On 04/13/2010 09:18 PM, Janito Ferreira Filho wrote: > is it a reasonable implementation to use the block cache to store > altered blocks notified in the journal without actually writing them? > I'm thinking about getting write access to the overwritten blocks when > the journal replays the log in read only mode, and then discard > everything when the file system is unmounted. That doesn't really sound like a good solution, and would also waste quite a bit of non-pageable memory. [...] > It would probably be more efficient if there were a block remapping > scheme. Does the block cache support any kind of block triggers? Would > it be possible to implement something like that (maybe in the future)? There is no support for this yet, but this would definitely be the way to go. Bye, Axel.