How are fedora kernels modified from vanilla kernels

[Matthew Miller]

On Sun, Oct 24, 2004 at 05:36:10PM -0700, Brian Mury wrote:
> On Sun, 2004-24-10 at 20:23 -0400, Matthew Miller wrote:
> > So, significantly less is changed -- but thirty thousand lines of changes
> > might still count as "heavily".
> So what exactly is different, anyway? What am I missing out on by
> running a generic kernel.org kernel? Anything worth having?

Well, that's hard to answer succinctly for that many lines of patching,
especially since I'm not a kernel hacker. :)

But one short answer is: there's a new development model with the 2.6 series
of kernels, and an important part of that model is that distribution kernel
branches are the "actually stable", and the official 2.6 releases subject to
surprising changes. (The kernel developers point out that this is less a
change and more a reflection of the way it's been in reality anyway.)

More on this here: <http://kerneltrap.org/node/view/3513>

A quick look at the patches shows that the big (as in K) ones are:
ext3-online-resize (44K), exec-shield (44K), netdump (49K), mlock (49K),
ext3-reservations (54K), 4g4g (126K), modsign-mpilib (204K), and tux (337K).

Tux (the in-kernel static-content web server) accounts for 45% of the total
lines of added code.

Module signing (there was something about this on one of these lists awhile
ago) is another 28%.

And in fact, the eight patch files listed above acconut for more than 85% of
the lines changed.

So actually, investigation may be swaying me back to the "it's lightly
patched" side -- plus a handful of moderately big additions.

Take a look at the source RPM -- many of the patches actually have short
comments explaining what they are, or else are pretty obvious from the name,
or can be searched on.



-- 
Matthew Miller           mattdm at mattdm.org        <http://www.mattdm.org/>
Boston University Linux      ------>                <http://linux.bu.edu/>