rculfhash: add assertions on node alignment
authorMathieu Desnoyers <mathieu.desnoyers@efficios.com>
Thu, 14 Feb 2013 16:36:43 +0000 (11:36 -0500)
committerMathieu Desnoyers <mathieu.desnoyers@efficios.com>
Fri, 15 Feb 2013 01:01:06 +0000 (20:01 -0500)
I've had a report of someone running into issues with the RCU lock-free
hash table by embedding the struct cds_lfht_node into a packed structure
by mistake, thus not respecting alignment requirements stated in
urcu/rculfhash.h. Assertions on "replace" and "add" operations should
catch this, but I notice that we should add assertions on the
REMOVAL_OWNER_FLAG to cover all possible misalignments.

Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
rculfhash.c

index b8fbb6481d9291a6226f2200f8374b810ffd34af..7080bb62ec528306b7cd42b05df5a1de7c788fc9 100644 (file)
@@ -832,13 +832,16 @@ void _cds_lfht_gc_bucket(struct cds_lfht_node *bucket, struct cds_lfht_node *nod
 
        assert(!is_bucket(bucket));
        assert(!is_removed(bucket));
+       assert(!is_removal_owner(bucket));
        assert(!is_bucket(node));
        assert(!is_removed(node));
+       assert(!is_removal_owner(node));
        for (;;) {
                iter_prev = bucket;
                /* We can always skip the bucket node initially */
                iter = rcu_dereference(iter_prev->next);
                assert(!is_removed(iter));
+               assert(!is_removal_owner(iter));
                assert(iter_prev->reverse_hash <= node->reverse_hash);
                /*
                 * We should never be called with bucket (start of chain)
@@ -859,6 +862,7 @@ void _cds_lfht_gc_bucket(struct cds_lfht_node *bucket, struct cds_lfht_node *nod
                        iter = next;
                }
                assert(!is_removed(iter));
+               assert(!is_removal_owner(iter));
                if (is_bucket(iter))
                        new_next = flag_bucket(clear_flag(next));
                else
@@ -879,8 +883,10 @@ int _cds_lfht_replace(struct cds_lfht *ht, unsigned long size,
                return -ENOENT;
 
        assert(!is_removed(old_node));
+       assert(!is_removal_owner(old_node));
        assert(!is_bucket(old_node));
        assert(!is_removed(new_node));
+       assert(!is_removal_owner(new_node));
        assert(!is_bucket(new_node));
        assert(new_node != old_node);
        for (;;) {
@@ -955,6 +961,7 @@ void _cds_lfht_add(struct cds_lfht *ht,
 
        assert(!is_bucket(node));
        assert(!is_removed(node));
+       assert(!is_removal_owner(node));
        bucket = lookup_bucket(ht, size, hash);
        for (;;) {
                uint32_t chain_len = 0;
@@ -1015,7 +1022,9 @@ void _cds_lfht_add(struct cds_lfht *ht,
        insert:
                assert(node != clear_flag(iter));
                assert(!is_removed(iter_prev));
+               assert(!is_removal_owner(iter_prev));
                assert(!is_removed(iter));
+               assert(!is_removal_owner(iter));
                assert(iter_prev != node);
                if (!bucket_flag)
                        node->next = clear_flag(iter);
@@ -1035,6 +1044,7 @@ void _cds_lfht_add(struct cds_lfht *ht,
 
        gc_node:
                assert(!is_removed(iter));
+               assert(!is_removal_owner(iter));
                if (is_bucket(iter))
                        new_next = flag_bucket(clear_flag(next));
                else
@@ -1697,6 +1707,7 @@ int cds_lfht_delete_bucket(struct cds_lfht *ht)
                if (!is_bucket(node))
                        return -EPERM;
                assert(!is_removed(node));
+               assert(!is_removal_owner(node));
        } while (!is_end(node));
        /*
         * size accessed without rcu_dereference because hash table is
This page took 0.026153 seconds and 4 git commands to generate.