[LU-16341] unable to handle kernel NULL in qmt_site_recalc_cb

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.16.0
• Severity: 3
• Assignee: Sergey Cheremencev
• Reporter: Sergey Cheremencev

Description

BUG: unable to handle kernel NULL pointer dereference at 00000000000000f8
...
RIP: 0010:qmt_site_recalc_cb+0x2ec/0x780 [lquota]
...
[ffffa5564118fda0] cfs_hash_for_each_tight at ffffffffc0c72c81 [libcfs]
[ffffa5564118fe08] qmt_pool_recalc at ffffffffc142dec7 [lquota]
[ffffa5564118ff10] kthread at ffffffffb45043a6
[ffffa5564118ff50] ret_from_fork at ffffffffb4e00255 

The panic occurres due to an empty qit_lqes array after qmt_pool_lqes_lookup_spec. Sometimes it is possible if global lqe is not enforced.

Issue Links

• is duplicated by: LU-16448 sanity-quota test_3b: crashes in qsd_reint_qpool
• is related to: LU-17974 sanity-quota test_3b: (qmt_lock.c:957:qmt_id_lock_notify()) ASSERTION( lqe->lqe_is_global )
• is related to: LU-16725 crash at qmt_free_lqe_gd+0xa/0x1f0
• is related to: LU-16772 Protect lqe_glbl_data in qmt_site_recalc_cb with mutex