diff mbox series

[net] tcp: fix TCP_SKB_CB(skb)->tcp_tw_isn not being used

Message ID 1609192760-4505-1-git-send-email-yangpc@wangsu.com (mailing list archive)
State New, archived
Delegated to: Netdev Maintainers
Headers show
Series [net] tcp: fix TCP_SKB_CB(skb)->tcp_tw_isn not being used | expand

Commit Message

Pengcheng Yang Dec. 28, 2020, 9:59 p.m. UTC 
TCP_SKB_CB(skb)->tcp_tw_isn contains an ISN, chosen by
tcp_timewait_state_process() , when SYN is received in TIMEWAIT state.
But tcp_tw_isn is not used because it is overwritten by
tcp_v4_restore_cb() after commit eeea10b83a13 ("tcp: add
tcp_v4_fill_cb()/tcp_v4_restore_cb()").

To fix this case, we record tcp_tw_isn before tcp_v4_restore_cb() and
then set it in tcp_v4_fill_cb(). V6 does the same.

Fixes: eeea10b83a13 ("tcp: add tcp_v4_fill_cb()/tcp_v4_restore_cb()")
Reported-by: chenc <chenc9@wangsu.com>
Signed-off-by: Pengcheng Yang <yangpc@wangsu.com>
---
 net/ipv4/tcp_ipv4.c | 14 ++++++++------
 net/ipv6/tcp_ipv6.c | 14 ++++++++------
 2 files changed, 16 insertions(+), 12 deletions(-)

Comments

Jakub Kicinski Jan. 14, 2021, 4:53 p.m. UTC | #1 
On Tue, 29 Dec 2020 05:59:20 +0800 Pengcheng Yang wrote:
> TCP_SKB_CB(skb)->tcp_tw_isn contains an ISN, chosen by
> tcp_timewait_state_process() , when SYN is received in TIMEWAIT state.
> But tcp_tw_isn is not used because it is overwritten by
> tcp_v4_restore_cb() after commit eeea10b83a13 ("tcp: add
> tcp_v4_fill_cb()/tcp_v4_restore_cb()").
> 
> To fix this case, we record tcp_tw_isn before tcp_v4_restore_cb() and
> then set it in tcp_v4_fill_cb(). V6 does the same.
> 
> Fixes: eeea10b83a13 ("tcp: add tcp_v4_fill_cb()/tcp_v4_restore_cb()")
> Reported-by: chenc <chenc9@wangsu.com>
> Signed-off-by: Pengcheng Yang <yangpc@wangsu.com>

Please fix the date and resend. This patch came in last night, 
but it has a date of December 28th.
Eric Dumazet Jan. 14, 2021, 5:11 p.m. UTC | #2 
On Thu, Jan 14, 2021 at 5:53 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Tue, 29 Dec 2020 05:59:20 +0800 Pengcheng Yang wrote:
> > TCP_SKB_CB(skb)->tcp_tw_isn contains an ISN, chosen by
> > tcp_timewait_state_process() , when SYN is received in TIMEWAIT state.
> > But tcp_tw_isn is not used because it is overwritten by
> > tcp_v4_restore_cb() after commit eeea10b83a13 ("tcp: add
> > tcp_v4_fill_cb()/tcp_v4_restore_cb()").
> >
> > To fix this case, we record tcp_tw_isn before tcp_v4_restore_cb() and
> > then set it in tcp_v4_fill_cb(). V6 does the same.
> >
> > Fixes: eeea10b83a13 ("tcp: add tcp_v4_fill_cb()/tcp_v4_restore_cb()")
> > Reported-by: chenc <chenc9@wangsu.com>
> > Signed-off-by: Pengcheng Yang <yangpc@wangsu.com>
>
> Please fix the date and resend. This patch came in last night,
> but it has a date of December 28th.

Not this whole madness about tcp_v4_fill_cb()/tcp_v4_restore_cb()
could be reverted
now we have an RB tree for out-of-order packets.
Pengcheng Yang Jan. 15, 2021, 3:03 a.m. UTC | #3 
On Fri, Jan 15, 2021 at 1:12 AM Eric Dumazet <edumazet@google.com> wrote:
> 
> On Thu, Jan 14, 2021 at 5:53 PM Jakub Kicinski <kuba@kernel.org> wrote:
> >
> > On Tue, 29 Dec 2020 05:59:20 +0800 Pengcheng Yang wrote:
> > > TCP_SKB_CB(skb)->tcp_tw_isn contains an ISN, chosen by
> > > tcp_timewait_state_process() , when SYN is received in TIMEWAIT state.
> > > But tcp_tw_isn is not used because it is overwritten by
> > > tcp_v4_restore_cb() after commit eeea10b83a13 ("tcp: add
> > > tcp_v4_fill_cb()/tcp_v4_restore_cb()").
> > >
> > > To fix this case, we record tcp_tw_isn before tcp_v4_restore_cb() and
> > > then set it in tcp_v4_fill_cb(). V6 does the same.
> > >
> > > Fixes: eeea10b83a13 ("tcp: add tcp_v4_fill_cb()/tcp_v4_restore_cb()")
> > > Reported-by: chenc <chenc9@wangsu.com>
> > > Signed-off-by: Pengcheng Yang <yangpc@wangsu.com>
> >
> > Please fix the date and resend. This patch came in last night,
> > but it has a date of December 28th.
> 
> Not this whole madness about tcp_v4_fill_cb()/tcp_v4_restore_cb()
> could be reverted

This makes isn always been 0 in tcp_conn_request(), because tcp_tw_isn
is always initialized to 0 in tcp_v4_fill_cb() and tcp_tw_isn becomes meaningless.
Do we need to make tcp_tw_isn work again?

> now we have an RB tree for out-of-order packets.

I don't quite understand how setting tcp_tw_isn for SYN would affect
out_of_order_queue?
diff mbox series

Patch

diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 58207c7..c8cceaa 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -1892,7 +1892,7 @@  static void tcp_v4_restore_cb(struct sk_buff *skb)
 }
 
 static void tcp_v4_fill_cb(struct sk_buff *skb, const struct iphdr *iph,
-			   const struct tcphdr *th)
+			   const struct tcphdr *th, __u32 isn)
 {
 	/* This is tricky : We move IPCB at its correct location into TCP_SKB_CB()
 	 * barrier() makes sure compiler wont play fool^Waliasing games.
@@ -1906,7 +1906,7 @@  static void tcp_v4_fill_cb(struct sk_buff *skb, const struct iphdr *iph,
 				    skb->len - th->doff * 4);
 	TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
 	TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
-	TCP_SKB_CB(skb)->tcp_tw_isn = 0;
+	TCP_SKB_CB(skb)->tcp_tw_isn = isn;
 	TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph);
 	TCP_SKB_CB(skb)->sacked	 = 0;
 	TCP_SKB_CB(skb)->has_rxtstamp =
@@ -1927,6 +1927,7 @@  int tcp_v4_rcv(struct sk_buff *skb)
 	const struct tcphdr *th;
 	bool refcounted;
 	struct sock *sk;
+	__u32 isn = 0;
 	int ret;
 
 	if (skb->pkt_type != PACKET_HOST)
@@ -1993,7 +1994,7 @@  int tcp_v4_rcv(struct sk_buff *skb)
 		if (!tcp_filter(sk, skb)) {
 			th = (const struct tcphdr *)skb->data;
 			iph = ip_hdr(skb);
-			tcp_v4_fill_cb(skb, iph, th);
+			tcp_v4_fill_cb(skb, iph, th, isn);
 			nsk = tcp_check_req(sk, skb, req, false, &req_stolen);
 		}
 		if (!nsk) {
@@ -2038,7 +2039,7 @@  int tcp_v4_rcv(struct sk_buff *skb)
 		goto discard_and_relse;
 	th = (const struct tcphdr *)skb->data;
 	iph = ip_hdr(skb);
-	tcp_v4_fill_cb(skb, iph, th);
+	tcp_v4_fill_cb(skb, iph, th, isn);
 
 	skb->dev = NULL;
 
@@ -2075,7 +2076,7 @@  int tcp_v4_rcv(struct sk_buff *skb)
 	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 		goto discard_it;
 
-	tcp_v4_fill_cb(skb, iph, th);
+	tcp_v4_fill_cb(skb, iph, th, isn);
 
 	if (tcp_checksum_complete(skb)) {
 csum_error:
@@ -2103,7 +2104,7 @@  int tcp_v4_rcv(struct sk_buff *skb)
 		goto discard_it;
 	}
 
-	tcp_v4_fill_cb(skb, iph, th);
+	tcp_v4_fill_cb(skb, iph, th, isn);
 
 	if (tcp_checksum_complete(skb)) {
 		inet_twsk_put(inet_twsk(sk));
@@ -2121,6 +2122,7 @@  int tcp_v4_rcv(struct sk_buff *skb)
 		if (sk2) {
 			inet_twsk_deschedule_put(inet_twsk(sk));
 			sk = sk2;
+			isn = TCP_SKB_CB(skb)->tcp_tw_isn;
 			tcp_v4_restore_cb(skb);
 			refcounted = false;
 			goto process;
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 0e1509b..a2ff19d 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1557,7 +1557,7 @@  static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 }
 
 static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
-			   const struct tcphdr *th)
+			   const struct tcphdr *th, __u32 isn)
 {
 	/* This is tricky: we move IP6CB at its correct location into
 	 * TCP_SKB_CB(). It must be done after xfrm6_policy_check(), because
@@ -1573,7 +1573,7 @@  static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
 				    skb->len - th->doff*4);
 	TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
 	TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
-	TCP_SKB_CB(skb)->tcp_tw_isn = 0;
+	TCP_SKB_CB(skb)->tcp_tw_isn = isn;
 	TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
 	TCP_SKB_CB(skb)->sacked = 0;
 	TCP_SKB_CB(skb)->has_rxtstamp =
@@ -1589,6 +1589,7 @@  INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 	const struct ipv6hdr *hdr;
 	bool refcounted;
 	struct sock *sk;
+	__u32 isn = 0;
 	int ret;
 	struct net *net = dev_net(skb->dev);
 
@@ -1652,7 +1653,7 @@  INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 		if (!tcp_filter(sk, skb)) {
 			th = (const struct tcphdr *)skb->data;
 			hdr = ipv6_hdr(skb);
-			tcp_v6_fill_cb(skb, hdr, th);
+			tcp_v6_fill_cb(skb, hdr, th, isn);
 			nsk = tcp_check_req(sk, skb, req, false, &req_stolen);
 		}
 		if (!nsk) {
@@ -1695,7 +1696,7 @@  INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 		goto discard_and_relse;
 	th = (const struct tcphdr *)skb->data;
 	hdr = ipv6_hdr(skb);
-	tcp_v6_fill_cb(skb, hdr, th);
+	tcp_v6_fill_cb(skb, hdr, th, isn);
 
 	skb->dev = NULL;
 
@@ -1730,7 +1731,7 @@  INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 		goto discard_it;
 
-	tcp_v6_fill_cb(skb, hdr, th);
+	tcp_v6_fill_cb(skb, hdr, th, isn);
 
 	if (tcp_checksum_complete(skb)) {
 csum_error:
@@ -1757,7 +1758,7 @@  INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 		goto discard_it;
 	}
 
-	tcp_v6_fill_cb(skb, hdr, th);
+	tcp_v6_fill_cb(skb, hdr, th, isn);
 
 	if (tcp_checksum_complete(skb)) {
 		inet_twsk_put(inet_twsk(sk));
@@ -1780,6 +1781,7 @@  INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 			struct inet_timewait_sock *tw = inet_twsk(sk);
 			inet_twsk_deschedule_put(tw);
 			sk = sk2;
+			isn = TCP_SKB_CB(skb)->tcp_tw_isn;
 			tcp_v6_restore_cb(skb);
 			refcounted = false;
 			goto process;