Really nasty testing of networking errors can be done on a single machine
disconnected from the network. It requires some hacked up kernel drivers
that look like a connection to a network device and software to drive it.
There is some talk in the TCP community on having a standard interface
for writing this kind of stuff. David and I have already offered to have
Linux support such an effort.
>Hardware errors can be simulated by software as well. Imagine a
>kernel thread (kerrord?) which would mess with pending I/O requests to
>simulate hardware errors?
Yes. I do this quite regularly to test networking code, although
I don't have a really nice clean interface for it yet. Perhaps a
standardized interface to introduce various types of hardware
errors could be useful.
-- Eric Schenk www: http://www.dna.lth.se/~erics Dept. of Comp. Sci., Lund University email: Eric.Schenk@dna.lth.se Box 118, S-221 00 LUND, Sweden fax: +46-46 13 10 21 ph: +46-46 222 96 38