1
0
Fork 0

net/mlx4_core: Fix deadlock when switching between polling and event fw commands

When switching from polling-based fw commands to event-based fw
commands, there is a race condition which could cause a fw command
in another task to hang: that task will keep waiting for the polling
sempahore, but may never be able to acquire it. This is due to
mlx4_cmd_use_events, which "down"s the sempahore back to 0.

During driver initialization, this is not a problem, since no other
tasks which invoke FW commands are active.

However, there is a problem if the driver switches to polling mode
and then back to event mode during normal operation.

The "test_interrupts" feature does exactly that.
Running "ethtool -t <eth device> offline" causes the PF driver to
temporarily switch to polling mode, and then back to event mode.
(Note that for VF drivers, such switching is not performed).

Fix this by adding a read-write semaphore for protection when
switching between modes.

Fixes: 225c7b1fee ("IB/mlx4: Add a driver Mellanox ConnectX InfiniBand adapters")
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
hifive-unleashed-5.1
Jack Morgenstein 2016-09-20 14:39:42 +03:00 committed by David S. Miller
parent 30353bfc43
commit a7e1f04905
2 changed files with 19 additions and 6 deletions

View File

@ -785,17 +785,23 @@ int __mlx4_cmd(struct mlx4_dev *dev, u64 in_param, u64 *out_param,
return mlx4_cmd_reset_flow(dev, op, op_modifier, -EIO);
if (!mlx4_is_mfunc(dev) || (native && mlx4_is_master(dev))) {
int ret;
if (dev->persist->state & MLX4_DEVICE_STATE_INTERNAL_ERROR)
return mlx4_internal_err_ret_value(dev, op,
op_modifier);
down_read(&mlx4_priv(dev)->cmd.switch_sem);
if (mlx4_priv(dev)->cmd.use_events)
return mlx4_cmd_wait(dev, in_param, out_param,
out_is_imm, in_modifier,
op_modifier, op, timeout);
ret = mlx4_cmd_wait(dev, in_param, out_param,
out_is_imm, in_modifier,
op_modifier, op, timeout);
else
return mlx4_cmd_poll(dev, in_param, out_param,
out_is_imm, in_modifier,
op_modifier, op, timeout);
ret = mlx4_cmd_poll(dev, in_param, out_param,
out_is_imm, in_modifier,
op_modifier, op, timeout);
up_read(&mlx4_priv(dev)->cmd.switch_sem);
return ret;
}
return mlx4_slave_cmd(dev, in_param, out_param, out_is_imm,
in_modifier, op_modifier, op, timeout);
@ -2454,6 +2460,7 @@ int mlx4_cmd_init(struct mlx4_dev *dev)
int flags = 0;
if (!priv->cmd.initialized) {
init_rwsem(&priv->cmd.switch_sem);
mutex_init(&priv->cmd.slave_cmd_mutex);
sema_init(&priv->cmd.poll_sem, 1);
priv->cmd.use_events = 0;
@ -2583,6 +2590,7 @@ int mlx4_cmd_use_events(struct mlx4_dev *dev)
if (!priv->cmd.context)
return -ENOMEM;
down_write(&priv->cmd.switch_sem);
for (i = 0; i < priv->cmd.max_cmds; ++i) {
priv->cmd.context[i].token = i;
priv->cmd.context[i].next = i + 1;
@ -2606,6 +2614,7 @@ int mlx4_cmd_use_events(struct mlx4_dev *dev)
down(&priv->cmd.poll_sem);
priv->cmd.use_events = 1;
up_write(&priv->cmd.switch_sem);
return err;
}
@ -2618,6 +2627,7 @@ void mlx4_cmd_use_polling(struct mlx4_dev *dev)
struct mlx4_priv *priv = mlx4_priv(dev);
int i;
down_write(&priv->cmd.switch_sem);
priv->cmd.use_events = 0;
for (i = 0; i < priv->cmd.max_cmds; ++i)
@ -2626,6 +2636,7 @@ void mlx4_cmd_use_polling(struct mlx4_dev *dev)
kfree(priv->cmd.context);
up(&priv->cmd.poll_sem);
up_write(&priv->cmd.switch_sem);
}
struct mlx4_cmd_mailbox *mlx4_alloc_cmd_mailbox(struct mlx4_dev *dev)

View File

@ -46,6 +46,7 @@
#include <linux/interrupt.h>
#include <linux/spinlock.h>
#include <net/devlink.h>
#include <linux/rwsem.h>
#include <linux/mlx4/device.h>
#include <linux/mlx4/driver.h>
@ -627,6 +628,7 @@ struct mlx4_cmd {
struct mutex slave_cmd_mutex;
struct semaphore poll_sem;
struct semaphore event_sem;
struct rw_semaphore switch_sem;
int max_cmds;
spinlock_t context_lock;
int free_head;