On Sat, Feb 02, 2019 at 08:14:04AM -0800, Christoph Hellwig wrote: > On Fri, Feb 01, 2019 at 11:33:48AM -0800, Darrick J. Wong wrote: > > Maybe we can just drop this patch entirely? If we leak any unlinked > > inodes they'll get cleaned up at next mount... wait, we never did merge > > Eric's patch to reap unlinked inodes at mount time, did we? > > Which patch? We must recovery unlinked inodes during each mount that > requires log recovery and have done that since the very beginning. I /think/ the patch was "xfs: don't require a dirty log on snapshots". > I don't see how we could ever see unlinked inodes during a clean > mount. /me had thought it was related to something about root filesystems that are mounted ro and then remounted rw, but <shrug> I can't recall ATM. Anyway, I'm still not sure we need all this mutex+counter infrastructure for a debugging assert... --D
