PATCH [7/27] Fix timeout/cancelled MAD handling

[Hal Rosenstock <halr@voltaire.com>]

Fixes an issue processing a sent MAD after it has timed out or 
been canceled.  The race occurs when a response MAD matches with
the send request.  The request could time out or be canceled after
the response MAD matches with the request, but before the request
completion can be processed.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>

This patch depends on patch 6/27.

-- 
 mad.c      |   14 ++++++++++++--
 mad_priv.h |    1 +
 2 files changed, 13 insertions(+), 2 deletions(-)
diff -uprN linux-2.6.13-rc2-mm1/drivers/infiniband6/core/mad.c linux-2.6.13-rc2-mm1/drivers/infiniband7/core/mad.c
-- linux-2.6.13-rc2-mm1/drivers/infiniband6/core/mad.c	2005-07-09 15:12:32.000000000 -0400
+++ linux-2.6.13-rc2-mm1/drivers/infiniband7/core/mad.c	2005-07-09 16:48:45.000000000 -0400
@@ -341,6 +341,7 @@ struct ib_mad_agent *ib_register_mad_age
 	spin_lock_init(&mad_agent_priv->lock);
 	INIT_LIST_HEAD(&mad_agent_priv->send_list);
 	INIT_LIST_HEAD(&mad_agent_priv->wait_list);
+	INIT_LIST_HEAD(&mad_agent_priv->done_list);
 	INIT_WORK(&mad_agent_priv->timed_work, timeout_sends, mad_agent_priv);
 	INIT_LIST_HEAD(&mad_agent_priv->local_list);
 	INIT_WORK(&mad_agent_priv->local_work, local_completions,
@@ -1559,6 +1560,16 @@ find_send_req(struct ib_mad_agent_privat
 	return NULL;
 }
 
+static void ib_mark_req_done(struct ib_mad_send_wr_private *mad_send_wr)
+{
+	mad_send_wr->timeout = 0;
+	if (mad_send_wr->refcount == 1) {
+		list_del(&mad_send_wr->agent_list);
+		list_add_tail(&mad_send_wr->agent_list,
+			      &mad_send_wr->mad_agent_priv->done_list);
+	}
+}
+
 static void ib_mad_complete_recv(struct ib_mad_agent_private *mad_agent_priv,
 				 struct ib_mad_recv_wc *mad_recv_wc)
 {
@@ -1580,8 +1591,7 @@ static void ib_mad_complete_recv(struct 
 				wake_up(&mad_agent_priv->wait);
 			return;
 		}
-		/* Timeout = 0 means that we won't wait for a response */
-		mad_send_wr->timeout = 0;
+		ib_mark_req_done(mad_send_wr);
 		spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
 
 		/* Defined behavior is to complete response before request */
diff -uprN linux-2.6.13-rc2-mm1/drivers/infiniband6/core/mad_priv.h linux-2.6.13-rc2-mm1/drivers/infiniband7/core/mad_priv.h
-- linux-2.6.13-rc2-mm1/drivers/infiniband6/core/mad_priv.h	2005-07-09 15:09:53.000000000 -0400
+++ linux-2.6.13-rc2-mm1/drivers/infiniband7/core/mad_priv.h	2005-07-09 16:48:05.000000000 -0400
@@ -92,6 +92,7 @@ struct ib_mad_agent_private {
 	spinlock_t lock;
 	struct list_head send_list;
 	struct list_head wait_list;
+	struct list_head done_list;
 	struct work_struct timed_work;
 	unsigned long timeout;
 	struct list_head local_list;


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/