* [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt()
@ 2016-10-20 6:35 Cong Wang
2016-10-20 11:12 ` Marcelo Ricardo Leitner
` (2 more replies)
0 siblings, 3 replies; 7+ messages in thread
From: Cong Wang @ 2016-10-20 6:35 UTC (permalink / raw)
To: netdev; +Cc: eric.dumazet, sploving1, Cong Wang, Marcelo Ricardo Leitner
Baozeng reported this deadlock case:
CPU0 CPU1
---- ----
lock([ 165.136033] sk_lock-AF_INET6);
lock([ 165.136033] rtnl_mutex);
lock([ 165.136033] sk_lock-AF_INET6);
lock([ 165.136033] rtnl_mutex);
Similar to commit 87e9f0315952
("ipv4: fix a potential deadlock in mcast getsockopt() path")
this is due to we still have a case, ipv6_sock_mc_close(),
where we acquire sk_lock before rtnl_lock. Close this deadlock
with the similar solution, that is always acquire rtnl lock first.
Fixes: baf606d9c9b1 ("ipv4,ipv6: grab rtnl before locking the socket")
Reported-by: Baozeng Ding <sploving1@gmail.com>
Tested-by: Baozeng Ding <sploving1@gmail.com>
Cc: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
---
include/net/addrconf.h | 1 +
net/ipv6/ipv6_sockglue.c | 3 ++-
net/ipv6/mcast.c | 17 ++++++++++++-----
3 files changed, 15 insertions(+), 6 deletions(-)
diff --git a/include/net/addrconf.h b/include/net/addrconf.h
index f2d0727..8f998af 100644
--- a/include/net/addrconf.h
+++ b/include/net/addrconf.h
@@ -174,6 +174,7 @@ int ipv6_sock_mc_join(struct sock *sk, int ifindex,
const struct in6_addr *addr);
int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
const struct in6_addr *addr);
+void __ipv6_sock_mc_close(struct sock *sk);
void ipv6_sock_mc_close(struct sock *sk);
bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
const struct in6_addr *src_addr);
diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c
index 5330262..636ec56 100644
--- a/net/ipv6/ipv6_sockglue.c
+++ b/net/ipv6/ipv6_sockglue.c
@@ -120,6 +120,7 @@ struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
static bool setsockopt_needs_rtnl(int optname)
{
switch (optname) {
+ case IPV6_ADDRFORM:
case IPV6_ADD_MEMBERSHIP:
case IPV6_DROP_MEMBERSHIP:
case IPV6_JOIN_ANYCAST:
@@ -198,7 +199,7 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
}
fl6_free_socklist(sk);
- ipv6_sock_mc_close(sk);
+ __ipv6_sock_mc_close(sk);
/*
* Sock is moving from IPv6 to IPv4 (sk_prot), so
diff --git a/net/ipv6/mcast.c b/net/ipv6/mcast.c
index 75c1fc5..14a3903 100644
--- a/net/ipv6/mcast.c
+++ b/net/ipv6/mcast.c
@@ -276,16 +276,14 @@ static struct inet6_dev *ip6_mc_find_dev_rcu(struct net *net,
return idev;
}
-void ipv6_sock_mc_close(struct sock *sk)
+void __ipv6_sock_mc_close(struct sock *sk)
{
struct ipv6_pinfo *np = inet6_sk(sk);
struct ipv6_mc_socklist *mc_lst;
struct net *net = sock_net(sk);
- if (!rcu_access_pointer(np->ipv6_mc_list))
- return;
+ ASSERT_RTNL();
- rtnl_lock();
while ((mc_lst = rtnl_dereference(np->ipv6_mc_list)) != NULL) {
struct net_device *dev;
@@ -303,8 +301,17 @@ void ipv6_sock_mc_close(struct sock *sk)
atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
kfree_rcu(mc_lst, rcu);
-
}
+}
+
+void ipv6_sock_mc_close(struct sock *sk)
+{
+ struct ipv6_pinfo *np = inet6_sk(sk);
+
+ if (!rcu_access_pointer(np->ipv6_mc_list))
+ return;
+ rtnl_lock();
+ __ipv6_sock_mc_close(sk);
rtnl_unlock();
}
--
2.1.0
^ permalink raw reply related [flat|nested] 7+ messages in thread
* Re: [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt()
2016-10-20 6:35 [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt() Cong Wang
@ 2016-10-20 11:12 ` Marcelo Ricardo Leitner
2016-10-20 13:47 ` Eric Dumazet
2016-10-21 15:29 ` David Miller
2 siblings, 0 replies; 7+ messages in thread
From: Marcelo Ricardo Leitner @ 2016-10-20 11:12 UTC (permalink / raw)
To: Cong Wang; +Cc: netdev, eric.dumazet, sploving1
On Wed, Oct 19, 2016 at 11:35:12PM -0700, Cong Wang wrote:
> Baozeng reported this deadlock case:
>
> CPU0 CPU1
> ---- ----
> lock([ 165.136033] sk_lock-AF_INET6);
> lock([ 165.136033] rtnl_mutex);
> lock([ 165.136033] sk_lock-AF_INET6);
> lock([ 165.136033] rtnl_mutex);
>
> Similar to commit 87e9f0315952
> ("ipv4: fix a potential deadlock in mcast getsockopt() path")
> this is due to we still have a case, ipv6_sock_mc_close(),
> where we acquire sk_lock before rtnl_lock. Close this deadlock
> with the similar solution, that is always acquire rtnl lock first.
>
> Fixes: baf606d9c9b1 ("ipv4,ipv6: grab rtnl before locking the socket")
> Reported-by: Baozeng Ding <sploving1@gmail.com>
> Tested-by: Baozeng Ding <sploving1@gmail.com>
> Cc: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
> ---
> include/net/addrconf.h | 1 +
> net/ipv6/ipv6_sockglue.c | 3 ++-
> net/ipv6/mcast.c | 17 ++++++++++++-----
> 3 files changed, 15 insertions(+), 6 deletions(-)
>
> diff --git a/include/net/addrconf.h b/include/net/addrconf.h
> index f2d0727..8f998af 100644
> --- a/include/net/addrconf.h
> +++ b/include/net/addrconf.h
> @@ -174,6 +174,7 @@ int ipv6_sock_mc_join(struct sock *sk, int ifindex,
> const struct in6_addr *addr);
> int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
> const struct in6_addr *addr);
> +void __ipv6_sock_mc_close(struct sock *sk);
> void ipv6_sock_mc_close(struct sock *sk);
> bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
> const struct in6_addr *src_addr);
> diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c
> index 5330262..636ec56 100644
> --- a/net/ipv6/ipv6_sockglue.c
> +++ b/net/ipv6/ipv6_sockglue.c
> @@ -120,6 +120,7 @@ struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
> static bool setsockopt_needs_rtnl(int optname)
> {
> switch (optname) {
> + case IPV6_ADDRFORM:
> case IPV6_ADD_MEMBERSHIP:
> case IPV6_DROP_MEMBERSHIP:
> case IPV6_JOIN_ANYCAST:
> @@ -198,7 +199,7 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
> }
>
> fl6_free_socklist(sk);
> - ipv6_sock_mc_close(sk);
> + __ipv6_sock_mc_close(sk);
Considering we already took rtnl lock and the way __ipv6_sock_mc_close()
is written, we don't need that check
if (!rcu_access_pointer(np->ipv6_mc_list))
here too as the while() in there does it already.
LGTM, thanks
Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
>
> /*
> * Sock is moving from IPv6 to IPv4 (sk_prot), so
> diff --git a/net/ipv6/mcast.c b/net/ipv6/mcast.c
> index 75c1fc5..14a3903 100644
> --- a/net/ipv6/mcast.c
> +++ b/net/ipv6/mcast.c
> @@ -276,16 +276,14 @@ static struct inet6_dev *ip6_mc_find_dev_rcu(struct net *net,
> return idev;
> }
>
> -void ipv6_sock_mc_close(struct sock *sk)
> +void __ipv6_sock_mc_close(struct sock *sk)
> {
> struct ipv6_pinfo *np = inet6_sk(sk);
> struct ipv6_mc_socklist *mc_lst;
> struct net *net = sock_net(sk);
>
> - if (!rcu_access_pointer(np->ipv6_mc_list))
> - return;
> + ASSERT_RTNL();
>
> - rtnl_lock();
> while ((mc_lst = rtnl_dereference(np->ipv6_mc_list)) != NULL) {
> struct net_device *dev;
>
> @@ -303,8 +301,17 @@ void ipv6_sock_mc_close(struct sock *sk)
>
> atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
> kfree_rcu(mc_lst, rcu);
> -
> }
> +}
> +
> +void ipv6_sock_mc_close(struct sock *sk)
> +{
> + struct ipv6_pinfo *np = inet6_sk(sk);
> +
> + if (!rcu_access_pointer(np->ipv6_mc_list))
> + return;
> + rtnl_lock();
> + __ipv6_sock_mc_close(sk);
> rtnl_unlock();
> }
>
> --
> 2.1.0
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt()
2016-10-20 6:35 [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt() Cong Wang
2016-10-20 11:12 ` Marcelo Ricardo Leitner
@ 2016-10-20 13:47 ` Eric Dumazet
2016-10-20 21:35 ` Cong Wang
2016-10-21 15:29 ` David Miller
2 siblings, 1 reply; 7+ messages in thread
From: Eric Dumazet @ 2016-10-20 13:47 UTC (permalink / raw)
To: Cong Wang; +Cc: netdev, sploving1, Marcelo Ricardo Leitner
On Wed, 2016-10-19 at 23:35 -0700, Cong Wang wrote:
> Baozeng reported this deadlock case:
...
> +
> +void ipv6_sock_mc_close(struct sock *sk)
> +{
> + struct ipv6_pinfo *np = inet6_sk(sk);
> +
> + if (!rcu_access_pointer(np->ipv6_mc_list))
> + return;
I wonder if rcu_dereference_protected(..., lockdep_sock_is_held(sk))
could be used instead, to get lockdep support ?
> + rtnl_lock();
> + __ipv6_sock_mc_close(sk);
> rtnl_unlock();
> }
>
Thanks !
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt()
2016-10-20 13:47 ` Eric Dumazet
@ 2016-10-20 21:35 ` Cong Wang
2016-10-20 22:10 ` Eric Dumazet
0 siblings, 1 reply; 7+ messages in thread
From: Cong Wang @ 2016-10-20 21:35 UTC (permalink / raw)
To: Eric Dumazet
Cc: Linux Kernel Network Developers, Baozeng Ding, Marcelo Ricardo Leitner
On Thu, Oct 20, 2016 at 6:47 AM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> On Wed, 2016-10-19 at 23:35 -0700, Cong Wang wrote:
>> Baozeng reported this deadlock case:
>
> ...
>
>> +
>> +void ipv6_sock_mc_close(struct sock *sk)
>> +{
>> + struct ipv6_pinfo *np = inet6_sk(sk);
>> +
>> + if (!rcu_access_pointer(np->ipv6_mc_list))
>> + return;
>
> I wonder if rcu_dereference_protected(..., lockdep_sock_is_held(sk))
> could be used instead, to get lockdep support ?
Maybe, but this "problem" exists without my patch too, right?
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt()
2016-10-20 21:35 ` Cong Wang
@ 2016-10-20 22:10 ` Eric Dumazet
2016-10-20 23:43 ` Cong Wang
0 siblings, 1 reply; 7+ messages in thread
From: Eric Dumazet @ 2016-10-20 22:10 UTC (permalink / raw)
To: Cong Wang
Cc: Linux Kernel Network Developers, Baozeng Ding, Marcelo Ricardo Leitner
On Thu, 2016-10-20 at 14:35 -0700, Cong Wang wrote:
> On Thu, Oct 20, 2016 at 6:47 AM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> > On Wed, 2016-10-19 at 23:35 -0700, Cong Wang wrote:
> >> Baozeng reported this deadlock case:
> >
> > ...
> >
> >> +
> >> +void ipv6_sock_mc_close(struct sock *sk)
> >> +{
> >> + struct ipv6_pinfo *np = inet6_sk(sk);
> >> +
> >> + if (!rcu_access_pointer(np->ipv6_mc_list))
> >> + return;
> >
> > I wonder if rcu_dereference_protected(..., lockdep_sock_is_held(sk))
> > could be used instead, to get lockdep support ?
>
> Maybe, but this "problem" exists without my patch too, right?
I used 'I wonder if' to say that we might have some better way to code
this test nowadays, but this can be done in a separate patch of course.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt()
2016-10-20 22:10 ` Eric Dumazet
@ 2016-10-20 23:43 ` Cong Wang
0 siblings, 0 replies; 7+ messages in thread
From: Cong Wang @ 2016-10-20 23:43 UTC (permalink / raw)
To: Eric Dumazet
Cc: Linux Kernel Network Developers, Baozeng Ding, Marcelo Ricardo Leitner
On Thu, Oct 20, 2016 at 3:10 PM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
>
> I used 'I wonder if' to say that we might have some better way to code
> this test nowadays, but this can be done in a separate patch of course.
OK, let's keep one bugfix in one patch. ;)
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt()
2016-10-20 6:35 [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt() Cong Wang
2016-10-20 11:12 ` Marcelo Ricardo Leitner
2016-10-20 13:47 ` Eric Dumazet
@ 2016-10-21 15:29 ` David Miller
2 siblings, 0 replies; 7+ messages in thread
From: David Miller @ 2016-10-21 15:29 UTC (permalink / raw)
To: xiyou.wangcong; +Cc: netdev, eric.dumazet, sploving1, marcelo.leitner
From: Cong Wang <xiyou.wangcong@gmail.com>
Date: Wed, 19 Oct 2016 23:35:12 -0700
> Baozeng reported this deadlock case:
>
> CPU0 CPU1
> ---- ----
> lock([ 165.136033] sk_lock-AF_INET6);
> lock([ 165.136033] rtnl_mutex);
> lock([ 165.136033] sk_lock-AF_INET6);
> lock([ 165.136033] rtnl_mutex);
>
> Similar to commit 87e9f0315952
> ("ipv4: fix a potential deadlock in mcast getsockopt() path")
> this is due to we still have a case, ipv6_sock_mc_close(),
> where we acquire sk_lock before rtnl_lock. Close this deadlock
> with the similar solution, that is always acquire rtnl lock first.
>
> Fixes: baf606d9c9b1 ("ipv4,ipv6: grab rtnl before locking the socket")
> Reported-by: Baozeng Ding <sploving1@gmail.com>
> Tested-by: Baozeng Ding <sploving1@gmail.com>
> Cc: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Applied, thanks.
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2016-10-21 15:29 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-10-20 6:35 [Patch net v2] ipv6: fix a potential deadlock in do_ipv6_setsockopt() Cong Wang
2016-10-20 11:12 ` Marcelo Ricardo Leitner
2016-10-20 13:47 ` Eric Dumazet
2016-10-20 21:35 ` Cong Wang
2016-10-20 22:10 ` Eric Dumazet
2016-10-20 23:43 ` Cong Wang
2016-10-21 15:29 ` David Miller
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.