Christoph, Sorry about the delay. Travel got in the way. >> On the wire between controller and target there's only CRC. If I want to >> write a "bad" CRC to disk, I have switch the controller to CRC mode. The >> controller can't convert a "bad" IP checksum to a "bad" CRC. The PI test >> tooling relies heavily on being able to write "bad" things to disk and >> read them back to validate that we detect the error. > > But how do you even toggle the flag? There is no no code to do that. > And if you already have a special kernel module for that it really > should just use a passthrough request to take care of that. A passthrough command to the controller? > Note that unlike the NOCHECK flag which I just cleaned up because they > were unused, this one actually does get in the way of the architecture > of the whole series :( We could add a per-bip csum_type but it would > feel really weird. Why would it feel weird? That's how it currently works. The qualification tool issues a flurry of commands injecting errors at various places in the stack to identify that the right entity (block layer, controller, storage device) catch a bad checksum, reference tag, etc. Being able to enable/disable checking at each place in the stack is important. I also have code for target that does the same thing in the reverse direction. -- Martin K. Petersen Oracle Linux Engineering
