[dpdk-dev] [PATCH 8/8] ipc: fix net/mlx5 memleak

Burakov, Anatoly anatoly.burakov at intel.com
Tue Apr 23 10:09:10 CEST 2019


On 22-Apr-19 6:51 PM, Yongseok Koh wrote:
> 
>> On Apr 17, 2019, at 7:44 AM, Herakliusz Lipiec <herakliusz.lipiec at intel.com> wrote:
>>
>> When sending multiple requests, rte_mp_request_sync
>> can succeed sending a few of those requests, but then
>> fail on a later one and in the end return with rc=-1.
>> The upper layers - e.g. device hotplug - currently
>> handles this case as if no messages were sent and no
>> memory for response buffers was allocated, which is
>> not true. Fixed by always freeing reply message buffers.
>>
>> Fixes: 9a8ab29b84d3 ("net/mlx5: replace IPC socket with EAL API")
>> Fixes: c18cf501a7af ("net/mlx5: enable secondary process to register DMA memory")
>> Cc: yskoh at mellanox.com
>> Signed-off-by: Herakliusz Lipiec <herakliusz.lipiec at intel.com>
>> ---
>> drivers/net/mlx5/mlx5_mp.c | 4 +++-
>> 1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/mlx5/mlx5_mp.c b/drivers/net/mlx5/mlx5_mp.c
>> index cea74adb6..c9915b1d5 100644
>> --- a/drivers/net/mlx5/mlx5_mp.c
>> +++ b/drivers/net/mlx5/mlx5_mp.c
>> @@ -258,6 +258,7 @@ mlx5_mp_req_mr_create(struct rte_eth_dev *dev, uintptr_t addr)
>> 	if (ret) {
>> 		DRV_LOG(ERR, "port %u request to primary process failed",
>> 			dev->data->port_id);
>> +		free(mp_rep.msgs);
>> 		return -rte_errno;
>> 	}
>> 	assert(mp_rep.nb_received == 1);
>> @@ -295,7 +296,8 @@ mlx5_mp_req_verbs_cmd_fd(struct rte_eth_dev *dev)
>> 	if (ret) {
>> 		DRV_LOG(ERR, "port %u request to primary process failed",
>> 			dev->data->port_id);
>> -		return -rte_errno;
>> +		ret = -rte_errno;
>> +		goto exit;
> 
> These two requests will be made by a secondary process targeting to the primary.
> Then, there's only one request in this case and we don't need to take care of that.
> Right?
> 
> Same comment for mlx4.

Hi Yongseok,

mp_rep.msgs is potentially allocated regardless of whether you're in 
primary or secondary, and whether the call to mp_request_sync succeeded 
or failed. Hence, need to free in all cases.

See this patch: http://patches.dpdk.org/patch/52868/
and this bug: https://bugs.dpdk.org/show_bug.cgi?id=228

> 
> Thanks,
> Yongseok
> 
>> 	}
>> 	assert(mp_rep.nb_received == 1);
>> 	mp_res = &mp_rep.msgs[0];
>> -- 
>> 2.17.2
>>
> 
> 


-- 
Thanks,
Anatoly


More information about the dev mailing list