afs: Panic on afs_conn refcount imbalance

An undercounted afs_conn can easily cause a panic and/or memory
corruption later on, since we put an rx_connection reference with each
afs_conn reference. Panic as soon as we detect this, as this indicates
a serious bug.

Change-Id: I251fd3303114d0822b8cf70805a8a447986a7762
Reviewed-on: http://gerrit.openafs.org/6413
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

diff --git a/src/afs/afs_conn.c b/src/afs/afs_conn.c
index 26e239e0892617708a9118ec624c1ebb654a653f..fcb42609aefd44c43f741a6a35613805055667fd 100644 (file)
--- a/src/afs/afs_conn.c
+++ b/src/afs/afs_conn.c
@@ -601,22 +601,8 @@ afs_PutConn(struct afs_conn *ac, struct rx_connection *rxconn,
     AFS_STATCNT(afs_PutConn);
     ac->refCount--;
     if (ac->refCount < 0) {
-       static int warned = 0;
-       /* So, someone is 'put'ing more refs than they got. From now on, we
-        * have no idea if the structure is actually still in use, so just
-        * set the refcount to a really negative number to make it unlikely
-        * that the count will ever reach 0 and the conn gets freed. This
-        * leaks memory, but the alternative is panicing, or risking memory
-        * corruption. */
-       ac->refCount = -10000;
-       if (!warned) {
-           warned = 1;
-           afs_warn("afs_PutConn: negative refCount with 0x%lx; this should "
-                    "not ever happen! Trying to carry on anyway, but please "
-                    "report this issue\n",
-                    (unsigned long)(uintptrsz)ac);
-       }
-       return;
+       osi_Panic("afs_PutConn: refcount imbalance 0x%lx %d",
+                 (unsigned long)(uintptrsz)ac, (int)ac->refCount);
     }
     ac->parent->refCount--;
     rx_PutConnection(rxconn);