Try recompiling the kernel and make sure that you comment out the SMP=1
line in the top of the /usr/src/linux/Makefile. I forgot to do that the
first time and my kernel would lock up all over the place, appearantly
during network access. Then I remembered about that silly little SMP
definition, removed it, recompiled and now everything seems to work
fine.
(the only other thing I did was move to 2.1.39 but the 39 patch doesn't
contain any network diffs, mostly just some EATA and Ultrastore and some
filesystem changes).
So I think that maybe a lot of people having 2.1.38 lockup are like me
and
forgot to comment out the SMP define on their uniprocessor machine.
Could somebody please refresh my memory as to why CONFIG_SMP is not part
of
the make config process? I know that a lot of files depend on this
definition but it would certainly be a lot more of an intuitive approach
to
either: A) put the CONFIG_SMP in the make config stuff, or B) change the
SMP stuff so that SMP=1 and CPUS=1 works just as well as SMP=1 and
CPUS>1
(well except for the expected performance difference)
I already have an outstanding question as to why B) isn't the case. I
still don't understand why a uniprocessor should be treated any
differently
than multiple processor (except that you could fine tune some scheduling
things for better performance).
Still trying to learn but the kernel appears to be growing faster than
my
brain.
- Jeff Wiegley