Skip to content

Commit

Permalink
blk-iolatency: deal with small samples
Browse files Browse the repository at this point in the history
There is logic to keep cgroups that haven't done a lot of IO in the most
recent scale window from being punished for over-active higher priority
groups.  However for things like ssd's where the windows are pretty
short we'll end up with small numbers of samples, so 5% of samples will
come out to 0 if there aren't enough.  Make the floor 1 sample to keep
us from improperly bailing out of scaling down.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
  • Loading branch information
Josef Bacik authored and Jens Axboe committed Sep 28, 2018
1 parent 9f60511 commit 22ed8a9
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion block/blk-iolatency.c
Original file line number Diff line number Diff line change
Expand Up @@ -366,7 +366,7 @@ static void check_scale_change(struct iolatency_grp *iolat)
* scale down event.
*/
samples_thresh = lat_info->nr_samples * 5;
samples_thresh = div64_u64(samples_thresh, 100);
samples_thresh = max(1ULL, div64_u64(samples_thresh, 100));
if (iolat->nr_samples <= samples_thresh)
return;
}
Expand Down

0 comments on commit 22ed8a9

Please sign in to comment.