[PATCH] Bluetooth: hci_sync: fix memory leak in hci_update_adv_data()

[PATCH] Bluetooth: hci_sync: fix memory leak in hci_update_adv_data()

[Zhengchao Shao]

When hci_cmd_sync_queue() failed in hci_update_adv_data(), inst_ptr is
not freed, which will cause memory leak. Add release process to error
path.

Fixes: 651cd3d65b0f ("Bluetooth: convert hci_update_adv_data to hci_sync")
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
---
 net/bluetooth/hci_sync.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/net/bluetooth/hci_sync.c b/net/bluetooth/hci_sync.c
index 9e2d7e4b850c..1485501bd72f 100644
--- a/net/bluetooth/hci_sync.c
+++ b/net/bluetooth/hci_sync.c
@@ -6197,10 +6197,15 @@ static int _update_adv_data_sync(struct hci_dev *hdev, void *data)
 int hci_update_adv_data(struct hci_dev *hdev, u8 instance)
 {
 	u8 *inst_ptr = kmalloc(1, GFP_KERNEL);
+	int ret;
 
 	if (!inst_ptr)
 		return -ENOMEM;
 
 	*inst_ptr = instance;
-	return hci_cmd_sync_queue(hdev, _update_adv_data_sync, inst_ptr, NULL);
+	ret = hci_cmd_sync_queue(hdev, _update_adv_data_sync, inst_ptr, NULL);
+	if (ret)
+		kfree(inst_ptr);
+
+	return ret;
 }
-- 
2.34.1

Re: [PATCH] Bluetooth: hci_sync: fix memory leak in hci_update_adv_data()

[Luiz Augusto von Dentz]

Hi Zhengchao,

On Sun, Jan 8, 2023 at 5:17 PM Zhengchao Shao <shaozhengchao@huawei.com> wrote:
>
> When hci_cmd_sync_queue() failed in hci_update_adv_data(), inst_ptr is
> not freed, which will cause memory leak. Add release process to error
> path.
>
> Fixes: 651cd3d65b0f ("Bluetooth: convert hci_update_adv_data to hci_sync")
> Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
> ---
>  net/bluetooth/hci_sync.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/net/bluetooth/hci_sync.c b/net/bluetooth/hci_sync.c
> index 9e2d7e4b850c..1485501bd72f 100644
> --- a/net/bluetooth/hci_sync.c
> +++ b/net/bluetooth/hci_sync.c
> @@ -6197,10 +6197,15 @@ static int _update_adv_data_sync(struct hci_dev *hdev, void *data)
>  int hci_update_adv_data(struct hci_dev *hdev, u8 instance)
>  {
>         u8 *inst_ptr = kmalloc(1, GFP_KERNEL);
> +       int ret;
>
>         if (!inst_ptr)
>                 return -ENOMEM;
>
>         *inst_ptr = instance;
> -       return hci_cmd_sync_queue(hdev, _update_adv_data_sync, inst_ptr, NULL);
> +       ret = hci_cmd_sync_queue(hdev, _update_adv_data_sync, inst_ptr, NULL);
> +       if (ret)
> +               kfree(inst_ptr);
> +
> +       return ret;
>  }

While this is correct I do wonder why we haven't used ERR_PTR/PTR_ERR
like we did with the likes of abort_conn_sync, that way we don't have
to allocate anything which makes things a lot simpler.

> --
> 2.34.1
>


-- 
Luiz Augusto von Dentz

Re: [PATCH] Bluetooth: hci_sync: fix memory leak in hci_update_adv_data()

[shaozhengchao]

On 2023/1/10 5:11, Luiz Augusto von Dentz wrote:
> Hi Zhengchao,
> 
> On Sun, Jan 8, 2023 at 5:17 PM Zhengchao Shao <shaozhengchao@huawei.com> wrote:
>>
>> When hci_cmd_sync_queue() failed in hci_update_adv_data(), inst_ptr is
>> not freed, which will cause memory leak. Add release process to error
>> path.
>>
>> Fixes: 651cd3d65b0f ("Bluetooth: convert hci_update_adv_data to hci_sync")
>> Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
>> ---
>>   net/bluetooth/hci_sync.c | 7 ++++++-
>>   1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/net/bluetooth/hci_sync.c b/net/bluetooth/hci_sync.c
>> index 9e2d7e4b850c..1485501bd72f 100644
>> --- a/net/bluetooth/hci_sync.c
>> +++ b/net/bluetooth/hci_sync.c
>> @@ -6197,10 +6197,15 @@ static int _update_adv_data_sync(struct hci_dev *hdev, void *data)
>>   int hci_update_adv_data(struct hci_dev *hdev, u8 instance)
>>   {
>>          u8 *inst_ptr = kmalloc(1, GFP_KERNEL);
>> +       int ret;
>>
>>          if (!inst_ptr)
>>                  return -ENOMEM;
>>
>>          *inst_ptr = instance;
>> -       return hci_cmd_sync_queue(hdev, _update_adv_data_sync, inst_ptr, NULL);
>> +       ret = hci_cmd_sync_queue(hdev, _update_adv_data_sync, inst_ptr, NULL);
>> +       if (ret)
>> +               kfree(inst_ptr);
>> +
>> +       return ret;
>>   }
> 
> While this is correct I do wonder why we haven't used ERR_PTR/PTR_ERR
> like we did with the likes of abort_conn_sync, that way we don't have
> to allocate anything which makes things a lot simpler.
> 

Hi Luiz:
	Thank you for your advice. I think it is better to use
ERR_PTR/PTR_ERR to replace allocation. I will send V2.

Zhengchao Shao
>> --
>> 2.34.1
>>
> 
>