Status of 5200B and 440 Kernel Support

Status of 5200B and 440 Kernel Support

[Albrecht Dreß]

[-- Attachment #1: Type: text/plain, Size: 640 bytes --]

Hi all,

sorry if this is a dumb question, I'm new in the list...

For a new project, I am looking through the available PPC based SoC's.   
Good candidates are probably the Freescale 5200B or the AMCC 440EP (I  
need a FPU).

For a decision, I need some further information about the Kernel  
support (source code available, if possible in stock kernels, *not*  
only vendor binary modules!) of the included devices, in particular
- MTD/NAND controller
- Ethernet
- UART, IIC and SPI
- USB host
- local external bus (not PCI) DMA

Any information or pointers would be really appreciated!

Thanks in advance,
Albrecht.

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: Status of 5200B and 440 Kernel Support

[Josh Boyer]

On Sun, 24 Feb 2008 11:17:39 +0100
Albrecht Dre=C3=9F <albrecht.dress@arcor.de> wrote:

> For a decision, I need some further information about the Kernel =20
> support (source code available, if possible in stock kernels, *not* =20
> only vendor binary modules!) of the included devices, in particular

For 440EP in arch/powerpc:

> - MTD/NAND controller

NOR flash works.  NAND needs some additional code.

> - Ethernet

Works.

> - UART, IIC and SPI

UART works.  IIC has recent patches.  No idea about SPI.

> - USB host

Works.

> - local external bus (not PCI) DMA

DMA to the EBC for what?

josh

Re: Status of 5200B and 440 Kernel Support

[Albrecht Dreß]

[-- Attachment #1: Type: text/plain, Size: 1016 bytes --]

Hi Josh,

thanks a lot for your answers!

Am 24.02.08 14:38 schrieb(en) Josh Boyer:
> > - MTD/NAND controller
> 
> NOR flash works.  NAND needs some additional code.

Hmm - does it mean it's currently under development?  I couldn't find  
anything in make xconfig for stock 2.6.24 (running on a PowerMac, but  
of course with the 44x enabled).

> UART works.  IIC has recent patches.  No idea about SPI.

SPI is probably not too hard.  I did some hacks for a slave in an  
arm920 based environment for that.  Here I would need master mode, so I  
guess some of the infrastructure form drivers/spi can be used.

> > - local external bus (not PCI) DMA
> 
> DMA to the EBC for what?

To load data from an external measurement device which should be  
connected to the cpu's external bus at a fixed address.  It will  
produce "bursts" which should be shifted out of it without cpu  
intervention if possible, as it does not have a big fifo, and I must  
not lose data.

Thanks,
Albrecht.

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

2.6.25-rc3 Sequoia build - section mismatch(es)

[Steve Heflin]

Upon building the default Sequoia (AMCC-440EPx) configuration, the 
MODPOST step warns about section mismatches:
    WARNING: modpost: Found 5 section mismatch(es).
    To see full details build your kernel with:
    'make CONFIG_DEBUG_SECTION_MISMATCH=y'

After using the 

2.6.25-rc3 Sequoia build - section mismatch(es)

[Steve Heflin]

Upon building the default Sequoia (AMCC-440EPx) configuration, the 
MODPOST step warns about section mismatches:
    WARNING: modpost: Found 5 section mismatch(es).
    To see full details build your kernel with:
    'make CONFIG_DEBUG_SECTION_MISMATCH=y'

After using the CONFIG_DEBUG_SECTION_MISMATCH option, the section 
mismatches are cases in one of the following:
  1) functions in the device initialization (.devinit) 
calls  routines which are contained in the .text section;
  2) functions in the device init section calls routines in the 
device exit section;
  3) data structures in the .data section contain pointers to 
functions in the .devinit section or .devexit section; (note that 
these data structures are used during initialization)

My question is, will this cause the code to malfunction?

thanks,
Steve