[PATCH net-next v2 0/4] net: prevent mixing incompatible memory providers in skbs

[PATCH net-next v2 0/4] net: prevent mixing incompatible memory providers in skbs

[Mina Almasry]

Enforce that all fragments in a nonlinear skb come from compatible
memory providers (and for devmem, the exact same binding).

Patch 1 fixes skb->unreadable propagation in skb_split() and skb_shift().
Patch 2 stores mp_priv directly inside struct net_iov.
Patch 3 adds skb_has_net_iovs(), skb_mem_provider_compat(), and
skb_frags_can_coalesce().
Patch 4 enforces provider/binding compatibility across TX, TCP, GRO,
IP defrag, skb_shift(), and skb_try_coalesce().

v1 -> v2:
- Generalize the check across TX, TCP collapse, GRO, IP defrag,
  skb_shift(), and skb_try_coalesce() by storing mp_priv directly in
  struct net_iov and introducing skb_frags_can_coalesce().
- Fix skb->unreadable propagation in skb_split() and skb_shift().
- Add KUnit coverage for cross-provider and cross-binding rejection.
- v1: https://lore.kernel.org/netdev/20261008034027.1403188-1-almasrymina@google.com/


Mina Almasry (4):
  net: fix skb->unreadable propagation in skb_split() and skb_shift()
  netmem: store memory provider private pointer directly in struct
    net_iov
  net: add skb_frags_can_coalesce() helper
  net: prevent mixing frags from different memory providers or bindings

 include/linux/skbuff.h   | 58 ++++++++++++++++++++++++++++++++++++++++
 include/net/netmem.h     | 15 +++++++++--
 include/net/tcp.h        |  5 ++--
 io_uring/zcrx.c          |  2 +-
 net/core/datagram.c      |  4 +--
 net/core/dev.c           | 14 +++-------
 net/core/devmem.c        |  3 ++-
 net/core/devmem.h        | 25 +++++++++++++++--
 net/core/gro.c           | 16 +++++------
 net/core/net_test.c      | 34 +++++++++++++++++------
 net/core/skbuff.c        | 34 ++++++++++++-----------
 net/ipv4/inet_fragment.c |  3 +++
 net/ipv4/tcp.c           |  2 +-
 net/ipv4/tcp_output.c    |  3 +++
 14 files changed, 163 insertions(+), 55 deletions(-)


base-commit: 4dd16e31135ea8437a720222df34e8745f385d6d
-- 
2.56.0.385.gac671c819a-goog

[PATCH net-next v2 1/4] net: fix skb->unreadable propagation in skb_split() and skb_shift()

[Mina Almasry]

skb_split() and skb_shift() do not accurately maintain skb->unreadable
when moving fragments between skbs:

- skb_split_inside_header() moves all fragments out of skb into skb1
  (leaving skb with nr_frags == 0 and only host-readable linear memory),
  but leaves skb->unreadable set on skb.
- skb_split_no_header() unconditionally copies skb->unreadable to skb1
  even if 0 fragments were moved to skb1 (k == 0), and leaves
  skb->unreadable set on skb when splitting at skb_headlen(skb) moves
  all fragments into skb1 (!skb_shinfo(skb)->nr_frags).
- skb_shift() neither propagates skb->unreadable to target skb when
  shifting unreadable fragments into an empty skb, nor clears
  skb->unreadable after all fragments are shifted out of skb.

Fix all three cases so skb->unreadable accurately tracks whether the
remaining fragments in each skb are unreadable.

This issue is theoretical and was found by manual code inspection and
LLM review; it was not triggered in production. Verified with the
net_core KUnit test suite; not tested on real hardware.

Fixes: 65249feb6b3d ("net: add support for skbs with unreadable frags")
Signed-off-by: Mina Almasry <almasrymina@google.com>
---
 net/core/skbuff.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 48f6764117243..dc761ceb4d204 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -4225,6 +4225,7 @@ static inline void skb_split_inside_header(struct sk_buff *skb,
 
 	skb_shinfo(skb1)->nr_frags = skb_shinfo(skb)->nr_frags;
 	skb1->unreadable	   = skb->unreadable;
+	skb->unreadable		   = 0;
 	skb_shinfo(skb)->nr_frags  = 0;
 	skb1->data_len		   = skb->data_len;
 	skb1->len		   += skb1->data_len;
@@ -4273,7 +4274,10 @@ static inline void skb_split_no_header(struct sk_buff *skb,
 	}
 	skb_shinfo(skb1)->nr_frags = k;
 
-	skb1->unreadable = skb->unreadable;
+	if (k)
+		skb1->unreadable = skb->unreadable;
+	if (!skb_shinfo(skb)->nr_frags)
+		skb->unreadable = 0;
 }
 
 /**
@@ -4447,6 +4451,9 @@ int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen)
 	skb->ip_summed = CHECKSUM_PARTIAL;
 
 	skb_shinfo(tgt)->flags |= skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
+	tgt->unreadable |= skb->unreadable;
+	if (!skb_shinfo(skb)->nr_frags && !skb_has_frag_list(skb))
+		skb->unreadable = 0;
 
 	skb_len_add(skb, -shiftlen);
 	skb_len_add(tgt, shiftlen);
-- 
2.56.0.385.gac671c819a-goog

[PATCH net-next v2 2/4] netmem: store memory provider private pointer directly in struct net_iov

[Mina Almasry]

Resolving the memory provider instance from a struct net_iov currently
chases niov->owner (struct net_iov_area) and then dereferences the
provider-specific container struct (such as net_devmem_dmabuf_binding
or io_zcrx_area->ifq).

Alias void *mp_priv over the unused _pp_mapping_pad slot inside struct
net_iov (keeping sizeof(struct net_iov) == 64 bytes unchanged) and populate
it in net_iov_init() for both TX and RX net_iovs so callers can obtain the
provider instance pointer in a single load from the net_iov cacheline.

Signed-off-by: Mina Almasry <almasrymina@google.com>
---
 include/net/netmem.h | 15 +++++++++++++--
 io_uring/zcrx.c      |  2 +-
 net/core/devmem.c    |  3 ++-
 net/core/devmem.h    |  3 +--
 net/core/net_test.c  | 10 +++++-----
 5 files changed, 22 insertions(+), 11 deletions(-)

diff --git a/include/net/netmem.h b/include/net/netmem.h
index fb693bc1702a4..c869be9c4d266 100644
--- a/include/net/netmem.h
+++ b/include/net/netmem.h
@@ -95,11 +95,20 @@ enum net_iov_type {
  *		supported.
  */
 struct net_iov {
-	struct netmem_desc desc;
+	union {
+		struct netmem_desc desc;
+		struct {
+			unsigned long _pad[3];
+			void *mp_priv;
+		};
+	};
 	enum net_iov_type type;
 	struct net_iov_area *owner;
 };
 
+static_assert(offsetof(struct net_iov, mp_priv) ==
+	      offsetof(struct netmem_desc, _pp_mapping_pad));
+
 struct net_iov_area {
 	/* Array of net_iovs for this area. */
 	struct net_iov *niovs;
@@ -120,10 +129,12 @@ static inline unsigned int net_iov_idx(const struct net_iov *niov)
  */
 static inline void net_iov_init(struct net_iov *niov,
 				struct net_iov_area *owner,
-				enum net_iov_type type)
+				enum net_iov_type type,
+				void *mp_priv)
 {
 	niov->owner = owner;
 	niov->type = type;
+	niov->mp_priv = mp_priv;
 }
 
 /* netmem */
diff --git a/io_uring/zcrx.c b/io_uring/zcrx.c
index 6d7abb1636e7f..6ec618d325606 100644
--- a/io_uring/zcrx.c
+++ b/io_uring/zcrx.c
@@ -562,7 +562,7 @@ static int __zcrx_create_area(struct io_zcrx_ifq *ifq,
 	for (i = 0; i < nr_iovs; i++) {
 		struct net_iov *niov = &area->nia.niovs[i];
 
-		net_iov_init(niov, &area->nia, NET_IOV_IOURING);
+		net_iov_init(niov, &area->nia, NET_IOV_IOURING, ifq);
 		area->freelist[i] = i;
 		atomic_set(&area->user_refs[i], 0);
 	}
diff --git a/net/core/devmem.c b/net/core/devmem.c
index 32fafbfa1359e..fe99c005cbc5e 100644
--- a/net/core/devmem.c
+++ b/net/core/devmem.c
@@ -323,7 +323,8 @@ net_devmem_bind_dmabuf(struct net_device *dev, void *vdev,
 		nr_niovs = len >> niov_shift;
 		for (i = 0; i < nr_niovs; i++, niov_idx++) {
 			niov = &binding->area.niovs[niov_idx];
-			net_iov_init(niov, &binding->area, NET_IOV_DMABUF);
+			net_iov_init(niov, &binding->area, NET_IOV_DMABUF,
+				     binding);
 			if (net_mp_niov_set_dma_addr(niov, dma_addr)) {
 				err = -EFAULT;
 				NL_SET_ERR_MSG(extack,
diff --git a/net/core/devmem.h b/net/core/devmem.h
index 8fa089776610a..942b8caf36e99 100644
--- a/net/core/devmem.h
+++ b/net/core/devmem.h
@@ -102,8 +102,7 @@ int net_devmem_bind_dmabuf_to_queue(struct net_device *dev, u32 rxq_idx,
 static inline struct net_devmem_dmabuf_binding *
 net_devmem_iov_binding(const struct net_iov *niov)
 {
-	return container_of(net_iov_owner(niov),
-			    struct net_devmem_dmabuf_binding, area);
+	return niov->mp_priv;
 }
 
 static inline u32 net_devmem_iov_binding_id(const struct net_iov *niov)
diff --git a/net/core/net_test.c b/net/core/net_test.c
index 2b5080c226636..63124dbf3c9a8 100644
--- a/net/core/net_test.c
+++ b/net/core/net_test.c
@@ -857,7 +857,7 @@ static void netmem_ref_test_net_iov(struct kunit *test)
 			      PERCPU_REF_INIT_ATOMIC, GFP_KERNEL);
 	KUNIT_ASSERT_EQ(test, err, 0);
 
-	net_iov_init(&niov, &binding.area, NET_IOV_DMABUF);
+	net_iov_init(&niov, &binding.area, NET_IOV_DMABUF, &binding);
 	netmem = net_iov_to_netmem(&niov);
 	base_ref = atomic_long_read(&binding.ref.data->count);
 
@@ -1145,7 +1145,7 @@ static void netmem_ref_test_release_data(struct kunit *test)
 			percpu_ref_init(&binding.ref, dummy_percpu_ref_release,
 					PERCPU_REF_INIT_ATOMIC, GFP_KERNEL),
 			0);
-	net_iov_init(&niov, &binding.area, NET_IOV_DMABUF);
+	net_iov_init(&niov, &binding.area, NET_IOV_DMABUF, &binding);
 	netmem = net_iov_to_netmem(&niov);
 	netmem_set_pp(netmem, pool);
 	netmem_or_pp_magic(netmem, PP_SIGNATURE);
@@ -1206,7 +1206,7 @@ static void netmem_ref_test_skb_segment(struct kunit *test)
 			percpu_ref_init(&binding.ref, dummy_percpu_ref_release,
 					PERCPU_REF_INIT_ATOMIC, GFP_KERNEL),
 			0);
-	net_iov_init(&niov, &binding.area, NET_IOV_DMABUF);
+	net_iov_init(&niov, &binding.area, NET_IOV_DMABUF, &binding);
 	niov_nm = net_iov_to_netmem(&niov);
 	netmem_set_pp(niov_nm, pool);
 	netmem_or_pp_magic(niov_nm, PP_SIGNATURE);
@@ -1508,7 +1508,7 @@ static void netmem_ref_test_try_coalesce(struct kunit *test)
 			percpu_ref_init(&binding.ref, dummy_percpu_ref_release,
 					PERCPU_REF_INIT_ATOMIC, GFP_KERNEL),
 			0);
-	net_iov_init(&niov, &binding.area, NET_IOV_DMABUF);
+	net_iov_init(&niov, &binding.area, NET_IOV_DMABUF, &binding);
 	niov_nm = net_iov_to_netmem(&niov);
 	netmem_set_pp(niov_nm, pool);
 	netmem_or_pp_magic(niov_nm, PP_SIGNATURE);
@@ -1654,7 +1654,7 @@ static void netmem_ref_test_tcp_devmem_recv(struct kunit *test)
 		KUNIT_ASSERT_EQ(test, ret, 0);
 	}
 	binding.area.niovs = &niov;
-	net_iov_init(&niov, &binding.area, NET_IOV_DMABUF);
+	net_iov_init(&niov, &binding.area, NET_IOV_DMABUF, &binding);
 	niov_nm = net_iov_to_netmem(&niov);
 	netmem_set_pp(niov_nm, pool);
 	netmem_or_pp_magic(niov_nm, PP_SIGNATURE);
-- 
2.56.0.385.gac671c819a-goog

[PATCH net-next v2 3/4] net: add skb_frags_can_coalesce() helper

[Mina Almasry]

Add skb_has_net_iovs(skb), assuming skb->unreadable and net_iov-backed
frags are equivalent for now (with a comment noting the two can be
decoupled in the future if readable net_iovs are introduced).

Also add skb_mem_provider_compat() and skb_frags_can_coalesce() so all
multi-skb coalescing, shifting, and GRO paths can share a single check
across pp_recycle, skb_zcopy(), and fragment memory providers:
- empty 0-length skbs are compatible without mixing readable payload
- page vs net_iov is rejected via skb_has_net_iovs() (0 pointer chases)
- page vs page is allowed via !skb_has_net_iovs() (0 pointer chases)
- net_iov vs net_iov compares mp_priv (walking frag_list if nr_frags is 0)

Signed-off-by: Mina Almasry <almasrymina@google.com>
---
 include/linux/skbuff.h | 58 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 58 insertions(+)

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 6787fcf6ac3e4..1e6e919c656d5 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -1882,6 +1882,16 @@ static inline bool skb_frags_readable(const struct sk_buff *skb)
 	return !skb->unreadable;
 }
 
+/* Return true if frags in this skb are backed by net_iovs.
+ * Assumes skb->unreadable and net_iov-backed frags are equivalent for now;
+ * the two can be decoupled in the future if readable net_iovs in skbs are
+ * introduced.
+ */
+static inline bool skb_has_net_iovs(const struct sk_buff *skb)
+{
+	return skb->unreadable;
+}
+
 static inline void skb_mark_not_on_list(struct sk_buff *skb)
 {
 	skb->next = NULL;
@@ -3763,6 +3773,54 @@ static inline void *skb_frag_address(const skb_frag_t *frag)
 	return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
 }
 
+static inline void *skb_mem_provider_priv(const struct sk_buff *skb)
+{
+	netmem_ref netmem;
+
+	if (!skb_has_net_iovs(skb) || !skb_shinfo(skb)->nr_frags)
+		return NULL;
+
+	netmem = skb_frag_netmem(&skb_shinfo(skb)->frags[0]);
+	return netmem_to_net_iov(netmem)->mp_priv;
+}
+
+/* Return true if the memory providers of @skb1 and @skb2 are compatible
+ * for coalescing. For any struct page-backed provider (the buddy allocator
+ * or a future provider handing out struct pages), coalescing is assumed
+ * feasible. For net_iov-backed providers, the exact mp_priv pointer must
+ * match.
+ */
+static inline bool skb_mem_provider_compat(const struct sk_buff *skb1,
+					   const struct sk_buff *skb2)
+{
+	void *priv1, *priv2;
+
+	if (!skb1->len || !skb2->len)
+		return true;
+
+	if (skb_has_net_iovs(skb1) != skb_has_net_iovs(skb2))
+		return false;
+
+	if (!skb_has_net_iovs(skb1))
+		return true;
+
+	priv1 = skb_mem_provider_priv(skb1);
+	priv2 = skb_mem_provider_priv(skb2);
+	return priv1 && priv1 == priv2;
+}
+
+static inline bool skb_frags_can_coalesce(struct sk_buff *to,
+					  struct sk_buff *from)
+{
+	if (to->pp_recycle != from->pp_recycle)
+		return false;
+
+	if (skb_zcopy(to) || skb_zcopy(from))
+		return false;
+
+	return skb_mem_provider_compat(to, from);
+}
+
 /**
  * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  * @frag: the paged fragment buffer
-- 
2.56.0.385.gac671c819a-goog

[PATCH net-next v2 4/4] net: prevent mixing frags from different memory providers or bindings

[Mina Almasry]

Enforce that all fragments in a nonlinear skb come from compatible
memory providers (and for devmem, the exact same binding):

- Add net_devmem_skb_binding() and use it in
  validate_xmit_unreadable_skb(), zerocopy_fill_skb_from_devmem(), and
  tcp_sendmsg_locked() so frags from different devmem bindings or memory
  providers cannot be mixed on TX.
- Use skb_frags_can_coalesce() across skb_try_coalesce(), skb_shift(),
  skb_gro_receive(), and skb_gro_receive_list().
- Use skb_mem_provider_compat() in tcp_skb_can_collapse(),
  tcp_skb_can_collapse_rx(), and inet_frag_queue_insert().
- Reject coalescing linear head payloads into unreadable skbs in
  skb_try_coalesce(), skb_gro_receive(), and
  tcp_can_coalesce_send_queue_head().

This issue is theoretical and was found by manual code inspection and
LLM review; it was not triggered in production. Verified with the
net_core KUnit test suite; not tested on real hardware.

Fixes: 65249feb6b3d ("net: add support for skbs with unreadable frags")
Fixes: bd61848900bf ("net: devmem: Implement TX path")
Signed-off-by: Mina Almasry <almasrymina@google.com>
---
 include/net/tcp.h        |  5 +++--
 net/core/datagram.c      |  4 ++--
 net/core/dev.c           | 14 ++++----------
 net/core/devmem.h        | 22 ++++++++++++++++++++++
 net/core/gro.c           | 16 ++++++----------
 net/core/net_test.c      | 24 +++++++++++++++++++++---
 net/core/skbuff.c        | 25 ++++++++++---------------
 net/ipv4/inet_fragment.c |  3 +++
 net/ipv4/tcp.c           |  2 +-
 net/ipv4/tcp_output.c    |  3 +++
 10 files changed, 75 insertions(+), 43 deletions(-)

diff --git a/include/net/tcp.h b/include/net/tcp.h
index 3560adddf8215..07a222c0fe2d7 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -1268,14 +1268,15 @@ static inline bool tcp_skb_can_collapse(const struct sk_buff *to,
 		      mptcp_skb_can_collapse(to, from) &&
 		      !skb_cmp_decrypted(to, from) &&
 		      skb_pure_zcopy_same(to, from) &&
-		      skb_frags_readable(to) == skb_frags_readable(from));
+		      skb_mem_provider_compat(to, from));
 }
 
 static inline bool tcp_skb_can_collapse_rx(const struct sk_buff *to,
 					   const struct sk_buff *from)
 {
 	return likely(mptcp_skb_can_collapse(to, from) &&
-		      !skb_cmp_decrypted(to, from));
+		      !skb_cmp_decrypted(to, from) &&
+		      skb_mem_provider_compat(to, from));
 }
 
 /* Events passed to congestion control interface */
diff --git a/net/core/datagram.c b/net/core/datagram.c
index ecc30a6f83023..1e7db61e25a56 100644
--- a/net/core/datagram.c
+++ b/net/core/datagram.c
@@ -712,8 +712,8 @@ zerocopy_fill_skb_from_devmem(struct sk_buff *skb, struct iov_iter *from,
 	size_t virt_addr, size, off;
 	struct net_iov *niov;
 
-	if (i && skb_frags_readable(skb))
-		return -EFAULT;
+	if (skb->len && net_devmem_skb_binding(skb) != binding)
+		return -EMSGSIZE;
 
 	/* Devmem filling works by taking an IOVEC from the user where the
 	 * iov_addrs are interpreted as an offset in bytes into the dma-buf to
diff --git a/net/core/dev.c b/net/core/dev.c
index 3ac700d37e596..c1ea3b26a5a11 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -4169,8 +4169,7 @@ static struct sk_buff *sk_validate_xmit_skb(struct sk_buff *skb,
 static struct sk_buff *validate_xmit_unreadable_skb(struct sk_buff *skb,
 						    struct net_device *dev)
 {
-	struct skb_shared_info *shinfo;
-	struct net_iov *niov;
+	struct net_devmem_dmabuf_binding *binding;
 
 	if (likely(skb_frags_readable(skb) ||
 		   dev->netmem_tx == NETMEM_TX_NO_DMA))
@@ -4179,14 +4178,9 @@ static struct sk_buff *validate_xmit_unreadable_skb(struct sk_buff *skb,
 	if (dev->netmem_tx == NETMEM_TX_NONE)
 		goto out_free;
 
-	shinfo = skb_shinfo(skb);
-
-	if (shinfo->nr_frags > 0) {
-		niov = netmem_to_net_iov(skb_frag_netmem(&shinfo->frags[0]));
-		if (net_is_devmem_iov(niov) &&
-		    READ_ONCE(net_devmem_iov_binding(niov)->dev) != dev)
-			goto out_free;
-	}
+	binding = net_devmem_skb_binding(skb);
+	if (binding && READ_ONCE(binding->dev) != dev)
+		goto out_free;
 
 out:
 	return skb;
diff --git a/net/core/devmem.h b/net/core/devmem.h
index 942b8caf36e99..90058d7350820 100644
--- a/net/core/devmem.h
+++ b/net/core/devmem.h
@@ -10,6 +10,7 @@
 #ifndef _NET_DEVMEM_H
 #define _NET_DEVMEM_H
 
+#include <linux/skbuff.h>
 #include <net/net_namespace.h>
 #include <net/netmem.h>
 #include <net/netdev_netlink.h>
@@ -105,6 +106,21 @@ net_devmem_iov_binding(const struct net_iov *niov)
 	return niov->mp_priv;
 }
 
+static inline struct net_devmem_dmabuf_binding *
+net_devmem_skb_binding(const struct sk_buff *skb)
+{
+	struct net_iov *niov;
+
+	if (!skb_has_net_iovs(skb) || !skb_shinfo(skb)->nr_frags)
+		return NULL;
+
+	niov = netmem_to_net_iov(skb_frag_netmem(&skb_shinfo(skb)->frags[0]));
+	if (!net_is_devmem_iov(niov))
+		return NULL;
+
+	return net_devmem_iov_binding(niov);
+}
+
 static inline u32 net_devmem_iov_binding_id(const struct net_iov *niov)
 {
 	return net_devmem_iov_binding(niov)->id;
@@ -222,6 +238,12 @@ net_devmem_iov_binding(const struct net_iov *niov)
 {
 	return NULL;
 }
+
+static inline struct net_devmem_dmabuf_binding *
+net_devmem_skb_binding(const struct sk_buff *skb)
+{
+	return NULL;
+}
 #endif
 
 #endif /* _NET_DEVMEM_H */
diff --git a/net/core/gro.c b/net/core/gro.c
index 1b033855a20db..62b8d532218e5 100644
--- a/net/core/gro.c
+++ b/net/core/gro.c
@@ -100,16 +100,7 @@ int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb)
 	struct sk_buff *lp;
 	int segs;
 
-	/* Do not splice page pool based packets w/ non-page pool
-	 * packets. This can result in reference count issues as page
-	 * pool pages will not decrement the reference count and will
-	 * instead be immediately returned to the pool or have frag
-	 * count decremented.
-	 */
-	if (p->pp_recycle != skb->pp_recycle)
-		return -ETOOMANYREFS;
-
-	if (skb_zcopy(p) || skb_zcopy(skb))
+	if (!skb_frags_can_coalesce(p, skb))
 		return -ETOOMANYREFS;
 
 	if (unlikely(p->len + len >= netif_get_gro_max_size(p->dev, p) ||
@@ -158,6 +149,8 @@ int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb)
 
 		NAPI_GRO_CB(skb)->free = NAPI_GRO_FREE;
 		goto done;
+	} else if (!skb_frags_readable(skb)) {
+		return -ETOOMANYREFS;
 	} else if (skb->head_frag) {
 		int nr_frags = pinfo->nr_frags;
 		skb_frag_t *frag = pinfo->frags + nr_frags;
@@ -233,6 +226,9 @@ int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb)
 
 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb)
 {
+	if (!skb_frags_can_coalesce(p, skb))
+		return -ETOOMANYREFS;
+
 	/* make sure to check flush flag and to not merge */
 	if (unlikely(p->len + skb->len >= 65536 ||
 		     NAPI_GRO_CB(skb)->flush))
diff --git a/net/core/net_test.c b/net/core/net_test.c
index 63124dbf3c9a8..99b664fc9186a 100644
--- a/net/core/net_test.c
+++ b/net/core/net_test.c
@@ -1485,7 +1485,7 @@ static void netmem_ref_test_try_coalesce(struct kunit *test)
 	};
 	struct sk_buff *to, *from, *clone;
 	struct net_devmem_dmabuf_binding binding = {};
-	struct net_iov niov = {};
+	struct net_iov niov = {}, niov2 = {};
 	struct page *pp_page, *reg_page;
 	netmem_ref pp_nm, niov_nm;
 	struct page_pool *pool;
@@ -1578,9 +1578,7 @@ static void netmem_ref_test_try_coalesce(struct kunit *test)
 	from = alloc_skb(0, GFP_KERNEL);
 	KUNIT_ASSERT_NOT_NULL(test, to);
 	KUNIT_ASSERT_NOT_NULL(test, from);
-	skb_put(to, skb_tailroom(to));
 	to->pp_recycle = 1;
-	to->unreadable = 1;
 	from->pp_recycle = 1;
 	skb_add_rx_frag_netmem(from, 0, niov_nm, 0, 64, PAGE_SIZE);
 	napi_pp_get_page(niov_nm);
@@ -1589,6 +1587,7 @@ static void netmem_ref_test_try_coalesce(struct kunit *test)
 	KUNIT_ASSERT_NOT_NULL(test, clone);
 	KUNIT_EXPECT_TRUE(test, skb_try_coalesce(to, from, &fragstolen,
 						 &delta));
+	KUNIT_EXPECT_TRUE(test, to->unreadable);
 	KUNIT_EXPECT_EQ(test, atomic_long_read(&niov.desc.pp_ref_count), 4L);
 	KUNIT_EXPECT_EQ(test, atomic_long_read(&binding.ref.data->count),
 			base_iov_ref);
@@ -1597,6 +1596,25 @@ static void netmem_ref_test_try_coalesce(struct kunit *test)
 	consume_skb(to);
 	KUNIT_EXPECT_EQ(test, atomic_long_read(&niov.desc.pp_ref_count), 2L);
 
+	/* 4. Reject coalescing net_iov skbs from different bindings. */
+	net_iov_init(&niov2, &binding.area, NET_IOV_DMABUF, &niov2);
+	to = alloc_skb(0, GFP_KERNEL);
+	from = alloc_skb(0, GFP_KERNEL);
+	KUNIT_ASSERT_NOT_NULL(test, to);
+	KUNIT_ASSERT_NOT_NULL(test, from);
+	to->pp_recycle = 1;
+	from->pp_recycle = 1;
+	skb_add_rx_frag_netmem(to, 0, niov_nm, 0, 32, PAGE_SIZE);
+	napi_pp_get_page(niov_nm);
+	skb_add_rx_frag_netmem(from, 0, net_iov_to_netmem(&niov2), 0, 32,
+			       PAGE_SIZE);
+	KUNIT_EXPECT_FALSE(test, skb_try_coalesce(to, from, &fragstolen,
+						  &delta));
+	skb_shinfo(from)->nr_frags = 0;
+	consume_skb(from);
+	consume_skb(to);
+	KUNIT_EXPECT_EQ(test, atomic_long_read(&niov.desc.pp_ref_count), 2L);
+
 	page_pool_put_full_page(pool, pp_page, false);
 	page_pool_put_full_page(pool, pp_page, false);
 	__free_page(reg_page);
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index dc761ceb4d204..c588ae785f337 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -4342,10 +4342,9 @@ int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen)
 
 	if (skb_headlen(skb))
 		return 0;
-	if (skb_zcopy(tgt) || skb_zcopy(skb))
+	if (!skb_frags_can_coalesce(tgt, skb))
 		return 0;
 
-	DEBUG_NET_WARN_ON_ONCE(tgt->pp_recycle != skb->pp_recycle);
 	DEBUG_NET_WARN_ON_ONCE(skb_cmp_decrypted(tgt, skb));
 
 	todo = shiftlen;
@@ -6301,17 +6300,8 @@ bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
 	 * tricky (due to potential race with the clone disappearing) and
 	 * rare, so not worth dealing with.
 	 */
-	if (to->pp_recycle != from->pp_recycle)
-		return false;
-
-	/* All frags in an skb must either all be struct page-backed or all
-	 * belong to the same memory provider instance; do not coalesce skbs
-	 * that would mix them.
-	 */
-	if (skb_frags_readable(from) != skb_frags_readable(to))
-		return false;
-
-	if (len <= skb_tailroom(to) && skb_frags_readable(from)) {
+	if (len <= skb_tailroom(to) && to->pp_recycle == from->pp_recycle &&
+	    skb_frags_readable(to) && skb_frags_readable(from)) {
 		if (len) {
 			if (unlikely(skb_copy_bits(from, 0,
 						   skb_tail_pointer(to),
@@ -6325,17 +6315,21 @@ bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
 		return true;
 	}
 
+	if (!skb_frags_can_coalesce(to, from))
+		return false;
+
 	to_shinfo = skb_shinfo(to);
 	from_shinfo = skb_shinfo(from);
 	if (to_shinfo->frag_list || from_shinfo->frag_list)
 		return false;
-	if (skb_zcopy(to) || skb_zcopy(from))
-		return false;
 
 	if (skb_headlen(from) != 0) {
 		struct page *page;
 		unsigned int offset;
 
+		if (!skb_frags_readable(from))
+			return false;
+
 		if (to_shinfo->nr_frags +
 		    from_shinfo->nr_frags >= MAX_SKB_FRAGS)
 			return false;
@@ -6367,6 +6361,7 @@ bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
 	to_shinfo->nr_frags += from_shinfo->nr_frags;
 	if (from_shinfo->nr_frags)
 		to_shinfo->flags |= from_shinfo->flags & SKBFL_SHARED_FRAG;
+	to->unreadable |= from->unreadable;
 
 	if (!skb_cloned(from))
 		from_shinfo->nr_frags = 0;
diff --git a/net/ipv4/inet_fragment.c b/net/ipv4/inet_fragment.c
index c18da91c1f031..a018ce6cf34ff 100644
--- a/net/ipv4/inet_fragment.c
+++ b/net/ipv4/inet_fragment.c
@@ -451,6 +451,9 @@ int inet_frag_queue_insert(struct inet_frag_queue *q, struct sk_buff *skb,
 	 * that reach the reassembled skb, whose head keeps the first
 	 * fragment's shinfo and whose frag_list is not GRO-shaped.
 	 */
+	if (last && !skb_mem_provider_compat(last, skb))
+		return IPFRAG_OVERLAP;
+
 	skb_gso_reset(skb);
 
 	/* RFC5722, Section 4, amended by Errata ID : 3089
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index b9de7c302ba83..67651b1936a8d 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -1251,7 +1251,7 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
 		trace_tcp_sendmsg_locked(sk, msg, skb, size_goal);
 
 		if (copy <= 0 || !tcp_skb_can_collapse_to(skb) ||
-		    unlikely(skb_frags_readable(skb) != !binding)) {
+		    unlikely(net_devmem_skb_binding(skb) != binding)) {
 			bool first_skb;
 
 new_segment:
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index cf3ddc055d48d..110dd1c775eb5 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -2657,6 +2657,9 @@ static bool tcp_can_coalesce_send_queue_head(struct sock *sk, int len)
 
 	skb = tcp_send_head(sk);
 	tcp_for_write_queue_from_safe(skb, next, sk) {
+		if (!skb_frags_readable(skb))
+			return false;
+
 		if (len <= skb->len)
 			break;
 
-- 
2.56.0.385.gac671c819a-goog

Re: [PATCH net-next v2 2/4] netmem: store memory provider private pointer directly in struct net_iov

[Stanislav Fomichev]

On 11/02, Mina Almasry wrote:

[..]

> Resolving the memory provider instance from a struct net_iov currently
> chases niov->owner (struct net_iov_area) and then dereferences the
> provider-specific container struct (such as net_devmem_dmabuf_binding
> or io_zcrx_area->ifq).

Chasing niov->owner is true, but I don't see another dereference you're
talking about? net_devmem_iov_binding does container_of to get from
area to the binding (and the area is the first member of the binding,
so it is a already a nop). And IIRC, iou has a similar arrangement.
So I'd vote to keep things as is unless you explain what I'm missing.

Re: [PATCH net-next v2 1/4] net: fix skb->unreadable propagation in skb_split() and skb_shift()

[Stanislav Fomichev]

On 11/02, Mina Almasry wrote:
> skb_split() and skb_shift() do not accurately maintain skb->unreadable
> when moving fragments between skbs:
> 
> - skb_split_inside_header() moves all fragments out of skb into skb1
>   (leaving skb with nr_frags == 0 and only host-readable linear memory),
>   but leaves skb->unreadable set on skb.
> - skb_split_no_header() unconditionally copies skb->unreadable to skb1
>   even if 0 fragments were moved to skb1 (k == 0), and leaves
>   skb->unreadable set on skb when splitting at skb_headlen(skb) moves
>   all fragments into skb1 (!skb_shinfo(skb)->nr_frags).
> - skb_shift() neither propagates skb->unreadable to target skb when
>   shifting unreadable fragments into an empty skb, nor clears
>   skb->unreadable after all fragments are shifted out of skb.
> 
> Fix all three cases so skb->unreadable accurately tracks whether the
> remaining fragments in each skb are unreadable.
> 
> This issue is theoretical and was found by manual code inspection and
> LLM review; it was not triggered in production. Verified with the
> net_core KUnit test suite; not tested on real hardware.
> 
> Fixes: 65249feb6b3d ("net: add support for skbs with unreadable frags")
> Signed-off-by: Mina Almasry <almasrymina@google.com>

Acked-by: Stanislav Fomichev <sdf@fomichev.me>

Re: [PATCH net-next v2 3/4] net: add skb_frags_can_coalesce() helper

[Stanislav Fomichev]

On 11/02, Mina Almasry wrote:
> Add skb_has_net_iovs(skb), assuming skb->unreadable and net_iov-backed
> frags are equivalent for now (with a comment noting the two can be
> decoupled in the future if readable net_iovs are introduced).
> 
> Also add skb_mem_provider_compat() and skb_frags_can_coalesce() so all
> multi-skb coalescing, shifting, and GRO paths can share a single check
> across pp_recycle, skb_zcopy(), and fragment memory providers:
> - empty 0-length skbs are compatible without mixing readable payload
> - page vs net_iov is rejected via skb_has_net_iovs() (0 pointer chases)
> - page vs page is allowed via !skb_has_net_iovs() (0 pointer chases)
> - net_iov vs net_iov compares mp_priv (walking frag_list if nr_frags is 0)

Maybe squash this and patch #4? Otherwise hard to follow :-(