Hi Ingo, Philip, Re: my previous message about raid and rc5des clobbering each other. After reading back over the linux-raid list archive, it would seem that what i'm seeing is more similar to the "out of memory" problems that other people have been seeing, rather an mmx clobber problem. keeping rc5des out of it, i tried some different chunk and filesystem stride sizes and did a "bonnie -s 400" at chunk=8, fs block=2048, stride=4, performance with regards to machine responsiveness: incredible, no ill-effects at all even with a bonnie 500MB benchmark. chunk=16, fs block=4096, stride=4. a bit worse. mouse jumpy, etc. a few processes started died with "out of memory for .." logged by kernel. chunk=32, fs block=4k, stride=8. bad. mouse and lots of X programmes inactive for long periods of time. eventually loads of programmes start dying off, including things like diald, sendmail, X.. machine is a bit hosed after this. so the problem would seem to be with chunk sizes perhaps. rc5des running constantly seems to aggravate it though. Anyway, it's working fine with chunk=8 for me at the moment, with rc5des. :) regards, Paul Jakma.
