On Thu, Jan 16, 2020 at 08:22:00PM +0800, yukuai (C) wrote: > > > ON 2020/1/10 0:46, Darrick J. Wong wrote: > > It sounds like a reasonable idea, though I was suggesting doing the > > snapshot-and-check in the xfs_db source, not fstests. > > The problem is that blocktrash do changed some bits of the attr block, > however, corrupt will still fail if the change is only inside the 'zero' > range. > > So, I think it's hard to fix the problem by doing the snapshot-and-check > in the xfs_db source. <nod> I'm a little concerned about having a static seed though, since the xfs_db rng isn't great. Does adding "-o 4" to the blocktrash command make it work reliably? That should be close enough to the start of the attrleaf block that we'll reliably corrupt *some* amount of stuff in the header. --D > Thanks! > Yu Kuai >
