On Sun, Mar 20, 2022 at 09:44:00AM -0700, Darrick J. Wong wrote: > From: Darrick J. Wong <djwong@xxxxxxxxxx> > > We've established in this patchset that the "alloc_set_aside" pool is > actually used to ensure that a bmbt split always succeeds so that the I don't think this is correct. See my previous comments about the reason for set_aside existing and why I think the bmbt portion of it isn't valid anymore. > filesystem won't run out of space mid-transaction and crash. That's exactly what transaction reservations and args->minleft is for, yes? Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx
