In the process of hammering the threaded MPM with flood (which is threaded too) on a SMP Linux 2.4.3 box, I got processes that were in a "disk sleep" state. They are not killable, attachable, or debuggable. Some documentation in the kernel refers to this D state as the "uninterruptible sleep" state. The only way to remove the processes is to reboot. Yuck.
Anyway, it seems to be a bug in the 2.4.3 kernel (maybe earlier?) kernel. However, some people reported that it was fixed in 2.4.3-ac5 (see google). I just upgraded to 2.4.6 (Mandrake's current cooker series kernel RPM) and I haven't been able to reproduce it yet. This seems worthy of a README as both Mandrake 8.0 and RedHat 7.1 are based on 2.4.3 - so I believe they will both exhibit this. -- justin
