On Tue, Apr 28, 2020 at 01:54:24PM -0700, Darrick J. Wong wrote: > > > + if (item->ri_type && item->ri_type->ra_pass2_fn) > > > + item->ri_type->ra_pass2_fn(log, item); > > > > Shouldn't we ensure eatly on that we always have a valid ->ri_type? > > Item sorting will bail out on unrecognized log item types, in which case > we will discard the transaction (and all its items) and abort the mount, > right? That should suffice to ensure that we always have a valid > ri_type long before we get to starting readahead for pass 2. Yes, I think so.
