[Cbe-oss-dev] [RFC 0/9] Axon drivers

Jean-Christophe Dubois jdubois at mc.com
Thu Dec 21 01:53:19 EST 2006

One more note:

You will see some ugly (#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,9)) in the 
code. This is because I have to support RH4 as an OS for the CAB host.

Sorry about that. Obviously this should go away.


On Wednesday 20 December 2006 12:13, jdubois at mc.com wrote:
> These patches are not meant for integration.
> They are provided here as a proposal and a base for discussion for an Axon
> driver framework to allow multiple drivers/services to share the resources.
> In addition the provided code has some limitation that prevent it to be
> integrated as is.
> 1) the drivers are provided without comment as requested by the Axon IP
> owner. 2) the drivers are tested on 2.6.16 (for the Cell part) as this is
> the only kernel version that can support the actual CAB device tree. I
> don't have yet another Axon based platform with a device tree able to
> support 2.6.19 or later. this mean the actual code is missing some features
> to support some new kernel feature like virtual IRQs or whatever bus
> abstraction there might now be for the internal Axon bus.
> Note that the proposed driver has been tested quite a bit on the CAB board
> and allow the Cell and the PCI-E host to access all the Axon resources on
> an equal stand. This mean a driver writen for the Cell on top of this
> framework should just work on the host. A typical example is the Ethernet
> over PCI driver where the same driver code (not binary obviously) run on
> the host and on the Cell.
> So this driver goes on top of the recent patches I sent for the 2.6.16
> kernel for the CAB that were on top of Arnd patches for FCAB.
> (http://www.kernel.org/pub/linux/kernel/people/arnd/patches/2.6.16-arnd2-ca
>b1) In particular one important feature is the support of the MBX direct
> interrupt as the indirect interrupt is unfortunately not reliable in the
> actual Axon.

More information about the cbe-oss-dev mailing list