[PATCH v4 0/4] Fix some mptcp syncookie process bugs

[PATCH v4 0/4] Fix some mptcp syncookie process bugs

[wujianguo106]

From: Jianguo Wu <wujianguo@chinatelecom.cn>

v3->v4:
patch1: using seq and sport/dport for hashing, and ignore network headers altogether,
        as suggest by Florian
patch2: no changes
patch3: no changes
patch4: no changes 

v2->v3:
patch1: directly use inet6_ehashfn() for IPv6
patch2: no changes
patch3: no changes
patch4: add Fixes tag.

v1->v2:
patch1: handle ipv6 sockets/addresses,
        always use 4-tuple drived hash and never look at skb->hash
patch2: no changes
patch3: split into 2 patches.
patch4: new added.


Jianguo Wu (4):
  mptcp: fix warning in __skb_flow_dissect() when do syn cookie for
    subflow join
  mptcp: remove redundant req destruct in subflow_check_req()
  mptcp: fix syncookie process if mptcp can not_accept new subflow
  mptcp: avoid processing packet if a subflow reset

 net/mptcp/options.c    |  2 ++
 net/mptcp/subflow.c    | 11 +++--------
 net/mptcp/syncookies.c | 16 +++++++++++++++-
 3 files changed, 20 insertions(+), 9 deletions(-)

-- 
1.8.3.1

[PATCH v4 1/4] mptcp: fix warning in __skb_flow_dissect() when do syn cookie for subflow join

[wujianguo106]

From: Jianguo Wu <wujianguo@chinatelecom.cn>

I got the following warning message while doing the test:

[   55.552626] TCP: request_sock_subflow: Possible SYN flooding on port 8099. Sending cookies.  Check SNMP counters.
[   55.553024] ------------[ cut here ]------------
[   55.553027] WARNING: CPU: 0 PID: 10 at net/core/flow_dissector.c:984 __skb_flow_dissect+0x280/0x1650
...
[   55.553117] CPU: 0 PID: 10 Comm: ksoftirqd/0 Not tainted 5.12.0+ #18
[   55.553121] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 02/27/2020
[   55.553124] RIP: 0010:__skb_flow_dissect+0x280/0x1650
...
[   55.553133] RSP: 0018:ffffb79580087770 EFLAGS: 00010246
[   55.553137] RAX: 0000000000000000 RBX: ffffffff8ddb58e0 RCX: ffffb79580087888
[   55.553139] RDX: ffffffff8ddb58e0 RSI: ffff8f7e4652b600 RDI: 0000000000000000
[   55.553141] RBP: ffffb79580087858 R08: 0000000000000000 R09: 0000000000000008
[   55.553143] R10: 000000008c622965 R11: 00000000d3313a5b R12: ffff8f7e4652b600
[   55.553146] R13: ffff8f7e465c9062 R14: 0000000000000000 R15: ffffb79580087888
[   55.553149] FS:  0000000000000000(0000) GS:ffff8f7f75e00000(0000) knlGS:0000000000000000
[   55.553152] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   55.553154] CR2: 00007f73d1d19000 CR3: 0000000135e10004 CR4: 00000000003706f0
[   55.553160] Call Trace:
[   55.553166]  ? __sha256_final+0x67/0xd0
[   55.553173]  ? sha256+0x7e/0xa0
[   55.553177]  __skb_get_hash+0x57/0x210
[   55.553182]  subflow_init_req_cookie_join_save+0xac/0xc0
[   55.553189]  subflow_check_req+0x474/0x550
[   55.553195]  ? ip_route_output_key_hash+0x67/0x90
[   55.553200]  ? xfrm_lookup_route+0x1d/0xa0
[   55.553207]  subflow_v4_route_req+0x8e/0xd0
[   55.553212]  tcp_conn_request+0x31e/0xab0
[   55.553218]  ? selinux_socket_sock_rcv_skb+0x116/0x210
[   55.553224]  ? tcp_rcv_state_process+0x179/0x6d0
[   55.553229]  tcp_rcv_state_process+0x179/0x6d0
[   55.553235]  tcp_v4_do_rcv+0xaf/0x220
[   55.553239]  tcp_v4_rcv+0xce4/0xd80
[   55.553243]  ? ip_route_input_rcu+0x246/0x260
[   55.553248]  ip_protocol_deliver_rcu+0x35/0x1b0
[   55.553253]  ip_local_deliver_finish+0x44/0x50
[   55.553258]  ip_local_deliver+0x6c/0x110
[   55.553262]  ? ip_rcv_finish_core.isra.19+0x5a/0x400
[   55.553267]  ip_rcv+0xd1/0xe0
...

After debugging, I found in __skb_flow_dissect(), skb->dev and skb->sk are both NULL,
then net is NULL, and trigger WARN_ON_ONCE(!net), actually net is always NULL in this
code path, as skb->dev is set to NULL in tcp_v4_rcv(), and skb->sk is never set.

Code snippet in __skb_flow_dissect() that trigger warning:
  975         if (skb) {
  976                 if (!net) {
  977                         if (skb->dev)
  978                                 net = dev_net(skb->dev);
  979                         else if (skb->sk)
  980                                 net = sock_net(skb->sk);
  981                 }
  982         }
  983
  984         WARN_ON_ONCE(!net);

So, using seq and transport header derived hash.

Fixes: 9466a1ccebbe("mptcp: enable JOIN requests even if cookies are in use").
Suggested-by: Paolo Abeni <pabeni@redhat.com>
Suggested-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
---
 net/mptcp/syncookies.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)

diff --git a/net/mptcp/syncookies.c b/net/mptcp/syncookies.c
index abe0fd099746..37127781aee9 100644
--- a/net/mptcp/syncookies.c
+++ b/net/mptcp/syncookies.c
@@ -37,7 +37,21 @@ struct join_entry {
 
 static u32 mptcp_join_entry_hash(struct sk_buff *skb, struct net *net)
 {
-	u32 i = skb_get_hash(skb) ^ net_hash_mix(net);
+	static u32 mptcp_join_hash_secret __read_mostly;
+	struct tcphdr *th = tcp_hdr(skb);
+	u32 seq, i;
+
+	net_get_random_once(&mptcp_join_hash_secret,
+			    sizeof(mptcp_join_hash_secret));
+
+	if (th->syn)
+		seq = TCP_SKB_CB(skb)->seq;
+	else
+		seq = TCP_SKB_CB(skb)->seq - 1;
+
+	i = jhash_3words(seq, net_hash_mix(net),
+			 (__force __u32)th->source << 16 | (__force __u32)th->dest,
+			 mptcp_join_hash_secret);
 
 	return i % ARRAY_SIZE(join_entries);
 }
-- 
1.8.3.1

[PATCH v4 2/4] mptcp: remove redundant req destruct in subflow_check_req()

[wujianguo106]

From: Jianguo Wu <wujianguo@chinatelecom.cn>

In subflow_check_req(), if subflow sport is mismatch, will put msk,
destroy token, and destruct req, then return -EPERM, which can be
done by subflow_req_destructor() via:
  tcp_conn_request()
    |--__reqsk_free()
      |--subflow_req_destructor()
So we should remove these redundant code, otherwise will call
tcp_v4_reqsk_destructor() twice, and may double free inet_rsk(req)->ireq_opt.

Fixes: 5bc56388c74f ("mptcp: add port number check for MP_JOIN")
Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
---
 net/mptcp/subflow.c | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index 6b1cd4257edf..75ed530706c0 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -213,11 +213,6 @@ static int subflow_check_req(struct request_sock *req,
 				 ntohs(inet_sk(sk_listener)->inet_sport),
 				 ntohs(inet_sk((struct sock *)subflow_req->msk)->inet_sport));
 			if (!mptcp_pm_sport_in_anno_list(subflow_req->msk, sk_listener)) {
-				sock_put((struct sock *)subflow_req->msk);
-				mptcp_token_destroy_request(req);
-				tcp_request_sock_ops.destructor(req);
-				subflow_req->msk = NULL;
-				subflow_req->mp_join = 0;
 				SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_MISMATCHPORTSYNRX);
 				return -EPERM;
 			}
-- 
1.8.3.1

[PATCH v4 3/4] mptcp: fix syncookie process if mptcp can not_accept new subflow

[wujianguo106]

From: Jianguo Wu <wujianguo@chinatelecom.cn>

Lots of "TCP: tcp_fin: Impossible, sk->sk_state=7" in client side
when doing stress testing.

There are at least two cases may trigger this warning:
1.mptcp is in syncookie, and server recv MP_JOIN SYN request,
  in subflow_check_req(), the mptcp_can_accept_new_subflow()
  return false, so subflow_init_req_cookie_join_save() isn't
  called, i.e. not store the data present in the MP_JOIN syn
  request and the random nonce in hash table - join_entries[],
  but still send synack. When recv 3rd-ack,
  mptcp_token_join_cookie_init_state() will return false, and
  3rd-ack is dropped, then if mptcp conn is closed by client,
  client will send a DATA_FIN and a MPTCP FIN, the DATA_FIN
  doesn't have MP_CAPABLE or MP_JOIN,
  so mptcp_subflow_init_cookie_req() will return 0, and pass
  the cookie check, MP_JOIN request is fallback to normal TCP.
  Server will send a TCP FIN if closed, in client side,
  when process TCP FIN, it will do reset, the code path is:
    tcp_data_queue()->mptcp_incoming_options()
      ->check_fully_established()->mptcp_subflow_reset().
  mptcp_subflow_reset() will set sock state to TCP_CLOSE,
  so tcp_fin will hit TCP_CLOSE, and print the warning.
2.mptcp is in syncookie, and server recv 3rd-ack, in
  mptcp_subflow_init_cookie_req(), mptcp_can_accept_new_subflow()
  return false, and subflow_req->mp_join is not set to 1,
  so in subflow_syn_recv_sock() will not reset the MP_JOIN
  subflow, but fallback to normal TCP, and then the same thing
  happens when server will send a TCP FIN if closed.

For case1, subflow_check_req() return -EPERM,
then tcp_conn_request() will drop MP_JOIN SYN.

For case2, let subflow_syn_recv_sock() call mptcp_can_accept_new_subflow(),
and do fatal fallback, send reset.

Fixes: 9466a1ccebbe("mptcp: enable JOIN requests even if cookies are in use")
Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
---
 net/mptcp/subflow.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index 75ed530706c0..6d98e19a20aa 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -224,6 +224,8 @@ static int subflow_check_req(struct request_sock *req,
 		if (unlikely(req->syncookie)) {
 			if (mptcp_can_accept_new_subflow(subflow_req->msk))
 				subflow_init_req_cookie_join_save(subflow_req, skb);
+			else
+				return -EPERM;
 		}
 
 		pr_debug("token=%u, remote_nonce=%u msk=%p", subflow_req->token,
@@ -263,9 +265,7 @@ int mptcp_subflow_init_cookie_req(struct request_sock *req,
 		if (!mptcp_token_join_cookie_init_state(subflow_req, skb))
 			return -EINVAL;
 
-		if (mptcp_can_accept_new_subflow(subflow_req->msk))
-			subflow_req->mp_join = 1;
-
+		subflow_req->mp_join = 1;
 		subflow_req->ssn_offset = TCP_SKB_CB(skb)->seq - 1;
 	}
 
-- 
1.8.3.1

[PATCH v4 4/4] mptcp: avoid processing packet if a subflow reset

[wujianguo106]

From: Jianguo Wu <wujianguo@chinatelecom.cn>

If check_fully_established() causes a subflow reset, it should not
continue to process the packet in tcp_data_queue().

setting:
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;

so that the following check will drop the pkt in
tcp_data_queue():
  if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq) {
	__kfree_skb(skb);
	return;
  }

Fixes: d582484726c4 ("mptcp: fix fallback for MP_JOIN subflows")
Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
---
 net/mptcp/options.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/net/mptcp/options.c b/net/mptcp/options.c
index 1aec01686c1a..01a1bf60b187 100644
--- a/net/mptcp/options.c
+++ b/net/mptcp/options.c
@@ -926,6 +926,8 @@ static bool check_fully_established(struct mptcp_sock *msk, struct sock *ssk,
 	return true;
 
 reset:
+	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;
+
 	mptcp_subflow_reset(ssk);
 	return false;
 }
-- 
1.8.3.1

Re: [PATCH v4 4/4] mptcp: avoid processing packet if a subflow reset

[Florian Westphal]

wujianguo106@163.com <wujianguo106@163.com> wrote:
> diff --git a/net/mptcp/options.c b/net/mptcp/options.c
> index 1aec01686c1a..01a1bf60b187 100644
> --- a/net/mptcp/options.c
> +++ b/net/mptcp/options.c
> @@ -926,6 +926,8 @@ static bool check_fully_established(struct mptcp_sock *msk, struct sock *ssk,
>  	return true;
>  
>  reset:
> +	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;

Magic.  Please at least add a comment here.

Re: [PATCH v4 4/4] mptcp: avoid processing packet if a subflow reset

[Jianguo Wu]

Hi Florian,

On 2021/6/15 23:19, Florian Westphal wrote:
> wujianguo106@163.com <wujianguo106@163.com> wrote:
>> diff --git a/net/mptcp/options.c b/net/mptcp/options.c
>> index 1aec01686c1a..01a1bf60b187 100644
>> --- a/net/mptcp/options.c
>> +++ b/net/mptcp/options.c
>> @@ -926,6 +926,8 @@ static bool check_fully_established(struct mptcp_sock *msk, struct sock *ssk,
>>  	return true;
>>  
>>  reset:
>> +	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;
> 
> Magic.  Please at least add a comment here.
> 

Ok, will add a comment, thanks!