mlx4_core: Fix potential deadlock in mlx4_eq_int()
The slave_state_lock spinlock is used in both interrupt context and
process context, hence irq locking must be used. Found by lockdep.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Roland Dreier <roland@purestorage.com>
diff --git a/drivers/net/ethernet/mellanox/mlx4/cmd.c b/drivers/net/ethernet/mellanox/mlx4/cmd.c
index e791e70..fdc5f23 100644
--- a/drivers/net/ethernet/mellanox/mlx4/cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx4/cmd.c
@@ -1498,6 +1498,7 @@
u32 reply;
u8 is_going_down = 0;
int i;
+ unsigned long flags;
slave_state[slave].comm_toggle ^= 1;
reply = (u32) slave_state[slave].comm_toggle << 31;
@@ -1576,12 +1577,12 @@
mlx4_warn(dev, "Bad comm cmd:%d from slave:%d\n", cmd, slave);
goto reset_slave;
}
- spin_lock(&priv->mfunc.master.slave_state_lock);
+ spin_lock_irqsave(&priv->mfunc.master.slave_state_lock, flags);
if (!slave_state[slave].is_slave_going_down)
slave_state[slave].last_cmd = cmd;
else
is_going_down = 1;
- spin_unlock(&priv->mfunc.master.slave_state_lock);
+ spin_unlock_irqrestore(&priv->mfunc.master.slave_state_lock, flags);
if (is_going_down) {
mlx4_warn(dev, "Slave is going down aborting command(%d)"
" executing from slave:%d\n",
@@ -1597,10 +1598,10 @@
reset_slave:
/* cleanup any slave resources */
mlx4_delete_all_resources_for_slave(dev, slave);
- spin_lock(&priv->mfunc.master.slave_state_lock);
+ spin_lock_irqsave(&priv->mfunc.master.slave_state_lock, flags);
if (!slave_state[slave].is_slave_going_down)
slave_state[slave].last_cmd = MLX4_COMM_CMD_RESET;
- spin_unlock(&priv->mfunc.master.slave_state_lock);
+ spin_unlock_irqrestore(&priv->mfunc.master.slave_state_lock, flags);
/*with slave in the middle of flr, no need to clean resources again.*/
inform_slave_state:
memset(&slave_state[slave].event_eq, 0,
diff --git a/drivers/net/ethernet/mellanox/mlx4/eq.c b/drivers/net/ethernet/mellanox/mlx4/eq.c
index c509a86..3cb7727 100644
--- a/drivers/net/ethernet/mellanox/mlx4/eq.c
+++ b/drivers/net/ethernet/mellanox/mlx4/eq.c
@@ -407,6 +407,7 @@
struct mlx4_slave_state *slave_state = priv->mfunc.master.slave_state;
int i;
int err;
+ unsigned long flags;
mlx4_dbg(dev, "mlx4_handle_slave_flr\n");
@@ -418,10 +419,10 @@
mlx4_delete_all_resources_for_slave(dev, i);
/*return the slave to running mode*/
- spin_lock(&priv->mfunc.master.slave_state_lock);
+ spin_lock_irqsave(&priv->mfunc.master.slave_state_lock, flags);
slave_state[i].last_cmd = MLX4_COMM_CMD_RESET;
slave_state[i].is_slave_going_down = 0;
- spin_unlock(&priv->mfunc.master.slave_state_lock);
+ spin_unlock_irqrestore(&priv->mfunc.master.slave_state_lock, flags);
/*notify the FW:*/
err = mlx4_cmd(dev, 0, i, 0, MLX4_CMD_INFORM_FLR_DONE,
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED);
@@ -446,6 +447,7 @@
u8 update_slave_state;
int i;
enum slave_port_gen_event gen_event;
+ unsigned long flags;
while ((eqe = next_eqe_sw(eq, dev->caps.eqe_factor))) {
/*
@@ -653,13 +655,13 @@
} else
update_slave_state = 1;
- spin_lock(&priv->mfunc.master.slave_state_lock);
+ spin_lock_irqsave(&priv->mfunc.master.slave_state_lock, flags);
if (update_slave_state) {
priv->mfunc.master.slave_state[flr_slave].active = false;
priv->mfunc.master.slave_state[flr_slave].last_cmd = MLX4_COMM_CMD_FLR;
priv->mfunc.master.slave_state[flr_slave].is_slave_going_down = 1;
}
- spin_unlock(&priv->mfunc.master.slave_state_lock);
+ spin_unlock_irqrestore(&priv->mfunc.master.slave_state_lock, flags);
queue_work(priv->mfunc.master.comm_wq,
&priv->mfunc.master.slave_flr_event_work);
break;