On Wed, Jun 16, 2010 at 09:20:09AM +1000, Dave Chinner wrote: > On Tue, Jun 15, 2010 at 08:55:38PM +1000, Nick Piggin wrote: > > On Tue, Jun 15, 2010 at 02:28:22PM +0400, Evgeniy Polyakov wrote: > > > On Tue, Jun 15, 2010 at 04:36:43PM +1000, Dave Chinner (david@xxxxxxxxxxxxx) wrote: > > > Per-mapping sorting will not do anything good in this case, even if > > > files were previously created in a good facion being placed closely and > > > so on, and only block layer will find a correlation between adjacent > > > blocks in different files. But with existing queue management it has > > > quite a small opportunity, and that's what I think Andrew is arguing > > > about. > > > > The solution is not to sort pages on their way to be submitted either, > > really. > > > > What I do in fsblock is to maintain a block-nr sorted tree of dirty > > blocks. This works nicely because fsblock dirty state is properly > > synchronized with page dirty state. > > How does this work with delayed allocation where there is no block > number associated with the page until writeback calls the allocation > routine? It doesn't. I have been thinking about how best to make that work. The mm/writeback is not in a good position to know what to do, so the fs would have to help. So either an fs callback, or the fs would have to insert the blocks (or some marker) into the tree itself. It's relatively easy to do for a single file (just walk the radix-tree and do delalloc conversions), but between multiple files is harder (current code has the same problem though). -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>