm not sure if anyone else has found it to be a problem but some boards
> we recently had built have an IXP425 on them which reports a Product
> Revision value 0x2 (lowest four bits of CP15, reg 0). This is only a
> problem because the file

There is a specification update from Intel, containing
a proposed fixed for B1 stepping CPUs.

http://www.intel.com/design/network/products/npfamily/docs/ixp4xx.htm?prod=IXP425#specupdt

best regards
Christian