On Thu, Apr 4, 2019 at 11:24 PM Zirconium Hacker <jared.e.vb@xxxxxxxxx> wrote: > > I've actually had a check running for the past 10 hours or so. > It got far enough, so I stopped it. The output is attached. What do you get for: # btrfs insp dump-s -f /dev/ # btrfs insp dump-t -b 2629033574400 /dev/ The csum test doesn't complain so that suggests any corruption happens before csum is calculated and written. It could be a bug, however we also have a wanted transid that is lower than what's found (finds a transid in the future). That's a bug no matter what, either the drive firmware has some bug or Btrfs, and I would expect the firmware can't be responsible for such a huge transid difference. But yeah maybe Qu has an idea, also he knows about qgroups, I don't know anything about them other than it can make things very slow. There are some fixes on the way soon but it doesn't really help today. -- Chris Murphy
