[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20080822.032733.155458803.davem@davemloft.net>
Date: Fri, 22 Aug 2008 03:27:33 -0700 (PDT)
From: David Miller <davem@...emloft.net>
To: herbert@...dor.apana.org.au
Cc: jarkao2@...il.com, netdev@...r.kernel.org, denys@...p.net.lb
Subject: Re: [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock().
From: Herbert Xu <herbert@...dor.apana.org.au>
Date: Fri, 22 Aug 2008 20:07:08 +1000
> Yes the ingress hierachy has a single root, i.e., it's a tree. But
> that has nothing to do with what I was talking about. I'm talking
> about the list at dev->rx_queue.qdisc_sleeping->list which is
> certainly not guaranteed to be empty.
It is guarenteed to be empty.
Only root qdiscs go to rx_queue.qdisc_sleeping, and such qdiscs
will always have an empty list.
> If you look at qdisc_create you'll find that every time we create
> a non-root ingress qdisc we add it to that list (we have to,
> otherwise qdisc_lookup doesn't work at all for ingress qdiscs).
We don't allow non-root ingress qdiscs. All ingress qdiscs
always take the device graft path, and always have a parent
of TC_H_INGRESS, and always operate on rx_queue.
> So when somebody on the TX side does a qdisc_lookup they may be
> walking the RX list without any protection. Similarly, if somebody
> on the ingress side does qdisc_lookup they may walk the TX lists
> without protection.
Ingress data paths do not do qdisc_lookup(). There is only root
allowed for ingress and thus rx_queue's non-default qdiscs.
Add some assertions and run some test tc commands if you don't believe
me :)
--
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