On Mon, Jun 15, 2020 at 09:43:51AM -0700, Darrick J. Wong wrote: > Also: Why not set PF_MEMALLOC_NOFS at the start of the freeze call > chain? Because there's no guarantee that we are always going to do this final work in the freeze syscall context? i.e. the state is specific to the context in which we are running the transaction, not the task context it is running in... Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx