diff mbox series

[net-next,v3,7/7] mld: add mc_lock for protecting per-interface mld data

Message ID 20210325161657.10517-8-ap420073@gmail.com
State Not Applicable, archived
Delegated to: Simon Wunderlich
Headers show
Series mld: change context from atomic to sleepable | expand

Commit Message

Taehee Yoo March 25, 2021, 4:16 p.m. UTC
The purpose of this lock is to avoid a bottleneck in the query/report
event handler logic.

By previous patches, almost all mld data is protected by RTNL.
So, the query and report event handler, which is data path logic
acquires RTNL too. Therefore if a lot of query and report events
are received, it uses RTNL for a long time.
So it makes the control-plane bottleneck because of using RTNL.
In order to avoid this bottleneck, mc_lock is added.

mc_lock protect only per-interface mld data and per-interface mld
data is used in the query/report event handler logic.
So, no longer rtnl_lock is needed in the query/report event handler logic.
Therefore bottleneck will be disappeared by mc_lock.

Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
---
v3:
 - Initial patch

 include/net/if_inet6.h |   1 +
 net/ipv6/mcast.c       | 309 +++++++++++++++++++++++++----------------
 2 files changed, 194 insertions(+), 116 deletions(-)

Comments

Eric Dumazet March 30, 2021, 11:59 a.m. UTC | #1
On 3/25/21 5:16 PM, Taehee Yoo wrote:
> The purpose of this lock is to avoid a bottleneck in the query/report
> event handler logic.
> 
> By previous patches, almost all mld data is protected by RTNL.
> So, the query and report event handler, which is data path logic
> acquires RTNL too. Therefore if a lot of query and report events
> are received, it uses RTNL for a long time.
> So it makes the control-plane bottleneck because of using RTNL.
> In order to avoid this bottleneck, mc_lock is added.
> 
> mc_lock protect only per-interface mld data and per-interface mld
> data is used in the query/report event handler logic.
> So, no longer rtnl_lock is needed in the query/report event handler logic.
> Therefore bottleneck will be disappeared by mc_lock.
> 

What testsuite have you run exactly to validate this monster patch ?

Have you used CONFIG_LOCKDEP=y / CONFIG_DEBUG_ATOMIC_SLEEP=y ?

> Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
> Signed-off-by: Taehee Yoo <ap420073@gmail.com>

[...]

>  /*
> - *	device multicast group del
> + * device multicast group del
>   */
>  int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
>  {
> @@ -943,8 +967,9 @@ int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
>  
>  	ASSERT_RTNL();
>  
> +	mutex_lock(&idev->mc_lock);
>  	for (map = &idev->mc_list;
> -	     (ma = rtnl_dereference(*map));
> +	     (ma = mc_dereference(*map, idev));
>  	     map = &ma->next) {
>  		if (ipv6_addr_equal(&ma->mca_addr, addr)) {
>  			if (--ma->mca_users == 0) {

This can be called with rcu_bh held, thus :

BUG: sleeping function called from invalid context at kernel/locking/mutex.c:928
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 4624, name: kworker/1:2
4 locks held by kworker/1:2/4624:
 #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: arch_atomic64_set arch/x86/include/asm/atomic64_64.h:34 [inline]
 #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: atomic64_set include/asm-generic/atomic-instrumented.h:856 [inline]
 #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: atomic_long_set include/asm-generic/atomic-long.h:41 [inline]
 #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: set_work_data kernel/workqueue.c:616 [inline]
 #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: set_work_pool_and_clear_pending kernel/workqueue.c:643 [inline]
 #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: process_one_work+0x871/0x1600 kernel/workqueue.c:2246
 #1: ffffc90009adfda8 ((addr_chk_work).work){+.+.}-{0:0}, at: process_one_work+0x8a5/0x1600 kernel/workqueue.c:2250
 #2: ffffffff8d66d328 (rtnl_mutex){+.+.}-{3:3}, at: addrconf_verify_work+0xa/0x20 net/ipv6/addrconf.c:4572
 #3: ffffffff8bf74300 (rcu_read_lock_bh){....}-{1:2}, at: addrconf_verify_rtnl+0x2b/0x1150 net/ipv6/addrconf.c:4459
Preemption disabled at:
[<ffffffff87b39f41>] local_bh_disable include/linux/bottom_half.h:19 [inline]
[<ffffffff87b39f41>] rcu_read_lock_bh include/linux/rcupdate.h:727 [inline]
[<ffffffff87b39f41>] addrconf_verify_rtnl+0x41/0x1150 net/ipv6/addrconf.c:4461
CPU: 1 PID: 4624 Comm: kworker/1:2 Not tainted 5.12.0-rc4-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: ipv6_addrconf addrconf_verify_work
Call Trace:
 __dump_stack lib/dump_stack.c:79 [inline]
 dump_stack+0x141/0x1d7 lib/dump_stack.c:120
 ___might_sleep.cold+0x1f1/0x237 kernel/sched/core.c:8328
 __mutex_lock_common kernel/locking/mutex.c:928 [inline]
 __mutex_lock+0xa9/0x1120 kernel/locking/mutex.c:1096
 __ipv6_dev_mc_dec+0x5f/0x340 net/ipv6/mcast.c:970
 addrconf_leave_solict net/ipv6/addrconf.c:2182 [inline]
 addrconf_leave_solict net/ipv6/addrconf.c:2174 [inline]
 __ipv6_ifa_notify+0x5b6/0xa90 net/ipv6/addrconf.c:6077
 ipv6_ifa_notify net/ipv6/addrconf.c:6100 [inline]
 ipv6_del_addr+0x463/0xae0 net/ipv6/addrconf.c:1294
 addrconf_verify_rtnl+0xd59/0x1150 net/ipv6/addrconf.c:4488
 addrconf_verify_work+0xf/0x20 net/ipv6/addrconf.c:4573
 process_one_work+0x98d/0x1600 kernel/workqueue.c:2275
 worker_thread+0x64c/0x1120 kernel/workqueue.c:2421
 kthread+0x3b1/0x4a0 kernel/kthread.c:292
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294
Eric Dumazet March 30, 2021, 12:24 p.m. UTC | #2
On 3/30/21 1:59 PM, Eric Dumazet wrote:
> 
> 
> On 3/25/21 5:16 PM, Taehee Yoo wrote:
>> The purpose of this lock is to avoid a bottleneck in the query/report
>> event handler logic.
>>
>> By previous patches, almost all mld data is protected by RTNL.
>> So, the query and report event handler, which is data path logic
>> acquires RTNL too. Therefore if a lot of query and report events
>> are received, it uses RTNL for a long time.
>> So it makes the control-plane bottleneck because of using RTNL.
>> In order to avoid this bottleneck, mc_lock is added.
>>
>> mc_lock protect only per-interface mld data and per-interface mld
>> data is used in the query/report event handler logic.
>> So, no longer rtnl_lock is needed in the query/report event handler logic.
>> Therefore bottleneck will be disappeared by mc_lock.
>>
> 
> What testsuite have you run exactly to validate this monster patch ?
> 
> Have you used CONFIG_LOCKDEP=y / CONFIG_DEBUG_ATOMIC_SLEEP=y ?
> 
>> Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
>> Signed-off-by: Taehee Yoo <ap420073@gmail.com>
> 
> [...]
> 
>>  /*
>> - *	device multicast group del
>> + * device multicast group del
>>   */
>>  int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
>>  {
>> @@ -943,8 +967,9 @@ int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
>>  
>>  	ASSERT_RTNL();
>>  
>> +	mutex_lock(&idev->mc_lock);
>>  	for (map = &idev->mc_list;
>> -	     (ma = rtnl_dereference(*map));
>> +	     (ma = mc_dereference(*map, idev));
>>  	     map = &ma->next) {
>>  		if (ipv6_addr_equal(&ma->mca_addr, addr)) {
>>  			if (--ma->mca_users == 0) {
> 
> This can be called with rcu_bh held, thus :
> 
> BUG: sleeping function called from invalid context at kernel/locking/mutex.c:928
> in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 4624, name: kworker/1:2
> 4 locks held by kworker/1:2/4624:
>  #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: arch_atomic64_set arch/x86/include/asm/atomic64_64.h:34 [inline]
>  #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: atomic64_set include/asm-generic/atomic-instrumented.h:856 [inline]
>  #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: atomic_long_set include/asm-generic/atomic-long.h:41 [inline]
>  #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: set_work_data kernel/workqueue.c:616 [inline]
>  #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: set_work_pool_and_clear_pending kernel/workqueue.c:643 [inline]
>  #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: process_one_work+0x871/0x1600 kernel/workqueue.c:2246
>  #1: ffffc90009adfda8 ((addr_chk_work).work){+.+.}-{0:0}, at: process_one_work+0x8a5/0x1600 kernel/workqueue.c:2250
>  #2: ffffffff8d66d328 (rtnl_mutex){+.+.}-{3:3}, at: addrconf_verify_work+0xa/0x20 net/ipv6/addrconf.c:4572
>  #3: ffffffff8bf74300 (rcu_read_lock_bh){....}-{1:2}, at: addrconf_verify_rtnl+0x2b/0x1150 net/ipv6/addrconf.c:4459
> Preemption disabled at:
> [<ffffffff87b39f41>] local_bh_disable include/linux/bottom_half.h:19 [inline]
> [<ffffffff87b39f41>] rcu_read_lock_bh include/linux/rcupdate.h:727 [inline]
> [<ffffffff87b39f41>] addrconf_verify_rtnl+0x41/0x1150 net/ipv6/addrconf.c:4461
> CPU: 1 PID: 4624 Comm: kworker/1:2 Not tainted 5.12.0-rc4-syzkaller #0
> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
> Workqueue: ipv6_addrconf addrconf_verify_work
> Call Trace:
>  __dump_stack lib/dump_stack.c:79 [inline]
>  dump_stack+0x141/0x1d7 lib/dump_stack.c:120
>  ___might_sleep.cold+0x1f1/0x237 kernel/sched/core.c:8328
>  __mutex_lock_common kernel/locking/mutex.c:928 [inline]
>  __mutex_lock+0xa9/0x1120 kernel/locking/mutex.c:1096
>  __ipv6_dev_mc_dec+0x5f/0x340 net/ipv6/mcast.c:970
>  addrconf_leave_solict net/ipv6/addrconf.c:2182 [inline]
>  addrconf_leave_solict net/ipv6/addrconf.c:2174 [inline]
>  __ipv6_ifa_notify+0x5b6/0xa90 net/ipv6/addrconf.c:6077
>  ipv6_ifa_notify net/ipv6/addrconf.c:6100 [inline]
>  ipv6_del_addr+0x463/0xae0 net/ipv6/addrconf.c:1294
>  addrconf_verify_rtnl+0xd59/0x1150 net/ipv6/addrconf.c:4488
>  addrconf_verify_work+0xf/0x20 net/ipv6/addrconf.c:4573
>  process_one_work+0x98d/0x1600 kernel/workqueue.c:2275
>  worker_thread+0x64c/0x1120 kernel/workqueue.c:2421
>  kthread+0x3b1/0x4a0 kernel/kthread.c:292
>  ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294
> 

I will test this fix:

diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
index 120073ffb666b18678e3145d91dac59fa865a592..8f3883f4cb4a15a0749b8f0fe00061e483ea26ca 100644
--- a/net/ipv6/addrconf.c
+++ b/net/ipv6/addrconf.c
@@ -4485,7 +4485,9 @@ static void addrconf_verify_rtnl(void)
                            age >= ifp->valid_lft) {
                                spin_unlock(&ifp->lock);
                                in6_ifa_hold(ifp);
+                               rcu_read_unlock_bh();
                                ipv6_del_addr(ifp);
+                               rcu_read_lock_bh();
                                goto restart;
                        } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
                                spin_unlock(&ifp->lock);
Taehee Yoo March 30, 2021, 3:01 p.m. UTC | #3
On 3/30/21 9:24 PM, Eric Dumazet wrote:
 >
 >
 > On 3/30/21 1:59 PM, Eric Dumazet wrote:
 >>
 >>
 >> On 3/25/21 5:16 PM, Taehee Yoo wrote:
 >>> The purpose of this lock is to avoid a bottleneck in the query/report
 >>> event handler logic.
 >>>
 >>> By previous patches, almost all mld data is protected by RTNL.
 >>> So, the query and report event handler, which is data path logic
 >>> acquires RTNL too. Therefore if a lot of query and report events
 >>> are received, it uses RTNL for a long time.
 >>> So it makes the control-plane bottleneck because of using RTNL.
 >>> In order to avoid this bottleneck, mc_lock is added.
 >>>
 >>> mc_lock protect only per-interface mld data and per-interface mld
 >>> data is used in the query/report event handler logic.
 >>> So, no longer rtnl_lock is needed in the query/report event handler 
logic.
 >>> Therefore bottleneck will be disappeared by mc_lock.
 >>>
 >>
 >> What testsuite have you run exactly to validate this monster patch ?
 >>

I've been using an application, which calls setsockopt() with the below 
options.
IPV6_ADD_MEMBERSHIP
IPV6_DROP_MEMBERSHIP
MCAST_JOIN_SOURCE_GROUP
MCAST_LEAVE_SOURCE_GROUP
MCAST_BLOCK_SOURCE
MCAST_UNBLOCK_SOURCE
MCAST_MSFILTER
And checks out  /proc/net/mcfilter6 and /proc/net/igmp6.

 >> Have you used CONFIG_LOCKDEP=y / CONFIG_DEBUG_ATOMIC_SLEEP=y ?
 >>

Yes, I'm using both configs.

 >>> Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
 >>> Signed-off-by: Taehee Yoo <ap420073@gmail.com>
 >>
 >> [...]
 >>
 >>>   /*
 >>> - *	device multicast group del
 >>> + * device multicast group del
 >>>    */
 >>>   int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct 
in6_addr *addr)
 >>>   {
 >>> @@ -943,8 +967,9 @@ int __ipv6_dev_mc_dec(struct inet6_dev *idev, 
const struct in6_addr *addr)
 >>>
 >>>   	ASSERT_RTNL();
 >>>
 >>> +	mutex_lock(&idev->mc_lock);
 >>>   	for (map = &idev->mc_list;
 >>> -	     (ma = rtnl_dereference(*map));
 >>> +	     (ma = mc_dereference(*map, idev));
 >>>   	     map = &ma->next) {
 >>>   		if (ipv6_addr_equal(&ma->mca_addr, addr)) {
 >>>   			if (--ma->mca_users == 0) {
 >>
 >> This can be called with rcu_bh held, thus :
 >>
 >> BUG: sleeping function called from invalid context at 
kernel/locking/mutex.c:928
 >> in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 4624, name: 
kworker/1:2
 >> 4 locks held by kworker/1:2/4624:
 >>   #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, 
at: arch_atomic64_set arch/x86/include/asm/atomic64_64.h:34 [inline]
 >>   #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, 
at: atomic64_set include/asm-generic/atomic-instrumented.h:856 [inline]
 >>   #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, 
at: atomic_long_set include/asm-generic/atomic-long.h:41 [inline]
 >>   #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, 
at: set_work_data kernel/workqueue.c:616 [inline]
 >>   #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, 
at: set_work_pool_and_clear_pending kernel/workqueue.c:643 [inline]
 >>   #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, 
at: process_one_work+0x871/0x1600 kernel/workqueue.c:2246
 >>   #1: ffffc90009adfda8 ((addr_chk_work).work){+.+.}-{0:0}, at: 
process_one_work+0x8a5/0x1600 kernel/workqueue.c:2250
 >>   #2: ffffffff8d66d328 (rtnl_mutex){+.+.}-{3:3}, at: 
addrconf_verify_work+0xa/0x20 net/ipv6/addrconf.c:4572
 >>   #3: ffffffff8bf74300 (rcu_read_lock_bh){....}-{1:2}, at: 
addrconf_verify_rtnl+0x2b/0x1150 net/ipv6/addrconf.c:4459
 >> Preemption disabled at:
 >> [<ffffffff87b39f41>] local_bh_disable include/linux/bottom_half.h:19 
[inline]
 >> [<ffffffff87b39f41>] rcu_read_lock_bh include/linux/rcupdate.h:727 
[inline]
 >> [<ffffffff87b39f41>] addrconf_verify_rtnl+0x41/0x1150 
net/ipv6/addrconf.c:4461
 >> CPU: 1 PID: 4624 Comm: kworker/1:2 Not tainted 5.12.0-rc4-syzkaller #0
 >> Hardware name: Google Google Compute Engine/Google Compute Engine, 
BIOS Google 01/01/2011
 >> Workqueue: ipv6_addrconf addrconf_verify_work
 >> Call Trace:
 >>   __dump_stack lib/dump_stack.c:79 [inline]
 >>   dump_stack+0x141/0x1d7 lib/dump_stack.c:120
 >>   ___might_sleep.cold+0x1f1/0x237 kernel/sched/core.c:8328
 >>   __mutex_lock_common kernel/locking/mutex.c:928 [inline]
 >>   __mutex_lock+0xa9/0x1120 kernel/locking/mutex.c:1096
 >>   __ipv6_dev_mc_dec+0x5f/0x340 net/ipv6/mcast.c:970
 >>   addrconf_leave_solict net/ipv6/addrconf.c:2182 [inline]
 >>   addrconf_leave_solict net/ipv6/addrconf.c:2174 [inline]
 >>   __ipv6_ifa_notify+0x5b6/0xa90 net/ipv6/addrconf.c:6077
 >>   ipv6_ifa_notify net/ipv6/addrconf.c:6100 [inline]
 >>   ipv6_del_addr+0x463/0xae0 net/ipv6/addrconf.c:1294
 >>   addrconf_verify_rtnl+0xd59/0x1150 net/ipv6/addrconf.c:4488
 >>   addrconf_verify_work+0xf/0x20 net/ipv6/addrconf.c:4573
 >>   process_one_work+0x98d/0x1600 kernel/workqueue.c:2275
 >>   worker_thread+0x64c/0x1120 kernel/workqueue.c:2421
 >>   kthread+0x3b1/0x4a0 kernel/kthread.c:292
 >>   ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294
 >>
 >
 > I will test this fix:

Thanks a lot!

 >
 > diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
 > index 
120073ffb666b18678e3145d91dac59fa865a592..8f3883f4cb4a15a0749b8f0fe00061e483ea26ca 
100644
 > --- a/net/ipv6/addrconf.c
 > +++ b/net/ipv6/addrconf.c
 > @@ -4485,7 +4485,9 @@ static void addrconf_verify_rtnl(void)
 >                              age >= ifp->valid_lft) {
 >                                  spin_unlock(&ifp->lock);
 >                                  in6_ifa_hold(ifp);
 > +                               rcu_read_unlock_bh();
 >                                  ipv6_del_addr(ifp);
 > +                               rcu_read_lock_bh();
 >                                  goto restart;
 >                          } else if (ifp->prefered_lft == 
INFINITY_LIFE_TIME) {
 >                                  spin_unlock(&ifp->lock);
 >
diff mbox series

Patch

diff --git a/include/net/if_inet6.h b/include/net/if_inet6.h
index 882e0f88756f..71bb4cc4d05d 100644
--- a/include/net/if_inet6.h
+++ b/include/net/if_inet6.h
@@ -190,6 +190,7 @@  struct inet6_dev {
 
 	spinlock_t		mc_query_lock;	/* mld query queue lock */
 	spinlock_t		mc_report_lock;	/* mld query report lock */
+	struct mutex		mc_lock;	/* mld global lock */
 
 	struct ifacaddr6	*ac_list;
 	rwlock_t		lock;
diff --git a/net/ipv6/mcast.c b/net/ipv6/mcast.c
index 3ad754388933..49b0cebfdcdc 100644
--- a/net/ipv6/mcast.c
+++ b/net/ipv6/mcast.c
@@ -111,6 +111,8 @@  int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
 /*
  *	socket join on multicast group
  */
+#define mc_dereference(e, idev) \
+	rcu_dereference_protected(e, lockdep_is_held(&(idev)->mc_lock))
 
 #define for_each_pmc_rtnl(np, pmc)				\
 	for (pmc = rtnl_dereference((np)->ipv6_mc_list);	\
@@ -122,10 +124,10 @@  int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
 	     pmc;						\
 	     pmc = rcu_dereference(pmc->next))
 
-#define for_each_psf_rtnl(mc, psf)				\
-	for (psf = rtnl_dereference((mc)->mca_sources);		\
+#define for_each_psf_mclock(mc, psf)				\
+	for (psf = mc_dereference((mc)->mca_sources, mc->idev);	\
 	     psf;						\
-	     psf = rtnl_dereference(psf->sf_next))
+	     psf = mc_dereference(psf->sf_next, mc->idev))
 
 #define for_each_psf_rcu(mc, psf)				\
 	for (psf = rcu_dereference((mc)->mca_sources);		\
@@ -133,14 +135,14 @@  int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
 	     psf = rcu_dereference(psf->sf_next))
 
 #define for_each_psf_tomb(mc, psf)				\
-	for (psf = rtnl_dereference((mc)->mca_tomb);		\
+	for (psf = mc_dereference((mc)->mca_tomb, mc->idev);	\
 	     psf;						\
-	     psf = rtnl_dereference(psf->sf_next))
+	     psf = mc_dereference(psf->sf_next, mc->idev))
 
-#define for_each_mc_rtnl(idev, mc)				\
-	for (mc = rtnl_dereference((idev)->mc_list);		\
+#define for_each_mc_mclock(idev, mc)				\
+	for (mc = mc_dereference((idev)->mc_list, idev);	\
 	     mc;						\
-	     mc = rtnl_dereference(mc->next))
+	     mc = mc_dereference(mc->next, idev))
 
 #define for_each_mc_rcu(idev, mc)				\
 	for (mc = rcu_dereference((idev)->mc_list);             \
@@ -148,9 +150,9 @@  int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
 	     mc = rcu_dereference(mc->next))
 
 #define for_each_mc_tomb(idev, mc)				\
-	for (mc = rtnl_dereference((idev)->mc_tomb);		\
+	for (mc = mc_dereference((idev)->mc_tomb, idev);	\
 	     mc;						\
-	     mc = rtnl_dereference(mc->next))
+	     mc = mc_dereference(mc->next, idev))
 
 static int unsolicited_report_interval(struct inet6_dev *idev)
 {
@@ -268,11 +270,12 @@  int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
 			if (dev) {
 				struct inet6_dev *idev = __in6_dev_get(dev);
 
-				(void) ip6_mc_leave_src(sk, mc_lst, idev);
+				ip6_mc_leave_src(sk, mc_lst, idev);
 				if (idev)
 					__ipv6_dev_mc_dec(idev, &mc_lst->addr);
-			} else
-				(void) ip6_mc_leave_src(sk, mc_lst, NULL);
+			} else {
+				ip6_mc_leave_src(sk, mc_lst, NULL);
+			}
 
 			atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 			kfree_rcu(mc_lst, rcu);
@@ -329,11 +332,12 @@  void __ipv6_sock_mc_close(struct sock *sk)
 		if (dev) {
 			struct inet6_dev *idev = __in6_dev_get(dev);
 
-			(void) ip6_mc_leave_src(sk, mc_lst, idev);
+			ip6_mc_leave_src(sk, mc_lst, idev);
 			if (idev)
 				__ipv6_dev_mc_dec(idev, &mc_lst->addr);
-		} else
-			(void) ip6_mc_leave_src(sk, mc_lst, NULL);
+		} else {
+			ip6_mc_leave_src(sk, mc_lst, NULL);
+		}
 
 		atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
 		kfree_rcu(mc_lst, rcu);
@@ -376,6 +380,7 @@  int ip6_mc_source(int add, int omode, struct sock *sk,
 
 	err = -EADDRNOTAVAIL;
 
+	mutex_lock(&idev->mc_lock);
 	for_each_pmc_rtnl(inet6, pmc) {
 		if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
 			continue;
@@ -469,6 +474,7 @@  int ip6_mc_source(int add, int omode, struct sock *sk,
 	/* update the interface list */
 	ip6_mc_add_src(idev, group, omode, 1, source, 1);
 done:
+	mutex_unlock(&idev->mc_lock);
 	if (leavegroup)
 		err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
 	return err;
@@ -529,25 +535,33 @@  int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
 			psin6 = (struct sockaddr_in6 *)list;
 			newpsl->sl_addr[i] = psin6->sin6_addr;
 		}
+		mutex_lock(&idev->mc_lock);
 		err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
-			newpsl->sl_count, newpsl->sl_addr, 0);
+				     newpsl->sl_count, newpsl->sl_addr, 0);
 		if (err) {
+			mutex_unlock(&idev->mc_lock);
 			sock_kfree_s(sk, newpsl, IP6_SFLSIZE(newpsl->sl_max));
 			goto done;
 		}
+		mutex_unlock(&idev->mc_lock);
 	} else {
 		newpsl = NULL;
-		(void) ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
+		mutex_lock(&idev->mc_lock);
+		ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
+		mutex_unlock(&idev->mc_lock);
 	}
 
+	mutex_lock(&idev->mc_lock);
 	psl = rtnl_dereference(pmc->sflist);
 	if (psl) {
-		(void) ip6_mc_del_src(idev, group, pmc->sfmode,
-			psl->sl_count, psl->sl_addr, 0);
+		ip6_mc_del_src(idev, group, pmc->sfmode,
+			       psl->sl_count, psl->sl_addr, 0);
 		atomic_sub(IP6_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
 		kfree_rcu(psl, rcu);
-	} else
-		(void) ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
+	} else {
+		ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
+	}
+	mutex_unlock(&idev->mc_lock);
 	rcu_assign_pointer(pmc->sflist, newpsl);
 	pmc->sfmode = gsf->gf_fmode;
 	err = 0;
@@ -650,6 +664,7 @@  bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
 	return rv;
 }
 
+/* called with mc_lock */
 static void igmp6_group_added(struct ifmcaddr6 *mc)
 {
 	struct net_device *dev = mc->idev->dev;
@@ -684,6 +699,7 @@  static void igmp6_group_added(struct ifmcaddr6 *mc)
 	mld_ifc_event(mc->idev);
 }
 
+/* called with mc_lock */
 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
 {
 	struct net_device *dev = mc->idev->dev;
@@ -711,6 +727,7 @@  static void igmp6_group_dropped(struct ifmcaddr6 *mc)
 
 /*
  * deleted ifmcaddr6 manipulation
+ * called with mc_lock
  */
 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 {
@@ -735,13 +752,13 @@  static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 		struct ip6_sf_list *psf;
 
 		rcu_assign_pointer(pmc->mca_tomb,
-				   rtnl_dereference(im->mca_tomb));
+				   mc_dereference(im->mca_tomb, idev));
 		rcu_assign_pointer(pmc->mca_sources,
-				   rtnl_dereference(im->mca_sources));
+				   mc_dereference(im->mca_sources, idev));
 		RCU_INIT_POINTER(im->mca_tomb, NULL);
 		RCU_INIT_POINTER(im->mca_sources, NULL);
 
-		for_each_psf_rtnl(pmc, psf)
+		for_each_psf_mclock(pmc, psf)
 			psf->sf_crcount = pmc->mca_crcount;
 	}
 
@@ -749,6 +766,7 @@  static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 	rcu_assign_pointer(idev->mc_tomb, pmc);
 }
 
+/* called with mc_lock */
 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 {
 	struct ip6_sf_list *psf, *sources, *tomb;
@@ -772,15 +790,15 @@  static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 		im->idev = pmc->idev;
 		if (im->mca_sfmode == MCAST_INCLUDE) {
 			tomb = rcu_replace_pointer(im->mca_tomb,
-						   rtnl_dereference(pmc->mca_tomb),
-						   lockdep_rtnl_is_held());
+						   mc_dereference(pmc->mca_tomb, pmc->idev),
+						   lockdep_is_held(&im->idev->mc_lock));
 			rcu_assign_pointer(pmc->mca_tomb, tomb);
 
 			sources = rcu_replace_pointer(im->mca_sources,
-						      rtnl_dereference(pmc->mca_sources),
-						      lockdep_rtnl_is_held());
+						      mc_dereference(pmc->mca_sources, pmc->idev),
+						      lockdep_is_held(&im->idev->mc_lock));
 			rcu_assign_pointer(pmc->mca_sources, sources);
-			for_each_psf_rtnl(im, psf)
+			for_each_psf_mclock(im, psf)
 				psf->sf_crcount = idev->mc_qrv;
 		} else {
 			im->mca_crcount = idev->mc_qrv;
@@ -791,28 +809,29 @@  static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
 	}
 }
 
+/* called with mc_lock */
 static void mld_clear_delrec(struct inet6_dev *idev)
 {
 	struct ifmcaddr6 *pmc, *nextpmc;
 
-	pmc = rtnl_dereference(idev->mc_tomb);
+	pmc = mc_dereference(idev->mc_tomb, idev);
 	RCU_INIT_POINTER(idev->mc_tomb, NULL);
 
 	for (; pmc; pmc = nextpmc) {
-		nextpmc = rtnl_dereference(pmc->next);
+		nextpmc = mc_dereference(pmc->next, idev);
 		ip6_mc_clear_src(pmc);
 		in6_dev_put(pmc->idev);
 		kfree_rcu(pmc, rcu);
 	}
 
 	/* clear dead sources, too */
-	for_each_mc_rtnl(idev, pmc) {
+	for_each_mc_mclock(idev, pmc) {
 		struct ip6_sf_list *psf, *psf_next;
 
-		psf = rtnl_dereference(pmc->mca_tomb);
+		psf = mc_dereference(pmc->mca_tomb, idev);
 		RCU_INIT_POINTER(pmc->mca_tomb, NULL);
 		for (; psf; psf = psf_next) {
-			psf_next = rtnl_dereference(psf->sf_next);
+			psf_next = mc_dereference(psf->sf_next, idev);
 			kfree_rcu(psf, rcu);
 		}
 	}
@@ -851,6 +870,7 @@  static void ma_put(struct ifmcaddr6 *mc)
 	}
 }
 
+/* called with mc_lock */
 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
 				   const struct in6_addr *addr,
 				   unsigned int mode)
@@ -902,10 +922,12 @@  static int __ipv6_dev_mc_inc(struct net_device *dev,
 		return -ENODEV;
 	}
 
-	for_each_mc_rtnl(idev, mc) {
+	mutex_lock(&idev->mc_lock);
+	for_each_mc_mclock(idev, mc) {
 		if (ipv6_addr_equal(&mc->mca_addr, addr)) {
 			mc->mca_users++;
 			ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
+			mutex_unlock(&idev->mc_lock);
 			in6_dev_put(idev);
 			return 0;
 		}
@@ -913,6 +935,7 @@  static int __ipv6_dev_mc_inc(struct net_device *dev,
 
 	mc = mca_alloc(idev, addr, mode);
 	if (!mc) {
+		mutex_unlock(&idev->mc_lock);
 		in6_dev_put(idev);
 		return -ENOMEM;
 	}
@@ -924,6 +947,7 @@  static int __ipv6_dev_mc_inc(struct net_device *dev,
 
 	mld_del_delrec(idev, mc);
 	igmp6_group_added(mc);
+	mutex_unlock(&idev->mc_lock);
 	ma_put(mc);
 	return 0;
 }
@@ -935,7 +959,7 @@  int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
 EXPORT_SYMBOL(ipv6_dev_mc_inc);
 
 /*
- *	device multicast group del
+ * device multicast group del
  */
 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 {
@@ -943,8 +967,9 @@  int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 
 	ASSERT_RTNL();
 
+	mutex_lock(&idev->mc_lock);
 	for (map = &idev->mc_list;
-	     (ma = rtnl_dereference(*map));
+	     (ma = mc_dereference(*map, idev));
 	     map = &ma->next) {
 		if (ipv6_addr_equal(&ma->mca_addr, addr)) {
 			if (--ma->mca_users == 0) {
@@ -952,14 +977,17 @@  int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
 
 				igmp6_group_dropped(ma);
 				ip6_mc_clear_src(ma);
+				mutex_unlock(&idev->mc_lock);
 
 				ma_put(ma);
 				return 0;
 			}
+			mutex_unlock(&idev->mc_lock);
 			return 0;
 		}
 	}
 
+	mutex_unlock(&idev->mc_lock);
 	return -ENOENT;
 }
 
@@ -1019,6 +1047,7 @@  bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
 	return rv;
 }
 
+/* called with mc_lock */
 static void mld_gq_start_work(struct inet6_dev *idev)
 {
 	unsigned long tv = prandom_u32() % idev->mc_maxdelay;
@@ -1028,6 +1057,7 @@  static void mld_gq_start_work(struct inet6_dev *idev)
 		in6_dev_hold(idev);
 }
 
+/* called with mc_lock */
 static void mld_gq_stop_work(struct inet6_dev *idev)
 {
 	idev->mc_gq_running = 0;
@@ -1035,6 +1065,7 @@  static void mld_gq_stop_work(struct inet6_dev *idev)
 		__in6_dev_put(idev);
 }
 
+/* called with mc_lock */
 static void mld_ifc_start_work(struct inet6_dev *idev, unsigned long delay)
 {
 	unsigned long tv = prandom_u32() % delay;
@@ -1043,6 +1074,7 @@  static void mld_ifc_start_work(struct inet6_dev *idev, unsigned long delay)
 		in6_dev_hold(idev);
 }
 
+/* called with mc_lock */
 static void mld_ifc_stop_work(struct inet6_dev *idev)
 {
 	idev->mc_ifc_count = 0;
@@ -1050,6 +1082,7 @@  static void mld_ifc_stop_work(struct inet6_dev *idev)
 		__in6_dev_put(idev);
 }
 
+/* called with mc_lock */
 static void mld_dad_start_work(struct inet6_dev *idev, unsigned long delay)
 {
 	unsigned long tv = prandom_u32() % delay;
@@ -1080,6 +1113,7 @@  static void mld_report_stop_work(struct inet6_dev *idev)
 
 /*
  * IGMP handling (alias multicast ICMPv6 messages)
+ * called with mc_lock
  */
 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
 {
@@ -1103,7 +1137,9 @@  static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
 	ma->mca_flags |= MAF_TIMER_RUNNING;
 }
 
-/* mark EXCLUDE-mode sources */
+/* mark EXCLUDE-mode sources
+ * called with mc_lock
+ */
 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
 			     const struct in6_addr *srcs)
 {
@@ -1111,7 +1147,7 @@  static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
 	int i, scount;
 
 	scount = 0;
-	for_each_psf_rtnl(pmc, psf) {
+	for_each_psf_mclock(pmc, psf) {
 		if (scount == nsrcs)
 			break;
 		for (i = 0; i < nsrcs; i++) {
@@ -1132,6 +1168,7 @@  static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
 	return true;
 }
 
+/* called with mc_lock */
 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
 			    const struct in6_addr *srcs)
 {
@@ -1144,7 +1181,7 @@  static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
 	/* mark INCLUDE-mode sources */
 
 	scount = 0;
-	for_each_psf_rtnl(pmc, psf) {
+	for_each_psf_mclock(pmc, psf) {
 		if (scount == nsrcs)
 			break;
 		for (i = 0; i < nsrcs; i++) {
@@ -1370,7 +1407,7 @@  static void __mld_query_work(struct sk_buff *skb)
 	int len, err;
 
 	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
-		goto out;
+		goto kfree_skb;
 
 	/* compute payload length excluding extension headers */
 	len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
@@ -1387,11 +1424,11 @@  static void __mld_query_work(struct sk_buff *skb)
 	    ipv6_hdr(skb)->hop_limit != 1 ||
 	    !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
 	    IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
-		goto out;
+		goto kfree_skb;
 
-	idev = __in6_dev_get(skb->dev);
+	idev = in6_dev_get(skb->dev);
 	if (!idev)
-		goto out;
+		goto kfree_skb;
 
 	mld = (struct mld_msg *)icmp6_hdr(skb);
 	group = &mld->mld_mca;
@@ -1442,11 +1479,11 @@  static void __mld_query_work(struct sk_buff *skb)
 	}
 
 	if (group_type == IPV6_ADDR_ANY) {
-		for_each_mc_rtnl(idev, ma) {
+		for_each_mc_mclock(idev, ma) {
 			igmp6_group_queried(ma, max_delay);
 		}
 	} else {
-		for_each_mc_rtnl(idev, ma) {
+		for_each_mc_mclock(idev, ma) {
 			if (!ipv6_addr_equal(group, &ma->mca_addr))
 				continue;
 			if (ma->mca_flags & MAF_TIMER_RUNNING) {
@@ -1468,6 +1505,8 @@  static void __mld_query_work(struct sk_buff *skb)
 	}
 
 out:
+	in6_dev_put(idev);
+kfree_skb:
 	consume_skb(skb);
 }
 
@@ -1495,10 +1534,10 @@  static void mld_query_work(struct work_struct *work)
 	}
 	spin_unlock_bh(&idev->mc_query_lock);
 
-	rtnl_lock();
+	mutex_lock(&idev->mc_lock);
 	while ((skb = __skb_dequeue(&q)))
 		__mld_query_work(skb);
-	rtnl_unlock();
+	mutex_unlock(&idev->mc_lock);
 
 	if (!rework)
 		in6_dev_put(idev);
@@ -1530,22 +1569,22 @@  int igmp6_event_report(struct sk_buff *skb)
 
 static void __mld_report_work(struct sk_buff *skb)
 {
-	struct ifmcaddr6 *ma;
 	struct inet6_dev *idev;
+	struct ifmcaddr6 *ma;
 	struct mld_msg *mld;
 	int addr_type;
 
 	/* Our own report looped back. Ignore it. */
 	if (skb->pkt_type == PACKET_LOOPBACK)
-		goto out;
+		goto kfree_skb;
 
 	/* send our report if the MC router may not have heard this report */
 	if (skb->pkt_type != PACKET_MULTICAST &&
 	    skb->pkt_type != PACKET_BROADCAST)
-		goto out;
+		goto kfree_skb;
 
 	if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
-		goto out;
+		goto kfree_skb;
 
 	mld = (struct mld_msg *)icmp6_hdr(skb);
 
@@ -1553,17 +1592,17 @@  static void __mld_report_work(struct sk_buff *skb)
 	addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
 	if (addr_type != IPV6_ADDR_ANY &&
 	    !(addr_type&IPV6_ADDR_LINKLOCAL))
-		goto out;
+		goto kfree_skb;
 
-	idev = __in6_dev_get(skb->dev);
+	idev = in6_dev_get(skb->dev);
 	if (!idev)
-		goto out;
+		goto kfree_skb;
 
 	/*
 	 *	Cancel the work for this group
 	 */
 
-	for_each_mc_rtnl(idev, ma) {
+	for_each_mc_mclock(idev, ma) {
 		if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
 			if (cancel_delayed_work(&ma->mca_work))
 				refcount_dec(&ma->mca_refcnt);
@@ -1573,7 +1612,8 @@  static void __mld_report_work(struct sk_buff *skb)
 		}
 	}
 
-out:
+	in6_dev_put(idev);
+kfree_skb:
 	consume_skb(skb);
 }
 
@@ -1600,10 +1640,10 @@  static void mld_report_work(struct work_struct *work)
 	}
 	spin_unlock_bh(&idev->mc_report_lock);
 
-	rtnl_lock();
+	mutex_lock(&idev->mc_lock);
 	while ((skb = __skb_dequeue(&q)))
 		__mld_report_work(skb);
-	rtnl_unlock();
+	mutex_unlock(&idev->mc_lock);
 
 	if (!rework)
 		in6_dev_put(idev);
@@ -1659,7 +1699,7 @@  mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
 	struct ip6_sf_list *psf;
 	int scount = 0;
 
-	for_each_psf_rtnl(pmc, psf) {
+	for_each_psf_mclock(pmc, psf) {
 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
 			continue;
 		scount++;
@@ -1833,6 +1873,7 @@  static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
 
 #define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
 
+/* called with mc_lock */
 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
 				int type, int gdeleted, int sdeleted,
 				int crsend)
@@ -1878,12 +1919,12 @@  static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
 	}
 	first = 1;
 	psf_prev = NULL;
-	for (psf = rtnl_dereference(*psf_list);
+	for (psf = mc_dereference(*psf_list, idev);
 	     psf;
 	     psf = psf_next) {
 		struct in6_addr *psrc;
 
-		psf_next = rtnl_dereference(psf->sf_next);
+		psf_next = mc_dereference(psf->sf_next, idev);
 
 		if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
 			psf_prev = psf;
@@ -1931,10 +1972,10 @@  static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
 				if (psf_prev)
 					rcu_assign_pointer(psf_prev->sf_next,
-							   rtnl_dereference(psf->sf_next));
+							   mc_dereference(psf->sf_next, idev));
 				else
 					rcu_assign_pointer(*psf_list,
-							   rtnl_dereference(psf->sf_next));
+							   mc_dereference(psf->sf_next, idev));
 				kfree_rcu(psf, rcu);
 				continue;
 			}
@@ -1964,13 +2005,14 @@  static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
 	return skb;
 }
 
+/* called with mc_lock */
 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
 {
 	struct sk_buff *skb = NULL;
 	int type;
 
 	if (!pmc) {
-		for_each_mc_rtnl(idev, pmc) {
+		for_each_mc_mclock(idev, pmc) {
 			if (pmc->mca_flags & MAF_NOREPORT)
 				continue;
 			if (pmc->mca_sfcount[MCAST_EXCLUDE])
@@ -1992,23 +2034,24 @@  static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
 
 /*
  * remove zero-count source records from a source filter list
+ * called with mc_lock
  */
-static void mld_clear_zeros(struct ip6_sf_list __rcu **ppsf)
+static void mld_clear_zeros(struct ip6_sf_list __rcu **ppsf, struct inet6_dev *idev)
 {
 	struct ip6_sf_list *psf_prev, *psf_next, *psf;
 
 	psf_prev = NULL;
-	for (psf = rtnl_dereference(*ppsf);
+	for (psf = mc_dereference(*ppsf, idev);
 	     psf;
 	     psf = psf_next) {
-		psf_next = rtnl_dereference(psf->sf_next);
+		psf_next = mc_dereference(psf->sf_next, idev);
 		if (psf->sf_crcount == 0) {
 			if (psf_prev)
 				rcu_assign_pointer(psf_prev->sf_next,
-						   rtnl_dereference(psf->sf_next));
+						   mc_dereference(psf->sf_next, idev));
 			else
 				rcu_assign_pointer(*ppsf,
-						   rtnl_dereference(psf->sf_next));
+						   mc_dereference(psf->sf_next, idev));
 			kfree_rcu(psf, rcu);
 		} else {
 			psf_prev = psf;
@@ -2016,6 +2059,7 @@  static void mld_clear_zeros(struct ip6_sf_list __rcu **ppsf)
 	}
 }
 
+/* called with mc_lock */
 static void mld_send_cr(struct inet6_dev *idev)
 {
 	struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
@@ -2024,10 +2068,10 @@  static void mld_send_cr(struct inet6_dev *idev)
 
 	/* deleted MCA's */
 	pmc_prev = NULL;
-	for (pmc = rtnl_dereference(idev->mc_tomb);
+	for (pmc = mc_dereference(idev->mc_tomb, idev);
 	     pmc;
 	     pmc = pmc_next) {
-		pmc_next = rtnl_dereference(pmc->next);
+		pmc_next = mc_dereference(pmc->next, idev);
 		if (pmc->mca_sfmode == MCAST_INCLUDE) {
 			type = MLD2_BLOCK_OLD_SOURCES;
 			dtype = MLD2_BLOCK_OLD_SOURCES;
@@ -2041,8 +2085,8 @@  static void mld_send_cr(struct inet6_dev *idev)
 			}
 			pmc->mca_crcount--;
 			if (pmc->mca_crcount == 0) {
-				mld_clear_zeros(&pmc->mca_tomb);
-				mld_clear_zeros(&pmc->mca_sources);
+				mld_clear_zeros(&pmc->mca_tomb, idev);
+				mld_clear_zeros(&pmc->mca_sources, idev);
 			}
 		}
 		if (pmc->mca_crcount == 0 &&
@@ -2059,7 +2103,7 @@  static void mld_send_cr(struct inet6_dev *idev)
 	}
 
 	/* change recs */
-	for_each_mc_rtnl(idev, pmc) {
+	for_each_mc_mclock(idev, pmc) {
 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
 			type = MLD2_BLOCK_OLD_SOURCES;
 			dtype = MLD2_ALLOW_NEW_SOURCES;
@@ -2181,6 +2225,7 @@  static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
 	goto out;
 }
 
+/* called with mc_lock */
 static void mld_send_initial_cr(struct inet6_dev *idev)
 {
 	struct sk_buff *skb;
@@ -2191,7 +2236,7 @@  static void mld_send_initial_cr(struct inet6_dev *idev)
 		return;
 
 	skb = NULL;
-	for_each_mc_rtnl(idev, pmc) {
+	for_each_mc_mclock(idev, pmc) {
 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
 			type = MLD2_CHANGE_TO_EXCLUDE;
 		else
@@ -2204,6 +2249,7 @@  static void mld_send_initial_cr(struct inet6_dev *idev)
 
 void ipv6_mc_dad_complete(struct inet6_dev *idev)
 {
+	mutex_lock(&idev->mc_lock);
 	idev->mc_dad_count = idev->mc_qrv;
 	if (idev->mc_dad_count) {
 		mld_send_initial_cr(idev);
@@ -2212,6 +2258,7 @@  void ipv6_mc_dad_complete(struct inet6_dev *idev)
 			mld_dad_start_work(idev,
 					   unsolicited_report_interval(idev));
 	}
+	mutex_unlock(&idev->mc_lock);
 }
 
 static void mld_dad_work(struct work_struct *work)
@@ -2219,8 +2266,7 @@  static void mld_dad_work(struct work_struct *work)
 	struct inet6_dev *idev = container_of(to_delayed_work(work),
 					      struct inet6_dev,
 					      mc_dad_work);
-
-	rtnl_lock();
+	mutex_lock(&idev->mc_lock);
 	mld_send_initial_cr(idev);
 	if (idev->mc_dad_count) {
 		idev->mc_dad_count--;
@@ -2228,10 +2274,11 @@  static void mld_dad_work(struct work_struct *work)
 			mld_dad_start_work(idev,
 					   unsolicited_report_interval(idev));
 	}
-	rtnl_unlock();
+	mutex_unlock(&idev->mc_lock);
 	in6_dev_put(idev);
 }
 
+/* called with mc_lock */
 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
 	const struct in6_addr *psfsrc)
 {
@@ -2239,7 +2286,7 @@  static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
 	int rv = 0;
 
 	psf_prev = NULL;
-	for_each_psf_rtnl(pmc, psf) {
+	for_each_psf_mclock(pmc, psf) {
 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
 			break;
 		psf_prev = psf;
@@ -2255,16 +2302,16 @@  static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
 		/* no more filters for this source */
 		if (psf_prev)
 			rcu_assign_pointer(psf_prev->sf_next,
-					   rtnl_dereference(psf->sf_next));
+					   mc_dereference(psf->sf_next, idev));
 		else
 			rcu_assign_pointer(pmc->mca_sources,
-					   rtnl_dereference(psf->sf_next));
+					   mc_dereference(psf->sf_next, idev));
 
 		if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
 		    !mld_in_v1_mode(idev)) {
 			psf->sf_crcount = idev->mc_qrv;
 			rcu_assign_pointer(psf->sf_next,
-					   rtnl_dereference(pmc->mca_tomb));
+					   mc_dereference(pmc->mca_tomb, idev));
 			rcu_assign_pointer(pmc->mca_tomb, psf);
 			rv = 1;
 		} else {
@@ -2274,6 +2321,7 @@  static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
 	return rv;
 }
 
+/* called with mc_lock */
 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
 			  int delta)
@@ -2285,7 +2333,7 @@  static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
 	if (!idev)
 		return -ENODEV;
 
-	for_each_mc_rtnl(idev, pmc) {
+	for_each_mc_mclock(idev, pmc) {
 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
 			break;
 	}
@@ -2294,9 +2342,8 @@  static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
 
 	sf_markstate(pmc);
 	if (!delta) {
-		if (!pmc->mca_sfcount[sfmode]) {
+		if (!pmc->mca_sfcount[sfmode])
 			return -EINVAL;
-		}
 
 		pmc->mca_sfcount[sfmode]--;
 	}
@@ -2317,16 +2364,19 @@  static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
 		pmc->mca_sfmode = MCAST_INCLUDE;
 		pmc->mca_crcount = idev->mc_qrv;
 		idev->mc_ifc_count = pmc->mca_crcount;
-		for_each_psf_rtnl(pmc, psf)
+		for_each_psf_mclock(pmc, psf)
 			psf->sf_crcount = 0;
 		mld_ifc_event(pmc->idev);
-	} else if (sf_setstate(pmc) || changerec)
+	} else if (sf_setstate(pmc) || changerec) {
 		mld_ifc_event(pmc->idev);
+	}
+
 	return err;
 }
 
 /*
  * Add multicast single-source filter to the interface list
+ * called with mc_lock
  */
 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
 	const struct in6_addr *psfsrc)
@@ -2334,7 +2384,7 @@  static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
 	struct ip6_sf_list *psf, *psf_prev;
 
 	psf_prev = NULL;
-	for_each_psf_rtnl(pmc, psf) {
+	for_each_psf_mclock(pmc, psf) {
 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
 			break;
 		psf_prev = psf;
@@ -2355,12 +2405,13 @@  static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
 	return 0;
 }
 
+/* called with mc_lock */
 static void sf_markstate(struct ifmcaddr6 *pmc)
 {
 	struct ip6_sf_list *psf;
 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
 
-	for_each_psf_rtnl(pmc, psf) {
+	for_each_psf_mclock(pmc, psf) {
 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
 			psf->sf_oldin = mca_xcount ==
 				psf->sf_count[MCAST_EXCLUDE] &&
@@ -2371,6 +2422,7 @@  static void sf_markstate(struct ifmcaddr6 *pmc)
 	}
 }
 
+/* called with mc_lock */
 static int sf_setstate(struct ifmcaddr6 *pmc)
 {
 	struct ip6_sf_list *psf, *dpsf;
@@ -2379,7 +2431,7 @@  static int sf_setstate(struct ifmcaddr6 *pmc)
 	int new_in, rv;
 
 	rv = 0;
-	for_each_psf_rtnl(pmc, psf) {
+	for_each_psf_mclock(pmc, psf) {
 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
 				!psf->sf_count[MCAST_INCLUDE];
@@ -2398,10 +2450,12 @@  static int sf_setstate(struct ifmcaddr6 *pmc)
 				if (dpsf) {
 					if (prev)
 						rcu_assign_pointer(prev->sf_next,
-								   rtnl_dereference(dpsf->sf_next));
+								   mc_dereference(dpsf->sf_next,
+										  pmc->idev));
 					else
 						rcu_assign_pointer(pmc->mca_tomb,
-								   rtnl_dereference(dpsf->sf_next));
+								   mc_dereference(dpsf->sf_next,
+										  pmc->idev));
 					kfree_rcu(dpsf, rcu);
 				}
 				psf->sf_crcount = qrv;
@@ -2424,7 +2478,7 @@  static int sf_setstate(struct ifmcaddr6 *pmc)
 					continue;
 				*dpsf = *psf;
 				rcu_assign_pointer(dpsf->sf_next,
-						   rtnl_dereference(pmc->mca_tomb));
+						   mc_dereference(pmc->mca_tomb, pmc->idev));
 				rcu_assign_pointer(pmc->mca_tomb, dpsf);
 			}
 			dpsf->sf_crcount = qrv;
@@ -2436,6 +2490,7 @@  static int sf_setstate(struct ifmcaddr6 *pmc)
 
 /*
  * Add multicast source filter list to the interface list
+ * called with mc_lock
  */
 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
@@ -2448,7 +2503,7 @@  static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
 	if (!idev)
 		return -ENODEV;
 
-	for_each_mc_rtnl(idev, pmc) {
+	for_each_mc_mclock(idev, pmc) {
 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
 			break;
 	}
@@ -2484,7 +2539,7 @@  static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
 
 		pmc->mca_crcount = idev->mc_qrv;
 		idev->mc_ifc_count = pmc->mca_crcount;
-		for_each_psf_rtnl(pmc, psf)
+		for_each_psf_mclock(pmc, psf)
 			psf->sf_crcount = 0;
 		mld_ifc_event(idev);
 	} else if (sf_setstate(pmc)) {
@@ -2493,21 +2548,22 @@  static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
 	return err;
 }
 
+/* called with mc_lock */
 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
 {
 	struct ip6_sf_list *psf, *nextpsf;
 
-	for (psf = rtnl_dereference(pmc->mca_tomb);
+	for (psf = mc_dereference(pmc->mca_tomb, pmc->idev);
 	     psf;
 	     psf = nextpsf) {
-		nextpsf = rtnl_dereference(psf->sf_next);
+		nextpsf = mc_dereference(psf->sf_next, pmc->idev);
 		kfree_rcu(psf, rcu);
 	}
 	RCU_INIT_POINTER(pmc->mca_tomb, NULL);
-	for (psf = rtnl_dereference(pmc->mca_sources);
+	for (psf = mc_dereference(pmc->mca_sources, pmc->idev);
 	     psf;
 	     psf = nextpsf) {
-		nextpsf = rtnl_dereference(psf->sf_next);
+		nextpsf = mc_dereference(psf->sf_next, pmc->idev);
 		kfree_rcu(psf, rcu);
 	}
 	RCU_INIT_POINTER(pmc->mca_sources, NULL);
@@ -2516,7 +2572,7 @@  static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
 	pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
 }
 
-
+/* called with mc_lock */
 static void igmp6_join_group(struct ifmcaddr6 *ma)
 {
 	unsigned long delay;
@@ -2546,19 +2602,27 @@  static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
 
 	psl = rtnl_dereference(iml->sflist);
 
+	if (idev)
+		mutex_lock(&idev->mc_lock);
+
 	if (!psl) {
 		/* any-source empty exclude case */
 		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
 	} else {
 		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
-				psl->sl_count, psl->sl_addr, 0);
+				     psl->sl_count, psl->sl_addr, 0);
 		RCU_INIT_POINTER(iml->sflist, NULL);
 		atomic_sub(IP6_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
 		kfree_rcu(psl, rcu);
 	}
+
+	if (idev)
+		mutex_unlock(&idev->mc_lock);
+
 	return err;
 }
 
+/* called with mc_lock */
 static void igmp6_leave_group(struct ifmcaddr6 *ma)
 {
 	if (mld_in_v1_mode(ma->idev)) {
@@ -2578,10 +2642,10 @@  static void mld_gq_work(struct work_struct *work)
 					      struct inet6_dev,
 					      mc_gq_work);
 
-	rtnl_lock();
+	mutex_lock(&idev->mc_lock);
 	mld_send_report(idev, NULL);
 	idev->mc_gq_running = 0;
-	rtnl_unlock();
+	mutex_unlock(&idev->mc_lock);
 
 	in6_dev_put(idev);
 }
@@ -2592,7 +2656,7 @@  static void mld_ifc_work(struct work_struct *work)
 					      struct inet6_dev,
 					      mc_ifc_work);
 
-	rtnl_lock();
+	mutex_lock(&idev->mc_lock);
 	mld_send_cr(idev);
 
 	if (idev->mc_ifc_count) {
@@ -2601,10 +2665,11 @@  static void mld_ifc_work(struct work_struct *work)
 			mld_ifc_start_work(idev,
 					   unsolicited_report_interval(idev));
 	}
-	rtnl_unlock();
+	mutex_unlock(&idev->mc_lock);
 	in6_dev_put(idev);
 }
 
+/* called with mc_lock */
 static void mld_ifc_event(struct inet6_dev *idev)
 {
 	if (mld_in_v1_mode(idev))
@@ -2619,14 +2684,14 @@  static void mld_mca_work(struct work_struct *work)
 	struct ifmcaddr6 *ma = container_of(to_delayed_work(work),
 					    struct ifmcaddr6, mca_work);
 
-	rtnl_lock();
+	mutex_lock(&ma->idev->mc_lock);
 	if (mld_in_v1_mode(ma->idev))
 		igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
 	else
 		mld_send_report(ma->idev, ma);
 	ma->mca_flags |=  MAF_LAST_REPORTER;
 	ma->mca_flags &= ~MAF_TIMER_RUNNING;
-	rtnl_unlock();
+	mutex_unlock(&ma->idev->mc_lock);
 
 	ma_put(ma);
 }
@@ -2639,8 +2704,10 @@  void ipv6_mc_unmap(struct inet6_dev *idev)
 
 	/* Install multicast list, except for all-nodes (already installed) */
 
-	for_each_mc_rtnl(idev, i)
+	mutex_lock(&idev->mc_lock);
+	for_each_mc_mclock(idev, i)
 		igmp6_group_dropped(i);
+	mutex_unlock(&idev->mc_lock);
 }
 
 void ipv6_mc_remap(struct inet6_dev *idev)
@@ -2649,14 +2716,15 @@  void ipv6_mc_remap(struct inet6_dev *idev)
 }
 
 /* Device going down */
-
 void ipv6_mc_down(struct inet6_dev *idev)
 {
 	struct ifmcaddr6 *i;
 
+	mutex_lock(&idev->mc_lock);
 	/* Withdraw multicast list */
-	for_each_mc_rtnl(idev, i)
+	for_each_mc_mclock(idev, i)
 		igmp6_group_dropped(i);
+	mutex_unlock(&idev->mc_lock);
 
 	/* Should stop work after group drop. or we will
 	 * start work again in mld_ifc_event()
@@ -2687,10 +2755,12 @@  void ipv6_mc_up(struct inet6_dev *idev)
 	/* Install multicast list, except for all-nodes (already installed) */
 
 	ipv6_mc_reset(idev);
-	for_each_mc_rtnl(idev, i) {
+	mutex_lock(&idev->mc_lock);
+	for_each_mc_mclock(idev, i) {
 		mld_del_delrec(idev, i);
 		igmp6_group_added(i);
 	}
+	mutex_unlock(&idev->mc_lock);
 }
 
 /* IPv6 device initialization. */
@@ -2709,6 +2779,7 @@  void ipv6_mc_init_dev(struct inet6_dev *idev)
 	skb_queue_head_init(&idev->mc_report_queue);
 	spin_lock_init(&idev->mc_query_lock);
 	spin_lock_init(&idev->mc_report_lock);
+	mutex_init(&idev->mc_lock);
 	ipv6_mc_reset(idev);
 }
 
@@ -2722,7 +2793,9 @@  void ipv6_mc_destroy_dev(struct inet6_dev *idev)
 
 	/* Deactivate works */
 	ipv6_mc_down(idev);
+	mutex_lock(&idev->mc_lock);
 	mld_clear_delrec(idev);
+	mutex_unlock(&idev->mc_lock);
 	mld_clear_query(idev);
 	mld_clear_report(idev);
 
@@ -2736,12 +2809,14 @@  void ipv6_mc_destroy_dev(struct inet6_dev *idev)
 	if (idev->cnf.forwarding)
 		__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
 
-	while ((i = rtnl_dereference(idev->mc_list))) {
-		rcu_assign_pointer(idev->mc_list, rtnl_dereference(i->next));
+	mutex_lock(&idev->mc_lock);
+	while ((i = mc_dereference(idev->mc_list, idev))) {
+		rcu_assign_pointer(idev->mc_list, mc_dereference(i->next, idev));
 
 		ip6_mc_clear_src(i);
 		ma_put(i);
 	}
+	mutex_unlock(&idev->mc_lock);
 }
 
 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
@@ -2750,12 +2825,14 @@  static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
 
 	ASSERT_RTNL();
 
+	mutex_lock(&idev->mc_lock);
 	if (mld_in_v1_mode(idev)) {
-		for_each_mc_rtnl(idev, pmc)
+		for_each_mc_mclock(idev, pmc)
 			igmp6_join_group(pmc);
 	} else {
 		mld_send_report(idev, NULL);
 	}
+	mutex_unlock(&idev->mc_lock);
 }
 
 static int ipv6_mc_netdev_event(struct notifier_block *this,