Skip to contentSkip to conversation

tls: fix RX desync on overlapping skbs

PATCH Applied Maximilian Immanuel Brandtner 2 participants 3 messages Aug 13 – Aug 18

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 tls: fix RX desync on overlapping skbs

The TCP receive queue can hold adjacent skbs whose sequence ranges overlap. The tls fast-path reads the record header with skb_copy_bits() by byte offset, which assumes skbs do not overlap, so a header split across the overlap is misread and the connection aborts (-EMSGSIZE/-EINVAL). tls_strp_check_queue_ok() detects such overlaps but only ran after the header was parsed, never covering the header itself.

Observed with parallel kTLS connections on:
- ConnectX-7 + IPsec crypto offload + GRO
- VirtIO (8 queues) + GRO

Fixes 84c61fe1a75b ("tls: rx: do not use the standard strparser")

---
v2:
- move the stm.offset addition into tls_strp_check_queue_ok() No functional change from v1
Tested on Linux kernel 7.2-rc6

---
 net/tls/tls_strp.c | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)
net/tls/tls_strp.c +11 -5
index 61b10c697ecc..6cc222008d95 100644
@@ -430,9 +430,10 @@ static int tls_strp_read_copy(struct tls_strparser *strp, bool qshort)
 	return 0;
 }
 
-static bool tls_strp_check_queue_ok(struct tls_strparser *strp)
+static bool tls_strp_check_queue_ok(struct tls_strparser *strp,
+				    unsigned int len)
 {
-	unsigned int len = strp->stm.offset + strp->stm.full_len;
+	unsigned int remaining = strp->stm.offset + len;
 	struct sk_buff *first, *skb;
 	u32 seq;
 
@@ -443,9 +444,9 @@ static bool tls_strp_check_queue_ok(struct tls_strparser *strp)
 	/* Make sure there's no duplicate data in the queue,
 	 * and the decrypted status matches.
 	 */
-	while (skb->len < len) {
+	while (skb->len < remaining) {
 		seq += skb->len;
-		len -= skb->len;
+		remaining -= skb->len;
 		skb = skb->next;
 
 		if (TCP_SKB_CB(skb)->seq != seq)
@@ -525,6 +526,11 @@ static int tls_strp_read_sock(struct tls_strparser *strp)
 
 	tls_strp_load_anchor_with_queue(strp, inq);
 	if (!strp->stm.full_len) {
+		if (inq < TLS_HEADER_SIZE)
+			return tls_strp_read_copy(strp, true);
+		if (!tls_strp_check_queue_ok(strp, TLS_HEADER_SIZE))
+			return tls_strp_read_copy(strp, false);
+
 		sz = tls_rx_msg_size(strp, strp->anchor);
 		if (sz < 0)
 			return sz;
@@ -535,7 +541,7 @@ static int tls_strp_read_sock(struct tls_strparser *strp)
 			return tls_strp_read_copy(strp, true);
 	}
 
-	if (!tls_strp_check_queue_ok(strp))
+	if (!tls_strp_check_queue_ok(strp, strp->stm.full_len))
 		return tls_strp_read_copy(strp, false);
 
 	WRITE_ONCE(strp->msg_ready, 1);
Signature · 1 line
2.55.0
#1 lore raw

I just saw that for the v2 of this patch sashiko reported a supposed potential null dereference which I believe to be an erroneous finding.

Sashiko claims that in tls_strp_check_queue_ok() after first = skb_shinfo(strp->anchor)->frag_list; first may be NULL which would result in a null dereference on seq = TCP_SKB_CB(first)->seq;

tls_strp_check_queue_ok(strp, TLS_HEADER_SIZE) is only called if inq >= TLS_HEADER_SIZE. tcp_inq() counts real payload bytes so [copied_seq, rcv_nxt) holds at least TLS_HEADER_SIZE bytes which live in sk-

Quoted 1 line

sk_receive_queue. tcp_recv_skb(sk, copied_seq) returns the skb

covering copied_seq. tls_strp_load_anchor_with_queue() stores it in anchor->frag_list.

Therefore, skb_shinfo(strp->anchor)->frag_list != NULL for the check introduced by this patch.

Kind regards,
Max Brandtner

#2 lore raw
In reply to Maximilian Immanuel Brandtner
This patch was applied to netdev/net.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:

Quoted 8 lines · On Thu, 13 Aug 2026 14:09:44 +0200 you wrote:

The TCP receive queue can hold adjacent skbs whose sequence ranges overlap. The tls fast-path reads the record header with skb_copy_bits() by byte offset, which assumes skbs do not overlap, so a header split across the overlap is misread and the connection aborts (-EMSGSIZE/-EINVAL). tls_strp_check_queue_ok() detects such overlaps but only ran after the header was parsed, never covering the header itself.

[...]

Here is the summary with links:
- [net,v2] tls: fix RX desync on overlapping skbs

    https://git.kernel.org/netdev/net/c/9466ef3ec972

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html