On Wed 08-11-17 14:00:58, Josef Bacik wrote: > From: Josef Bacik <jbacik@xxxxxx> > > Provide a mechanism for file systems to indicate how much dirty metadata they > are holding. This introduces a few things > > 1) Zone stats for dirty metadata, which is the same as the NR_FILE_DIRTY. > 2) WB stat for dirty metadata. This way we know if we need to try and call into > the file system to write out metadata. This could potentially be used in the > future to make balancing of dirty pages smarter. > > Signed-off-by: Josef Bacik <jbacik@xxxxxx> ... > diff --git a/mm/vmscan.c b/mm/vmscan.c > index 13d711dd8776..0281abd62e87 100644 > --- a/mm/vmscan.c > +++ b/mm/vmscan.c > @@ -3827,7 +3827,8 @@ static unsigned long node_pagecache_reclaimable(struct pglist_data *pgdat) > > /* If we can't clean pages, remove dirty pages from consideration */ > if (!(node_reclaim_mode & RECLAIM_WRITE)) > - delta += node_page_state(pgdat, NR_FILE_DIRTY); > + delta += node_page_state(pgdat, NR_FILE_DIRTY) + > + node_page_state(pgdat, NR_METADATA_DIRTY); > > /* Watch for any possible underflows due to delta */ > if (unlikely(delta > nr_pagecache_reclaimable)) Do you expect your metadata pages to be accounted in NR_FILE_PAGES? Otherwise this doesn't make sense. And even if they would, this function is about kswapd / direct page reclaim and I don't think you've added smarts there to writeout metadata. So if your metadata pages are going to show up in NR_FILE_PAGES, you need to subtract NR_METADATA_DIRTY from reclaimable pages always. It would be good to see btrfs counterpart to these patches so that we can answer questions like this easily... Honza -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
