[Cluster-devel] [PATCH][next] fs: dlm: Fix memory leak of object mh

[Colin Ian King]

On 26/05/2021 15:19, Alexander Ahring Oder Aring wrote:
> Hi,
> 
> On Wed, May 26, 2021 at 9:40 AM Colin King <colin.king at canonical.com> wrote:
>>
>> From: Colin Ian King <colin.king at canonical.com>
>>
>> There is an error return path that is not kfree'ing mh after
>> it has been successfully allocates.  Fix this by free'ing it.
>>
>> Addresses-Coverity: ("Resource leak")
>> Fixes: a070a91cf140 ("fs: dlm: add more midcomms hooks")
>> Signed-off-by: Colin Ian King <colin.king at canonical.com>
>> ---
>>  fs/dlm/rcom.c | 1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff --git a/fs/dlm/rcom.c b/fs/dlm/rcom.c
>> index 085f21966c72..19298edc1573 100644
>> --- a/fs/dlm/rcom.c
>> +++ b/fs/dlm/rcom.c
>> @@ -393,6 +393,7 @@ static void receive_rcom_lookup(struct dlm_ls *ls, struct dlm_rcom *rc_in)
>>         if (rc_in->rc_id == 0xFFFFFFFF) {
>>                 log_error(ls, "receive_rcom_lookup dump from %d", nodeid);
>>                 dlm_dump_rsb_name(ls, rc_in->rc_buf, len);
>> +               kfree(mh);
>>                 return;
> 
> This seems to be a bigger issue, we cannot revert the buffer
> allocation with a kfree, we cannot revert it at all. We should avoid
> any error handling between create_rcom() and send_rcom(). In general
> between get_buffer/commit_buffer.
> 
> I don't see a problem with moving the error handling before
> create_rcom(). That should fix the issue.

Good point, I'll send a V2 in a while

> 
> - Alex
>