ipv4: ipmr: fix invalid cache resolving when adding a non-matching entry
authorPatrick McHardy <kaber@trash.net>
Thu, 15 Apr 2010 11:29:28 +0000 (13:29 +0200)
committerPatrick McHardy <kaber@trash.net>
Thu, 15 Apr 2010 11:31:29 +0000 (13:31 +0200)
The patch to convert struct mfc_cache to list_heads (ipv4: ipmr: convert
struct mfc_cache to struct list_head) introduced a bug when adding new
cache entries that don't match any unresolved entries.

The unres queue is searched for a matching entry, which is then resolved.
When no matching entry is present, the iterator points to the head of the
list, but is treated as a matching entry. Use a seperate variable to
indicate that a matching entry was found.

Signed-off-by: Patrick McHardy <kaber@trash.net>
net/ipv4/ipmr.c

index 5df5fd7..0643fb6 100644 (file)
@@ -1089,12 +1089,14 @@ static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
         *      Check to see if we resolved a queued list. If so we
         *      need to send on the frames and tidy up.
         */
+       found = false;
        spin_lock_bh(&mfc_unres_lock);
        list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
                if (uc->mfc_origin == c->mfc_origin &&
                    uc->mfc_mcastgrp == c->mfc_mcastgrp) {
                        list_del(&uc->list);
                        atomic_dec(&mrt->cache_resolve_queue_len);
+                       found = true;
                        break;
                }
        }
@@ -1102,7 +1104,7 @@ static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
                del_timer(&mrt->ipmr_expire_timer);
        spin_unlock_bh(&mfc_unres_lock);
 
-       if (uc) {
+       if (found) {
                ipmr_cache_resolve(net, mrt, uc, c);
                ipmr_cache_free(uc);
        }