On 8/11/25 11:13 PM, Rosemarie O'Riorden wrote:
> Previously, if there was a logical switch with a virtual port that had a
> virtual parent port configured, and then the parent port was deleted and
> recreated, the parent port would not be reclaimed by the chassis and
> thus would not behave as the parent port, breaking traffic.
> 
> This is because virtual ports were not being updated if their parents
> were recently created or deleted.
> 
> So long as virtual ports have a parent port configured, there should be
> no issue deleting and recreating the parent. It should continue working
> as the parent as soon as it's up again.
> 
> With this change, the parent port will function as the parent even after
> deletion and recreation. The chassis will properly reclaim the virtual
> port.
> 
> Fixes: b337750e45be ("northd: Incremental processing of VIF changes in 
> 'northd' node.")
> Reported-at: https://issues.redhat.com/browse/FDP-710
> Co-authored-by: Mohammad Heib <[email protected]>
> Signed-off-by: Mohammad Heib <[email protected]>
> Signed-off-by: Rosemarie O'Riorden <[email protected]>
> ---
> v4:
>   - Added Fixes tag.
>   - Combined two ssets for deleted and created ports into one.
>   - Nest looping through list under condition that sset is not empty.
>   - Only destory sset if not empty, otherwise unnecessary.
>   - Spacing, ordering, and other syntax nits.
> 
> v3:
>   - Fixed algorithm to use an sset and loop through created ports, only
>     updating them if their virtual parent was recently created or
>     deleted.
>   - Added case for deleted parents, as just mentioned.
>   - Fixed send_garp() syntax in test and remove redefinition.
>   - Added wait for port binding to disappear in test.
>   - Added more virtual parents to test.
>   - Fixed various syntax issues/nits.
> 
> v2: Fixed memory leak by freeing op_smap and tokstr.
> ---
>  northd/northd.c | 45 +++++++++++++++++++++++++
>  tests/ovn.at    | 89 +++++++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 134 insertions(+)
> 
> diff --git a/northd/northd.c b/northd/northd.c
> index 2cb69f9aa..742d39a72 100644
> --- a/northd/northd.c
> +++ b/northd/northd.c
> @@ -4559,7 +4559,10 @@ ls_handle_lsp_changes(struct ovsdb_idl_txn 
> *ovnsb_idl_txn,
>      bool ls_had_only_router_ports = (!vector_is_empty(&od->router_ports)
>              && (vector_len(&od->router_ports) == hmap_count(&od->ports)));
>  
> +    struct ovs_list existing_virtual_ports;
>      struct ovn_port *op;
> +
> +    ovs_list_init(&existing_virtual_ports);
>      HMAP_FOR_EACH (op, dp_node, &od->ports) {
>          op->visited = false;
>      }
> @@ -4629,6 +4632,8 @@ ls_handle_lsp_changes(struct ovsdb_idl_txn 
> *ovnsb_idl_txn,
>                  delete_fdb_entries(ni->sbrec_fdb_by_dp_and_port,
>                                     od->tunnel_key, old_tunnel_key);
>              }
> +        } else if (!strcmp(op->nbsp->type, "virtual")) {
> +            ovs_list_push_back(&existing_virtual_ports, &op->list);
>          }
>          op->visited = true;
>      }
> @@ -4675,6 +4680,46 @@ ls_handle_lsp_changes(struct ovsdb_idl_txn 
> *ovnsb_idl_txn,
>          }
>      }
>  
> +    /* Update old virtual ports that have newly created or newly deleted
> +     * VIF as parent port. This code handles cases where the virtual port was
> +     * created before the parent port or when the parent port was recreated.
> +     */
> +    struct hmapx_node *hmapx_node;
> +
> +    struct sset created_or_deleted_ports;
> +    sset_init(&created_or_deleted_ports);
> +
> +    HMAPX_FOR_EACH (hmapx_node, &trk_lsps->created) {
> +        op = hmapx_node->data;
> +        sset_add(&created_or_deleted_ports, op->nbsp->name);
> +    }
> +
> +    HMAPX_FOR_EACH (hmapx_node, &trk_lsps->deleted) {
> +        op = hmapx_node->data;
> +        sset_add(&created_or_deleted_ports, op->nbsp->name);
> +    }
> +
> +    if (!sset_is_empty(&created_or_deleted_ports)) {
> +        LIST_FOR_EACH_POP (op, list, &existing_virtual_ports) {
> +            const char *virtual_parents =
> +                        smap_get_def(&op->nbsp->options,
> +                                     "virtual-parents", "");

nit: I'd move the line to the left a bit and just split in 2 instead of 3.

> +            char *tokstr = xstrdup(virtual_parents);
> +            char *save_ptr = NULL;
> +            char *vparent;
> +
> +            for (vparent = strtok_r(tokstr, ",", &save_ptr); vparent != NULL;
> +                 vparent = strtok_r(NULL, ",", &save_ptr)) {
> +                if (sset_find(&created_or_deleted_ports, vparent)) {
> +                    add_op_to_northd_tracked_ports(&trk_lsps->updated, op);
> +                    break;
> +                }
> +            }
> +            free(tokstr);
> +        }
> +        sset_destroy(&created_or_deleted_ports);

While technically current implementation of sset doesn't allocate any memory
for the empty set, it's not guaranteed, so we should still call the destroy()
after we called init().  So, this line move this out of the if statement.

Other than that the patch looks good to me.  Thanks!

Maybe maintainers can apply the following diff while picking up the fix:

diff --git a/northd/northd.c b/northd/northd.c
index 742d39a72..e380e0bd3 100644
--- a/northd/northd.c
+++ b/northd/northd.c
@@ -4702,8 +4702,7 @@ ls_handle_lsp_changes(struct ovsdb_idl_txn *ovnsb_idl_txn,
     if (!sset_is_empty(&created_or_deleted_ports)) {
         LIST_FOR_EACH_POP (op, list, &existing_virtual_ports) {
             const char *virtual_parents =
-                        smap_get_def(&op->nbsp->options,
-                                     "virtual-parents", "");
+                smap_get_def(&op->nbsp->options, "virtual-parents", "");
             char *tokstr = xstrdup(virtual_parents);
             char *save_ptr = NULL;
             char *vparent;
@@ -4717,8 +4716,8 @@ ls_handle_lsp_changes(struct ovsdb_idl_txn *ovnsb_idl_txn,
             }
             free(tokstr);
         }
-        sset_destroy(&created_or_deleted_ports);
     }
+    sset_destroy(&created_or_deleted_ports);
 
     return true;
 
---

With that,

Acked-by: Ilya Maximets <[email protected]>
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to