Several versions of this ebuild ago (aka lolo-sources) it was found that the O(1) scheduler in conjunction with some other patches were causing strange problems on SMP systems. The decision was to remove Ingo's scheduler however it has managed to appear again in the patch and should/needs to be removed again because of the same SMP problems. I'm still working on tracking down the appropriate conflicts and can say that both ac-sources and ck-sources (most recent versions) show the problem but a 2.4.19 vanilla with only the O(1) patch does not... can we pull O(1) untill something is figured out? Reproducible: Always Steps to Reproduce: 1. 2. 3.
You are right, I did pull it to resolve process priority issues. The newly implemented O(1)sched does not seem to exhibit priority issues on several smp boxes that I have running. Nor have others reported SMP issues. I'd like more info on which issues you are noticing please. FYI - The most recent stable gentoo-sources (2.4.20-r2) does not contain O(1). Only pfeifer-sources-2.4.20-pre7 and gentoo-sources-2.4.20-r3 contain it once again and both are marked ~x86. So, it has not of yet been re-worked into our stable branch. However, the direction the {gentoo/pfeifer}-sources is going will include O(1) unless I can see some drastic reason as to not include it. Thanks, Jay
Alright, after 3 days of rebuilding kernels I've managed to resolve my problems... this also means it is a problem specific to me. Very strange but it involves building the kernel with certain options built in instead of as modules. I'd go ahead and close this bug because in my tizzy of building I deleted the config that was causing me problems and can't reproduce it. Sorry for the bother, however keep this in mind because its possible in the future someone may come up with the same problem! Thanks for all the help!
closed as per user.
