On Sat, Oct 11, 2008 at 12:51:52PM +0200, Nick Piggin wrote: > On Fri, Oct 10, 2008 at 11:32:56PM +0530, Aneesh Kumar K.V wrote: > > The range_cyclic writeback mode use the address_space > > writeback_index as the start index for writeback. With > > delayed allocation we were updating writeback_index > > wrongly resulting in highly fragmented file. Number of > > extents reduced from 4000 to 27 for a 3GB file with > > the below patch. > > > > The patch also removes the range_cont writeback mode > > added for ext4 delayed allocation. Instead we add > > two new flags in writeback_control which control > > the behaviour of write_cache_pages. > > The mm/page-writeback.c changes look OK, although it loks like you've > got rid of range_cont? Should we do a patch to get rid of it entirely > from the tree first? > > I don't mind rediffing my patchset on top of this, but this seems smaller > and not strictly a bugfix so I would prefer to go the other way if you > agree. > > Seems like it could be broken up into several patches (eg. pagevec_lookup). > > The results look very nice. I actually tried to do that. But to do that and also achieve a working bisect kernel, I will have to do the patches in below way a) Introduce ext4_write_cache_pages b) remove range_cont from write_cache_pages c) Introduce the new flags to writeback_control d) switch ext4 to use write_cache_pages. I thought that involved lot of code which are later getting removed. So i went for a single patch. -aneesh -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
