Re: Testing md modules?

On Thursday February 26, nathapl@cs.okstate.edu wrote:

I'm to the point now where my Reed-Solomon code appears to be working, at 
least for some simple tests I've thrown at it.  I'd like to test it more 
exhaustively - does anybody have any scripts or programs that work 
deterministicly on a block device to test it for correctness?  What has 
been done in the past to make sure the previous RAID modules worked, beside 
sticking the code out there and waiting for bug reports?

untar a complete kernel source, and compile it.
Lots of times in parallel.

It certainly isn't guaranteed to find all problems, but it is easy to
do and will exercise a lot of code.

You should randomly fail drives and start rebuilds while that is
happening.

Also diff -r of two kernel trees after exploding them.

And make sure you are short of memory so that lots of reads go to
disk.

NeilBrown