bpf, sockmap: reject a packet-modifying SK_SKB stream parser

A BPF_PROG_TYPE_SK_SKB stream parser runs on strparser's message head, which can chain skbs through frag_list. A parser that resizes the skb frees the frag_list segments that strparser still tracks through skb_nextp, leading to a use-after-free.

A stream parser is only meant to measure the next message, not to modify the packet, so reject a packet-modifying parser at attach time.

v5:
- target bpf-next instead of bpf
- add Reviewed-by tag (Jiayuan Chen)

v4:
- https://lore.kernel.org/all/[email protected]/

v3:
- https://lore.kernel.org/all/[email protected]/

v2:
- https://lore.kernel.org/all/[email protected]/

v1:
- https://lore.kernel.org/all/[email protected]/

Sechang Lim (3):
selftests/bpf: don't modify the skb in the strparser parser prog bpf, sockmap: reject a packet-modifying SK_SKB stream parser selftests/bpf: test rejection of a packet-modifying SK_SKB stream
parser

 net/core/sock_map.c                           | 20 ++++++++++++
 .../selftests/bpf/prog_tests/sockmap_strp.c   | 31 +++++++++++++++++++
 .../selftests/bpf/progs/sockmap_parse_prog.c  | 22 -------------
 .../selftests/bpf/progs/test_sockmap_strp.c   |  7 +++++
 4 files changed, 58 insertions(+), 22 deletions(-)

2.43.0

sockmap_parse_prog.c is attached as an SK_SKB stream parser and modifies the skb: it calls bpf_skb_pull_data() and writes a byte into the packet. A stream parser runs on strparser's message head and must not modify it. A resize frees the frag_list segments strparser still tracks, leading to a use-after-free.

Make the parser read-only. It only needs to return the message length, which keeps it attaching once packet-modifying parsers are rejected.

---
 .../selftests/bpf/progs/sockmap_parse_prog.c  | 22 -------------------
 1 file changed, 22 deletions(-)

index c9abfe3a11af..56e9aebf05f2 100644
@@ -5,28 +5,6 @@
 SEC("sk_skb1")
 int bpf_prog1(struct __sk_buff *skb)
 {
-	void *data_end = (void *)(long) skb->data_end;
-	void *data = (void *)(long) skb->data;
-	__u8 *d = data;
-	int err;
-
-	if (data + 10 > data_end) {
-		err = bpf_skb_pull_data(skb, 10);
-		if (err)
-			return SK_DROP;
-
-		data_end = (void *)(long)skb->data_end;
-		data = (void *)(long)skb->data;
-		if (data + 10 > data_end)
-			return SK_DROP;
-	}
-
-	/* This write/read is a bit pointless but tests the verifier and
-	 * strparser handler for read/write pkt data and access into sk
-	 * fields.
-	 */
-	d = data;
-	d[7] = 1;
 	return skb->len;
 }
 

2.43.0

sk_psock_strp_parse() runs the BPF_PROG_TYPE_SK_SKB stream-parser program to find the length of the next message. strparser assembles a message out of several received skbs by chaining them onto the head's frag_list and recording where to append the next one in strp->skb_nextp:

	*strp->skb_nextp = skb;
	strp->skb_nextp = &skb->next;

and then calls the parser on the head:

	len = (*strp->cb.parse_msg)(strp, head);

The parser is only meant to inspect the skb, but the program may call bpf_skb_change_tail() -- or the sibling bpf_skb_pull_data(), bpf_skb_change_head(), bpf_skb_adjust_room(), all allowed for SK_SKB. Once the head carries a frag_list these go

	... -> skb_ensure_writable -> pskb_may_pull -> __pskb_pull_tail

and __pskb_pull_tail() frees the frag_list skbs that strparser still tracks through skb_nextp:

	while ((list = skb_shinfo(skb)->frag_list) != insp) {
		skb_shinfo(skb)->frag_list = list->next;
		consume_skb(list);
	}

strp->skb_nextp now points into a freed sk_buff. The next segment of the same message arrives in __strp_recv(), which links it with *strp->skb_nextp = skb, an 8-byte write into the freed skb. The free and the write happen in different __strp_recv() calls, so the message has to span at least three segments before it triggers.

  BUG: KASAN: slab-use-after-free in __strp_recv+0x447/0xda0
  Write of size 8 at addr ffff88810db86140 by task repro/349

  Call Trace:
   <IRQ>
   __strp_recv+0x447/0xda0
   __tcp_read_sock+0x13d/0x590
   tcp_bpf_strp_read_sock+0x195/0x320
   strp_data_ready+0x267/0x340
   sk_psock_strp_data_ready+0x1ce/0x350
   tcp_data_queue+0x1364/0x2fd0
   tcp_rcv_established+0xe07/0x1640
   [...]

  Allocated by task 349:
   skb_clone+0x17b/0x210
   __strp_recv+0x2c3/0xda0
   __tcp_read_sock+0x13d/0x590
   [...]

  Freed by task 349:
   kmem_cache_free+0x150/0x570
   __pskb_pull_tail+0x57b/0xc20
   skb_ensure_writable+0x236/0x260
   __bpf_skb_change_tail+0x1d4/0x590
   sk_skb_change_tail+0x2a/0x40
   bpf_prog_1b285dcd6c41373e+0x27/0x30
   bpf_prog_run_pin_on_cpu+0xf3/0x260
   sk_psock_strp_parse+0x118/0x1e0
   __strp_recv+0x4f6/0xda0
   [...]

The same resize also leaves the head's length inconsistent with its frags, so a later __pskb_pull_tail() can instead hit the BUG_ON(skb_copy_bits(...)) in net/core/skbuff.c.

A stream parser is only meant to measure the next message, not to modify the packet. Reject a parser whose program can change packet data (prog->aux->changes_pkt_data) at attach time. The check is shared by sock_map_prog_update() and sock_map_link_update_prog(), which between them cover prog attach, link create and link update. Verdict programs are unaffected and may still modify the skb.

---
 net/core/sock_map.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

index 99e3789492a0..c60ba6d292f9 100644
@@ -1515,6 +1515,17 @@ static int sock_map_prog_link_lookup(struct bpf_map *map, struct bpf_prog ***ppr
 	return 0;
 }
 
+static int sock_map_prog_attach_check(enum bpf_attach_type attach_type,
+				      struct bpf_prog *prog)
+{
+	/* A stream parser must not modify the skb, only measure it. */
+	if (prog && attach_type == BPF_SK_SKB_STREAM_PARSER &&
+	    prog->aux->changes_pkt_data)
+		return -EINVAL;
+
+	return 0;
+}
+
 /* Handle the following four cases:
  * prog_attach: prog != NULL, old == NULL, link == NULL
  * prog_detach: prog == NULL, old != NULL, link == NULL

@@ -1533,6 +1544,10 @@ static int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog,
 	if (ret)
 		return ret;
 
+	ret = sock_map_prog_attach_check(which, prog);
+	if (ret)
+		return ret;
+
 	/* for prog_attach/prog_detach/link_attach, return error if a bpf_link
 	 * exists for that prog.
 	 */

@@ -1776,6 +1791,11 @@ static int sock_map_link_update_prog(struct bpf_link *link,
 		ret = -EINVAL;
 		goto out;
 	}
+
+	ret = sock_map_prog_attach_check(link->attach_type, prog);
+	if (ret)
+		goto out;
+
 	if (!sockmap_link->map) {
 		ret = -ENOLINK;
 		goto out;

2.43.0

Verify that attaching an SK_SKB stream parser that can modify the packet is rejected, while a read-only parser still attaches.

---
 .../selftests/bpf/prog_tests/sockmap_strp.c   | 31 +++++++++++++++++++
 .../selftests/bpf/progs/test_sockmap_strp.c   |  7 +++++
 2 files changed, 38 insertions(+)

index 621b3b71888e..1d7231728eaf 100644
@@ -431,6 +431,35 @@ static void test_sockmap_strp_verdict(int family, int sotype)
 	test_sockmap_strp__destroy(strp);
 }
 
+static void test_sockmap_strp_parser_reject(void)
+{
+	struct test_sockmap_strp *strp = NULL;
+	int parser_mod, parser_ro, link;
+	int err, map;
+
+	strp = test_sockmap_strp__open_and_load();
+	if (!ASSERT_OK_PTR(strp, "test_sockmap_strp__open_and_load"))
+		return;
+
+	map = bpf_map__fd(strp->maps.sock_map);
+	parser_mod = bpf_program__fd(strp->progs.prog_skb_parser_resize);
+	parser_ro = bpf_program__fd(strp->progs.prog_skb_parser);
+
+	err = bpf_prog_attach(parser_mod, map, BPF_SK_SKB_STREAM_PARSER, 0);
+	ASSERT_ERR(err, "bpf_prog_attach parser_mod");
+
+	link = bpf_link_create(parser_ro, map, BPF_SK_SKB_STREAM_PARSER, NULL);
+	if (!ASSERT_GE(link, 0, "bpf_link_create parser_ro"))
+		goto out;
+
+	err = bpf_link_update(link, parser_mod, NULL);
+	ASSERT_ERR(err, "bpf_link_update parser_mod");
+out:
+	if (link >= 0)
+		close(link);
+	test_sockmap_strp__destroy(strp);
+}
+
 void test_sockmap_strp(void)
 {
 	if (test__start_subtest("sockmap strp tcp pass"))

@@ -451,4 +480,6 @@ void test_sockmap_strp(void)
 		test_sockmap_strp_multiple_pkt(AF_INET, SOCK_STREAM);
 	if (test__start_subtest("sockmap strp tcp dispatch"))
 		test_sockmap_strp_dispatch_pkt(AF_INET, SOCK_STREAM);
+	if (test__start_subtest("sockmap strp parser reject pkt mod"))
+		test_sockmap_strp_parser_reject();
 }

index dde3d5bec515..fe88fa6d40bc 100644
@@ -50,4 +50,11 @@ int prog_skb_parser_partial(struct __sk_buff *skb)
 	return 10;
 }
 
+SEC("sk_skb/stream_parser")
+int prog_skb_parser_resize(struct __sk_buff *skb)
+{
+	bpf_skb_change_tail(skb, skb->len, 0);
+	return skb->len;
+}
+
 char _license[] SEC("license") = "GPL";

2.43.0

Hi Sechang, all,

This series broke test_maps (test_sockmap subtest) on the bpf tree. Currently on BPF CI the test fails on bpf, but passes on bpf-next (it doesn't have the series yet).

test_maps fails with:

+ taskset 0xF ./test_maps
[ 8.352378] clocksource: Watchdog remote CPU 2 read timed out
Failed sockmap unexpected timeout

See test_maps.c:995 in test_sockmap(): the 30s select() times out and test_maps exits 1. Note there is no "Failed stream parser bpf prog attach" message, the parser attaches fine.

The series was merged into bpf on 2026-06-26 00:42 UTC

CI runs:
last good (pre-merge, 06-25):
https://github.com/kernel-patches/bpf/actions/runs/28158326456
first bad (post-merge, 06-26):
https://github.com/kernel-patches/bpf/actions/runs/28210181858
recent bad (06-30):
https://github.com/kernel-patches/bpf/actions/runs/28475936023

Confirmed locally reverting the 3 commits and rebuilding makes test_sockmap pass again.

Could you please help investigate?

Thanks!

I'll work on this.