On Fri, 1 Dec 2000, Erwin Authried wrote:
> porting back sometimes difficult. I'm thinking of MTD and JFFS, for example.
> I have done some work for MTD so that the current CVS version ís usable
> for 2.0, but only the most important modules have been ported back. The
> latest JFFS sources are not usable for 2.0, as far as I know. I'm aware
2.0 and 2.4 have vastly different VFS's so it's difficult to port back and
forth. It should be easy to locate bugfixes in the 2.4 version though to
port into 2.0. But there's a lot of stuff pending that are optimizations
that rely on 2.4 (dentries instead of jffs in-core reps etc) which would
be pointless to try to port back to 2.0. That's why inode.c is a different
file for the versions and not one with ifdef's.. OTOH, GC improvements
should be easily back-ported (though the GC is a kernel thread in the
2.2/2.4 version, I don't think the MTD guys did that change in 2.0.
We (at Axis) still use the 2.0 version of JFFS but will switch over to 2.4
eventually and thus probably won't do any new features on the 2.0 branch
As for the uClinux 2.4 port, that's probably a good idea. There really is
a lot of new neat stuff in 2.4. For example the much more advanced routing
with filters and shaping, USB, better VFS and pagecache, slab allocator,
better IDE system etc etc (some was in 2.2 already of course)...
In our case we found it easier to build an MMU for the CPU than to port
uclinux to 2.4 though :)
This message resent by the firstname.lastname@example.org list server http://www.uClinux.com/
This archive was generated by hypermail 2b30 : Sun Apr 07 2002 - 00:01:39 EST