Curt, > > The above stats are probably useful. I'm not so convinced about the stats > > below - it looks like it should be simple enough to get them by enabling > > some trace points and processing output (or if we are missing some > > tracepoints, it would be worthwhile to add them). > > For these specifically, I'd agree with you. In general, though, I > think that having generally available aggregated stats is really > useful, in a different way than tracepoints are. If there comes such useful aggregated stats in future, they may go to vmstat and/or /debug/bdi/<dev>/stats. Then we make the writeback stats a simple uniform interface rather than a hybrid one. > > > >> periodic writeback 377 The above one can go to the "work" column, "periodic" row of the writeback stats table :) I'm in particular interested in the "work" column of the "try_to_free_pages" row. I also suspect there could be many short-lived background work, hence there lots of them. > >> single inode wait 0 > >> writeback_wb wait 1 Thanks, Fengguang -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>