rhashtable: Fix rhlist duplicates insertion
authorPaul Blakey <paulb@mellanox.com>
Sun, 4 Mar 2018 15:29:48 +0000 (17:29 +0200)
committerDavid S. Miller <davem@davemloft.net>
Wed, 7 Mar 2018 15:44:02 +0000 (10:44 -0500)
When inserting duplicate objects (those with the same key),
current rhlist implementation messes up the chain pointers by
updating the bucket pointer instead of prev next pointer to the
newly inserted node. This causes missing elements on removal and
travesal.

Fix that by properly updating pprev pointer to point to
the correct rhash_head next pointer.

Issue: 1241076
Change-Id: I86b2c140bcb4aeb10b70a72a267ff590bb2b17e7
Fixes: ca26893f05e8 ('rhashtable: Add rhlist interface')
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/rhashtable.h
lib/rhashtable.c

index c9df252..668a21f 100644 (file)
@@ -766,8 +766,10 @@ slow_path:
                if (!key ||
                    (params.obj_cmpfn ?
                     params.obj_cmpfn(&arg, rht_obj(ht, head)) :
-                    rhashtable_compare(&arg, rht_obj(ht, head))))
+                    rhashtable_compare(&arg, rht_obj(ht, head)))) {
+                       pprev = &head->next;
                        continue;
+               }
 
                data = rht_obj(ht, head);
 
index 3825c30..47de025 100644 (file)
@@ -506,8 +506,10 @@ static void *rhashtable_lookup_one(struct rhashtable *ht,
                if (!key ||
                    (ht->p.obj_cmpfn ?
                     ht->p.obj_cmpfn(&arg, rht_obj(ht, head)) :
-                    rhashtable_compare(&arg, rht_obj(ht, head))))
+                    rhashtable_compare(&arg, rht_obj(ht, head)))) {
+                       pprev = &head->next;
                        continue;
+               }
 
                if (!ht->rhlist)
                        return rht_obj(ht, head);