[PATCH 6.6.y 1/3] net: dst: add four helpers to annotate data-races around dst->dev

[PATCH 6.6.y 1/3] net: dst: add four helpers to annotate data-races around dst->dev

[Ruohan Lan]

From: Eric Dumazet <edumazet@google.com>

[ Upstream commit 88fe14253e181878c2ddb51a298ae8c468a63010 ]

dst->dev is read locklessly in many contexts,
and written in dst_dev_put().

Fixing all the races is going to need many changes.

We probably will have to add full RCU protection.

Add three helpers to ease this painful process.

static inline struct net_device *dst_dev(const struct dst_entry *dst)
{
       return READ_ONCE(dst->dev);
}

static inline struct net_device *skb_dst_dev(const struct sk_buff *skb)
{
       return dst_dev(skb_dst(skb));
}

static inline struct net *skb_dst_dev_net(const struct sk_buff *skb)
{
       return dev_net(skb_dst_dev(skb));
}

static inline struct net *skb_dst_dev_net_rcu(const struct sk_buff *skb)
{
       return dev_net_rcu(skb_dst_dev(skb));
}

Fixes: 4a6ce2b6f2ec ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20250630121934.3399505-7-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ Minor context conflict resolved. ]
Signed-off-by: Ruohan Lan <ruohanlan@aliyun.com>
---
 include/net/dst.h | 20 ++++++++++++++++++++
 net/core/dst.c    |  4 ++--
 net/core/sock.c   |  8 ++++----
 3 files changed, 26 insertions(+), 6 deletions(-)

diff --git a/include/net/dst.h b/include/net/dst.h
index 55d1be268d24..ea3b050f8b38 100644
--- a/include/net/dst.h
+++ b/include/net/dst.h
@@ -581,6 +581,26 @@ static inline struct net_device *skb_dst_dev_rcu(const struct sk_buff *skb)
 	return dst_dev_rcu(skb_dst(skb));
 }
 
+static inline struct net_device *dst_dev(const struct dst_entry *dst)
+{
+	return READ_ONCE(dst->dev);
+}
+
+static inline struct net_device *skb_dst_dev(const struct sk_buff *skb)
+{
+	return dst_dev(skb_dst(skb));
+}
+
+static inline struct net *skb_dst_dev_net(const struct sk_buff *skb)
+{
+	return dev_net(skb_dst_dev(skb));
+}
+
+static inline struct net *skb_dst_dev_net_rcu(const struct sk_buff *skb)
+{
+	return dev_net_rcu(skb_dst_dev(skb));
+}
+
 struct dst_entry *dst_blackhole_check(struct dst_entry *dst, u32 cookie);
 void dst_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 			       struct sk_buff *skb, u32 mtu, bool confirm_neigh);
diff --git a/net/core/dst.c b/net/core/dst.c
index 2513665696f6..ac67706e5f87 100644
--- a/net/core/dst.c
+++ b/net/core/dst.c
@@ -152,7 +152,7 @@ void dst_dev_put(struct dst_entry *dst)
 		dst->ops->ifdown(dst, dev);
 	WRITE_ONCE(dst->input, dst_discard);
 	WRITE_ONCE(dst->output, dst_discard_out);
-	dst->dev = blackhole_netdev;
+	WRITE_ONCE(dst->dev, blackhole_netdev);
 	netdev_ref_replace(dev, blackhole_netdev, &dst->dev_tracker,
 			   GFP_ATOMIC);
 }
@@ -265,7 +265,7 @@ unsigned int dst_blackhole_mtu(const struct dst_entry *dst)
 {
 	unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 
-	return mtu ? : dst->dev->mtu;
+	return mtu ? : dst_dev(dst)->mtu;
 }
 EXPORT_SYMBOL_GPL(dst_blackhole_mtu);
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 8e4c87a39dc8..0e52847c57f8 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -2450,8 +2450,8 @@ static u32 sk_dst_gso_max_size(struct sock *sk, struct dst_entry *dst)
 		   !ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr));
 #endif
 	/* pairs with the WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */
-	max_size = is_ipv6 ? READ_ONCE(dst->dev->gso_max_size) :
-			READ_ONCE(dst->dev->gso_ipv4_max_size);
+	max_size = is_ipv6 ? READ_ONCE(dst_dev(dst)->gso_max_size) :
+			READ_ONCE(dst_dev(dst)->gso_ipv4_max_size);
 	if (max_size > GSO_LEGACY_MAX_SIZE && !sk_is_tcp(sk))
 		max_size = GSO_LEGACY_MAX_SIZE;
 
@@ -2462,7 +2462,7 @@ void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
 {
 	u32 max_segs = 1;
 
-	sk->sk_route_caps = dst->dev->features;
+	sk->sk_route_caps = dst_dev(dst)->features;
 	if (sk_is_tcp(sk))
 		sk->sk_route_caps |= NETIF_F_GSO;
 	if (sk->sk_route_caps & NETIF_F_GSO)
@@ -2476,7 +2476,7 @@ void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
 			sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
 			sk->sk_gso_max_size = sk_dst_gso_max_size(sk, dst);
 			/* pairs with the WRITE_ONCE() in netif_set_gso_max_segs() */
-			max_segs = max_t(u32, READ_ONCE(dst->dev->gso_max_segs), 1);
+			max_segs = max_t(u32, READ_ONCE(dst_dev(dst)->gso_max_segs), 1);
 		}
 	}
 	sk->sk_gso_max_segs = max_segs;
-- 
2.43.0

FAILED: patch "[PATCH] ip: orphan prefetched skbs before multicast forwarding" failed to apply to 6.6-stable tree

[gregkh]

The patch below does not apply to the 6.6-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable@vger.kernel.org>.

To reproduce the conflict and resubmit, you may use the following commands:

git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.6.y
git checkout FETCH_HEAD
git cherry-pick -x e36ce6e78fe3fc3c071a26750783b7ba081ce10d
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<stable@vger.kernel.org>' --in-reply-to '2026090318-pusher-lucid-87ba@gregkh' --subject-prefix 'PATCH 6.6.y' 'HEAD^..'

Possible dependencies:



thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From e36ce6e78fe3fc3c071a26750783b7ba081ce10d Mon Sep 17 00:00:00 2001
From: Zhiling Zou <zhilinz@nebusec.ai>
Date: Thu, 13 Aug 2026 00:36:38 +0800
Subject: [PATCH] ip: orphan prefetched skbs before multicast forwarding

IPv4 and IPv6 input preserve an skb->sk association installed by
bpf_sk_assign() so that local delivery can use the selected socket under
RCU. Both address families can also prefetch a socket in UDP early demux.
In both paths (BPF and UDP early demux) a reference is not guaranteed to
be held on the socket.

When a multicast packet is not locally deliverable, IPv6 hands the
original skb to ip6_mr_input(). IPv4's ip_mr_input() similarly keeps the
original skb when local delivery is not needed. Either path can put the
skb on an unresolved multicast route queue or forward it after the
receive-side RCU section ends.

After the prefetched socket is destroyed, a later skb free invokes
sock_pfree() and dereferences the stale skb->sk. Orphan the skb before
each non-local multicast forwarding path. Local delivery retains the
original skb; the existing skb_clone() calls provide multicast forwarding
with a socket-free clone.

Fixes: cf7fbe660f2d ("bpf: Add socket assign support")
Fixes: 08842c43d016 ("udp: no longer touch sk->sk_refcnt in early demux")
Cc: stable@vger.kernel.org
Reported-by: Vega <vega@nebusec.ai>
Signed-off-by: Zhiling Zou <zhilinz@nebusec.ai>
Reported-by: Vega <vega@nebusec.ai>
Signed-off-by: Zhiling Zou <zhilinz@nebusec.ai>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/0c52eb3d7532aaf8bccf37e0f7c922143c639735.1786552223.git.zhilinz@nebusec.ai
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

diff --git a/net/ipv4/ipmr.c b/net/ipv4/ipmr.c
index 1d9a4ac14fce..e5f2b1c6150d 100644
--- a/net/ipv4/ipmr.c
+++ b/net/ipv4/ipmr.c
@@ -2213,6 +2213,9 @@ int ip_mr_input(struct sk_buff *skb)
 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
 		goto dont_forward;
 
+	if (!local)
+		skb_orphan(skb);
+
 	mrt = ipmr_rt_fib_lookup(net, skb);
 	if (IS_ERR(mrt)) {
 		kfree_skb(skb);
diff --git a/net/ipv6/ip6_input.c b/net/ipv6/ip6_input.c
index 8972863c93ee..d332ec60f915 100644
--- a/net/ipv6/ip6_input.c
+++ b/net/ipv6/ip6_input.c
@@ -622,6 +622,7 @@ int ip6_mc_input(struct sk_buff *skb)
 		if (deliver) {
 			skb2 = skb_clone(skb, GFP_ATOMIC);
 		} else {
+			skb_orphan(skb);
 			skb2 = skb;
 			skb = NULL;
 		}

[PATCH 6.6.y 1/3] net: dst: add four helpers to annotate data-races around dst->dev

[Sasha Levin]

From: Eric Dumazet <edumazet@google.com>

[ Upstream commit 88fe14253e181878c2ddb51a298ae8c468a63010 ]

dst->dev is read locklessly in many contexts,
and written in dst_dev_put().

Fixing all the races is going to need many changes.

We probably will have to add full RCU protection.

Add three helpers to ease this painful process.

static inline struct net_device *dst_dev(const struct dst_entry *dst)
{
       return READ_ONCE(dst->dev);
}

static inline struct net_device *skb_dst_dev(const struct sk_buff *skb)
{
       return dst_dev(skb_dst(skb));
}

static inline struct net *skb_dst_dev_net(const struct sk_buff *skb)
{
       return dev_net(skb_dst_dev(skb));
}

static inline struct net *skb_dst_dev_net_rcu(const struct sk_buff *skb)
{
       return dev_net_rcu(skb_dst_dev(skb));
}

Fixes: 4a6ce2b6f2ec ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20250630121934.3399505-7-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: e36ce6e78fe3 ("ip: orphan prefetched skbs before multicast forwarding")
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 include/net/dst.h    | 20 ++++++++++++++++++++
 net/core/dst.c       |  4 ++--
 net/core/sock.c      |  8 ++++----
 net/ipv6/ip6_input.c |  4 ++--
 4 files changed, 28 insertions(+), 8 deletions(-)

diff --git a/include/net/dst.h b/include/net/dst.h
index 55d1be268d243..1983945dfcdf1 100644
--- a/include/net/dst.h
+++ b/include/net/dst.h
@@ -569,6 +569,11 @@ static inline void skb_dst_update_pmtu_no_confirm(struct sk_buff *skb, u32 mtu)
 		dst->ops->update_pmtu(dst, NULL, skb, mtu, false);
 }
 
+static inline struct net_device *dst_dev(const struct dst_entry *dst)
+{
+	return READ_ONCE(dst->dev);
+}
+
 static inline struct net_device *dst_dev_rcu(const struct dst_entry *dst)
 {
 	/* In the future, use rcu_dereference(dst->dev) */
@@ -576,11 +581,26 @@ static inline struct net_device *dst_dev_rcu(const struct dst_entry *dst)
 	return READ_ONCE(dst->dev);
 }
 
+static inline struct net_device *skb_dst_dev(const struct sk_buff *skb)
+{
+	return dst_dev(skb_dst(skb));
+}
+
 static inline struct net_device *skb_dst_dev_rcu(const struct sk_buff *skb)
 {
 	return dst_dev_rcu(skb_dst(skb));
 }
 
+static inline struct net *skb_dst_dev_net(const struct sk_buff *skb)
+{
+	return dev_net(skb_dst_dev(skb));
+}
+
+static inline struct net *skb_dst_dev_net_rcu(const struct sk_buff *skb)
+{
+	return dev_net_rcu(skb_dst_dev(skb));
+}
+
 struct dst_entry *dst_blackhole_check(struct dst_entry *dst, u32 cookie);
 void dst_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 			       struct sk_buff *skb, u32 mtu, bool confirm_neigh);
diff --git a/net/core/dst.c b/net/core/dst.c
index 5ed8cb10748f5..f65e6be7e2571 100644
--- a/net/core/dst.c
+++ b/net/core/dst.c
@@ -153,7 +153,7 @@ void dst_dev_put(struct dst_entry *dst)
 		dst->ops->ifdown(dst, dev);
 	WRITE_ONCE(dst->input, dst_discard);
 	WRITE_ONCE(dst->output, dst_discard_out);
-	dst->dev = blackhole_netdev;
+	WRITE_ONCE(dst->dev, blackhole_netdev);
 	netdev_ref_replace(dev, blackhole_netdev, &dst->dev_tracker,
 			   GFP_ATOMIC);
 }
@@ -266,7 +266,7 @@ unsigned int dst_blackhole_mtu(const struct dst_entry *dst)
 {
 	unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 
-	return mtu ? : dst->dev->mtu;
+	return mtu ? : dst_dev(dst)->mtu;
 }
 EXPORT_SYMBOL_GPL(dst_blackhole_mtu);
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 846e95805c199..2f82dc45f93ec 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -2454,8 +2454,8 @@ static u32 sk_dst_gso_max_size(struct sock *sk, struct dst_entry *dst)
 		   !ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr));
 #endif
 	/* pairs with the WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */
-	max_size = is_ipv6 ? READ_ONCE(dst->dev->gso_max_size) :
-			READ_ONCE(dst->dev->gso_ipv4_max_size);
+	max_size = is_ipv6 ? READ_ONCE(dst_dev(dst)->gso_max_size) :
+			READ_ONCE(dst_dev(dst)->gso_ipv4_max_size);
 	if (max_size > GSO_LEGACY_MAX_SIZE && !sk_is_tcp(sk))
 		max_size = GSO_LEGACY_MAX_SIZE;
 
@@ -2466,7 +2466,7 @@ void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
 {
 	u32 max_segs = 1;
 
-	sk->sk_route_caps = dst->dev->features;
+	sk->sk_route_caps = dst_dev(dst)->features;
 	if (sk_is_tcp(sk))
 		sk->sk_route_caps |= NETIF_F_GSO;
 	if (sk->sk_route_caps & NETIF_F_GSO)
@@ -2480,7 +2480,7 @@ void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
 			sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
 			sk->sk_gso_max_size = sk_dst_gso_max_size(sk, dst);
 			/* pairs with the WRITE_ONCE() in netif_set_gso_max_segs() */
-			max_segs = max_t(u32, READ_ONCE(dst->dev->gso_max_segs), 1);
+			max_segs = max_t(u32, READ_ONCE(dst_dev(dst)->gso_max_segs), 1);
 		}
 	}
 	sk->sk_gso_max_segs = max_segs;
diff --git a/net/ipv6/ip6_input.c b/net/ipv6/ip6_input.c
index 133610a49da6b..54146e280c2b7 100644
--- a/net/ipv6/ip6_input.c
+++ b/net/ipv6/ip6_input.c
@@ -569,9 +569,9 @@ int ip6_mc_input(struct sk_buff *skb)
 			/* unknown RA - process it normally */
 		}
 
-		if (deliver)
+		if (deliver) {
 			skb2 = skb_clone(skb, GFP_ATOMIC);
-		else {
+		} else {
 			skb2 = skb;
 			skb = NULL;
 		}
-- 
2.53.0

[PATCH 6.6.y 2/3] ipv6: ip6_mc_input() and ip6_mr_input() cleanups

[Sasha Levin]

From: Eric Dumazet <edumazet@google.com>

[ Upstream commit 46a94e44b9ec4d8d20108a1748c410d71a8dc759 ]

Both functions are always called under RCU.

We remove the extra rcu_read_lock()/rcu_read_unlock().

We use skb_dst_dev_net_rcu() instead of skb_dst_dev_net().

We use dev_net_rcu() instead of dev_net().

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20250630121934.3399505-11-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: e36ce6e78fe3 ("ip: orphan prefetched skbs before multicast forwarding")
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 net/ipv6/ip6_input.c | 25 +++++++++----------------
 net/ipv6/ip6mr.c     |  9 ++++-----
 2 files changed, 13 insertions(+), 21 deletions(-)

diff --git a/net/ipv6/ip6_input.c b/net/ipv6/ip6_input.c
index 54146e280c2b7..e775a5a766692 100644
--- a/net/ipv6/ip6_input.c
+++ b/net/ipv6/ip6_input.c
@@ -497,38 +497,32 @@ EXPORT_SYMBOL_GPL(ip6_input);
 
 int ip6_mc_input(struct sk_buff *skb)
 {
+	struct net_device *dev = skb->dev;
 	int sdif = inet6_sdif(skb);
 	const struct ipv6hdr *hdr;
-	struct net_device *dev;
 	bool deliver;
 
-	__IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
-			 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
-			 skb->len);
+	__IP6_UPD_PO_STATS(skb_dst_dev_net_rcu(skb),
+			   __in6_dev_get_safely(dev), IPSTATS_MIB_INMCAST,
+			   skb->len);
 
 	/* skb->dev passed may be master dev for vrfs. */
 	if (sdif) {
-		rcu_read_lock();
-		dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif);
+		dev = dev_get_by_index_rcu(dev_net_rcu(dev), sdif);
 		if (!dev) {
-			rcu_read_unlock();
 			kfree_skb(skb);
 			return -ENODEV;
 		}
-	} else {
-		dev = skb->dev;
 	}
 
 	hdr = ipv6_hdr(skb);
 	deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
-	if (sdif)
-		rcu_read_unlock();
 
 #ifdef CONFIG_IPV6_MROUTE
 	/*
 	 *      IPv6 multicast router mode is now supported ;)
 	 */
-	if (atomic_read(&dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding) &&
+	if (atomic_read(&dev_net_rcu(skb->dev)->ipv6.devconf_all->mc_forwarding) &&
 	    !(ipv6_addr_type(&hdr->daddr) &
 	      (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
 	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
@@ -576,15 +570,14 @@ int ip6_mc_input(struct sk_buff *skb)
 			skb = NULL;
 		}
 
-		if (skb2) {
+		if (skb2)
 			ip6_mr_input(skb2);
-		}
 	}
 out:
 #endif
-	if (likely(deliver))
+	if (likely(deliver)) {
 		ip6_input(skb);
-	else {
+	} else {
 		/* discard */
 		kfree_skb(skb);
 	}
diff --git a/net/ipv6/ip6mr.c b/net/ipv6/ip6mr.c
index 18451dfeae682..cd2116862a4b4 100644
--- a/net/ipv6/ip6mr.c
+++ b/net/ipv6/ip6mr.c
@@ -2225,21 +2225,20 @@ static void ip6_mr_forward(struct net *net, struct mr_table *mrt,
 
 int ip6_mr_input(struct sk_buff *skb)
 {
+	struct net_device *dev = skb->dev;
+	struct net *net = dev_net_rcu(dev);
 	struct mfc6_cache *cache;
-	struct net *net = dev_net(skb->dev);
 	struct mr_table *mrt;
 	struct flowi6 fl6 = {
-		.flowi6_iif	= skb->dev->ifindex,
+		.flowi6_iif	= dev->ifindex,
 		.flowi6_mark	= skb->mark,
 	};
 	int err;
-	struct net_device *dev;
 
 	/* skb->dev passed in is the master dev for vrfs.
 	 * Get the proper interface that does have a vif associated with it.
 	 */
-	dev = skb->dev;
-	if (netif_is_l3_master(skb->dev)) {
+	if (netif_is_l3_master(dev)) {
 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
 		if (!dev) {
 			kfree_skb(skb);
-- 
2.53.0

[PATCH 6.6.y 1/3] net: dst: add four helpers to annotate data-races around dst->dev

[Sasha Levin]

From: Eric Dumazet <edumazet@google.com>

[ Upstream commit 88fe14253e181878c2ddb51a298ae8c468a63010 ]

dst->dev is read locklessly in many contexts,
and written in dst_dev_put().

Fixing all the races is going to need many changes.

We probably will have to add full RCU protection.

Add three helpers to ease this painful process.

static inline struct net_device *dst_dev(const struct dst_entry *dst)
{
       return READ_ONCE(dst->dev);
}

static inline struct net_device *skb_dst_dev(const struct sk_buff *skb)
{
       return dst_dev(skb_dst(skb));
}

static inline struct net *skb_dst_dev_net(const struct sk_buff *skb)
{
       return dev_net(skb_dst_dev(skb));
}

static inline struct net *skb_dst_dev_net_rcu(const struct sk_buff *skb)
{
       return dev_net_rcu(skb_dst_dev(skb));
}

Fixes: 4a6ce2b6f2ec ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20250630121934.3399505-7-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: e36ce6e78fe3 ("ip: orphan prefetched skbs before multicast forwarding")
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 include/net/dst.h    | 20 ++++++++++++++++++++
 net/core/dst.c       |  4 ++--
 net/core/sock.c      |  8 ++++----
 net/ipv6/ip6_input.c |  4 ++--
 4 files changed, 28 insertions(+), 8 deletions(-)

diff --git a/include/net/dst.h b/include/net/dst.h
index 55d1be268d243..1983945dfcdf1 100644
--- a/include/net/dst.h
+++ b/include/net/dst.h
@@ -569,6 +569,11 @@ static inline void skb_dst_update_pmtu_no_confirm(struct sk_buff *skb, u32 mtu)
 		dst->ops->update_pmtu(dst, NULL, skb, mtu, false);
 }
 
+static inline struct net_device *dst_dev(const struct dst_entry *dst)
+{
+	return READ_ONCE(dst->dev);
+}
+
 static inline struct net_device *dst_dev_rcu(const struct dst_entry *dst)
 {
 	/* In the future, use rcu_dereference(dst->dev) */
@@ -576,11 +581,26 @@ static inline struct net_device *dst_dev_rcu(const struct dst_entry *dst)
 	return READ_ONCE(dst->dev);
 }
 
+static inline struct net_device *skb_dst_dev(const struct sk_buff *skb)
+{
+	return dst_dev(skb_dst(skb));
+}
+
 static inline struct net_device *skb_dst_dev_rcu(const struct sk_buff *skb)
 {
 	return dst_dev_rcu(skb_dst(skb));
 }
 
+static inline struct net *skb_dst_dev_net(const struct sk_buff *skb)
+{
+	return dev_net(skb_dst_dev(skb));
+}
+
+static inline struct net *skb_dst_dev_net_rcu(const struct sk_buff *skb)
+{
+	return dev_net_rcu(skb_dst_dev(skb));
+}
+
 struct dst_entry *dst_blackhole_check(struct dst_entry *dst, u32 cookie);
 void dst_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 			       struct sk_buff *skb, u32 mtu, bool confirm_neigh);
diff --git a/net/core/dst.c b/net/core/dst.c
index 5ed8cb10748f5..f65e6be7e2571 100644
--- a/net/core/dst.c
+++ b/net/core/dst.c
@@ -153,7 +153,7 @@ void dst_dev_put(struct dst_entry *dst)
 		dst->ops->ifdown(dst, dev);
 	WRITE_ONCE(dst->input, dst_discard);
 	WRITE_ONCE(dst->output, dst_discard_out);
-	dst->dev = blackhole_netdev;
+	WRITE_ONCE(dst->dev, blackhole_netdev);
 	netdev_ref_replace(dev, blackhole_netdev, &dst->dev_tracker,
 			   GFP_ATOMIC);
 }
@@ -266,7 +266,7 @@ unsigned int dst_blackhole_mtu(const struct dst_entry *dst)
 {
 	unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 
-	return mtu ? : dst->dev->mtu;
+	return mtu ? : dst_dev(dst)->mtu;
 }
 EXPORT_SYMBOL_GPL(dst_blackhole_mtu);
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 846e95805c199..2f82dc45f93ec 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -2454,8 +2454,8 @@ static u32 sk_dst_gso_max_size(struct sock *sk, struct dst_entry *dst)
 		   !ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr));
 #endif
 	/* pairs with the WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */
-	max_size = is_ipv6 ? READ_ONCE(dst->dev->gso_max_size) :
-			READ_ONCE(dst->dev->gso_ipv4_max_size);
+	max_size = is_ipv6 ? READ_ONCE(dst_dev(dst)->gso_max_size) :
+			READ_ONCE(dst_dev(dst)->gso_ipv4_max_size);
 	if (max_size > GSO_LEGACY_MAX_SIZE && !sk_is_tcp(sk))
 		max_size = GSO_LEGACY_MAX_SIZE;
 
@@ -2466,7 +2466,7 @@ void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
 {
 	u32 max_segs = 1;
 
-	sk->sk_route_caps = dst->dev->features;
+	sk->sk_route_caps = dst_dev(dst)->features;
 	if (sk_is_tcp(sk))
 		sk->sk_route_caps |= NETIF_F_GSO;
 	if (sk->sk_route_caps & NETIF_F_GSO)
@@ -2480,7 +2480,7 @@ void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
 			sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
 			sk->sk_gso_max_size = sk_dst_gso_max_size(sk, dst);
 			/* pairs with the WRITE_ONCE() in netif_set_gso_max_segs() */
-			max_segs = max_t(u32, READ_ONCE(dst->dev->gso_max_segs), 1);
+			max_segs = max_t(u32, READ_ONCE(dst_dev(dst)->gso_max_segs), 1);
 		}
 	}
 	sk->sk_gso_max_segs = max_segs;
diff --git a/net/ipv6/ip6_input.c b/net/ipv6/ip6_input.c
index 133610a49da6b..54146e280c2b7 100644
--- a/net/ipv6/ip6_input.c
+++ b/net/ipv6/ip6_input.c
@@ -569,9 +569,9 @@ int ip6_mc_input(struct sk_buff *skb)
 			/* unknown RA - process it normally */
 		}
 
-		if (deliver)
+		if (deliver) {
 			skb2 = skb_clone(skb, GFP_ATOMIC);
-		else {
+		} else {
 			skb2 = skb;
 			skb = NULL;
 		}
-- 
2.53.0

[PATCH 6.6.y 3/3] ip: orphan prefetched skbs before multicast forwarding

[Sasha Levin]

From: Zhiling Zou <zhilinz@nebusec.ai>

[ Upstream commit e36ce6e78fe3fc3c071a26750783b7ba081ce10d ]

IPv4 and IPv6 input preserve an skb->sk association installed by
bpf_sk_assign() so that local delivery can use the selected socket under
RCU. Both address families can also prefetch a socket in UDP early demux.
In both paths (BPF and UDP early demux) a reference is not guaranteed to
be held on the socket.

When a multicast packet is not locally deliverable, IPv6 hands the
original skb to ip6_mr_input(). IPv4's ip_mr_input() similarly keeps the
original skb when local delivery is not needed. Either path can put the
skb on an unresolved multicast route queue or forward it after the
receive-side RCU section ends.

After the prefetched socket is destroyed, a later skb free invokes
sock_pfree() and dereferences the stale skb->sk. Orphan the skb before
each non-local multicast forwarding path. Local delivery retains the
original skb; the existing skb_clone() calls provide multicast forwarding
with a socket-free clone.

Fixes: cf7fbe660f2d ("bpf: Add socket assign support")
Fixes: 08842c43d016 ("udp: no longer touch sk->sk_refcnt in early demux")
Cc: stable@vger.kernel.org
Reported-by: Vega <vega@nebusec.ai>
Signed-off-by: Zhiling Zou <zhilinz@nebusec.ai>
Reported-by: Vega <vega@nebusec.ai>
Signed-off-by: Zhiling Zou <zhilinz@nebusec.ai>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/0c52eb3d7532aaf8bccf37e0f7c922143c639735.1786552223.git.zhilinz@nebusec.ai
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 net/ipv4/ipmr.c      | 3 +++
 net/ipv6/ip6_input.c | 1 +
 2 files changed, 4 insertions(+)

diff --git a/net/ipv4/ipmr.c b/net/ipv4/ipmr.c
index af9412a507cf3..93d08e343a256 100644
--- a/net/ipv4/ipmr.c
+++ b/net/ipv4/ipmr.c
@@ -2139,6 +2139,9 @@ int ip_mr_input(struct sk_buff *skb)
 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
 		goto dont_forward;
 
+	if (!local)
+		skb_orphan(skb);
+
 	mrt = ipmr_rt_fib_lookup(net, skb);
 	if (IS_ERR(mrt)) {
 		kfree_skb(skb);
diff --git a/net/ipv6/ip6_input.c b/net/ipv6/ip6_input.c
index e775a5a766692..9b6eaae0c6c80 100644
--- a/net/ipv6/ip6_input.c
+++ b/net/ipv6/ip6_input.c
@@ -566,6 +566,7 @@ int ip6_mc_input(struct sk_buff *skb)
 		if (deliver) {
 			skb2 = skb_clone(skb, GFP_ATOMIC);
 		} else {
+			skb_orphan(skb);
 			skb2 = skb;
 			skb = NULL;
 		}
-- 
2.53.0

[PATCH 6.6.y 2/3] ipv6: ip6_mc_input() and ip6_mr_input() cleanups

[Sasha Levin]

From: Eric Dumazet <edumazet@google.com>

[ Upstream commit 46a94e44b9ec4d8d20108a1748c410d71a8dc759 ]

Both functions are always called under RCU.

We remove the extra rcu_read_lock()/rcu_read_unlock().

We use skb_dst_dev_net_rcu() instead of skb_dst_dev_net().

We use dev_net_rcu() instead of dev_net().

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20250630121934.3399505-11-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: e36ce6e78fe3 ("ip: orphan prefetched skbs before multicast forwarding")
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 net/ipv6/ip6_input.c | 25 +++++++++----------------
 net/ipv6/ip6mr.c     |  9 ++++-----
 2 files changed, 13 insertions(+), 21 deletions(-)

diff --git a/net/ipv6/ip6_input.c b/net/ipv6/ip6_input.c
index 54146e280c2b7..e775a5a766692 100644
--- a/net/ipv6/ip6_input.c
+++ b/net/ipv6/ip6_input.c
@@ -497,38 +497,32 @@ EXPORT_SYMBOL_GPL(ip6_input);
 
 int ip6_mc_input(struct sk_buff *skb)
 {
+	struct net_device *dev = skb->dev;
 	int sdif = inet6_sdif(skb);
 	const struct ipv6hdr *hdr;
-	struct net_device *dev;
 	bool deliver;
 
-	__IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
-			 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST,
-			 skb->len);
+	__IP6_UPD_PO_STATS(skb_dst_dev_net_rcu(skb),
+			   __in6_dev_get_safely(dev), IPSTATS_MIB_INMCAST,
+			   skb->len);
 
 	/* skb->dev passed may be master dev for vrfs. */
 	if (sdif) {
-		rcu_read_lock();
-		dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif);
+		dev = dev_get_by_index_rcu(dev_net_rcu(dev), sdif);
 		if (!dev) {
-			rcu_read_unlock();
 			kfree_skb(skb);
 			return -ENODEV;
 		}
-	} else {
-		dev = skb->dev;
 	}
 
 	hdr = ipv6_hdr(skb);
 	deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL);
-	if (sdif)
-		rcu_read_unlock();
 
 #ifdef CONFIG_IPV6_MROUTE
 	/*
 	 *      IPv6 multicast router mode is now supported ;)
 	 */
-	if (atomic_read(&dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding) &&
+	if (atomic_read(&dev_net_rcu(skb->dev)->ipv6.devconf_all->mc_forwarding) &&
 	    !(ipv6_addr_type(&hdr->daddr) &
 	      (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
 	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
@@ -576,15 +570,14 @@ int ip6_mc_input(struct sk_buff *skb)
 			skb = NULL;
 		}
 
-		if (skb2) {
+		if (skb2)
 			ip6_mr_input(skb2);
-		}
 	}
 out:
 #endif
-	if (likely(deliver))
+	if (likely(deliver)) {
 		ip6_input(skb);
-	else {
+	} else {
 		/* discard */
 		kfree_skb(skb);
 	}
diff --git a/net/ipv6/ip6mr.c b/net/ipv6/ip6mr.c
index 18451dfeae682..cd2116862a4b4 100644
--- a/net/ipv6/ip6mr.c
+++ b/net/ipv6/ip6mr.c
@@ -2225,21 +2225,20 @@ static void ip6_mr_forward(struct net *net, struct mr_table *mrt,
 
 int ip6_mr_input(struct sk_buff *skb)
 {
+	struct net_device *dev = skb->dev;
+	struct net *net = dev_net_rcu(dev);
 	struct mfc6_cache *cache;
-	struct net *net = dev_net(skb->dev);
 	struct mr_table *mrt;
 	struct flowi6 fl6 = {
-		.flowi6_iif	= skb->dev->ifindex,
+		.flowi6_iif	= dev->ifindex,
 		.flowi6_mark	= skb->mark,
 	};
 	int err;
-	struct net_device *dev;
 
 	/* skb->dev passed in is the master dev for vrfs.
 	 * Get the proper interface that does have a vif associated with it.
 	 */
-	dev = skb->dev;
-	if (netif_is_l3_master(skb->dev)) {
+	if (netif_is_l3_master(dev)) {
 		dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
 		if (!dev) {
 			kfree_skb(skb);
-- 
2.53.0

[PATCH 6.6.y 3/3] ip: orphan prefetched skbs before multicast forwarding

[Sasha Levin]

From: Zhiling Zou <zhilinz@nebusec.ai>

[ Upstream commit e36ce6e78fe3fc3c071a26750783b7ba081ce10d ]

IPv4 and IPv6 input preserve an skb->sk association installed by
bpf_sk_assign() so that local delivery can use the selected socket under
RCU. Both address families can also prefetch a socket in UDP early demux.
In both paths (BPF and UDP early demux) a reference is not guaranteed to
be held on the socket.

When a multicast packet is not locally deliverable, IPv6 hands the
original skb to ip6_mr_input(). IPv4's ip_mr_input() similarly keeps the
original skb when local delivery is not needed. Either path can put the
skb on an unresolved multicast route queue or forward it after the
receive-side RCU section ends.

After the prefetched socket is destroyed, a later skb free invokes
sock_pfree() and dereferences the stale skb->sk. Orphan the skb before
each non-local multicast forwarding path. Local delivery retains the
original skb; the existing skb_clone() calls provide multicast forwarding
with a socket-free clone.

Fixes: cf7fbe660f2d ("bpf: Add socket assign support")
Fixes: 08842c43d016 ("udp: no longer touch sk->sk_refcnt in early demux")
Cc: stable@vger.kernel.org
Reported-by: Vega <vega@nebusec.ai>
Signed-off-by: Zhiling Zou <zhilinz@nebusec.ai>
Reported-by: Vega <vega@nebusec.ai>
Signed-off-by: Zhiling Zou <zhilinz@nebusec.ai>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/0c52eb3d7532aaf8bccf37e0f7c922143c639735.1786552223.git.zhilinz@nebusec.ai
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 net/ipv4/ipmr.c      | 3 +++
 net/ipv6/ip6_input.c | 1 +
 2 files changed, 4 insertions(+)

diff --git a/net/ipv4/ipmr.c b/net/ipv4/ipmr.c
index af9412a507cf3..93d08e343a256 100644
--- a/net/ipv4/ipmr.c
+++ b/net/ipv4/ipmr.c
@@ -2139,6 +2139,9 @@ int ip_mr_input(struct sk_buff *skb)
 	if (IPCB(skb)->flags & IPSKB_FORWARDED)
 		goto dont_forward;
 
+	if (!local)
+		skb_orphan(skb);
+
 	mrt = ipmr_rt_fib_lookup(net, skb);
 	if (IS_ERR(mrt)) {
 		kfree_skb(skb);
diff --git a/net/ipv6/ip6_input.c b/net/ipv6/ip6_input.c
index e775a5a766692..9b6eaae0c6c80 100644
--- a/net/ipv6/ip6_input.c
+++ b/net/ipv6/ip6_input.c
@@ -566,6 +566,7 @@ int ip6_mc_input(struct sk_buff *skb)
 		if (deliver) {
 			skb2 = skb_clone(skb, GFP_ATOMIC);
 		} else {
+			skb_orphan(skb);
 			skb2 = skb;
 			skb = NULL;
 		}
-- 
2.53.0