Date:	Mon, 23 Dec 2013 17:30:36 -0500
From:	Jamal Hadi Salim <jhs@...atatu.com>
To:	Cong Wang <xiyou.wangcong@...il.com>
CC:	Linux Kernel Network Developers <netdev@...r.kernel.org>,
	"David S. Miller" <davem@...emloft.net>
Subject: Re: [PATCH net-next v4 3/8] net_sched: mirred: remove action when
 the target device is gone

On 12/23/13 17:14, Cong Wang wrote:
> On Mon, Dec 23, 2013 at 12:41 PM, Jamal Hadi Salim <jhs@...atatu.com> wrote:
>>
>> Now if you have such a graph:
>
> Making such an abstraction only helps misunderstanding, really...
>

So I used a graph that maps to a netdevice, qdisc, filter and actions
like you asked to - there is still a misunderstanding?

>> What you cannot do is, on your own as kernel, decide you want to delete
>> one action that is _bound_ to a filter because one of its attributes is
>> gone berserk. It doesnt matter whether such an action is mirred or foo
>> or whether D and E dont exist. You can put a big hole where the town
>> used to be and leave roads leading to the town.
>
> Again, since (non-shared) actions attached to a filter are chained by
> a doubly linked list, why not?
>
>> It will still be referenced by things preeceding it (primarily the
>> classifier which keeps an action chain intact), which is bad when the
>> next packet arrives.
>
> You know, there is a head for such linked list for the filter to refer,
> so what's the problem with deleting a node from a linked list if
> we lock it properly? In non-shared case, no filter can refer to any
> action without the head of the chain, right? So what is the problem
> of "referenced by things preeceding it" when they are linked and locked
> properly?
>

Huh? This has nothing to do with whether you are capable to remove a
node from a list. It is about you making a decision that the node should
be removed. You dont have sufficient information to make such a decision.
Remember earlier i said you are making a macroscopic decision by looking
at microscope? I just saw a white blood cell, let me chop that toe.

>> You let the user/control program which is monitoring things do that
>> and "reroute" around the problem. If you insist on putting a little part
>> of the medula oblangata in the stomach, then:
>> the only correct way to do it is delete the filter.
>
> Apparent not, different actions can attached to the same filter
> and only mirred action has a target dev.
>

Parse error.

>> And you start doing that you are making some serious policy decisions
>> in the kernel and adding lots of complexity.
>>
>
> Why removing a filter when the netdev is removed is not policy?
> Actually it is, ONLY that it is not be able to shared with current
> implementation.
>

Of course it is. Since you understand that why do you think removing
an action which is part of that policy is a good idea?

> Since you love mechanism _so much_, why not make filters shared
> by other qdisc's and stop removing them when netdev is gone in kernel?
>
> Be realistic, Jamal, user-space is hard, you can't simply let user-space
> decide everything, especially when you don't provide a simple way to do it.
> Netlink is already hard to use, even with libnl, since it enforces a cache
> layer. Sit down and spend some time to write some libnl code, compare it
> with this patch.
>

You gotta be joking. I live through this every ither day.
Frankly, I think i will have no progress if i tried to convince you the
world is round. I am dropping from this discussion.

cheers,
jamal
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists