[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 06 Jun 2014 10:56:04 -0700
From: Eric Dumazet <eric.dumazet@...il.com>
To: Alexei Starovoitov <alexei.starovoitov@...il.com>
Cc: Alexey Preobrazhensky <preobr@...gle.com>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
Kostya Serebryany <kcc@...gle.com>,
Dmitry Vyukov <dvyukov@...gle.com>,
Lars Bull <larsbull@...gle.com>,
Eric Dumazet <edumazet@...gle.com>,
Bruce Curtis <brutus@...gle.com>,
Maciej Żenczykowski <maze@...gle.com>,
dormando <dormando@...ia.net>
Subject: Re: Potential race in ip4_datagram_release_cb
On Fri, 2014-06-06 at 10:44 -0700, Alexei Starovoitov wrote:
> On Fri, Jun 6, 2014 at 9:16 AM, Eric Dumazet <eric.dumazet@...il.com> wrote:
> > We should either :
> >
> > 1) use xchg() and no lock at all to change sk_dst_cache, as we did for
> > sk_rx_dst ( cf udp_sk_rx_dst_set() )
> >
> > 2) No longer use sk_dst_lock, and always use the socket lock
> > (sk->sk_lock.slock) instead.
>
> Probably needs some combination of both.
I do not think so.
If we use xchg(), then we do not need anything else.
If we use a spinlock, then xchg() seems useless.
Any combination is buggy.
In the past, UDP transmit was holding socket lock,
but we added a lockless path, and I suppose more people
use a UDP socket from different threads.
Then, Steffen added the ip4_datagram_release_cb() thing,
increasing the chance of mixing both 'protections'.
Setting sk_dst_cache should hardly be a hot path.
--
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