Something that struck me while wading through the whole ext3 journaling errors thread was a tongue-in-cheek suggestion to Peter that he picture the kernel as a state machine and start verifying it, to which Peter replied with a static analysis tool link.
I also recall a test script that someone cooked up to reproduce a raid6 bug that involved loopback devices etc
Finally, I recall a couple of vague mentions that it was possible to exercise the code by putting faulty virtual block devices under it and simulating faults, and that there was someone, or a group of someone that does a fair amount of reliability testing (maybe it was RedHat?)
For no really good reason, that made me wonder how the code is tested, and if there were any automated test suites.
Assuming there aren't, I pictured a modular system that includes three parts.
1) a raid testing engine that takes some sort of raid command language and uses it to build a specified raid out of loop devices and faulty block devices and then exercise it using a specified set of tasks (load, failures, resysncs, etc) all the while posting results back to an URL where a CGI just logs them or something simple
2) a script generator that can crank out command files for the testing engine that cover all the various permutations.
3) an image-building system that produces a bootable ISO based on knoppix with a kernel from a kernel source tree you specify that will hit a configurable URL for a script to run
Then I'm picturing a cycle where new code can be built into an ISO, the script can generate all the test permutations, and set VMWare (or similar) going in a loop, harvesting results in a crash-proof way.
I'm not that skilled at C code anymore, but this kind of system is mostly just duct-tape (imho), and I can do that. It might be fun to play around with if there's value to it. I've been meaning to learn how to customize knoppix and automate vmware tasks anyway, and we already have the raid6 test as an example script.
-Mike - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
