Re: [PATCH 07/14] net: sched: use reference counting action init

[Vlad Buslov <vladbu@mellanox.com>]

On Tue 15 May 2018 at 11:39, Jiri Pirko <jiri@resnulli.us> wrote:
> Tue, May 15, 2018 at 01:32:51PM CEST, vladbu@mellanox.com wrote:
>>
>>On Tue 15 May 2018 at 11:24, Jiri Pirko <jiri@resnulli.us> wrote:
>>> Mon, May 14, 2018 at 04:27:08PM CEST, vladbu@mellanox.com wrote:
>>>>Change action API to assume that action init function always takes
>>>>reference to action, even when overwriting existing action. This is
>>>>necessary because action API continues to use action pointer after init
>>>>function is done. At this point action becomes accessible for concurrent
>>>>modifications so user must always hold reference to it.
>>>>
>>>>Implement helper put list function to atomically release list of actions
>>>>after action API init code is done using them.
>>>>
>>>>Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
>>>>---
>>>> net/sched/act_api.c | 38 +++++++++++++++++---------------------
>>>> 1 file changed, 17 insertions(+), 21 deletions(-)
>>>>
>>>
>>> [...]
>>>
>>>
>>>>@@ -1196,8 +1190,7 @@ tca_action_gd(struct net *net, struct nlattr *nla, struct nlmsghdr *n,
>>>> 		return ret;
>>>> 	}
>>>> err:
>>>>-	if (event != RTM_GETACTION)
>>>
>>> Howcome you do this for RTM_GETACTION now too? Where is the related
>>> "get"?
>>
>>In patch 5. There is always a possibility of concurrent delete without
>>rtnl lock so all usages of action pointers were converted to hold
>>reference to action.
>
> So that means that if you run kernel in between, with patch 5 but
> without patch 7 and you do RTM_GETACTION, you leak a reference, right?

Right.

>
>
>>
>>>
>>>
>>>>-		tcf_action_destroy(&actions, 0);
>>>>+	tcf_action_put_lst(&actions);
>>>> 	return ret;
>>>> }
>>>> 
>>>
>>> [...]
>>