[Cbe-oss-dev] [patch 07/16] powerpc and x86_64: Initial port of PCIe endpoint device driver for Axon

[Arnd Bergmann]

On Thursday 13 March 2008, Murali Iyer wrote:
> On Thu, 2008-03-13 at 04:57 +0100, Arnd Bergmann wrote:
> 
> > > +void axon_initialize_dacs_sysfs(struct axon *axon_priv)
> > > +{
> > 
> > What does this function have to do with sysfs? It seems to just
> > set some generic variables per device that happen to also be exported
> > through sysfs.
> This is request from users.. some might belong to debugfs but several
> might have to stay. Team will consider making it easy to read.

I was just contesting the naming of the function, not its purpose.
Function names should tell the reader what the function does, not
what the caller wants to use it for.

> > The request size limits seem rather strange to me. User space should not
> > assume that you can map anything of this size into the IOMMU. On most
> > systems, the IOMMU has a total size of just a few megabytes, shared by
> > all devices, and it's certainly not just architecture dependent.
> I suspect it is very application specific. we may have to come across
> traditional limitations to satisfy the app.
> I am not the expert on this.. and need to defer this discussion until
> code for mainline is put out..

This is a user-visible ABI, so we can never change it once the code
goes into mainline. Therefore it's important to get this right
upfront.
One option would be to have all the sysfs interface code in a separate
patch and defer merging of that part.

Similar to the line of thought above, there should be no application
specific interfaces in generic code. An interface exported by the
driver should make sense for any application using it.

	Arnd <><