I'm done reading and while a complex subsystem and a bunch of changes they individually all seem sane to me (although a108557b could have side effects that are hard to spot).
For SRU considerations I think this includes potential change of behavior of formerly silently ignored errors now becoming visible (or with more detail) to the guest. But IMHO silently hiding I/O errors is asking for problems and data corruption (just as you've found it=, while reporting them is correct.
I'll later build a PPA for your testing as you seem to have a testcase with Disktest on your side that can reproduce the issue that made you aware.
I'm done reading and while a complex subsystem and a bunch of changes they individually all seem sane to me (although a108557b could have side effects that are hard to spot).
For SRU considerations I think this includes potential change of behavior of formerly silently ignored errors now becoming visible (or with more detail) to the guest. But IMHO silently hiding I/O errors is asking for problems and data corruption (just as you've found it=, while reporting them is correct.
I'll later build a PPA for your testing as you seem to have a testcase with Disktest on your side that can reproduce the issue that made you aware.