[Cbe-oss-dev] [patch 00/16] powerpc and x86_64: Initial port of PCIe endpoint device driver for Axon

[Arnd Bergmann]

On Saturday 08 March 2008, Murali Iyer wrote:
> NOT FOR INCLUSION IN MAINLINE.
> 
> This prototype device driver(s) developed to demonstrate high speed
> data transfer across PCIe bus for IBM's Axon Cell Companion chip connecting x86_64
> compatible host and CBE based Blade system.
> In general, this patch set provides remote DMA kind of interface and virtual
> ethernet interface across PCIe bus. For details see the each patches.

Hi Murali and Brett, great that you could finally send this driver to the public
lists! I'll start a new review, probably mentioning a few of the points that
I've already replied to you in private mails. I hope that finally some other
people can join in on the review.

When you send the next version, it may be good to also include
linuxppc-dev at ozlabs.org to get a broader audience, but for the first
round, sending only to cbe-oss-dev is appropriate.

> This patch set applies to 2.6.23.1 (yes not to the HEAD). There are several
> improvements needs to be made including porting to HEAD, breaking into multiple
> drivers for greater acceptability and usability, coding style including making sparse
> happy etc. etc.

Agreed. I think it makes most sense to include changes for review comments
only in the version based on the latest upstream kernel though, so you don't
have to maintain two streams of development.
Future submissions should be based on the latest head, so that people can
easily do compile tests on it.
 
> We are mailing this, so interested people could get early access to the code.
> Hopefully, in the coming weeks will start posting drivers in small chunks for
> review and hopefully for upstreaming.

Excellent. I think it's good to have this post as a reference of what the
drivers are all about, but posting smaller subsets should help the review
process. I don't think many volunteers can be bothered reading through all
the code at once, but a single separate driver typically gets important
review comments.

> Configure the following for Cell based systems to get axon.ko and apnet.ko
> 
> CONFIG_TB_AXON=m
> CONFIG_TB_AXON_MTRR_SMA=y
> CONFIG_APNET=m
> 
> and for x86_64 systems
> 
> CONFIG_TB_AXON=m
> CONFIG_TB_AXON_PCI=y
> CONFIG_TB_AXON_MTRR_SMA=y
> CONFIG_APNET=m
> 
> Some patches are very large and not sure this mailing list takes it.

The patches all came through fine, no problem.

Regarding the configuration options, one of the goals will be that every
configuration you can select actually compiles and makes sense at least
for a hypothetical system. So CONFIG_TB_AXON_PCI=y should build on any
machine with PCI support, independent of the architecture and there should
be no obvious bugs that would prevent it e.g. from running on a powerpc
host.

For the emails, please use meaningful subject lines for each patch, quilt
will do that if the first line in the patch file starts with 'Subject:'.
When you Cc me personally, please use arnd at arndb.de instead of my IBM
address.

Thanks,

	Arnd <><