Skip to contentSkip to conversation

mptcp: replace per-addr listener sockets

PATCH Superseded Florian Westphal 5 participants 14 messages Feb 17, 2022 – Feb 23, 2022

Subsystems (3)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v2 cover mptcp: replace per-addr listener sockets

Changes since v1:
- avoid uneeded sk refcount inc/dec pair
- use more descriptive name for function name
- fix typo in rfc number (Mathieu)
- don't increment listner refcount, caller assumes noref

Not changed:
- hook location, plain tcp sockets take precedence.
- per-netns listener

Global listener means other namespaces are affected in case of backlog overflow. Pernet listeners avoid that, so I don't see a strong argument for tcp refactoring that would allow global sk.

This series introduces a hidden per-netns pseudo "listener" socket to handle mptcp join requests with a valid token but that do not match an existing listening socket.

First patch extens existing helper to also restrict search to the given netns.

Second patch is a minor preparation patch: MPTCP Join requests packets that fail to find a suitable socket by means of standard address/port demultiplexing will be steered to a pseudo-listener, similar to TPROXY interception.

This pseudo listener isn't bound to an address or port (all zero), so we need to fetch the port number from the tcp header and not the listener sk.

Patch three adds a stub to the tcp demux code.
This has no functionality, its extra to make tcp datapath change stand out.

Fourth patch is the bulk work, it adds per netns listener and implements token-based socket demultiplexing.

Last patch zaps the per-address sockets from mptcp, they are not needed anymore.

Florian Westphal (5):
mptcp: check netns in mptcp_token_exists
mptcp: prefer ip address in syn skb instead of listen sk bound address
tcp: add mptcp join demultiplex hooks
mptcp: handle join requests via pernet listen socket
mptcp: remove per-address listening sockets

 include/net/mptcp.h    |  15 +++
 net/ipv4/tcp_ipv4.c    |   4 +
 net/ipv6/tcp_ipv6.c    |  23 +++--
 net/mptcp/ctrl.c       | 211 ++++++++++++++++++++++++++++++++++++++++-
 net/mptcp/pm_netlink.c |  82 +++-------------
 net/mptcp/protocol.c   |   2 +-
 net/mptcp/protocol.h   |   6 +-
 net/mptcp/subflow.c    |  12 ++-
 net/mptcp/token.c      |   5 +-
 9 files changed, 272 insertions(+), 88 deletions(-)
Signature · 1 line
2.34.1
#1 lore raw

PATCH v2 1/5 mptcp: check netns in mptcp_token_exists

This will be used by a followup patch to check if the token specified in mp_join option exists in the current netns.

At this time, tokens are unique across all namespaces, but we need to treat a token that exists in netns x as 'does not exist' when doing existence check from netns y.

---
 net/mptcp/protocol.h | 2 +-
 net/mptcp/subflow.c  | 4 +++-
 net/mptcp/token.c    | 5 +++--
 3 files changed, 7 insertions(+), 4 deletions(-)
net/mptcp/protocol.h +1 -1
index 18ca0248c084..c43ca46dbc27 100644
@@ -725,7 +725,7 @@ void mptcp_token_destroy_request(struct request_sock *req);
 int mptcp_token_new_connect(struct sock *sk);
 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
 			struct mptcp_sock *msk);
-bool mptcp_token_exists(u32 token);
+bool mptcp_token_exists(const struct net *net, u32 token);
 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
 					 long *s_num);
net/mptcp/subflow.c +3 -1
index e727d838da0e..be43077fe76e 100644
@@ -176,10 +176,12 @@ static int subflow_check_req(struct request_sock *req,
 		} while (subflow_req->local_key == 0);
 
 		if (unlikely(req->syncookie)) {
+			const struct net *net = read_pnet(&inet_rsk(req)->ireq_net);
+
 			mptcp_crypto_key_sha(subflow_req->local_key,
 					     &subflow_req->token,
 					     &subflow_req->idsn);
-			if (mptcp_token_exists(subflow_req->token)) {
+			if (mptcp_token_exists(net, subflow_req->token)) {
 				if (retries-- > 0)
 					goto again;
 				SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_TOKENFALLBACKINIT);
net/mptcp/token.c +3 -2
index f52ee7b26aed..0593c83385e0 100644
@@ -203,7 +203,7 @@ void mptcp_token_accept(struct mptcp_subflow_request_sock *req,
 	spin_unlock_bh(&bucket->lock);
 }
 
-bool mptcp_token_exists(u32 token)
+bool mptcp_token_exists(const struct net *net, u32 token)
 {
 	struct hlist_nulls_node *pos;
 	struct token_bucket *bucket;
@@ -216,7 +216,8 @@ bool mptcp_token_exists(u32 token)
 again:
 	sk_nulls_for_each_rcu(sk, pos, &bucket->msk_chain) {
 		msk = mptcp_sk(sk);
-		if (READ_ONCE(msk->token) == token)
+		if (READ_ONCE(msk->token) == token &&
+		    net_eq(sock_net(sk), net))
 			goto found;
 	}
 	if (get_nulls_value(pos) != (token & token_mask))
Signature · 1 line
2.34.1
#2 lore raw

PATCH v2 2/5 mptcp: prefer ip address in syn skb instead of listen sk bound address

In reply to Florian Westphal

Once we change mptcp to use tproxy-like scheme to steer mptcp join requests to a special pernet socket, the 'sk bound address' becomes meaningless because it will never be identical to the tcp dport/ip daddr of the on-wire packet.

Prepare for this: pass the skbuff and use the packet data instead of the address the listener socket is bound to.

---
 net/mptcp/pm_netlink.c | 17 +++++++++++++++--
 net/mptcp/protocol.h   |  2 +-
 net/mptcp/subflow.c    |  5 +++--
 3 files changed, 19 insertions(+), 5 deletions(-)
net/mptcp/pm_netlink.c +15 -2
index 56f5603c10f2..614b5d05aa62 100644
@@ -279,13 +279,26 @@ mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
 	return NULL;
 }
 
-bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk)
+static void skb_fetch_src_address(const struct sk_buff *skb,
+				  struct mptcp_addr_info *addr)
+{
+	addr->port = tcp_hdr(skb)->dest;
+	if (addr->family == AF_INET)
+		addr->addr.s_addr = ip_hdr(skb)->daddr;
+#if IS_ENABLED(CONFIG_MPTCP_IPV6)
+	else if (addr->family == AF_INET6)
+		addr->addr6 = ipv6_hdr(skb)->daddr;
+#endif
+}
+
+bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, int af, const struct sk_buff *skb)
 {
 	struct mptcp_pm_add_entry *entry;
 	struct mptcp_addr_info saddr;
 	bool ret = false;
 
-	local_address((struct sock_common *)sk, &saddr);
+	saddr.family = af;
+	skb_fetch_src_address(skb, &saddr);
 
 	spin_lock_bh(&msk->pm.lock);
 	list_for_each_entry(entry, &msk->pm.anno_list, list) {
net/mptcp/protocol.h +1 -1
index c43ca46dbc27..7bd064b68b51 100644
@@ -760,7 +760,7 @@ void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
-bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
+bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, int af, const struct sk_buff *skb);
 struct mptcp_pm_add_entry *
 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
 		       const struct mptcp_addr_info *addr, bool check_id);
net/mptcp/subflow.c +3 -2
index be43077fe76e..8be20f7b76df 100644
@@ -218,7 +218,8 @@ static int subflow_check_req(struct request_sock *req,
 			pr_debug("syn inet_sport=%d %d",
 				 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)) {
+			if (!mptcp_pm_sport_in_anno_list(subflow_req->msk,
+							 sk_listener->sk_family, skb)) {
 				SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_MISMATCHPORTSYNRX);
 				return -EPERM;
 			}
@@ -750,7 +751,7 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 				pr_debug("ack inet_sport=%d %d",
 					 ntohs(inet_sk(sk)->inet_sport),
 					 ntohs(inet_sk((struct sock *)owner)->inet_sport));
-				if (!mptcp_pm_sport_in_anno_list(owner, sk)) {
+				if (!mptcp_pm_sport_in_anno_list(owner, sk->sk_family, skb)) {
 					SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_MISMATCHPORTACKRX);
 					goto dispose_child;
 				}
Signature · 1 line
2.34.1
#3 lore raw

PATCH v2 3/5 tcp: add mptcp join demultiplex hooks

In reply to Florian Westphal

Split from the next patch to make core tcp changes more obvious:
add a dummy function that gets called after tcp socket demux came up empty.

This will be used by mptcp to check if a tcp syn contains an mptcp join option with a valid token (connection id).

If so, a hidden pernet mptcp listener socket is returned and packet resumes normally.

This patch series does not cover timewait sockets so far.

---
 include/net/mptcp.h | 5 +++++
 net/ipv4/tcp_ipv4.c | 4 ++++
 net/ipv6/tcp_ipv6.c | 4 ++++
 3 files changed, 13 insertions(+)
include/net/mptcp.h +5 -0
index 8b1afd6f5cc4..5ee422b56902 100644
@@ -197,6 +197,10 @@ static inline __be32 mptcp_reset_option(const struct sk_buff *skb)
 
 	return htonl(0u);
 }
+static inline struct sock *mptcp_handle_join(int af, struct sk_buff *skb)
+{
+	return NULL;
+}
 #else
 
 static inline void mptcp_init(void)
@@ -274,6 +278,7 @@ static inline int mptcp_subflow_init_cookie_req(struct request_sock *req,
 }
 
 static inline __be32 mptcp_reset_option(const struct sk_buff *skb)  { return htonl(0u); }
+static inline struct sock *mptcp_handle_join(int af, struct sk_buff *skb) { return NULL; }
 #endif /* CONFIG_MPTCP */
 
 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
net/ipv4/tcp_ipv4.c +4 -0
index 6873f46fc8ba..6e6675a09443 100644
@@ -2140,6 +2140,10 @@ int tcp_v4_rcv(struct sk_buff *skb)
 	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 		goto discard_it;
 
+	sk = mptcp_handle_join(AF_INET, skb);
+	if (sk)
+		goto process;
+
 	tcp_v4_fill_cb(skb, iph, th);
 
 	if (tcp_checksum_complete(skb)) {
net/ipv6/tcp_ipv6.c +4 -0
index 0c648bf07f39..788040db8e9e 100644
@@ -1782,6 +1782,10 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 		goto discard_it;
 
+	sk = mptcp_handle_join(AF_INET6, skb);
+	if (sk)
+		goto process;
+
 	tcp_v6_fill_cb(skb, hdr, th);
 
 	if (tcp_checksum_complete(skb)) {
Signature · 1 line
2.34.1
maintainer
#4 lore raw

Re: [PATCH mptcp-next v2 3/5] tcp: add mptcp join demultiplex hooks

Quoted 7 lines · Florian Westphal

Split from the next patch to make core tcp changes more obvious:
add a dummy function that gets called after tcp socket demux came up empty.

This will be used by mptcp to check if a tcp syn contains an mptcp join option with a valid token (connection id).

Quoted 6 lines

If so, a hidden pernet mptcp listener socket is returned and packet resumes normally.

This patch series does not cover timewait sockets so far.

Following-up todays mtg discussion WRT filtering incoming MPJ on existing endpoint.

If I read correctly, subflow_syn_recv_sock() will filter the incoming packets vs the annonced list, but only if the incoming packet's destination port is different from the listener port.

- incoming request targeting a bad address and/or port will reach the syn-ack before being rejected.

- incoming request targeting port 0 will likely be always accepted.

What about moving the anno_list check in __mptcp_handle_join()? Should address the 2 above points and keep the anno_list traversals to the bare minimum.

WDYT?

Cheers,

Paolo

#5 lore raw

Re: [PATCH mptcp-next v2 3/5] tcp: add mptcp join demultiplex hooks

Quoted 3 lines · Paolo Abeni

If I read correctly, subflow_syn_recv_sock() will filter the incoming packets vs the annonced list, but only if the incoming packet's destination port is different from the listener port.

Same as today, if it hit listener port then no issue...?
The listener port isn't in the announced list, so the check would fail.

Quoted 2 lines

- incoming request targeting a bad address and/or port will reach the syn-ack before being rejected.

Yes.

Quoted 1 line

- incoming request targeting port 0 will likely be always accepted.

No, port 0 is illegal, there is an explicit check that prevents such port from being wired up to the magic listener.

Quoted 3 lines

What about moving the anno_list check in __mptcp_handle_join()? Should address the 2 above points and keep the anno_list traversals to the bare minimum.

I can have a look.

maintainer
#6 lore raw

Re: [PATCH mptcp-next v2 3/5] tcp: add mptcp join demultiplex hooks

Quoted 7 lines · Florian Westphal
Quoted 3 lines · Paolo Abeni

If I read correctly, subflow_syn_recv_sock() will filter the incoming packets vs the annonced list, but only if the incoming packet's destination port is different from the listener port.

Same as today, if it hit listener port then no issue...?
The listener port isn't in the announced list, so the check would fail.

Side not not related to this series: It looks like that if we have 2 MPTCP listeners on different addresses and/or ports, and no 'signal' endpoints, the kernel will accept MPJ on either sockets for any mptcp connection, regardless of the originating listener. I'm unsure if that is expeted/fits the RFC fully.
@Mat: ^^^ WDYT?

Quoted 10 lines
Quoted 3 lines

- incoming request targeting a bad address and/or port will reach the syn-ack before being rejected.

Yes.

Quoted 1 line

- incoming request targeting port 0 will likely be always accepted.

No, port 0 is illegal, there is an explicit check that prevents such port from being wired up to the magic listener.

whoops, I missed that check, despite the huge comment. Sorry for the noise here.

Quoted 5 lines
Quoted 3 lines

What about moving the anno_list check in __mptcp_handle_join()? Should address the 2 above points and keep the anno_list traversals to the bare minimum.

I can have a look.

Thanks! It looks like some minor changes are needed to handle !ipv6 build.

Cheers,
Paolo

#7 lore raw

PATCH v2 4/5 mptcp: handle join requests via pernet listen socket

In reply to Florian Westphal

Currently mptcp adds kernel-based listener socket for all netlink-configured mptcp address endpoints.

This has caveats because kernel may interfere with unrelated programs that use same address/port pairs.

RFC 8684 says:
Demultiplexing subflow SYNs MUST be done using the token; this is unlike traditional TCP, where the destination port is used for demultiplexing SYN packets. Once a subflow is set up, demultiplexing packets is done using the 5-tuple, as in traditional TCP.

This patch deviates from this in that it retrains the existing checks of verifying the incoming requests destination vs. the list of announced addresses.

This can be relaxed later if deemed appropriate.

Furthermore, TCP-only listeners take precedence: An MPTCP peer MUST NOT announce address:port pairs that are already in use by a non-mptcp listener.
This could be changed, but it requires move of mptcp_handle_join() hook *before* the tcp port demux, i.e. an additional conditional in hotpath. As-is, the additional conditional (syn && !rst && ...) is in the 'no socket found' path.

The pernet 'listening' socket is hidden from userspace. It is not part of any hashes and not bound to any address or port.

TPROXY-like semantics apply: If tcp demux cannot find a port for a given packet, check if the packet is a syn packet with a valid join token.

If so, the pernet listener is returned and tcp processing resumes. Otherwise, handling is intentical.

This patch does not cover timewait sockets.

---
 include/net/mptcp.h  |  10 ++
 net/ipv6/tcp_ipv6.c  |  19 ++--
 net/mptcp/ctrl.c     | 211 ++++++++++++++++++++++++++++++++++++++++++-
 net/mptcp/protocol.c |   2 +-
 net/mptcp/protocol.h |   2 +-
 net/mptcp/subflow.c  |   3 +
 6 files changed, 233 insertions(+), 14 deletions(-)
include/net/mptcp.h +10 -0
index 5ee422b56902..49c188b978e1 100644
@@ -189,6 +189,7 @@ int mptcp_subflow_init_cookie_req(struct request_sock *req,
 				  struct sk_buff *skb);
 
 __be32 mptcp_get_reset_option(const struct sk_buff *skb);
+struct sock *__mptcp_handle_join(int af, struct sk_buff *skb);
 
 static inline __be32 mptcp_reset_option(const struct sk_buff *skb)
 {
@@ -199,6 +200,11 @@ static inline __be32 mptcp_reset_option(const struct sk_buff *skb)
 }
 static inline struct sock *mptcp_handle_join(int af, struct sk_buff *skb)
 {
+	const struct tcphdr *th = tcp_hdr(skb);
+
+	if (th->syn && !th->ack && !th->rst && !th->fin)
+		return __mptcp_handle_join(af, skb);
+
 	return NULL;
 }
 #else
@@ -283,9 +289,13 @@ static inline struct sock *mptcp_handle_join(int af, struct sk_buff *skb) { retu
 
 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 int mptcpv6_init(void);
+int mptcpv6_init_net(struct net *net);
+void mptcpv6_exit_net(struct net *net);
 void mptcpv6_handle_mapped(struct sock *sk, bool mapped);
 #elif IS_ENABLED(CONFIG_IPV6)
 static inline int mptcpv6_init(void) { return 0; }
+static inline int mptcpv6_init_net(struct net *net) { return 0; }
+static inline void mptcpv6_exit_net(struct net *net) { }
 static inline void mptcpv6_handle_mapped(struct sock *sk, bool mapped) { }
 #endif
 
net/ipv6/tcp_ipv6.c +11 -8
index 788040db8e9e..3b8608d35dcd 100644
@@ -2233,13 +2233,22 @@ static struct inet_protosw tcpv6_protosw = {
 
 static int __net_init tcpv6_net_init(struct net *net)
 {
-	return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
-				    SOCK_RAW, IPPROTO_TCP, net);
+	int err = inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
+				       SOCK_RAW, IPPROTO_TCP, net);
+	if (err)
+		return err;
+
+	err = mptcpv6_init_net(net);
+	if (err)
+		inet_ctl_sock_destroy(net->ipv6.tcp_sk);
+
+	return err;
 }
 
 static void __net_exit tcpv6_net_exit(struct net *net)
 {
 	inet_ctl_sock_destroy(net->ipv6.tcp_sk);
+	mptcpv6_exit_net(net);
 }
 
 static struct pernet_operations tcpv6_net_ops = {
@@ -2264,15 +2273,9 @@ int __init tcpv6_init(void)
 	if (ret)
 		goto out_tcpv6_protosw;
 
-	ret = mptcpv6_init();
-	if (ret)
-		goto out_tcpv6_pernet_subsys;
-
 out:
 	return ret;
 
-out_tcpv6_pernet_subsys:
-	unregister_pernet_subsys(&tcpv6_net_ops);
 out_tcpv6_protosw:
 	inet6_unregister_protosw(&tcpv6_protosw);
 out_tcpv6_protocol:
net/mptcp/ctrl.c +207 -4

250 lines not on this page · raw message

net/mptcp/protocol.c +1 -1
index 4599bde215b2..5b54e3c8efea 100644
@@ -3777,7 +3777,7 @@ static struct inet_protosw mptcp_v6_protosw = {
 	.flags		= INET_PROTOSW_ICSK,
 };
 
-int __init mptcp_proto_v6_init(void)
+int __net_init mptcp_proto_v6_init(void)
 {
 	int err;
 
net/mptcp/protocol.h +1 -1
index 7bd064b68b51..6a81e2a21301 100644
@@ -647,7 +647,7 @@ static inline bool mptcp_has_another_subflow(struct sock *ssk)
 
 void __init mptcp_proto_init(void);
 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
-int __init mptcp_proto_v6_init(void);
+int __net_init mptcp_proto_v6_init(void);
 #endif
 
 struct sock *mptcp_sk_clone(const struct sock *sk,
net/mptcp/subflow.c +3 -0
index 8be20f7b76df..4696d27a8994 100644
@@ -116,6 +116,9 @@ static void subflow_init_req(struct request_sock *req, const struct sock *sk_lis
 
 static bool subflow_use_different_sport(struct mptcp_sock *msk, const struct sock *sk)
 {
+	if (inet_sk(sk)->inet_sport == 0)
+		return true;
+
 	return inet_sk(sk)->inet_sport != inet_sk((struct sock *)msk)->inet_sport;
 }
 
Signature · 1 line
2.34.1
#8 lore raw

Re: [PATCH mptcp-next v2 4/5] mptcp: handle join requests via pernet listen socket

I love your patch! Yet something to improve:

Hi Florian,

I love your patch! Yet something to improve:

[auto build test ERROR on mptcp/export]
[cannot apply to linus/master v5.17-rc4 next-20220217]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Florian-Westphal/mptcp-replace-per-addr-listener-sockets/20220217-222844
base:   https://github.com/multipath-tcp/mptcp_net-next.git export
config: um-i386_defconfig (https://download.01.org/0day-ci/archive/20220218/202202180141.4Rse3JCH-lkp(a)intel.com/config)
compiler: gcc-9 (Debian 9.3.0-22) 9.3.0

reproduce (this is a W=1 build):
# https://github.com/0day-ci/linux/commit/044982520ba41e284eebe48421fad7feb55f2106
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Florian-Westphal/mptcp-replace-per-addr-listener-sockets/20220217-222844 git checkout 044982520ba41e284eebe48421fad7feb55f2106
# save the config file to linux build tree
mkdir build_dir
make W=1 O=build_dir ARCH=um SUBARCH=i386 SHELL=/bin/bash

If you fix the issue, kindly add following tag as appropriate

All errors (new ones prefixed by >>):

net/mptcp/ctrl.c: In function '__mptcp_handle_join':

>> net/mptcp/ctrl.c:251:17: error: 'struct mptcp_pernet' has no member named 'join6'; did you mean 'join4'?
     251 |   lsk = pernet->join6.sk;
         |                 ^~~~~
         |                 join4

vim +251 net/mptcp/ctrl.c

205
206 struct sock *__mptcp_handle_join(int af, struct sk_buff *skb)
207 {
208 struct mptcp_options_received mp_opt;
209 struct mptcp_pernet *pernet;
210 struct socket *ssock;
211 struct sock *lsk;
212 struct net *net;
213
214 /* paranoia check: don't allow 0 destination port,
215 * else __inet_inherit_port will insert the child socket
216 * into the phony hash slot of the pernet listener.
217 */
218 if (tcp_hdr(skb)->dest == 0)
219 return NULL;
220
221 mptcp_get_options(skb, &mp_opt);
222
223 if (!(mp_opt.suboptions & OPTIONS_MPTCP_MPJ))
224 return NULL;
225
226 net = dev_net(skb_dst(skb)->dev);
227 if (!mptcp_is_enabled(net))
228 return NULL;
229
230 /* RFC8684: If the token is unknown [..], the receiver will send
231 * back a reset (RST) signal, analogous to an unknown port in TCP,
232 * containing an MP_TCPRST option (Section 3.6) [..]
233 */
234 if (!mptcp_token_exists(net, mp_opt.token)) {
235 struct mptcp_ext *ext = skb_ext_add(skb, SKB_EXT_MPTCP);
236
237 if (ext) {
238 memset(ext, 0, sizeof(*ext));
239 ext->reset_reason = MPTCP_RST_EMPTCP;
240 }
241 return NULL;
242 }
243
244 pernet = mptcp_get_pernet(net);
245
246 switch (af) {
247 case AF_INET:
248 lsk = pernet->join4.sk;
249 break;
250 case AF_INET6:

 > 251			lsk = pernet->join6.sk;

252 break;
253 default:
254 WARN_ON_ONCE(1);
255 return NULL;
256 }
257
258 ssock = __mptcp_nmpc_socket(mptcp_sk(lsk));
259 if (WARN_ON(!ssock))
260 return NULL;
261
262 return ssock->sk;
263 }
264

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

#9 lore raw

Re: [PATCH mptcp-next v2 4/5] mptcp: handle join requests via pernet listen socket

In reply to Florian Westphal
I love your patch! Yet something to improve:

Hi Florian,

I love your patch! Yet something to improve:

[auto build test ERROR on mptcp/export]
[cannot apply to linus/master v5.17-rc4 next-20220217]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Florian-Westphal/mptcp-replace-per-addr-listener-sockets/20220217-222844
base:   https://github.com/multipath-tcp/mptcp_net-next.git export
config: mips-buildonly-randconfig-r002-20220217 (https://download.01.org/0day-ci/archive/20220218/202202180141.VfybVriM-lkp(a)intel.com/config)
compiler: clang version 15.0.0 (https://github.com/llvm/llvm-project d271fc04d5b97b12e6b797c6067d3c96a8d7470e)

reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# install mips cross compiling tool for clang build
# apt-get install binutils-mips-linux-gnu
# https://github.com/0day-ci/linux/commit/044982520ba41e284eebe48421fad7feb55f2106
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Florian-Westphal/mptcp-replace-per-addr-listener-sockets/20220217-222844 git checkout 044982520ba41e284eebe48421fad7feb55f2106
# save the config file to linux build tree
mkdir build_dir
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 O=build_dir ARCH=mips SHELL=/bin/bash net/mptcp/

If you fix the issue, kindly add following tag as appropriate

All errors (new ones prefixed by >>):

>> net/mptcp/ctrl.c:251:17: error: no member named 'join6' in 'struct mptcp_pernet'; did you mean 'join4'?
                   lsk = pernet->join6.sk;
                                 ^~~~~
                                 join4
   net/mptcp/ctrl.c:49:23: note: 'join4' declared here
           struct mptcp_join_sk join4;
                                ^

1 error generated.

vim +251 net/mptcp/ctrl.c

205
206 struct sock *__mptcp_handle_join(int af, struct sk_buff *skb)
207 {
208 struct mptcp_options_received mp_opt;
209 struct mptcp_pernet *pernet;
210 struct socket *ssock;
211 struct sock *lsk;
212 struct net *net;
213
214 /* paranoia check: don't allow 0 destination port,
215 * else __inet_inherit_port will insert the child socket
216 * into the phony hash slot of the pernet listener.
217 */
218 if (tcp_hdr(skb)->dest == 0)
219 return NULL;
220
221 mptcp_get_options(skb, &mp_opt);
222
223 if (!(mp_opt.suboptions & OPTIONS_MPTCP_MPJ))
224 return NULL;
225
226 net = dev_net(skb_dst(skb)->dev);
227 if (!mptcp_is_enabled(net))
228 return NULL;
229
230 /* RFC8684: If the token is unknown [..], the receiver will send
231 * back a reset (RST) signal, analogous to an unknown port in TCP,
232 * containing an MP_TCPRST option (Section 3.6) [..]
233 */
234 if (!mptcp_token_exists(net, mp_opt.token)) {
235 struct mptcp_ext *ext = skb_ext_add(skb, SKB_EXT_MPTCP);
236
237 if (ext) {
238 memset(ext, 0, sizeof(*ext));
239 ext->reset_reason = MPTCP_RST_EMPTCP;
240 }
241 return NULL;
242 }
243
244 pernet = mptcp_get_pernet(net);
245
246 switch (af) {
247 case AF_INET:
248 lsk = pernet->join4.sk;
249 break;
250 case AF_INET6:

 > 251			lsk = pernet->join6.sk;

252 break;
253 default:
254 WARN_ON_ONCE(1);
255 return NULL;
256 }
257
258 ssock = __mptcp_nmpc_socket(mptcp_sk(lsk));
259 if (WARN_ON(!ssock))
260 return NULL;
261
262 return ssock->sk;
263 }
264

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

#10 lore raw

Re: [PATCH mptcp-next v2 4/5] mptcp: handle join requests via pernet listen socket

In reply to Florian Westphal
Quoted 6 lines · Florian Westphal

Currently mptcp adds kernel-based listener socket for all netlink-configured mptcp address endpoints.

This has caveats because kernel may interfere with unrelated programs that use same address/port pairs.

Quoted 7 lines

RFC 8684 says:
Demultiplexing subflow SYNs MUST be done using the token; this is unlike traditional TCP, where the destination port is used for demultiplexing SYN packets. Once a subflow is set up, demultiplexing packets is done using the 5-tuple, as in traditional TCP.

This patch deviates from this in that it retrains the existing checks of

Typo: retains

Quoted 13 lines

verifying the incoming requests destination vs. the list of announced addresses.

This can be relaxed later if deemed appropriate.

Furthermore, TCP-only listeners take precedence: An MPTCP peer MUST NOT announce address:port pairs that are already in use by a non-mptcp listener.
This could be changed, but it requires move of mptcp_handle_join() hook *before* the tcp port demux, i.e. an additional conditional in hotpath. As-is, the additional conditional (syn && !rst && ...) is in the 'no socket found' path.

Quoted 8 lines

The pernet 'listening' socket is hidden from userspace. It is not part of any hashes and not bound to any address or port.

TPROXY-like semantics apply: If tcp demux cannot find a port for a given packet, check if the packet is a syn packet with a valid join token.

If so, the pernet listener is returned and tcp processing resumes. Otherwise, handling is intentical.

Typo: identical

Quoted 3 lines

This patch does not cover timewait sockets.

Can you elaborate on what properly covering timewait sockets would look like? Does the approach used by multipath-tcp.org (handling token lookup after inet_lookup_listener() / case TCP_TW_SYN in tcp_v4_rcv()) fit?

I have one more comment below, but other than that the code that's here is looking good and the big outstanding questions are the things you noted as not-yet-changed in the cover letter (MP_JOIN token lookup for established or TCP listening sockets). We did discuss that in today's meeting so I won't rehash here.

Quoted 10 lines

---

include/net/mptcp.h  |  10 ++
net/ipv6/tcp_ipv6.c  |  19 ++--
net/mptcp/ctrl.c     | 211 ++++++++++++++++++++++++++++++++++++++++++-
net/mptcp/protocol.c |   2 +-
net/mptcp/protocol.h |   2 +-
net/mptcp/subflow.c  |   3 +
6 files changed, 233 insertions(+), 14 deletions(-)

...

Quoted 4 lines
net/mptcp/ctrl.c +0 -0
index ae20b7d92e28..6358c803ba12 100644

...

Quoted 41 lines
net/mptcp/ctrl.c +37 -0
@@ -185,13 +203,173 @@ static void mptcp_pernet_del_table(struct mptcp_pernet *pernet) {}
 
 #endif /* CONFIG_SYSCTL */
 
+struct sock *__mptcp_handle_join(int af, struct sk_buff *skb)
+{
+	struct mptcp_options_received mp_opt;
+	struct mptcp_pernet *pernet;
+	struct socket *ssock;
+	struct sock *lsk;
+	struct net *net;
+
+	/* paranoia check: don't allow 0 destination port,
+	 * else __inet_inherit_port will insert the child socket
+	 * into the phony hash slot of the pernet listener.
+	 */
+	if (tcp_hdr(skb)->dest == 0)
+		return NULL;
+
+	mptcp_get_options(skb, &mp_opt);
+
+	if (!(mp_opt.suboptions & OPTIONS_MPTCP_MPJ))
+		return NULL;
+
+	net = dev_net(skb_dst(skb)->dev);
+	if (!mptcp_is_enabled(net))
+		return NULL;
+
+	/* RFC8684: If the token is unknown [..], the receiver will send
+	 * back a reset (RST) signal, analogous to an unknown port in TCP,
+	 * containing an MP_TCPRST option (Section 3.6) [..]
+	 */
+	if (!mptcp_token_exists(net, mp_opt.token)) {
+		struct mptcp_ext *ext = skb_ext_add(skb, SKB_EXT_MPTCP);
+
+		if (ext) {
+			memset(ext, 0, sizeof(*ext));
+			ext->reset_reason = MPTCP_RST_EMPTCP;
+		}
+		return NULL;
+	}

net/mptcp/ctrl.c static void mptcp_pernet_del_table(struct mptcp_pernet *pernet) {} · line 252

Quoted 10 lines
net/mptcp/ctrl.c +10 -0
+
+	pernet = mptcp_get_pernet(net);
+
+	switch (af) {
+	case AF_INET:
+		lsk = pernet->join4.sk;
+		break;
+	case AF_INET6:
+		lsk = pernet->join6.sk;
+		break;

As kbuild noted, missing an

#if IS_ENABLED(CONFIG_MPTCP_IPV6)

check here.

--
Mat Martineau
Intel

#11 lore raw

Re: [PATCH mptcp-next v2 4/5] mptcp: handle join requests via pernet listen socket

Quoted 4 lines · Mat Martineau
Quoted 2 lines

If so, the pernet listener is returned and tcp processing resumes. Otherwise, handling is intentical.

Typo: identical

Thanks, fixed both typos.

Quoted 5 lines
Quoted 2 lines

This patch does not cover timewait sockets.

Can you elaborate on what properly covering timewait sockets would look like? Does the approach used by multipath-tcp.org (handling token lookup after inet_lookup_listener() / case TCP_TW_SYN in tcp_v4_rcv()) fit?

Looks like its as simple as:

net/ipv4/tcp_ipv4.c +3 -0
@@ -2189,6 +2189,9 @@ int tcp_v4_rcv(struct sk_buff *skb)
 							iph->daddr, th->dest,
 							inet_iif(skb),
 							sdif);
+		if (!sk2)
+			sk2 = mptcp_handle_join(AF_INET, skb);
+
 		if (sk2) {
 			inet_twsk_deschedule_put(inet_twsk(sk));
 			sk = sk2;
net/ipv6/tcp_ipv6.c +3 -0
index 3b8608d35dcd..f2f2308c6bda 100644
@@ -1833,6 +1833,9 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 					    ntohs(th->dest),
 					    tcp_v6_iif_l3_slave(skb),
 					    sdif);
+		if (!sk2)
+			sk2 = mptcp_handle_join(AF_INET6, skb);
+
 		if (sk2) {
 			struct inet_timewait_sock *tw = inet_twsk(sk);
 			inet_twsk_deschedule_put(tw);

I think I'll squash this into the previous patch and will just remove the timewait-socket bit from this patch.

Quoted 9 lines

Review

Quoted 3 lines
(quoted diff) +3 -0
+	case AF_INET6:
+		lsk = pernet->join6.sk;
+		break;

As kbuild noted, missing an

#if IS_ENABLED(CONFIG_MPTCP_IPV6)

check here.

Added.

#12 lore raw

Re: [PATCH mptcp-next v2 4/5] mptcp: handle join requests via pernet listen socket

In reply to Florian Westphal
Quoted 6 lines · Florian Westphal

Currently mptcp adds kernel-based listener socket for all netlink-configured mptcp address endpoints.

This has caveats because kernel may interfere with unrelated programs that use same address/port pairs.

I assume that this refers to the potential race between a kernel listener and the application which Paolo had raised. I'm not sure if this change eliminates that possibility. Pasting Paolo's example below from the prior discussion:

"""
s1 = socket()
bind(s1, A)
listen(s1)
// at this point incoming MPTCP connection can be established on s1 // and ADD_ADDR sub-options could be sent back

s2 = socket()
bind(s2, B)
listen(s2)
"""

Over a new MPTCP connection following listen(s1), the PM can issue an ADD_ADDR with B. In light of this change there would be no listener created for B. But if the remote endpoint immediately established a subflow in response, then that would result in a subflow (connection) socket at B. It appears (and correct me if I'm wrong) that bind(s2, B) might fail at this point (?). In other words, it seems like the subflow creation could race with a subsequent bind() causing startup issues in the application.

If my assessment is correct, then I think this change isn't sidestepping the issue that motivated it in the first place. However, I do like the fact that we don't need code to manage kernel listeners, just so long as there are no other breaking corner cases and the TCP level changes are acceptable upstream.

Quoted 423 lines

RFC 8684 says:
Demultiplexing subflow SYNs MUST be done using the token; this is unlike traditional TCP, where the destination port is used for demultiplexing SYN packets. Once a subflow is set up, demultiplexing packets is done using the 5-tuple, as in traditional TCP.

This patch deviates from this in that it retrains the existing checks of verifying the incoming requests destination vs. the list of announced addresses.

This can be relaxed later if deemed appropriate.

Furthermore, TCP-only listeners take precedence: An MPTCP peer MUST NOT announce address:port pairs that are already in use by a non-mptcp listener.
This could be changed, but it requires move of mptcp_handle_join() hook *before* the tcp port demux, i.e. an additional conditional in hotpath. As-is, the additional conditional (syn && !rst && ...) is in the 'no socket found' path.

The pernet 'listening' socket is hidden from userspace. It is not part of any hashes and not bound to any address or port.

TPROXY-like semantics apply: If tcp demux cannot find a port for a given packet, check if the packet is a syn packet with a valid join token.

If so, the pernet listener is returned and tcp processing resumes. Otherwise, handling is intentical.

This patch does not cover timewait sockets.

---
 include/net/mptcp.h  |  10 ++
 net/ipv6/tcp_ipv6.c  |  19 ++--
 net/mptcp/ctrl.c     | 211 ++++++++++++++++++++++++++++++++++++++++++-
 net/mptcp/protocol.c |   2 +-
 net/mptcp/protocol.h |   2 +-
 net/mptcp/subflow.c  |   3 +
 6 files changed, 233 insertions(+), 14 deletions(-)
include/net/mptcp.h +10 -0
index 5ee422b56902..49c188b978e1 100644
@@ -189,6 +189,7 @@ int mptcp_subflow_init_cookie_req(struct request_sock *req,
 				  struct sk_buff *skb);
 
 __be32 mptcp_get_reset_option(const struct sk_buff *skb);
+struct sock *__mptcp_handle_join(int af, struct sk_buff *skb);
 
 static inline __be32 mptcp_reset_option(const struct sk_buff *skb)
 {
@@ -199,6 +200,11 @@ static inline __be32 mptcp_reset_option(const struct sk_buff *skb)
 }
 static inline struct sock *mptcp_handle_join(int af, struct sk_buff *skb)
 {
+	const struct tcphdr *th = tcp_hdr(skb);
+
+	if (th->syn && !th->ack && !th->rst && !th->fin)
+		return __mptcp_handle_join(af, skb);
+
 	return NULL;
 }
 #else
@@ -283,9 +289,13 @@ static inline struct sock *mptcp_handle_join(int af, struct sk_buff *skb) { retu
 
 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 int mptcpv6_init(void);
+int mptcpv6_init_net(struct net *net);
+void mptcpv6_exit_net(struct net *net);
 void mptcpv6_handle_mapped(struct sock *sk, bool mapped);
 #elif IS_ENABLED(CONFIG_IPV6)
 static inline int mptcpv6_init(void) { return 0; }
+static inline int mptcpv6_init_net(struct net *net) { return 0; }
+static inline void mptcpv6_exit_net(struct net *net) { }
 static inline void mptcpv6_handle_mapped(struct sock *sk, bool mapped) { }
 #endif
 
net/ipv6/tcp_ipv6.c +11 -8
index 788040db8e9e..3b8608d35dcd 100644
@@ -2233,13 +2233,22 @@ static struct inet_protosw tcpv6_protosw = {
 
 static int __net_init tcpv6_net_init(struct net *net)
 {
-	return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
-				    SOCK_RAW, IPPROTO_TCP, net);
+	int err = inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
+				       SOCK_RAW, IPPROTO_TCP, net);
+	if (err)
+		return err;
+
+	err = mptcpv6_init_net(net);
+	if (err)
+		inet_ctl_sock_destroy(net->ipv6.tcp_sk);
+
+	return err;
 }
 
 static void __net_exit tcpv6_net_exit(struct net *net)
 {
 	inet_ctl_sock_destroy(net->ipv6.tcp_sk);
+	mptcpv6_exit_net(net);
 }
 
 static struct pernet_operations tcpv6_net_ops = {
@@ -2264,15 +2273,9 @@ int __init tcpv6_init(void)
 	if (ret)
 		goto out_tcpv6_protosw;
 
-	ret = mptcpv6_init();
-	if (ret)
-		goto out_tcpv6_pernet_subsys;
-
 out:
 	return ret;
 
-out_tcpv6_pernet_subsys:
-	unregister_pernet_subsys(&tcpv6_net_ops);
 out_tcpv6_protosw:
 	inet6_unregister_protosw(&tcpv6_protosw);
 out_tcpv6_protocol:
net/mptcp/ctrl.c +207 -4
index ae20b7d92e28..6358c803ba12 100644
@@ -21,6 +21,12 @@ static int mptcp_pernet_id;
 static int mptcp_pm_type_max = __MPTCP_PM_TYPE_MAX;
 #endif
 
+struct mptcp_join_sk {
+	struct sock *sk;
+	struct inet_bind_bucket *tb;
+	struct inet_bind_hashbucket head;
+};
+
 struct mptcp_pernet {
 #ifdef CONFIG_SYSCTL
 	struct ctl_table_header *ctl_table_hdr;
@@ -32,6 +38,18 @@ struct mptcp_pernet {
 	u8 checksum_enabled;
 	u8 allow_join_initial_addr_port;
 	u8 pm_type;
+
+	/* pernet listener to handle mptcp join requests
+	 * based on the mptcp token.
+	 *
+	 * Has to be pernet because tcp uses
+	 * sock_net(sk_listener) to obtain the net namespace for
+	 * the syn/ack route lookup.
+	 */
+	struct mptcp_join_sk join4;
+#if IS_ENABLED(CONFIG_MPTCP_IPV6)
+	struct mptcp_join_sk join6;
+#endif
 };
 
 static struct mptcp_pernet *mptcp_get_pernet(const struct net *net)
@@ -185,13 +203,173 @@ static void mptcp_pernet_del_table(struct mptcp_pernet *pernet) {}
 
 #endif /* CONFIG_SYSCTL */
 
+struct sock *__mptcp_handle_join(int af, struct sk_buff *skb)
+{
+	struct mptcp_options_received mp_opt;
+	struct mptcp_pernet *pernet;
+	struct socket *ssock;
+	struct sock *lsk;
+	struct net *net;
+
+	/* paranoia check: don't allow 0 destination port,
+	 * else __inet_inherit_port will insert the child socket
+	 * into the phony hash slot of the pernet listener.
+	 */
+	if (tcp_hdr(skb)->dest == 0)
+		return NULL;
+
+	mptcp_get_options(skb, &mp_opt);
+
+	if (!(mp_opt.suboptions & OPTIONS_MPTCP_MPJ))
+		return NULL;
+
+	net = dev_net(skb_dst(skb)->dev);
+	if (!mptcp_is_enabled(net))
+		return NULL;
+
+	/* RFC8684: If the token is unknown [..], the receiver will send
+	 * back a reset (RST) signal, analogous to an unknown port in TCP,
+	 * containing an MP_TCPRST option (Section 3.6) [..]
+	 */
+	if (!mptcp_token_exists(net, mp_opt.token)) {
+		struct mptcp_ext *ext = skb_ext_add(skb, SKB_EXT_MPTCP);
+
+		if (ext) {
+			memset(ext, 0, sizeof(*ext));
+			ext->reset_reason = MPTCP_RST_EMPTCP;
+		}
+		return NULL;
+	}
+
+	pernet = mptcp_get_pernet(net);
+
+	switch (af) {
+	case AF_INET:
+		lsk = pernet->join4.sk;
+		break;
+	case AF_INET6:
+		lsk = pernet->join6.sk;
+		break;
+	default:
+		WARN_ON_ONCE(1);
+		return NULL;
+	}
+
+	ssock = __mptcp_nmpc_socket(mptcp_sk(lsk));
+	if (WARN_ON(!ssock))
+		return NULL;
+
+	return ssock->sk;
+}
+
+static struct socket *mptcp_create_join_listen_socket(struct net *net, int af)
+{
+	struct socket *s, *ssock;
+	int err;
+
+	err = sock_create_kern(net, af, SOCK_STREAM, IPPROTO_MPTCP, &s);
+	if (err)
+		return ERR_PTR(err);
+
+	ssock = __mptcp_nmpc_socket(mptcp_sk(s->sk));
+	if (!ssock) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	ssock->sk->sk_max_ack_backlog = SOMAXCONN;
+	inet_sk_state_store(ssock->sk, TCP_LISTEN);
+
+	s->sk->sk_max_ack_backlog = SOMAXCONN;
+	inet_sk_state_store(s->sk, TCP_LISTEN);
+
+	s->sk->sk_net_refcnt = 1;
+	get_net_track(net, &s->sk->ns_tracker, GFP_KERNEL);
+	sock_inuse_add(net, 1);
+
+	return s;
+out:
+	sock_release(s);
+	return ERR_PTR(err);
+}
+
+static int mptcp_init_join_sk(struct net *net, struct sock *sk, struct mptcp_join_sk *join_sk)
+{
+	struct socket *ssock = __mptcp_nmpc_socket(mptcp_sk(sk));
+	struct inet_hashinfo *table = ssock->sk->sk_prot->h.hashinfo;
+	struct inet_bind_bucket *tb;
+
+	spin_lock_init(&join_sk->head.lock);
+	INIT_HLIST_HEAD(&join_sk->head.chain);
+
+	/* Our "listen socket" isn't bound to any address or port.
+	 * Conceptually, SYN packet with mptcp join request are steered to
+	 * this pernet socket just like TPROXY steals arbitrary connection
+	 * requests to assign them to listening socket with different
+	 * address or port.
+	 *
+	 * The bind_bucket is needed for sake of __inet_inherit_port(),
+	 * so it can place the new child socket in the correct
+	 * bind_bucket slot.
+	 *
+	 * A phony head is used to hide this socket from normal sk loookup.
+	 */
+	tb = inet_bind_bucket_create(table->bind_bucket_cachep,
+				     net, &join_sk->head, 0, 0);
+	if (!tb)
+		return -ENOMEM;
+
+	inet_csk(ssock->sk)->icsk_bind_hash = tb;
+	return 0;
+}
+
 static int __net_init mptcp_net_init(struct net *net)
 {
 	struct mptcp_pernet *pernet = mptcp_get_pernet(net);
+	struct socket *sock;
+	int err;
 
 	mptcp_pernet_set_defaults(pernet);
 
-	return mptcp_pernet_new_table(net, pernet);
+	err = mptcp_pernet_new_table(net, pernet);
+	if (err)
+		return err;
+
+	sock = mptcp_create_join_listen_socket(net, AF_INET);
+	if (IS_ERR(sock)) {
+		err = PTR_ERR(sock);
+		goto out_table;
+	}
+
+	err = mptcp_init_join_sk(net, sock->sk, &pernet->join4);
+	if (err) {
+		sock_release(sock);
+		goto out_table;
+	}
+
+	/* struct sock is still reachable via sock->sk_socket backpointer */
+	pernet->join4.sk = sock->sk;
+	return err;
+
+out_table:
+	if (!net_eq(net, &init_net))
+		mptcp_pernet_del_table(pernet);
+	return err;
+}
+
+static void __net_exit mptcp_exit_join_sk(struct mptcp_join_sk *jsk)
+{
+	struct socket *ssock = __mptcp_nmpc_socket(mptcp_sk(jsk->sk));
+	struct inet_bind_bucket *tb;
+	struct inet_hashinfo *table;
+
+	table = ssock->sk->sk_prot->h.hashinfo;
+
+	tb = inet_csk(ssock->sk)->icsk_bind_hash;
+	inet_bind_bucket_destroy(table->bind_bucket_cachep, tb);
+
+	ssock = jsk->sk->sk_socket;
+	sock_release(ssock);
 }
 
 /* Note: the callback will only be called per extra netns */
@@ -200,6 +378,7 @@ static void __net_exit mptcp_net_exit(struct net *net)
 	struct mptcp_pernet *pernet = mptcp_get_pernet(net);
 
 	mptcp_pernet_del_table(pernet);
+	mptcp_exit_join_sk(&pernet->join4);
 }
 
 static struct pernet_operations mptcp_pernet_ops = {
@@ -219,12 +398,36 @@ void __init mptcp_init(void)
 }
 
 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
-int __init mptcpv6_init(void)
+int __net_init mptcpv6_init_net(struct net *net)
 {
+	struct mptcp_pernet *pernet = mptcp_get_pernet(net);
+	struct socket *sock;
 	int err;
 
-	err = mptcp_proto_v6_init();
+	if (net_eq(net, &init_net)) {
+		err = mptcp_proto_v6_init();
+		if (err)
+			return err;
+	}
 
-	return err;
+	sock = mptcp_create_join_listen_socket(net, AF_INET6);
+	if (IS_ERR(sock))
+		return PTR_ERR(sock);
+
+	err = mptcp_init_join_sk(net, sock->sk, &pernet->join6);
+	if (err) {
+		sock_release(sock);
+		return err;
+	}
+
+	pernet->join6.sk = sock->sk;
+	return 0;
+}
+
+void __net_exit mptcpv6_exit_net(struct net *net)
+{
+	struct mptcp_pernet *pernet = mptcp_get_pernet(net);
+
+	mptcp_exit_join_sk(&pernet->join6);
 }
 #endif
net/mptcp/protocol.c +1 -1
index 4599bde215b2..5b54e3c8efea 100644
@@ -3777,7 +3777,7 @@ static struct inet_protosw mptcp_v6_protosw = {
 	.flags		= INET_PROTOSW_ICSK,
 };
 
-int __init mptcp_proto_v6_init(void)
+int __net_init mptcp_proto_v6_init(void)
 {
 	int err;
 
net/mptcp/protocol.h +1 -1
index 7bd064b68b51..6a81e2a21301 100644
@@ -647,7 +647,7 @@ static inline bool mptcp_has_another_subflow(struct sock *ssk)
 
 void __init mptcp_proto_init(void);
 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
-int __init mptcp_proto_v6_init(void);
+int __net_init mptcp_proto_v6_init(void);
 #endif
 
 struct sock *mptcp_sk_clone(const struct sock *sk,
net/mptcp/subflow.c +3 -0
index 8be20f7b76df..4696d27a8994 100644
@@ -116,6 +116,9 @@ static void subflow_init_req(struct request_sock *req, const struct sock *sk_lis
 
 static bool subflow_use_different_sport(struct mptcp_sock *msk, const struct sock *sk)
 {
+	if (inet_sk(sk)->inet_sport == 0)
+		return true;
+
 	return inet_sk(sk)->inet_sport != inet_sk((struct sock *)msk)->inet_sport;
 }
 
#13 lore raw

PATCH v2 5/5 mptcp: remove per-address listening sockets

In reply to Florian Westphal

Not required anymore, syn packets with a join requests are redirected to pernet mptcp pseudo-listening socket.

---
 net/mptcp/pm_netlink.c | 65 ------------------------------------------
 1 file changed, 65 deletions(-)
net/mptcp/pm_netlink.c +0 -65
index 614b5d05aa62..8a26c7177054 100644
@@ -27,7 +27,6 @@ struct mptcp_pm_addr_entry {
 	struct mptcp_addr_info	addr;
 	u8			flags;
 	int			ifindex;
-	struct socket		*lsk;
 };
 
 struct mptcp_pm_add_entry {
@@ -954,57 +953,6 @@ static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
 	return ret;
 }
 
-static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
-					    struct mptcp_pm_addr_entry *entry)
-{
-	int addrlen = sizeof(struct sockaddr_in);
-	struct sockaddr_storage addr;
-	struct mptcp_sock *msk;
-	struct socket *ssock;
-	int backlog = 1024;
-	int err;
-
-	err = sock_create_kern(sock_net(sk), entry->addr.family,
-			       SOCK_STREAM, IPPROTO_MPTCP, &entry->lsk);
-	if (err)
-		return err;
-
-	msk = mptcp_sk(entry->lsk->sk);
-	if (!msk) {
-		err = -EINVAL;
-		goto out;
-	}
-
-	ssock = __mptcp_nmpc_socket(msk);
-	if (!ssock) {
-		err = -EINVAL;
-		goto out;
-	}
-
-	mptcp_info2sockaddr(&entry->addr, &addr, entry->addr.family);
-#if IS_ENABLED(CONFIG_MPTCP_IPV6)
-	if (entry->addr.family == AF_INET6)
-		addrlen = sizeof(struct sockaddr_in6);
-#endif
-	err = kernel_bind(ssock, (struct sockaddr *)&addr, addrlen);
-	if (err) {
-		pr_warn("kernel_bind error, err=%d", err);
-		goto out;
-	}
-
-	err = kernel_listen(ssock, backlog);
-	if (err) {
-		pr_warn("kernel_listen error, err=%d", err);
-		goto out;
-	}
-
-	return 0;
-
-out:
-	sock_release(entry->lsk);
-	return err;
-}
-
 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
 {
 	struct mptcp_pm_addr_entry *entry;
@@ -1050,7 +998,6 @@ int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
 	entry->addr.port = 0;
 	entry->ifindex = 0;
 	entry->flags = 0;
-	entry->lsk = NULL;
 	ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
 	if (ret < 0)
 		kfree(entry);
@@ -1258,19 +1205,9 @@ static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
 	}
 
 	*entry = addr;
-	if (entry->addr.port) {
-		ret = mptcp_pm_nl_create_listen_socket(skb->sk, entry);
-		if (ret) {
-			GENL_SET_ERR_MSG(info, "create listen socket error");
-			kfree(entry);
-			return ret;
-		}
-	}
 	ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
 	if (ret < 0) {
 		GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
-		if (entry->lsk)
-			sock_release(entry->lsk);
 		kfree(entry);
 		return ret;
 	}
@@ -1375,8 +1312,6 @@ static int mptcp_nl_remove_subflow_and_signal_addr(struct net *net,
 /* caller must ensure the RCU grace period is already elapsed */
 static void __mptcp_pm_release_addr_entry(struct mptcp_pm_addr_entry *entry)
 {
-	if (entry->lsk)
-		sock_release(entry->lsk);
 	kfree(entry);
 }
 
Signature · 1 line
2.34.1