Below review comments nned to be addressed by 3.7 code freeze:

Review comments from patch:
http://review.gluster.org/#/c/9769/24/xlators/cluster/dht/src/dht-common.c


File: xlators/cluster/dht/src/dht-common.c
Function: 	dht_aggregate_quota_xattr
Kaushal M		3:18 PM

From what I understand, dht_aggregate is called only from the 2nd brick onwards, and adds/aggregates information to the 1st bricks information. (I might be wrong about this though).
In that case, there is no actual need to aggregate the dir_count here and correct it later on. The correction later is just dividing the aggregated dir_count with subvol count, which will be the same as the dir_count of 1st brick (as directories are replicated on all DHT subvolumes).
If there was any particular reason you did it this way, can you explain it?

In case there is a real need for the correction, why not just start the count from 2 here.
On line 350, you have a comment explaining why you are using (brick_count + 1) as the divisor. You wouldn't need that if you started the count from 2 here.
OTOH, if you did start the count from 2 here, you'd need a comment explaining it here.




--------------------------
Atin:
Please include remove-object functionality in inode-quota.t before code freeze.