When destroying a cm_id from a context of a work queue and if the lap_state of this cm_id is IB_CM_LAP_SENT, we need to release the reference of this id that was taken upon the send of the lap message. Otherwise, if the expected apr message gets lost, it is only after a long time that the reference will be released, while during that the work handler thread is not available to process other things. Also, a sigle variable 'msg' was used to hold the last snet mad. However, this variable serves 2 state machines (CM and LAP) which can cause to memory leaks or to unexpected behavior. This patch adds another variable to hold the lap mad.
Signed-off-by: Moni Shoua <[email protected]> Signed-off-by: Amir Vadai <[email protected]> -- drivers/infiniband/core/cm.c | 25 ++++++++++++++++++++++--- 1 file changed, 22 insertions(+), 3 deletions(-) diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index 64e0903..3c00646 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -210,6 +210,7 @@ struct cm_id_private { atomic_t refcount; struct ib_mad_send_buf *msg; + struct ib_mad_send_buf *lap_msg; struct cm_timewait_info *timewait_info; /* todo: use alternate port on send failure */ struct cm_av av; @@ -839,6 +840,24 @@ static void cm_destroy_id(struct ib_cm_id *cm_id, int err) cm_id_priv = container_of(cm_id, struct cm_id_private, id); retest: spin_lock_irq(&cm_id_priv->lock); + + /* handle lap states first */ + switch (cm_id->lap_state) { + case IB_CM_LAP_UNINIT: + case IB_CM_LAP_IDLE: + break; + case IB_CM_LAP_SENT: + cm_id_priv->id.lap_state = IB_CM_LAP_IDLE; + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->lap_msg); + cm_id_priv->lap_msg = NULL; + break; + case IB_CM_LAP_RCVD: + case IB_CM_MRA_LAP_SENT: + case IB_CM_MRA_LAP_RCVD: + default: + break; + } + switch (cm_id->state) { case IB_CM_LISTEN: cm_id->state = IB_CM_IDLE; @@ -2615,7 +2634,7 @@ int ib_send_cm_lap(struct ib_cm_id *cm_id, } cm_id->lap_state = IB_CM_LAP_SENT; - cm_id_priv->msg = msg; + cm_id_priv->lap_msg = msg; out: spin_unlock_irqrestore(&cm_id_priv->lock, flags); return ret; @@ -2811,8 +2830,8 @@ static int cm_apr_handler(struct cm_work *work) goto out; } cm_id_priv->id.lap_state = IB_CM_LAP_IDLE; - ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); - cm_id_priv->msg = NULL; + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->lap_msg); + cm_id_priv->lap_msg = NULL; ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
