[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: question



On Wed, Dec 20, 2000 at 06:07:32AM -0500, Ingo Molnar wrote:
> 
> On Thu, 14 Dec 2000, Andi Kleen wrote:
> 
> > > Considering the fact that the network stack was more or less
> > > re-written for 2.4, I would tend to describe that as non-trivial.
> >
> > Sorry, but I have to correct such an incorrect statement.
> >
> > It has not been rewriten, just made SMP multithreaded with minimum
> > changes. The 2.4 stack is very similar to 2.2.
> 
> i suspect Michael ment 2.4 plus the zerocopy changes. The 2.4+ZC
> networking driver API is significantly different (and cleaned up).
> 'minimum changes' is a pretty big understatement - about ~100 thousand
> lines of code were changed from 2.2 -> 2.4, only in the net/ source
> hierarchy, not counting driver changes. [the net/ hierarchy has 200
> thousand lines of code, so every second line was changed.]

I bet you included new code like netfilter and decnet which have not much
impact on core networking but lots of LOCs in the diff run @)

> (but even that isnt strictly true, softnet itself redesigned a pretty
> important part of the networking code)

In a straightforward way to multithread it.

My point was only that a 2.2 based UP TUX port likely wouldn't be that hard
to do, if someone wanted to sink time into an old code base. Do you agree?

-Andi





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index] []