On Mon, Nov 27, 2023 at 07:33:25AM +0100, Christoph Hellwig wrote: > On Mon, Nov 27, 2023 at 10:31:31AM +0530, Ritesh Harjani wrote: > > The code change looks very obvious. But sorry that I have some queries > > which I would like to clarify - > > > > The dirty page we are trying to write can always belong to the dirty > > inode with inline data in it right? > > Yes. > > > So it is then the FS responsibility to un-inline the inode in the > > ->map_blocks call is it? > > I think they way it currently works for gfs2 is that writeback from the > page cache never goes back into the inline area. > > If we ever have a need to actually write back inline data we could > change this code to support it, but right now I just want to make the > assert more consistent. Question: Do we even /want/ writeback to be initiating transactions to log the inline data? I suppose for ext4/jbd2 that would be the least inefficient time to do that. --D
