[PATCH net 0/2] mlx5_core fixes for 3.18

[PATCH net 0/2] mlx5_core fixes for 3.18

[Eli Cohen]

Hi Dave,
the following two patches fix races to could lead to kernel panic in some cases.

Thanks,
Eli

Eli Cohen (2):
  net/mlx5_core: Fix race in create EQ
  net/mlx5_core: Fix race on driver load

 drivers/net/ethernet/mellanox/mlx5/core/eq.c   | 7 +++----
 drivers/net/ethernet/mellanox/mlx5/core/main.c | 4 ++--
 2 files changed, 5 insertions(+), 6 deletions(-)

-- 
2.1.2

[PATCH net 1/2] net/mlx5_core: Fix race in create EQ

[Eli Cohen]

After the EQ is created, it can possibly generate interrupts and the interrupt
handler is referencing eq->dev. It is therefore required to set eq->dev before
calling request_irq() so if an event is generated before request_irq() returns,
we will have a valid eq->dev field.

Signed-off-by: Eli Cohen <eli@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/eq.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eq.c b/drivers/net/ethernet/mellanox/mlx5/core/eq.c
index a278238a2db6..ad2c96a02a53 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eq.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eq.c
@@ -374,15 +374,14 @@ int mlx5_create_map_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq, u8 vecidx,
 	snprintf(eq->name, MLX5_MAX_EQ_NAME, "%s@pci:%s",
 		 name, pci_name(dev->pdev));
 	eq->eqn = out.eq_number;
+	eq->irqn = vecidx;
+	eq->dev = dev;
+	eq->doorbell = uar->map + MLX5_EQ_DOORBEL_OFFSET;
 	err = request_irq(table->msix_arr[vecidx].vector, mlx5_msix_handler, 0,
 			  eq->name, eq);
 	if (err)
 		goto err_eq;
 
-	eq->irqn = vecidx;
-	eq->dev = dev;
-	eq->doorbell = uar->map + MLX5_EQ_DOORBEL_OFFSET;
-
 	err = mlx5_debug_eq_add(dev, eq);
 	if (err)
 		goto err_irq;
-- 
2.1.2

[PATCH net 2/2] net/mlx5_core: Fix race on driver load

[Eli Cohen]

When events arrive at driver load, the event handler gets called even before
the spinlock and list are initialized. Fix this by moving the initialization
before EQs creation.

Signed-off-by: Eli Cohen <eli@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/main.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
index 88b2ffa0edfb..ecc6341e728a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
@@ -855,14 +855,14 @@ static int init_one(struct pci_dev *pdev,
 	dev->profile = &profile[prof_sel];
 	dev->event = mlx5_core_event;
 
+	INIT_LIST_HEAD(&priv->ctx_list);
+	spin_lock_init(&priv->ctx_lock);
 	err = mlx5_dev_init(dev, pdev);
 	if (err) {
 		dev_err(&pdev->dev, "mlx5_dev_init failed %d\n", err);
 		goto out;
 	}
 
-	INIT_LIST_HEAD(&priv->ctx_list);
-	spin_lock_init(&priv->ctx_lock);
 	err = mlx5_register_device(dev);
 	if (err) {
 		dev_err(&pdev->dev, "mlx5_register_device failed %d\n", err);
-- 
2.1.2

Re: [PATCH net 0/2] mlx5_core fixes for 3.18

[David Miller]

From: Eli Cohen <eli-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
Date: Thu,  6 Nov 2014 12:51:20 +0200

> the following two patches fix races to could lead to kernel panic in some cases.

Series applied, thanks Eli.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html