[LU-5411] cfs_hlist_for_each_entry() not working as intended

• Type: Bug
• Resolution: Duplicate
• Priority: Minor
• Severity: 3
• Assignee: WC Triage
• Reporter: Frank Zago (Inactive)

Description

When HAVE_HLIST_FOR_EACH_3ARG is defined, cfs_hlist_for_each_entry_safe (and similarly cfs_hlist_for_each_entry) are defined as this:

        pos = NULL; hlist_for_each_entry_safe(tpos, n, head, member)

However if we have some code like:

            for (i = 0; i < REMOTE_PERM_HASHSIZE; i++)
                    cfs_hlist_for_each_entry_safe(lrp, node, next, hash + i,
                                                  lrp_list)
                            free_ll_remote_perm(lrp);

we end up actually doing:

  
            for (i = 0; i < REMOTE_PERM_HASHSIZE; i++)
                pos = NULL;
            hlist_for_each_entry_safe(lrp, node, next, hash + i, lrp_list)
                            free_ll_remote_perm(lrp);

which is probably not what was intended.

pos = NULL can be safely removed since it's an internal cursor. However that will generate some "unused" warnings on platforms where HAVE_HLIST_FOR_EACH_3ARG is defined.

The other option is to add braces around the call, such as:

  
            for (i = 0; i < REMOTE_PERM_HASHSIZE; i++) {
                    cfs_hlist_for_each_entry_safe(lrp, node, next, hash + i,
                                                  lrp_list)
                            free_ll_remote_perm(lrp);
            }

However that won't prevent a similar bug from being introduced in the future.

Issue Links

• is related to: LU-3963 cleanup libcfs wrappers