Robert Love wrote: > Well, I'm a caring guy so I went and tried to duplicate (in much fear > for my data) your problem. Thank you - > I compiled 2.4.14+preempt+ext3. Booted, ran multiple dbench runs, > booted X, compiled a new kernel. (Un)fortunately, I could not reproduce > the problem (or any problem). Maybe my hardware brings out the best - > So I compiled 2.4.14+preempt, this time using ext2 like you. I also > enabled highmem and SMP, although I don't need those. Again I ran > multiple dbench runs, went into X, and here I am ... the kernel is > solid. I am using ext3 as well - > What was the last kernel you had no problems with when used with > preempt? hmm, let me go back through our correspondence.... Looks like we documented 2.4.12-ac3 as being the point where you came up with the fix (Oct 20) Unfortunately I wiped the disks before installing 7.2, so the past is lost. I had been concentrating on seeing how fast tux is, and what I had to do to get the ext3 dbench results back into the same ballpark as with ext2 - 2.4.14 was the first self compiled kernel after installing 7.2 that I can definitely say I applied the preempt patch. I will try this on my low end box, compiled with all the high end options - it showed the problem last time... > Is 2.4.14 running now? Without preempt? Fine? Well when I get back to work tomorrow I will see what it takes to get the box back up - I may end up just doing a reinstall, since it's a test box and nothing important on it yet. I'll surely build a vanilla 2.4.14 and test. > Can you attach your .config? Sure, attached - > > Trace; c019ed8a > > Trace; c019edef Perhaps these lines show that the compaq "smart controller" could be implicated? cu jjs