[Devel] [PATCH rh7] net: Use {get, put}_net() in inet_twsk_{alloc, free}()
Kirill Tkhai
ktkhai at odin.com
Fri May 29 03:20:56 PDT 2015
В Чт, 28/05/2015 в 18:56 +0300, Andrew Vagin пишет:
> On Wed, May 27, 2015 at 02:32:22PM +0300, Kirill Tkhai wrote:
> > hold_net() doesn't increment net refcounter if NETNS_REFCNT_DEBUG
> > is not defined. In this case inet_twdr_do_twkill_work() may happen
> > after network is destoyed and lead to memory corruption like this
> > happened in https://jira.sw.ru/browse/PSBM-33755.
> >
> > The patch replaces {hold,release}_net with {get,put}_net, which
> > makes refcounters be incremented always.
>
> In this case a netns will not be destroyed while it has time wait
> sockets. It isn't expected behaviour.
>
> I don't understand why do we need tcp_v4_kill_ve_sockets().
I've talked to Pavel. He says we do not need this function neither
in 2.6.32 nor in 3.10.
Appropriate removing patches -- I'll send them.
> >
> > Such object changes, which extends its lifetime, should be checked
> > for safety of refcouter waiters, who wait for 0 refcounter. I mean
> > the sequence of net device shutdown for example, see
> > netdev_wait_allrefs() for details.
> >
> > Luckily, we don't have the same issues with struct net, because nobody
> > waits for it.
> >
> > PCS6 also needs the same patch.
> >
> > Signed-off-by: Kirill Tkhai <ktkhai at odin.com>
> > ---
> > net/ipv4/inet_timewait_sock.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/net/ipv4/inet_timewait_sock.c b/net/ipv4/inet_timewait_sock.c
> > index e2bda07..526e3b5 100644
> > --- a/net/ipv4/inet_timewait_sock.c
> > +++ b/net/ipv4/inet_timewait_sock.c
> > @@ -103,7 +103,7 @@ void inet_twsk_free(struct inet_timewait_sock *tw)
> > #ifdef CONFIG_BEANCOUNTERS
> > put_beancounter(tw->tw_ub);
> > #endif
> > - release_net(twsk_net(tw));
> > + put_net(twsk_net(tw));
> > kmem_cache_free(tw->tw_prot->twsk_prot->twsk_slab, tw);
> > module_put(owner);
> > }
> > @@ -202,7 +202,7 @@ struct inet_timewait_sock *inet_twsk_alloc(const struct sock *sk, const int stat
> > tw->tw_ipv6only = 0;
> > tw->tw_transparent = inet->transparent;
> > tw->tw_prot = sk->sk_prot_creator;
> > - twsk_net_set(tw, hold_net(sock_net(sk)));
> > + twsk_net_set(tw, get_net(sock_net(sk)));
> > /*
> > * Because we use RCU lookups, we should not set tw_refcnt
> > * to a non null value before everything is setup for this
> >
More information about the Devel
mailing list