Re: [PATCH v2] ksmbd: prevent memory leak on error return

[Namjae Jeon]

2023-11-20 10:33 GMT+09:00, Zongmin Zhou <zhouzongmin@xxxxxxxxxx>:
>
> On 2023/11/19 22:17, Namjae Jeon wrote:
>> 2023-11-19 18:14 GMT+09:00, Pierre Mariani <pierre.mariani@xxxxxxxxx>:
>>> On 11/8/2023 5:17 PM, Zongmin Zhou wrote:
>>>> When allocated memory for 'new' failed,just return
>>>> will cause memory leak of 'ar'.
>>>>
>>>> v2: rollback iov_alloc_cnt when allocate memory failed.
>>>>
>>>> Fixes: 1819a9042999 ("ksmbd: reorganize ksmbd_iov_pin_rsp()")
>>>>
>>>> Reported-by: kernel test robot <lkp@xxxxxxxxx>
>>>> Reported-by: Dan Carpenter <error27@xxxxxxxxx>
>>>> Closes: https://lore.kernel.org/r/202311031837.H3yo7JVl-lkp@xxxxxxxxx/
>>>> Signed-off-by: Zongmin Zhou<zhouzongmin@xxxxxxxxxx>
>>>> ---
>>>>   fs/smb/server/ksmbd_work.c | 5 ++++-
>>>>   1 file changed, 4 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/fs/smb/server/ksmbd_work.c b/fs/smb/server/ksmbd_work.c
>>>> index a2ed441e837a..44bce4c56daf 100644
>>>> --- a/fs/smb/server/ksmbd_work.c
>>>> +++ b/fs/smb/server/ksmbd_work.c
>>>> @@ -123,8 +123,11 @@ static int __ksmbd_iov_pin_rsp(struct ksmbd_work
>>>> *work, void *ib, int len,
>>>>   		new = krealloc(work->iov,
>>>>   			       sizeof(struct kvec) * work->iov_alloc_cnt,
>>>>   			       GFP_KERNEL | __GFP_ZERO);
>>>> -		if (!new)
>>>> +		if (!new) {
>>>> +			kfree(ar);
>>>> +			work->iov_alloc_cnt -= 4;
>>>>   			return -ENOMEM;
>>>> +		}
>>>>   		work->iov = new;
>>>>   	}
>>>>
>>> A few lines above, ar is allocated inside the 'if (aux_size)' block.
>>> If aux_size is falsy, isn't it possible that ar will be NULL hence
>>> we should have 'if (ar) kfree(ar);'?
>> We need to initialize ar to NULL on that case. And Passing a NULL
>> pointer to kfree is safe, So NULL check before kfree() is not needed.
> Yes, ar should be initialized to NULL to avoid the case of  aux_size
> will be false.
> Since kfree(NULL) is safe.
> Should I  send another patch for this?
I would appreciate it if you could do that.

>
> Best regards!
>> Thanks.
>
>