[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 11 Mar 2015 18:31:42 -0700
From: Eric Dumazet <eric.dumazet@...il.com>
To: David Miller <davem@...emloft.net>
Cc: netdev@...r.kernel.org, salo@...gle.com
Subject: Re: [PATCH net-next] net: add real socket cookies
On Wed, 2015-03-11 at 21:21 -0400, David Miller wrote:
> From: Eric Dumazet <eric.dumazet@...il.com>
> Date: Wed, 11 Mar 2015 15:30:35 -0700
>
> > Note that sk_clone_lock() really does the clear for us :
> >
> > net/core/sock.c:1541: atomic64_set(&newsk->sk_cookie, 0);
> >
> > atomic_cmpxchg() is needed here in case a concurrent dumper already
> > caught the new socket.
>
> How is newsk visible at this moment?
>
> It is freshly allocated, and not present in any tables. Nothing even
> indirectly links to it at this point.
>
> Visibility only happens after we return from this function, in
> routines such as tcp_v4_syn_recv_sock() which do the hash table
> insertion.
Yes, but remember we use RCU with SLAB_DESTROY_BY_RCU thing,
and inet_diag_dump_one_icsk() uses RCU as well.
I wanted to be extra careful here. Note that with v2, this
atomic_cmpxchg() wont happen unless one dump operation previously
revealed the SYN_RECV to the world.
I have no strong opinion and definitely it is currently safe to use
atomic64_set() here.
--
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