[PATCH net v1] net/ipv6: mcast: fix circular locking dependency in __ipv6_dev_mc_inc()

From: Jiayuan Chen

Date: Tue Mar 17 2026 - 07:16:52 EST


From: Jiayuan Chen <jiayuan.chen@xxxxxxxxxx>

syzbot reported a possible circular locking dependency:

fs_reclaim --> sk_lock-AF_INET6 --> &idev->mc_lock

CPU0 CPU1
---- ----
lock(&idev->mc_lock)
lock(sk_lock-AF_INET6)
lock(&idev->mc_lock) // blocked
kzalloc(GFP_KERNEL)
fs_reclaim
...nbd I/O...
sk_lock-AF_INET6 // blocked -> DEADLOCK

__ipv6_dev_mc_inc() does GFP_KERNEL allocation inside mc_lock via
mca_alloc(). This can enter memory reclaim, which through nbd block
I/O may need sk_lock-AF_INET6. But sk_lock -> mc_lock already exists
via setsockopt -> __ipv6_sock_mc_join, so we have a deadlock.

Before commit 63ed8de4be81 ("mld: add mc_lock for protecting
per-interface mld data"), only RTNL was held during the allocation.
The lock ordering was always RTNL -> sk_lock (the nbd path doesn't
involve RTNL), so there was no circular dependency.

Split mca_alloc() into mca_alloc() + mca_init(): mca_alloc() does the
GFP_KERNEL allocation before mc_lock, mca_init() initializes under
mc_lock. If the address already exists, the pre-allocated memory is
simply freed. Also move inet6_ifmcaddr_notify() outside mc_lock since
it also does GFP_KERNEL allocation.

Fixes: 63ed8de4be81 ("mld: add mc_lock for protecting per-interface mld data")
Reported-by: syzbot+afbcf622635e98bf40d2@xxxxxxxxxxxxxxxxxxxxxxxxx
Closes: https://lore.kernel.org/all/69b7dc76.050a0220.248e02.0113.GAE@xxxxxxxxxx/T/
Cc: Jiayuan Chen <jiayuan.chen@xxxxxxxxx>
Signed-off-by: Jiayuan Chen <jiayuan.chen@xxxxxxxxxx>
---
net/ipv6/mcast.c | 36 +++++++++++++++++++-----------------
1 file changed, 19 insertions(+), 17 deletions(-)

diff --git a/net/ipv6/mcast.c b/net/ipv6/mcast.c
index 3330adcf26db..2dfa7ed54d17 100644
--- a/net/ipv6/mcast.c
+++ b/net/ipv6/mcast.c
@@ -860,18 +860,16 @@ static void ma_put(struct ifmcaddr6 *mc)
}
}

-static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
- const struct in6_addr *addr,
- unsigned int mode)
+static struct ifmcaddr6 *mca_alloc(void)
{
- struct ifmcaddr6 *mc;
+ return kzalloc_obj(struct ifmcaddr6);
+}

+static void mca_init(struct inet6_dev *idev, const struct in6_addr *addr,
+ unsigned int mode, struct ifmcaddr6 *mc)
+{
mc_assert_locked(idev);

- mc = kzalloc_obj(*mc);
- if (!mc)
- return NULL;
-
INIT_DELAYED_WORK(&mc->mca_work, mld_mca_work);

mc->mca_addr = *addr;
@@ -887,8 +885,6 @@ static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
mc->mca_flags |= MAF_NOREPORT;
-
- return mc;
}

static void inet6_ifmcaddr_notify(struct net_device *dev,
@@ -932,6 +928,7 @@ static void inet6_ifmcaddr_notify(struct net_device *dev,
static int __ipv6_dev_mc_inc(struct net_device *dev,
const struct in6_addr *addr, unsigned int mode)
{
+ struct ifmcaddr6 *mc_alloced;
struct inet6_dev *idev;
struct ifmcaddr6 *mc;

@@ -940,10 +937,17 @@ static int __ipv6_dev_mc_inc(struct net_device *dev,
if (!idev)
return -EINVAL;

+ mc_alloced = mca_alloc();
+ if (!mc_alloced) {
+ in6_dev_put(idev);
+ return -ENOMEM;
+ }
+
mutex_lock(&idev->mc_lock);

if (READ_ONCE(idev->dead)) {
mutex_unlock(&idev->mc_lock);
+ kfree(mc_alloced);
in6_dev_put(idev);
return -ENODEV;
}
@@ -953,26 +957,24 @@ static int __ipv6_dev_mc_inc(struct net_device *dev,
mc->mca_users++;
ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
mutex_unlock(&idev->mc_lock);
+ kfree(mc_alloced);
in6_dev_put(idev);
return 0;
}
}

- mc = mca_alloc(idev, addr, mode);
- if (!mc) {
- mutex_unlock(&idev->mc_lock);
- in6_dev_put(idev);
- return -ENOMEM;
- }
+ mca_init(idev, addr, mode, mc_alloced);
+ mc = mc_alloced;

rcu_assign_pointer(mc->next, idev->mc_list);
rcu_assign_pointer(idev->mc_list, mc);

mld_del_delrec(idev, mc);
igmp6_group_added(mc);
- inet6_ifmcaddr_notify(dev, mc, RTM_NEWMULTICAST);
mutex_unlock(&idev->mc_lock);

+ inet6_ifmcaddr_notify(dev, mc, RTM_NEWMULTICAST);
+
return 0;
}

--
2.43.0