[openib-general] Re: [PATCH] CMA and iWARP

[Tom Tucker]

On Sat, 2006-01-21 at 09:13 -0800, Sean Hefty wrote:
> >> I don't see much urgency in merging it now.  When svn diverges from
> >> what's upstream in the kernel, it makes my life harder because I have
> >> to figure out which patches belong upstream and sometimes merge things
> >> by hand (when they hit the divergent regions).
> >
> >The easy solution here is not to diverge. Unless the iWARP support
> >regresses IB functionality, it does no harm and creates a single
> >software core for both iWARP and IB developers to bring new drivers to
> >market.
> 
> Until iWarp is integrated with the kernel, 

It thought the approach was branch --> trunk --> kernel. What am I
missing here?

> the code will diverge however.  And I
> agree with Roland, merging diverged code upstream is a pain.  

No argument here. Merging code downstream is a pain too ;-)

> I'm definitely
> willing to re-organize the code to make it easier to maintain the code out of
> the tree.  Also, if we can isolate the IB/iWarp code into separate files, then
> it's not a big issue pushing changes upstream.

Making the code more modular is a good idea anyway. The provider and CM
are already in separate files. At some point, though there is a single
API and these files will have code for both transports (e.g.
ib_verbs.h). One way to modularize the CMA is to have transport CM's
register with the CMA and force all calls through function pointers ala
verbs.

> - Sean