On Fri, Mar 19, 2004 at 09:49:45AM +1100, Norman Gaywood wrote: > There have been a lot of threads about the stability of the FC1 SMP > kernel on this list. Most point to these bugzillas: > > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=109962 I wonder whether this patch from Andrea Arcangeli's tree helps with umount: http://www.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.23aa2/00_umount-against-unused-dirty-inodes-race-2 [This week I hacked mkinitrd to use genext2fs, just to avoid the other (related?) long-lived umount problem: "spinning in umount of loopback-mounted file". Now I can at least install a new kernel without hassle.] > I doubt this problem will be fixed. It's been around a long time and > there are now some shiny new toys (2.6 kernel) to distract the kernel > developers. I came across the above patch today as I've gone searching for alternatives to the FC1 kernels. It's going to be a busy weekend evaluating the alternatives as a base kernel on which to add local fixes and functionality. Working NPTL is a requirement. I'm currently looking at: - RHEL3 errata kernel - Andrea's 2.4.23aa2 - Franken-kernel based on 2.4.25-rmap15l,O(1)-sched,futex,nptl - 2.6.x kernel Much as I'd like to just start using 2.6 everywhere, it still feels a bit premature, judging by the traffic on LKML, so I'm looking for a solution to bridge the next few months. Regards, Bill Rugolsky
