Comment 2 for bug 1543430

Ugh, yeah, I see now that the patch fixing dealing with loopbacks at all "regresses" in a way, but not dealing correctly with the loop entries even though it appears to deal with the devmapper table just fine.

I'm looking at it right now, it's not pretty. There are basically two ways I can think of to deal with this. Either relying more on loopdev/device than S_ISREG on the buf that describes the file opened (we needed to stat the loopback device after picking it, otherwise no major/minor numbers to do the partition mappings), or hacking buf.st_mode to convince kpartx we're still dealing with a regular file.

I'm trying to see what might break if we just use loopdev or device everywhere. I will upload as soon as I'm reasonably confident this does not further break things, and I'll adjust the autopkgtest accordingly.