1) Today I had a small 'accident' with partitions and mke2fs. fdisk
claimed (as usual) that 'reread of partition table failed'. As usual I
ignored this. However, this once the actual re-read HAD failed, so upon
rebooting and mounting the partition... I got a panic.
My grievance is that my machine was completely _DEAD_ (well, my mate's
machine, so I'm not as bothered ;-). I've seen linux recover from device
driver crashes(!), gracefully disable the offending interrupt, and stagger
blindly forth.
Should mounting a hosed FS have the capability to kill our machines? For
the time being ignore the fact _how_ I created the knackered FS; it is
irrelevant to my point and anyone saying "that was stupid Chris" will get
a personal reply of 10,000 lines saying "I know".
2) Just to really put the cat amoung the pigeons... and this IS stupid and
WAS deliberate, but still..
(kernel 2.1.29)
# insmod ./serial.o
A module named 'serial' already exists
# mv serial.o other.o
# insmod ./other.o
# rmmod other
kernel: OOPS!!! OOPS!!! kfree of non kmalloc'ed! etc!
I'm a fan of even root not being able to cause kernel oopses aside from
eg. cat /dev/random > /dev/kcore.
I'm unsure if other modules behave the same, but surely a simple check on
behalf of this serial module could have gone along the lines of "ooh, the
IRQ's I want AND I/O port ranges, are already taken! better refuse to
load.."
Chris