Currently it is allowed to create hard-link across quota limit directories. This becomes hard to enforce quota.
This cannot be implemented in the current quota design: Consider below scenario Case-1) a) below data with hard-links pre-exists /d1/f1 /d2/f1 b) Set quota limit on /d1 and /d2 c) Now when creating hard-link /d1/f1 /d1/f2 can fail As we do fetch the inode_parent of f1 (f1's parent can be /d1 or /d2 in the dentry list) Case-2) a) Set quota limit on /d1 and /d2 b) create file /d1/f1 and link /d1/f2 c) rename /d1/f2 /d2/f2 d) Now when creating hard-link /d1/f1 /d1/f2 can fail As we do fetch the inode_parent of f1 (f1's parent can be /d1 or /d2 in the dentry list) So marking the bug as wontfix