[lustre-devel] Isn't there a memory leak here !!!

[Sudheendra Sampath]

Hi,

While doing transaction code analysis, I found the following in
lustre/target/tgt_lastrcvd.c.

    552 int tgt_new_client_cb_add(struct thandle *th, struct obd_export
*exp)
    553 {
    554         struct tgt_new_client_callback  *ccb;
    555         struct dt_txn_commit_cb         *dcb;
    556         int                              rc;
    557
*    558         OBD_ALLOC_PTR(ccb);*
    559         if (ccb == NULL)
    560                 return -ENOMEM;
    561
    562         ccb->lncc_exp = class_export_cb_get(exp);
    563
    564         dcb = &ccb->lncc_cb;
    565         dcb->dcb_func = tgt_cb_new_client;
    566         INIT_LIST_HEAD(&dcb->dcb_linkage);
    567         strlcpy(dcb->dcb_name, "tgt_cb_new_client",
sizeof(dcb->dcb_name));
    568
    569         rc = dt_trans_cb_add(th, dcb);
    570         if (rc) {
    571                 class_export_cb_put(exp);
*    572                 OBD_FREE_PTR(ccb);*
    573         }
    574         return rc;
    575 }

OBD_FREE_PTR() is in the condition block which means that the expectation
is dt_trans_cb_add() returns something "!= 0".

>From the code, osd_trans_cb_add() and osp_trans_cb_add() returns zero
value.  So, my point is OBD_FREE_PTR() should be outside the condition
block.

Please correct me if my understanding is incorrect.

-- 
Regards

Sudheendra Sampath
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-devel-lustre.org/attachments/20210120/12aa4c5b/attachment.html>