Re: [PATCH] net/mlx5: fix potential null dereference when enable shared FDB

Re: [PATCH] net/mlx5: fix potential null dereference when enable shared FDB

[Charles Han(韩春超)]

[-- Attachment #1: Type: text/plain, Size: 3452 bytes --]

-ENXIO indicates "No such device or address". I've found that in mlx5/core, if mlx5_get_flow_namespace() returns null, it basically returns -EOPNOTSUPP.

-----邮件原件-----
发件人: Przemek Kitszel <przemyslaw.kitszel@intel.com> 
发送时间: 2025年4月2日 19:02
收件人: Charles Han(韩春超) <hanchunchao@inspur.com>
抄送: netdev@vger.kernel.org; linux-rdma@vger.kernel.org; linux-kernel@vger.kernel.org; saeedm@nvidia.com; leon@kernel.org; tariqt@nvidia.com; andrew+netdev@lunn.ch; davem@davemloft.net; edumazet@google.com; kuba@kernel.org; pabeni@redhat.com; markzhang@nvidia.com; mbloch@nvidia.com
主题: Re: [PATCH] net/mlx5: fix potential null dereference when enable shared FDB

On 4/2/25 11:43, Charles Han wrote:
> mlx5_get_flow_namespace() may return a NULL pointer, dereferencing it 
> without NULL check may lead to NULL dereference.
> Add a NULL check for ns.
> 
> Fixes: db202995f503 ("net/mlx5: E-Switch, add logic to enable shared 
> FDB")
> Signed-off-by: Charles Han <hanchunchao@inspur.com>
> ---
>   .../net/ethernet/mellanox/mlx5/core/eswitch_offloads.c | 10 ++++++++++
>   drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c       |  5 +++++
>   2 files changed, 15 insertions(+)
> 
> diff --git 
> a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c 
> b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
> index a6a8eea5980c..dc58e4c2d786 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
> @@ -2667,6 +2667,11 @@ static int esw_set_slave_root_fdb(struct mlx5_core_dev *master,
>   	if (master) {
>   		ns = mlx5_get_flow_namespace(master,
>   					     MLX5_FLOW_NAMESPACE_FDB);
> +		if (!ns) {
> +			mlx5_core_warn(master, "Failed to get flow namespace\n");
> +			return -EOPNOTSUPP;

I would return -ENXIO in such cases, you were searching and not found that.

IOW it is obvious that dereferencing a null ptr is not supported.

If you agree, please apply the same comment for your other patch:
https://lore.kernel.org/netdev/20250402093221.3253-1-hanchunchao@inspur.com/T/#u

> +		}
> +
>   		root = find_root(&ns->node);
>   		mutex_lock(&root->chain_lock);
>   		MLX5_SET(set_flow_table_root_in, in, @@ -2679,6 +2684,11 @@ static 
> int esw_set_slave_root_fdb(struct mlx5_core_dev *master,
>   	} else {
>   		ns = mlx5_get_flow_namespace(slave,
>   					     MLX5_FLOW_NAMESPACE_FDB);
> +		if (!ns) {
> +			mlx5_core_warn(slave, "Failed to get flow namespace\n");
> +			return -EOPNOTSUPP;
> +		}
> +
>   		root = find_root(&ns->node);
>   		mutex_lock(&root->chain_lock);
>   		MLX5_SET(set_flow_table_root_in, in, table_id, diff --git 
> a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c 
> b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
> index a47c29571f64..18e59f6a0f2d 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
> @@ -186,6 +186,11 @@ static int mlx5_cmd_set_slave_root_fdb(struct mlx5_core_dev *master,
>   	} else {
>   		ns = mlx5_get_flow_namespace(slave,
>   					     MLX5_FLOW_NAMESPACE_FDB);
> +		if (!ns) {
> +			mlx5_core_warn(slave, "Failed to get flow namespace\n");
> +			return -EOPNOTSUPP;
> +		}
> +
>   		root = find_root(&ns->node);
>   		MLX5_SET(set_flow_table_root_in, in, table_id,
>   			 root->root_ft->id);


[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 3830 bytes --]

Re: [PATCH] net/mlx5: fix potential null dereference when enable shared FDB

[Tariq Toukan]

On 03/04/2025 12:52, Charles Han(韩春超) wrote:
> -ENXIO indicates "No such device or address". I've found that in mlx5/core, if mlx5_get_flow_namespace() returns null, it basically returns -EOPNOTSUPP.
> 

Please do not top-post.

+1.
If namespace is not found it's due to lack of support.


> -----邮件原件-----
> 发件人: Przemek Kitszel <przemyslaw.kitszel@intel.com>
> 发送时间: 2025年4月2日 19:02
> 收件人: Charles Han(韩春超) <hanchunchao@inspur.com>
> 抄送: netdev@vger.kernel.org; linux-rdma@vger.kernel.org; linux-kernel@vger.kernel.org; saeedm@nvidia.com; leon@kernel.org; tariqt@nvidia.com; andrew+netdev@lunn.ch; davem@davemloft.net; edumazet@google.com; kuba@kernel.org; pabeni@redhat.com; markzhang@nvidia.com; mbloch@nvidia.com
> 主题: Re: [PATCH] net/mlx5: fix potential null dereference when enable shared FDB
> 
> On 4/2/25 11:43, Charles Han wrote:
>> mlx5_get_flow_namespace() may return a NULL pointer, dereferencing it
>> without NULL check may lead to NULL dereference.
>> Add a NULL check for ns.
>>
>> Fixes: db202995f503 ("net/mlx5: E-Switch, add logic to enable shared
>> FDB")
>> Signed-off-by: Charles Han <hanchunchao@inspur.com>

Acked-by: Tariq Toukan <tariqt@nvidia.com>

>> ---
>>    .../net/ethernet/mellanox/mlx5/core/eswitch_offloads.c | 10 ++++++++++
>>    drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c       |  5 +++++
>>    2 files changed, 15 insertions(+)
>>
>> diff --git
>> a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
>> b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
>> index a6a8eea5980c..dc58e4c2d786 100644
>> --- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
>> +++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
>> @@ -2667,6 +2667,11 @@ static int esw_set_slave_root_fdb(struct mlx5_core_dev *master,
>>    	if (master) {
>>    		ns = mlx5_get_flow_namespace(master,
>>    					     MLX5_FLOW_NAMESPACE_FDB);
>> +		if (!ns) {
>> +			mlx5_core_warn(master, "Failed to get flow namespace\n");
>> +			return -EOPNOTSUPP;
> 
> I would return -ENXIO in such cases, you were searching and not found that.
> 
> IOW it is obvious that dereferencing a null ptr is not supported.
> 
> If you agree, please apply the same comment for your other patch:
> https://lore.kernel.org/netdev/20250402093221.3253-1-hanchunchao@inspur.com/T/#u
> 
>> +		}
>> +
>>    		root = find_root(&ns->node);
>>    		mutex_lock(&root->chain_lock);
>>    		MLX5_SET(set_flow_table_root_in, in, @@ -2679,6 +2684,11 @@ static
>> int esw_set_slave_root_fdb(struct mlx5_core_dev *master,
>>    	} else {
>>    		ns = mlx5_get_flow_namespace(slave,
>>    					     MLX5_FLOW_NAMESPACE_FDB);
>> +		if (!ns) {
>> +			mlx5_core_warn(slave, "Failed to get flow namespace\n");
>> +			return -EOPNOTSUPP;
>> +		}
>> +
>>    		root = find_root(&ns->node);
>>    		mutex_lock(&root->chain_lock);
>>    		MLX5_SET(set_flow_table_root_in, in, table_id, diff --git
>> a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
>> b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
>> index a47c29571f64..18e59f6a0f2d 100644
>> --- a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
>> +++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
>> @@ -186,6 +186,11 @@ static int mlx5_cmd_set_slave_root_fdb(struct mlx5_core_dev *master,
>>    	} else {
>>    		ns = mlx5_get_flow_namespace(slave,
>>    					     MLX5_FLOW_NAMESPACE_FDB);
>> +		if (!ns) {
>> +			mlx5_core_warn(slave, "Failed to get flow namespace\n");
>> +			return -EOPNOTSUPP;
>> +		}
>> +
>>    		root = find_root(&ns->node);
>>    		MLX5_SET(set_flow_table_root_in, in, table_id,
>>    			 root->root_ft->id);
> 

Re: [PATCH] net/mlx5: fix potential null dereference when enable shared FDB

[Tariq Toukan]

On 03/04/2025 17:03, Tariq Toukan wrote:
> 
> 
> On 03/04/2025 12:52, Charles Han(韩春超) wrote:
>> -ENXIO indicates "No such device or address". I've found that in mlx5/ 
>> core, if mlx5_get_flow_namespace() returns null, it basically returns 
>> -EOPNOTSUPP.
>>
> 
> Please do not top-post.
> 
> +1.
> If namespace is not found it's due to lack of support.
> 
> 
>> -----邮件原件-----
>> 发件人: Przemek Kitszel <przemyslaw.kitszel@intel.com>
>> 发送时间: 2025年4月2日 19:02
>> 收件人: Charles Han(韩春超) <hanchunchao@inspur.com>
>> 抄送: netdev@vger.kernel.org; linux-rdma@vger.kernel.org; linux- 
>> kernel@vger.kernel.org; saeedm@nvidia.com; leon@kernel.org; 
>> tariqt@nvidia.com; andrew+netdev@lunn.ch; davem@davemloft.net; 
>> edumazet@google.com; kuba@kernel.org; pabeni@redhat.com; 
>> markzhang@nvidia.com; mbloch@nvidia.com
>> 主题: Re: [PATCH] net/mlx5: fix potential null dereference when enable 
>> shared FDB
>>
>> On 4/2/25 11:43, Charles Han wrote:
>>> mlx5_get_flow_namespace() may return a NULL pointer, dereferencing it
>>> without NULL check may lead to NULL dereference.
>>> Add a NULL check for ns.
>>>
>>> Fixes: db202995f503 ("net/mlx5: E-Switch, add logic to enable shared
>>> FDB")
>>> Signed-off-by: Charles Han <hanchunchao@inspur.com>
> 
> Acked-by: Tariq Toukan <tariqt@nvidia.com>
> 

Re-visiting this...
See comment below.

>>> ---
>>>    .../net/ethernet/mellanox/mlx5/core/eswitch_offloads.c | 10 ++++++ 
>>> ++++
>>>    drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c       |  5 +++++
>>>    2 files changed, 15 insertions(+)
>>>
>>> diff --git
>>> a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
>>> b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
>>> index a6a8eea5980c..dc58e4c2d786 100644
>>> --- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
>>> +++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
>>> @@ -2667,6 +2667,11 @@ static int esw_set_slave_root_fdb(struct 
>>> mlx5_core_dev *master,
>>>        if (master) {
>>>            ns = mlx5_get_flow_namespace(master,
>>>                             MLX5_FLOW_NAMESPACE_FDB);
>>> +        if (!ns) {
>>> +            mlx5_core_warn(master, "Failed to get flow namespace\n");

Use esw_warn(), for all new instances.

[PATCH V2] net/mlx5: fix potential null dereference when enable shared FDB

[Charles Han]

mlx5_get_flow_namespace() may return a NULL pointer, dereferencing it
without NULL check may lead to NULL dereference.
Add a NULL check for ns.

Fixes: db202995f503 ("net/mlx5: E-Switch, add logic to enable shared FDB")
Signed-off-by: Charles Han <hanchunchao@inspur.com>
---
 .../net/ethernet/mellanox/mlx5/core/eswitch_offloads.c | 10 ++++++++++
 drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c       |  5 +++++
 2 files changed, 15 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
index a6a8eea5980c..5405134e74b6 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
@@ -2667,6 +2667,11 @@ static int esw_set_slave_root_fdb(struct mlx5_core_dev *master,
 	if (master) {
 		ns = mlx5_get_flow_namespace(master,
 					     MLX5_FLOW_NAMESPACE_FDB);
+		if (!ns) {
+			esw_warn(master, "Failed to get flow namespace\n");
+			return -EOPNOTSUPP;
+		}
+
 		root = find_root(&ns->node);
 		mutex_lock(&root->chain_lock);
 		MLX5_SET(set_flow_table_root_in, in,
@@ -2679,6 +2684,11 @@ static int esw_set_slave_root_fdb(struct mlx5_core_dev *master,
 	} else {
 		ns = mlx5_get_flow_namespace(slave,
 					     MLX5_FLOW_NAMESPACE_FDB);
+		if (!ns) {
+			esw_warn(slave, "Failed to get flow namespace\n");
+			return -EOPNOTSUPP;
+		}
+
 		root = find_root(&ns->node);
 		mutex_lock(&root->chain_lock);
 		MLX5_SET(set_flow_table_root_in, in, table_id,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
index a47c29571f64..18e59f6a0f2d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
@@ -186,6 +186,11 @@ static int mlx5_cmd_set_slave_root_fdb(struct mlx5_core_dev *master,
 	} else {
 		ns = mlx5_get_flow_namespace(slave,
 					     MLX5_FLOW_NAMESPACE_FDB);
+		if (!ns) {
+			mlx5_core_warn(slave, "Failed to get flow namespace\n");
+			return -EOPNOTSUPP;
+		}
+
 		root = find_root(&ns->node);
 		MLX5_SET(set_flow_table_root_in, in, table_id,
 			 root->root_ft->id);
-- 
2.43.0

Re: [PATCH V2] net/mlx5: fix potential null dereference when enable shared FDB

[Tariq Toukan]

On 07/04/2025 9:47, Charles Han wrote:
> mlx5_get_flow_namespace() may return a NULL pointer, dereferencing it
> without NULL check may lead to NULL dereference.
> Add a NULL check for ns.
> 
> Fixes: db202995f503 ("net/mlx5: E-Switch, add logic to enable shared FDB")
> Signed-off-by: Charles Han <hanchunchao@inspur.com>
> ---
>   .../net/ethernet/mellanox/mlx5/core/eswitch_offloads.c | 10 ++++++++++
>   drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c       |  5 +++++
>   2 files changed, 15 insertions(+)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
> index a6a8eea5980c..5405134e74b6 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
> @@ -2667,6 +2667,11 @@ static int esw_set_slave_root_fdb(struct mlx5_core_dev *master,
>   	if (master) {
>   		ns = mlx5_get_flow_namespace(master,
>   					     MLX5_FLOW_NAMESPACE_FDB);
> +		if (!ns) {
> +			esw_warn(master, "Failed to get flow namespace\n");
> +			return -EOPNOTSUPP;
> +		}
> +
>   		root = find_root(&ns->node);
>   		mutex_lock(&root->chain_lock);
>   		MLX5_SET(set_flow_table_root_in, in,
> @@ -2679,6 +2684,11 @@ static int esw_set_slave_root_fdb(struct mlx5_core_dev *master,
>   	} else {
>   		ns = mlx5_get_flow_namespace(slave,
>   					     MLX5_FLOW_NAMESPACE_FDB);
> +		if (!ns) {
> +			esw_warn(slave, "Failed to get flow namespace\n");
> +			return -EOPNOTSUPP;
> +		}
> +
>   		root = find_root(&ns->node);
>   		mutex_lock(&root->chain_lock);
>   		MLX5_SET(set_flow_table_root_in, in, table_id,
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
> index a47c29571f64..18e59f6a0f2d 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c
> @@ -186,6 +186,11 @@ static int mlx5_cmd_set_slave_root_fdb(struct mlx5_core_dev *master,
>   	} else {
>   		ns = mlx5_get_flow_namespace(slave,
>   					     MLX5_FLOW_NAMESPACE_FDB);
> +		if (!ns) {
> +			mlx5_core_warn(slave, "Failed to get flow namespace\n");
> +			return -EOPNOTSUPP;
> +		}
> +
>   		root = find_root(&ns->node);
>   		MLX5_SET(set_flow_table_root_in, in, table_id,
>   			 root->root_ft->id);

Thanks for your patch.

I wonder, did you fail on any of these, or just caught them while 
reading the code?