one more data point, maybe completely unrelated, but maybe pointing to
some severe buffer cache problems ?! ::
last week I created a 130MB MSDOS partition (reusing the swap partition
for a moment;) on a 4GB SCSI disk at NCR 53C810 in a PPro200/128MB box
with plain 2.0.29 because I needed real M$-Win31 for an hour...
after the dos job was finished I booted Linux again, mounted the msdos fs
in Linux and tried to fill up this file system with one big file using
cat /dev/zero > /msdos/bigfile
when the file system run full, the machine completely crashed (no kernel mgs,
no screen flashing; just completely halted).
trying to reboot showed a *completely* hosed 1GB root file system !!!
I tried to repair the thing for some time using e2fsck (needed several runs
because lost+found couldn't be expanded anymore and then e2fsck crashed;
so I had to move some suff out off lost+found and start over again 2 more times)
finally I re-installed the whole machine :-((
I'll try this again tomorrow after having made a full backup.
any ideas what should be tested or what may be the reason ?
Harald
-- All SCSI disks will from now on ___ _____ be required to send an email notice 0--,| /OOOOOOO\ 24 hours prior to complete hardware failure! <_/ / /OOOOOOOOOOO\ \ \/OOOOOOOOOOOOOOO\ \ OOOOOOOOOOOOOOOOO|// Harald Koenig, \/\/\/\/\/\/\/\/\/ Inst.f.Theoret.Astrophysik // / \\ \ koenig@tat.physik.uni-tuebingen.de ^^^^^ ^^^^^