If an error occurs, 'mlx4_en_destroy_netdev()' is called.
It then calls 'mlx4_en_free_resources()' which does the needed resources
cleanup.
So, doing some explicit kfree in the error handling path would lead to
some double kfree.
Simplify code to avoid such a case.
Fixes: 67f8b1dcb9ee ("net/mlx4_en: Refactor the XDP forwarding rings scheme")
Signed-off-by: Christophe JAILLET <[email protected]>
---
drivers/net/ethernet/mellanox/mlx4/en_netdev.c | 8 +-------
1 file changed, 1 insertion(+), 7 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
index e0adac4a9a19..9670b33fc9b1 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
@@ -3324,12 +3324,11 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
MAX_TX_RINGS, GFP_KERNEL);
if (!priv->tx_ring[t]) {
err = -ENOMEM;
- goto err_free_tx;
+ goto out;
}
priv->tx_cq[t] = kzalloc(sizeof(struct mlx4_en_cq *) *
MAX_TX_RINGS, GFP_KERNEL);
if (!priv->tx_cq[t]) {
- kfree(priv->tx_ring[t]);
err = -ENOMEM;
goto out;
}
@@ -3582,11 +3581,6 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
return 0;
-err_free_tx:
- while (t--) {
- kfree(priv->tx_ring[t]);
- kfree(priv->tx_cq[t]);
- }
out:
mlx4_en_destroy_netdev(dev);
return err;
--
2.17.0
On Thu, May 10, 2018 at 09:02:26AM +0200, Christophe JAILLET wrote:
> If an error occurs, 'mlx4_en_destroy_netdev()' is called.
> It then calls 'mlx4_en_free_resources()' which does the needed resources
> cleanup.
>
> So, doing some explicit kfree in the error handling path would lead to
> some double kfree.
Patch make sense but what's bothering me is that mlx4_en_free_resources
loops on the entire array, assuming !priv->tx_ring[t] means entry is
allocated but the existing code does not assume that, see [1]. So i looked
to see where tx_ring array is zeroed and didn't find it.
Am i missing something here.
>
> Simplify code to avoid such a case.
>
> Fixes: 67f8b1dcb9ee ("net/mlx4_en: Refactor the XDP forwarding rings scheme")
> Signed-off-by: Christophe JAILLET <[email protected]>
> ---
> drivers/net/ethernet/mellanox/mlx4/en_netdev.c | 8 +-------
> 1 file changed, 1 insertion(+), 7 deletions(-)
>
> diff --git a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
> index e0adac4a9a19..9670b33fc9b1 100644
> --- a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
> +++ b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
> @@ -3324,12 +3324,11 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
> MAX_TX_RINGS, GFP_KERNEL);
> if (!priv->tx_ring[t]) {
> err = -ENOMEM;
> - goto err_free_tx;
> + goto out;
> }
> priv->tx_cq[t] = kzalloc(sizeof(struct mlx4_en_cq *) *
> MAX_TX_RINGS, GFP_KERNEL);
> if (!priv->tx_cq[t]) {
> - kfree(priv->tx_ring[t]);
> err = -ENOMEM;
> goto out;
> }
> @@ -3582,11 +3581,6 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
>
> return 0;
>
> -err_free_tx:
> - while (t--) {
[1]
> - kfree(priv->tx_ring[t]);
> - kfree(priv->tx_cq[t]);
> - }
> out:
> mlx4_en_destroy_netdev(dev);
> return err;
> --
> 2.17.0
>
> --
> 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
On Thu, May 10, 2018 at 04:38:08PM +0300, Yuval Shaia wrote:
> On Thu, May 10, 2018 at 09:02:26AM +0200, Christophe JAILLET wrote:
> > If an error occurs, 'mlx4_en_destroy_netdev()' is called.
> > It then calls 'mlx4_en_free_resources()' which does the needed resources
> > cleanup.
> >
> > So, doing some explicit kfree in the error handling path would lead to
> > some double kfree.
>
> Patch make sense but what's bothering me is that mlx4_en_free_resources
> loops on the entire array, assuming !priv->tx_ring[t] means entry is
> allocated but the existing code does not assume that, see [1]. So i looked
> to see where tx_ring array is zeroed and didn't find it.
>
> Am i missing something here.
>
It's zeroed twice. alloc_etherdev_mqs() allocates zeroed memory and
then we do a memset(priv, 0, sizeof(struct mlx4_en_priv));
regards,
dan carpenter
Le 10/05/2018 à 15:38, Yuval Shaia a écrit :
> On Thu, May 10, 2018 at 09:02:26AM +0200, Christophe JAILLET wrote:
>> If an error occurs, 'mlx4_en_destroy_netdev()' is called.
>> It then calls 'mlx4_en_free_resources()' which does the needed resources
>> cleanup.
>>
>> So, doing some explicit kfree in the error handling path would lead to
>> some double kfree.
> Patch make sense but what's bothering me is that mlx4_en_free_resources
> loops on the entire array, assuming !priv->tx_ring[t] means entry is
> allocated but the existing code does not assume that, see [1]. So i looked
> to see where tx_ring array is zeroed and didn't find it.
>
> Am i missing something here.
My understanding is that the array is zeoed at line 3289, when the whole
priv struct is memset(0)'ed (also done in alloc_etherdev_mqs but leaving
an explicit memset help to remind that the struct is zeroed)
If speed matters here (and I don't think so), the memset could be saved
(the mlx4_en_priv struct is quite big after all), but at least a comment
should remind that it is initialized within alloc_etherdev_mqs.
CJ
>> Simplify code to avoid such a case.
>>
>> Fixes: 67f8b1dcb9ee ("net/mlx4_en: Refactor the XDP forwarding rings scheme")
>> Signed-off-by: Christophe JAILLET <[email protected]>
>> ---
>> drivers/net/ethernet/mellanox/mlx4/en_netdev.c | 8 +-------
>> 1 file changed, 1 insertion(+), 7 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
>> index e0adac4a9a19..9670b33fc9b1 100644
>> --- a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
>> +++ b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
>> @@ -3324,12 +3324,11 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
>> MAX_TX_RINGS, GFP_KERNEL);
>> if (!priv->tx_ring[t]) {
>> err = -ENOMEM;
>> - goto err_free_tx;
>> + goto out;
>> }
>> priv->tx_cq[t] = kzalloc(sizeof(struct mlx4_en_cq *) *
>> MAX_TX_RINGS, GFP_KERNEL);
>> if (!priv->tx_cq[t]) {
>> - kfree(priv->tx_ring[t]);
>> err = -ENOMEM;
>> goto out;
>> }
>> @@ -3582,11 +3581,6 @@ int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
>>
>> return 0;
>>
>> -err_free_tx:
>> - while (t--) {
> [1]
>
>> - kfree(priv->tx_ring[t]);
>> - kfree(priv->tx_cq[t]);
>> - }
>> out:
>> mlx4_en_destroy_netdev(dev);
>> return err;
>> --
>> 2.17.0
>>
>> --
>> 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
On 10/05/2018 5:18 PM, Dan Carpenter wrote:
> On Thu, May 10, 2018 at 04:38:08PM +0300, Yuval Shaia wrote:
>> On Thu, May 10, 2018 at 09:02:26AM +0200, Christophe JAILLET wrote:
>>> If an error occurs, 'mlx4_en_destroy_netdev()' is called.
>>> It then calls 'mlx4_en_free_resources()' which does the needed resources
>>> cleanup.
>>>
>>> So, doing some explicit kfree in the error handling path would lead to
>>> some double kfree.
>>
>> Patch make sense but what's bothering me is that mlx4_en_free_resources
>> loops on the entire array, assuming !priv->tx_ring[t] means entry is
>> allocated but the existing code does not assume that, see [1]. So i looked
>> to see where tx_ring array is zeroed and didn't find it.
>>
>> Am i missing something here.
>>
>
> It's zeroed twice. alloc_etherdev_mqs() allocates zeroed memory and
> then we do a memset(priv, 0, sizeof(struct mlx4_en_priv));
>
> regards,
> dan carpenter
>
We do zero (twice) on init, that's right. But I think Yuval's comment is
valid in case of the driver went into configuration change, or down/up,
that reallocates the rings. I'm double checking this.
On 10/05/2018 5:36 PM, Tariq Toukan wrote:
>
>
> On 10/05/2018 5:18 PM, Dan Carpenter wrote:
>> On Thu, May 10, 2018 at 04:38:08PM +0300, Yuval Shaia wrote:
>>> On Thu, May 10, 2018 at 09:02:26AM +0200, Christophe JAILLET wrote:
>>>> If an error occurs, 'mlx4_en_destroy_netdev()' is called.
>>>> It then calls 'mlx4_en_free_resources()' which does the needed
>>>> resources
>>>> cleanup.
>>>>
>>>> So, doing some explicit kfree in the error handling path would lead to
>>>> some double kfree.
>>>
>>> Patch make sense but what's bothering me is that mlx4_en_free_resources
>>> loops on the entire array, assuming !priv->tx_ring[t] means entry is
>>> allocated but the existing code does not assume that, see [1]. So i
>>> looked
>>> to see where tx_ring array is zeroed and didn't find it.
>>>
>>> Am i missing something here.
>>>
>>
>> It's zeroed twice. alloc_etherdev_mqs() allocates zeroed memory and
>> then we do a memset(priv, 0, sizeof(struct mlx4_en_priv));
>>
>> regards,
>> dan carpenter
>>
>
> We do zero (twice) on init, that's right. But I think Yuval's comment is
> valid in case of the driver went into configuration change, or down/up,
> that reallocates the rings. I'm double checking this.
Well, the flows in which we need to nullify the tx_rings pointer (if
any, I still need to investigate this) is not related to this init function.
Here we're safe.
Anyway, a V2 is already submitted, please use it for your next comments.
I think patch is OK.