Skip to contentSkip to conversation

mptcp: refactor first subflow init

PATCH Superseded Paolo Abeni 3 participants 25 messages Jan 17, 2023 – Jan 26, 2023

Subsystems (4)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Paolo Abeni

author maintainer
#0 lore raw

v2 cover mptcp: refactor first subflow init

This is the needed refactor for the selinux fixes, as discussed on the ML.

Patch the first 2 patches address old, currently not so relevant bugs which will become more serious once the refactor is applied.

Patches 3-7 are pre-reqs for the bulk changes, but also IMHO nice to have even stand-alone.

The main change, introduced by patch 8, consists in moving the first subflow initialization from the msk init callback into the mptcp syscall needing such data (namely: bind, listen, connect).

Patches 9, 10, 11 are not strictly needed, but are some nice to have follow-up, cleaning-up the related code.

Specifically patch 10 closes issues/290

Finally patches 12 && 13 address the LSM issue. They really target the LSM subtree, and are added here just to allow verify the fix in our tree before the LSM submission.

Sharing after little testing to get feedback and let the bot massage the new code: a couple of patches can have subtle effect, I would like to have syzkaller digest them for a while.

Paolo Abeni (13):
mptcp: fix locking for setsockopt corner-case
mptcp: fix locking for in-kernel listener creation.
mptcp: refactor passive socket initialization.
mptcp: drop unneeded argument
mptcp: drop legacy code.
mptcp: avoid unneeded __mptcp_nmpc_socket() usage
mptcp: move fastopen subflow check inside mptcp_sendmsg_fastopen()
mptcp: move first subflow allocation at mpc access time
mptcp: do not keep around the first subflow after disconnect.
mptcp: fastclose msk when cleaning unaccepted sockets
mptcp: refactor mptcp_stream_accept()
security, lsm: Introduce security_mptcp_add_subflow()
selinux: Implement mptcp_add_subflow hook

 include/linux/lsm_hook_defs.h |   1 +
 include/linux/lsm_hooks.h     |   9 ++
 include/linux/security.h      |   6 ++
 net/mptcp/options.c           |   9 +-
 net/mptcp/pm.c                |   4 +-
 net/mptcp/pm_netlink.c        |  14 +--
 net/mptcp/protocol.c          | 163 ++++++++++++++++++----------------
 net/mptcp/protocol.h          |   4 +-
 net/mptcp/sockopt.c           |  29 +++---
 net/mptcp/subflow.c           |  48 +++++++---
 security/security.c           |   5 ++
 security/selinux/hooks.c      |  16 ++++
 security/selinux/netlabel.c   |   8 +-
 13 files changed, 198 insertions(+), 118 deletions(-)
Signature · 1 line
2.39.0
author maintainer
#1 lore raw

PATCH v2 1/13 mptcp: fix locking for setsockopt corner-case

We need to call the __mptcp_nmpc_socket(), and later subflow socket access under the msk socket lock, or e.g. a racing connect() could change the socket status under the wood, with unexpected results.

Fixes 54635bd04701 ("mptcp: add TCP_FASTOPEN_CONNECT socket option")

---
this and the next patch are new v2. In both case the addressed issue is almost irrelevant until patch 8/12, but will cause lockdep splat after such change

---
 net/mptcp/sockopt.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
net/mptcp/sockopt.c +9 -2
index 9986681aaf40..8a9656248b0f 100644
@@ -760,14 +760,21 @@ static int mptcp_setsockopt_v4(struct mptcp_sock *msk, int optname,
 static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
 					  sockptr_t optval, unsigned int optlen)
 {
+	struct sock *sk = (struct sock *)msk;
 	struct socket *sock;
+	int ret = -EINVAL;
 
 	/* Limit to first subflow, before the connection establishment */
+	lock_sock(sk);
 	sock = __mptcp_nmpc_socket(msk);
 	if (!sock)
-		return -EINVAL;
+		goto unlock;
 
-	return tcp_setsockopt(sock->sk, level, optname, optval, optlen);
+	ret = tcp_setsockopt(sock->sk, level, optname, optval, optlen);
+
+unlock:
+	release_sock(sk);
+	return ret;
 }
 
 static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
Signature · 1 line
2.39.0
author maintainer
#2 lore raw

PATCH v2 2/13 mptcp: fix locking for in-kernel listener creation.

In reply to Paolo Abeni

For consistency, in mptcp_pm_nl_create_listen_socket(), we need to call the __mptcp_nmpc_socket() under the msk socket lock.

Note that as a side effect, mptcp_subflow_create_socket() needs a 'nested' lockdep annotation, as it will acquire the subflow (kernel) socket lock under the in-kernel listener msk socket lock.

The current lack of locking is almost harmless, because the relevant socket is not exposed to the user space, but in future we will add more complexity to the mentioned helper, let's play safe.

Fixes 1729cf186d8a ("mptcp: create the listening socket for new port")
---
 net/mptcp/pm_netlink.c | 10 ++++++----
 net/mptcp/subflow.c    |  2 +-
 2 files changed, 7 insertions(+), 5 deletions(-)
net/mptcp/pm_netlink.c +6 -4
index e82a112b8779..155916174841 100644
@@ -999,8 +999,8 @@ static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
 {
 	int addrlen = sizeof(struct sockaddr_in);
 	struct sockaddr_storage addr;
-	struct mptcp_sock *msk;
 	struct socket *ssock;
+	struct sock *newsk;
 	int backlog = 1024;
 	int err;
 
@@ -1009,11 +1009,13 @@ static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
 	if (err)
 		return err;
 
-	msk = mptcp_sk(entry->lsk->sk);
-	if (!msk)
+	newsk = entry->lsk->sk;
+	if (!newsk)
 		return -EINVAL;
 
-	ssock = __mptcp_nmpc_socket(msk);
+	lock_sock(newsk);
+	ssock = __mptcp_nmpc_socket(mptcp_sk(newsk));
+	release_sock(newsk);
 	if (!ssock)
 		return -EINVAL;
 
net/mptcp/subflow.c +1 -1
index ec54413fb31f..a3e5026bee5b 100644
@@ -1679,7 +1679,7 @@ int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
 	if (err)
 		return err;
 
-	lock_sock(sf->sk);
+	lock_sock_nested(sf->sk, SINGLE_DEPTH_NESTING);
 
 	/* the newly created socket has to be in the same cgroup as its parent */
 	mptcp_attach_cgroup(sk, sf->sk);
Signature · 1 line
2.39.0
author maintainer
#3 lore raw

PATCH v2 3/13 mptcp: refactor passive socket initialization.

In reply to Paolo Abeni

After commit 30e51b923e43 ("mptcp: fix unreleased socket in accept queue") unaccepted msk sockets go throu complete shutdown, we don't need anymore to delay inserting the first subflow into the subflow lists.

The reference counting deserve some extra care, as __mptcp_close() is unaware of the request socket linkage to the first subflow.

---
Notes:
- this schema assumes that the TCP code will never drop
a request socket from the receive queue after the 3whs. I tried to verify such assumption as strictily as I could, but more eyes more then welcome!
- this will cause pktdrill failure for close_before_accept.pkt, because the msk will become fully established before accept() - imho a good thing - and send out add_addr earlier.

The pktdrill test change should be easier.

---
 net/mptcp/protocol.c | 17 -----------------
 net/mptcp/subflow.c  | 31 ++++++++++++++++++++++++-------
 2 files changed, 24 insertions(+), 24 deletions(-)
net/mptcp/protocol.c +0 -17
index 003b44a79fce..d298d629b3b2 100644
@@ -825,7 +825,6 @@ static bool __mptcp_finish_join(struct mptcp_sock *msk, struct sock *ssk)
 	if (sk->sk_socket && !ssk->sk_socket)
 		mptcp_sock_graft(ssk, sk->sk_socket);
 
-	mptcp_propagate_sndbuf((struct sock *)msk, ssk);
 	mptcp_sockopt_sync_locked(msk, ssk);
 	return true;
 }
@@ -3753,22 +3752,6 @@ static int mptcp_stream_accept(struct socket *sock, struct socket *newsock,
 
 		lock_sock(newsk);
 
-		/* PM/worker can now acquire the first subflow socket
-		 * lock without racing with listener queue cleanup,
-		 * we can notify it, if needed.
-		 *
-		 * Even if remote has reset the initial subflow by now
-		 * the refcnt is still at least one.
-		 */
-		subflow = mptcp_subflow_ctx(msk->first);
-		list_add(&subflow->node, &msk->conn_list);
-		sock_hold(msk->first);
-		if (mptcp_is_fully_established(newsk))
-			mptcp_pm_fully_established(msk, msk->first, GFP_KERNEL);
-
-		mptcp_rcv_space_init(msk, msk->first);
-		mptcp_propagate_sndbuf(newsk, msk->first);
-
 		/* set ssk->sk_socket of accept()ed flows to mptcp socket.
 		 * This is needed so NOSPACE flag can be set from tcp stack.
 		 */
net/mptcp/subflow.c +24 -7
index a3e5026bee5b..5e6752cd280b 100644
@@ -749,6 +749,7 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 	struct mptcp_options_received mp_opt;
 	bool fallback, fallback_is_fatal;
 	struct sock *new_msk = NULL;
+	struct mptcp_sock *owner;
 	struct sock *child;
 
 	pr_debug("listener=%p, req=%p, conn=%p", listener, req, listener->conn);
@@ -823,6 +824,8 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 		ctx->setsockopt_seq = listener->setsockopt_seq;
 
 		if (ctx->mp_capable) {
+			owner = mptcp_sk(new_msk);
+
 			/* this can't race with mptcp_close(), as the msk is
 			 * not yet exposted to user-space
 			 */
@@ -831,14 +834,14 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 			/* record the newly created socket as the first msk
 			 * subflow, but don't link it yet into conn_list
 			 */
-			WRITE_ONCE(mptcp_sk(new_msk)->first, child);
+			WRITE_ONCE(owner->first, child);
 
 			/* new mpc subflow takes ownership of the newly
 			 * created mptcp socket
 			 */
 			mptcp_sk(new_msk)->setsockopt_seq = ctx->setsockopt_seq;
-			mptcp_pm_new_connection(mptcp_sk(new_msk), child, 1);
-			mptcp_token_accept(subflow_req, mptcp_sk(new_msk));
+			mptcp_pm_new_connection(owner, child, 1);
+			mptcp_token_accept(subflow_req, owner);
 			ctx->conn = new_msk;
 			new_msk = NULL;
 
@@ -846,15 +849,21 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 			 * uses the correct data
 			 */
 			mptcp_copy_inaddrs(ctx->conn, child);
+			mptcp_propagate_sndbuf(ctx->conn, child);
+
+			mptcp_rcv_space_init(owner, child);
+			list_add(&ctx->node, &owner->conn_list);
+			sock_hold(child);
 
 			/* with OoO packets we can reach here without ingress
 			 * mpc option
 			 */
-			if (mp_opt.suboptions & OPTION_MPTCP_MPC_ACK)
+			if (mp_opt.suboptions & OPTION_MPTCP_MPC_ACK) {
 				mptcp_subflow_fully_established(ctx, &mp_opt);
+				mptcp_pm_fully_established(owner, child, GFP_ATOMIC);
+				ctx->pm_notified = 1;
+			}
 		} else if (ctx->mp_join) {
-			struct mptcp_sock *owner;
-
 			owner = subflow_req->msk;
 			if (!owner) {
 				subflow_add_reset_reason(skb, MPTCP_RST_EPROHIBIT);
@@ -1836,9 +1845,17 @@ void mptcp_subflow_queue_clean(struct sock *listener_sk, struct sock *listener_s
 		sock_hold(sk);
 		lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
 		next = msk->dl_next;
-		msk->first = NULL;
 		msk->dl_next = NULL;
 
+		/* The upcoming mptcp_close is going to drop all the references
+		 * to the first subflow, ignoring that one of such reference is
+		 * owned by the request socket still in the accept queue and that
+		 * later inet_csk_listen_stop will drop it.
+		 * Acquire an extra reference here to avoid an UaF at that point.
+		 */
+		if (msk->first)
+			sock_hold(msk->first);
+
 		do_cancel_work = __mptcp_close(sk, 0);
 		release_sock(sk);
 		if (do_cancel_work) {
Signature · 1 line
2.39.0
author maintainer
#5 lore raw

PATCH v2 5/13 mptcp: drop legacy code.

In reply to Paolo Abeni

After the previous commits the PM worker can't race anymore with the unaccepted subflow close and disposal, as the msk keeps a reference to such subflow.

We can remove the now irrelevant and confusing checks explicitly preventing the mentioned race.

---
 net/mptcp/options.c | 7 +------
 net/mptcp/subflow.c | 7 +++----
 2 files changed, 4 insertions(+), 10 deletions(-)
net/mptcp/options.c +1 -6
index 99c4f9e9bb90..91d5b59540e9 100644
@@ -988,12 +988,7 @@ static bool check_fully_established(struct mptcp_sock *msk, struct sock *ssk,
 	mptcp_subflow_fully_established(subflow, mp_opt);
 
 check_notify:
-	/* if the subflow is not already linked into the conn_list, we can't
-	 * notify the PM: this subflow is still on the listener queue
-	 * and the PM possibly acquiring the subflow lock could race with
-	 * the listener close
-	 */
-	if (likely(subflow->pm_notified) || list_empty(&subflow->node))
+	if (likely(subflow->pm_notified))
 		return true;
 
 	subflow->pm_notified = 1;
net/mptcp/subflow.c +3 -4
index 7b91dc57049e..6f198d6e1b22 100644
@@ -1933,11 +1933,10 @@ static void subflow_ulp_release(struct sock *ssk)
 
 	sk = ctx->conn;
 	if (sk) {
-		/* if the msk has been orphaned, keep the ctx
-		 * alive, will be freed by __mptcp_close_ssk(),
-		 * when the subflow is still unaccepted
+		/* if the subflow has been closed by the TCP stack, keep
+		 * the ctx alive, will be freed by __mptcp_close_ssk()
 		 */
-		release = ctx->disposable || list_empty(&ctx->node);
+		release = ctx->disposable;
 		sock_put(sk);
 	}
 
Signature · 1 line
2.39.0
author maintainer
#6 lore raw

PATCH v2 6/13 mptcp: avoid unneeded __mptcp_nmpc_socket() usage

In reply to Paolo Abeni

In a few spots the mptcp code invokes the __mptcp_nmpc_socket() helper multiple times under the same socket lock scope. Additionally, in such places, the socket status ensure that threre is not an MP capable handshake running.

Under the above condition we can replace the later __mptcp_nmpc_socket() helper invocation with direct access to the msk->subflow pointer and better document such access is not supposed to fail with WARN().

---
 net/mptcp/protocol.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
net/mptcp/protocol.c +6 -3
index d298d629b3b2..fc13f1e45137 100644
@@ -3172,7 +3172,7 @@ static struct sock *mptcp_accept(struct sock *sk, int flags, int *err,
 	struct socket *listener;
 	struct sock *newsk;
 
-	listener = __mptcp_nmpc_socket(msk);
+	listener = msk->subflow;
 	if (WARN_ON_ONCE(!listener)) {
 		*err = -EINVAL;
 		return NULL;
@@ -3392,7 +3392,7 @@ static int mptcp_get_port(struct sock *sk, unsigned short snum)
 	struct mptcp_sock *msk = mptcp_sk(sk);
 	struct socket *ssock;
 
-	ssock = __mptcp_nmpc_socket(msk);
+	ssock = msk->subflow;
 	pr_debug("msk=%p, subflow=%p", msk, ssock);
 	if (WARN_ON_ONCE(!ssock))
 		return -EINVAL;
@@ -3738,7 +3738,10 @@ static int mptcp_stream_accept(struct socket *sock, struct socket *newsock,
 
 	pr_debug("msk=%p", msk);
 
-	ssock = __mptcp_nmpc_socket(msk);
+	/* buggy applications can call accept on socket states other then LISTEN
+	 * but no need to allocate the first subflow just to error out.
+	 */
+	ssock = msk->subflow;
 	if (!ssock)
 		return -EINVAL;
 
Signature · 1 line
2.39.0
author maintainer
#4 lore raw

PATCH v2 4/13 mptcp: drop unneeded argument

In reply to Paolo Abeni

After the previous commit every mptcp_pm_fully_established() is always invoked with a GFP_ATOMIC argument. We can drop it.

---
 net/mptcp/options.c  | 2 +-
 net/mptcp/pm.c       | 4 ++--
 net/mptcp/protocol.h | 2 +-
 net/mptcp/subflow.c  | 2 +-
 4 files changed, 5 insertions(+), 5 deletions(-)
net/mptcp/options.c +1 -1
index b30cea2fbf3f..99c4f9e9bb90 100644
@@ -1001,7 +1001,7 @@ static bool check_fully_established(struct mptcp_sock *msk, struct sock *ssk,
 		clear_3rdack_retransmission(ssk);
 		mptcp_pm_subflow_established(msk);
 	} else {
-		mptcp_pm_fully_established(msk, ssk, GFP_ATOMIC);
+		mptcp_pm_fully_established(msk, ssk);
 	}
 	return true;
 
net/mptcp/pm.c +2 -2
index 8e0cf6275e94..4ed4d29d9c11 100644
@@ -126,7 +126,7 @@ static bool mptcp_pm_schedule_work(struct mptcp_sock *msk,
 	return true;
 }
 
-void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp)
+void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk)
 {
 	struct mptcp_pm_data *pm = &msk->pm;
 	bool announce = false;
@@ -150,7 +150,7 @@ void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk,
 	spin_unlock_bh(&pm->lock);
 
 	if (announce)
-		mptcp_event(MPTCP_EVENT_ESTABLISHED, msk, ssk, gfp);
+		mptcp_event(MPTCP_EVENT_ESTABLISHED, msk, ssk, GFP_ATOMIC);
 }
 
 void mptcp_pm_connection_closed(struct mptcp_sock *msk)
net/mptcp/protocol.h +1 -1
index 871ec3e93314..5f1a30959b5c 100644
@@ -799,7 +799,7 @@ bool mptcp_pm_addr_families_match(const struct sock *sk,
 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
-void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
+void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
net/mptcp/subflow.c +1 -1
index 5e6752cd280b..7b91dc57049e 100644
@@ -860,7 +860,7 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 			 */
 			if (mp_opt.suboptions & OPTION_MPTCP_MPC_ACK) {
 				mptcp_subflow_fully_established(ctx, &mp_opt);
-				mptcp_pm_fully_established(owner, child, GFP_ATOMIC);
+				mptcp_pm_fully_established(owner, child);
 				ctx->pm_notified = 1;
 			}
 		} else if (ctx->mp_join) {
Signature · 1 line
2.39.0
author maintainer
#7 lore raw

PATCH v2 7/13 mptcp: move fastopen subflow check inside mptcp_sendmsg_fastopen()

In reply to Paolo Abeni

So that we can avoid a bunch of check in fastpath. Additionally we can specialize such check according to the specific fastopen method
- defer_connect vs MSG_FASTOPEN.

The latter bits will simplify the next patches.

---
 net/mptcp/protocol.c | 24 ++++++++++++++++++------
 1 file changed, 18 insertions(+), 6 deletions(-)
net/mptcp/protocol.c +18 -6
index fc13f1e45137..9c4c729bf271 100644
@@ -1694,13 +1694,27 @@ static void mptcp_set_nospace(struct sock *sk)
 
 static int mptcp_disconnect(struct sock *sk, int flags);
 
-static int mptcp_sendmsg_fastopen(struct sock *sk, struct sock *ssk, struct msghdr *msg,
+static int mptcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg,
 				  size_t len, int *copied_syn)
 {
 	unsigned int saved_flags = msg->msg_flags;
 	struct mptcp_sock *msk = mptcp_sk(sk);
+	struct sock *ssk;
 	int ret;
 
+	/* on flags based fastopen the mptcp is supposed to create the
+	 * first subflow right now. Otherwise we are in the defer_connect
+	 * path, and the first subflow must be already present.
+	 * Since the defer_connect flag is cleared after the first succsful
+	 * fastopen attempt, no need to check for additional subflow status.
+	 */
+	if (msg->msg_flags & MSG_FASTOPEN && !__mptcp_nmpc_socket(msk))
+		return -EINVAL;
+	if (!msk->first)
+		return -EINVAL;
+
+	ssk = msk->first;
+
 	lock_sock(ssk);
 	msg->msg_flags |= MSG_DONTWAIT;
 	msk->connect_flags = O_NONBLOCK;
@@ -1723,6 +1737,7 @@ static int mptcp_sendmsg_fastopen(struct sock *sk, struct sock *ssk, struct msgh
 	} else if (ret && ret != -EINPROGRESS) {
 		mptcp_disconnect(sk, 0);
 	}
+	inet_sk(sk)->defer_connect = 0;
 
 	return ret;
 }
@@ -1731,7 +1746,6 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 {
 	struct mptcp_sock *msk = mptcp_sk(sk);
 	struct page_frag *pfrag;
-	struct socket *ssock;
 	size_t copied = 0;
 	int ret = 0;
 	long timeo;
@@ -1741,12 +1755,10 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 
 	lock_sock(sk);
 
-	ssock = __mptcp_nmpc_socket(msk);
-	if (unlikely(ssock && (inet_sk(ssock->sk)->defer_connect ||
-			       msg->msg_flags & MSG_FASTOPEN))) {
+	if (unlikely(inet_sk(sk)->defer_connect || msg->msg_flags & MSG_FASTOPEN)) {
 		int copied_syn = 0;
 
-		ret = mptcp_sendmsg_fastopen(sk, ssock->sk, msg, len, &copied_syn);
+		ret = mptcp_sendmsg_fastopen(sk, msg, len, &copied_syn);
 		copied += copied_syn;
 		if (ret == -EINPROGRESS && copied_syn > 0)
 			goto out;
Signature · 1 line
2.39.0
author maintainer
#8 lore raw

PATCH v2 8/13 mptcp: move first subflow allocation at mpc access time

In reply to Paolo Abeni

In the long run this will simplify the mptcp code and will allow for more consistent behavior. Move the first subflow allocation out of the sock->init ops into the __mptcp_nmpc_socket() helper.

Since the first subflow creation can now happen after the first setsockopt() we additionally need to invoke mptcp_sockopt_sync() on it.

---

v1 -> v2:
- really remove subflow creation from init() (!!!)

---
 net/mptcp/pm_netlink.c |  4 +--
 net/mptcp/protocol.c   | 61 +++++++++++++++++++++++++-----------------
 net/mptcp/protocol.h   |  2 +-
 net/mptcp/sockopt.c    | 20 +++++++-------
 4 files changed, 51 insertions(+), 36 deletions(-)
net/mptcp/pm_netlink.c +2 -2
index 155916174841..d1d859517d91 100644
@@ -1016,8 +1016,8 @@ static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
 	lock_sock(newsk);
 	ssock = __mptcp_nmpc_socket(mptcp_sk(newsk));
 	release_sock(newsk);
-	if (!ssock)
-		return -EINVAL;
+	if (IS_ERR(ssock))
+		return PTR_ERR(ssock);
 
 	mptcp_info2sockaddr(&entry->addr, &addr, entry->addr.family);
 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
net/mptcp/protocol.c +37 -24

109 lines not on this page · raw message

net/mptcp/protocol.h +1 -1
index 5f1a30959b5c..3a055438c65e 100644
@@ -632,7 +632,7 @@ void __mptcp_subflow_send_ack(struct sock *ssk);
 void mptcp_subflow_reset(struct sock *ssk);
 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
-struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
+struct socket *__mptcp_nmpc_socket(struct mptcp_sock *msk);
 bool __mptcp_close(struct sock *sk, long timeout);
 void mptcp_cancel_work(struct sock *sk);
 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
net/mptcp/sockopt.c +11 -9
index 8a9656248b0f..9bddae9c5c58 100644
@@ -301,9 +301,9 @@ static int mptcp_setsockopt_sol_socket(struct mptcp_sock *msk, int optname,
 	case SO_BINDTOIFINDEX:
 		lock_sock(sk);
 		ssock = __mptcp_nmpc_socket(msk);
-		if (!ssock) {
+		if (IS_ERR(ssock)) {
 			release_sock(sk);
-			return -EINVAL;
+			return PTR_ERR(ssock);
 		}
 
 		ret = sock_setsockopt(ssock, SOL_SOCKET, optname, optval, optlen);
@@ -396,9 +396,9 @@ static int mptcp_setsockopt_v6(struct mptcp_sock *msk, int optname,
 	case IPV6_FREEBIND:
 		lock_sock(sk);
 		ssock = __mptcp_nmpc_socket(msk);
-		if (!ssock) {
+		if (IS_ERR(ssock)) {
 			release_sock(sk);
-			return -EINVAL;
+			return PTR_ERR(ssock);
 		}
 
 		ret = tcp_setsockopt(ssock->sk, SOL_IPV6, optname, optval, optlen);
@@ -693,9 +693,9 @@ static int mptcp_setsockopt_sol_ip_set_transparent(struct mptcp_sock *msk, int o
 	lock_sock(sk);
 
 	ssock = __mptcp_nmpc_socket(msk);
-	if (!ssock) {
+	if (IS_ERR(ssock)) {
 		release_sock(sk);
-		return -EINVAL;
+		return PTR_ERR(ssock);
 	}
 
 	issk = inet_sk(ssock->sk);
@@ -762,13 +762,15 @@ static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int
 {
 	struct sock *sk = (struct sock *)msk;
 	struct socket *sock;
-	int ret = -EINVAL;
+	int ret;
 
 	/* Limit to first subflow, before the connection establishment */
 	lock_sock(sk);
 	sock = __mptcp_nmpc_socket(msk);
-	if (!sock)
+	if (IS_ERR(sock)) {
+		ret = PTR_ERR(sock);
 		goto unlock;
+	}
 
 	ret = tcp_setsockopt(sock->sk, level, optname, optval, optlen);
 
@@ -872,7 +874,7 @@ static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int
 	}
 
 	ssock = __mptcp_nmpc_socket(msk);
-	if (!ssock)
+	if (IS_ERR(ssock))
 		goto out;
 
 	ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);
Signature · 1 line
2.39.0
author maintainer
#12 lore raw

PATCH v2 9/13 mptcp: do not keep around the first subflow after disconnect.

In reply to Paolo Abeni

After the previous patch the first subflow is allocated as needed at bind, connect, listen time. We don't need anymore to keep alive the first subflow after a disconnect just to be able to perform such syscall.

Overal this change makes the passive and active sockets consistent:
even passive sockets will be allowed to complete life cycle after disconnect.

---
 net/mptcp/protocol.c | 24 +++++-------------------
 1 file changed, 5 insertions(+), 19 deletions(-)
net/mptcp/protocol.c +5 -19
index 4f7a71561efd..b867e3eec5b9 100644
@@ -2371,11 +2371,9 @@ static void __mptcp_close_ssk(struct sock *sk, struct sock *ssk,
 			      unsigned int flags)
 {
 	struct mptcp_sock *msk = mptcp_sk(sk);
-	bool need_push, dispose_it;
+	bool need_push;
 
-	dispose_it = !msk->subflow || ssk != msk->subflow->sk;
-	if (dispose_it)
-		list_del(&subflow->node);
+	list_del(&subflow->node);
 
 	lock_sock_nested(ssk, SINGLE_DEPTH_NESTING);
 
@@ -2389,15 +2387,6 @@ static void __mptcp_close_ssk(struct sock *sk, struct sock *ssk,
 	}
 
 	need_push = (flags & MPTCP_CF_PUSH) && __mptcp_retransmit_pending_data(sk);
-	if (!dispose_it) {
-		tcp_disconnect(ssk, 0);
-		msk->subflow->state = SS_UNCONNECTED;
-		mptcp_subflow_ctx_reset(subflow);
-		release_sock(ssk);
-
-		goto out;
-	}
-
 	sock_orphan(ssk);
 	subflow->disposable = 1;
 
@@ -2424,10 +2413,11 @@ static void __mptcp_close_ssk(struct sock *sk, struct sock *ssk,
 
 	sock_put(ssk);
 
-	if (ssk == msk->first)
+	if (ssk == msk->first) {
 		msk->first = NULL;
+		mptcp_dispose_initial_subflow(msk);
+	}
 
-out:
 	if (ssk == msk->last_snd)
 		msk->last_snd = NULL;
 
@@ -3270,10 +3260,6 @@ static void mptcp_destroy(struct sock *sk)
 {
 	struct mptcp_sock *msk = mptcp_sk(sk);
 
-	/* clears msk->subflow, allowing the following to close
-	 * even the initial subflow
-	 */
-	mptcp_dispose_initial_subflow(msk);
 	mptcp_destroy_common(msk, 0);
 	sk_sockets_allocated_dec(sk);
 }
Signature · 1 line
2.39.0
author maintainer
#13 lore raw

PATCH v2 10/13 mptcp: fastclose msk when cleaning unaccepted sockets

In reply to Paolo Abeni

When cleaning up unaccepted mptcp socket still laying inside the listener queue at listener close time, such sockets will go through a regular close, waiting for a timeout before shutting down the subflows.

There is no need to keep the kernel resources in use for such a possibly long time: short-circuit to fast-close.

---
 net/mptcp/protocol.c | 7 +++++--
 net/mptcp/subflow.c  | 2 +-
 2 files changed, 6 insertions(+), 3 deletions(-)
net/mptcp/protocol.c +5 -2
index b867e3eec5b9..f742d558a1b8 100644
@@ -2977,10 +2977,13 @@ bool __mptcp_close(struct sock *sk, long timeout)
 		goto cleanup;
 	}
 
-	if (mptcp_check_readable(msk)) {
-		/* the msk has read data, do the MPTCP equivalent of TCP reset */
+	if (mptcp_check_readable(msk) || timeout < 0) {
+		/* If the msk has read data, or the caller explicitly ask it,
+		 * do the MPTCP equivalent of TCP reset, aka MTCP fastclose
+		 */
 		inet_sk_state_store(sk, TCP_CLOSE);
 		mptcp_do_fastclose(sk);
+		timeout = 0;
 	} else if (mptcp_close_state(sk)) {
 		__mptcp_wr_shutdown(sk);
 	}
net/mptcp/subflow.c +1 -1
index 6f198d6e1b22..a1f8bb745c1b 100644
@@ -1856,7 +1856,7 @@ void mptcp_subflow_queue_clean(struct sock *listener_sk, struct sock *listener_s
 		if (msk->first)
 			sock_hold(msk->first);
 
-		do_cancel_work = __mptcp_close(sk, 0);
+		do_cancel_work = __mptcp_close(sk, -1);
 		release_sock(sk);
 		if (do_cancel_work) {
 			/* lockdep will report a false positive ABBA deadlock
Signature · 1 line
2.39.0
author maintainer
#15 lore raw

PATCH v2 11/13 mptcp: refactor mptcp_stream_accept()

In reply to Paolo Abeni

Rewrite the mptcp socket accept op, partially open-codying inet_accept(), instead of indirectly calling it.

This way we can avoid acquiring the new socket lock twice and we can avoid a couple of indirect calls.

---
 net/mptcp/protocol.c | 27 +++++++++++++++++++--------
 1 file changed, 19 insertions(+), 8 deletions(-)
net/mptcp/protocol.c +19 -8
index f742d558a1b8..cf161deb64d8 100644
@@ -3748,6 +3748,7 @@ static int mptcp_stream_accept(struct socket *sock, struct socket *newsock,
 {
 	struct mptcp_sock *msk = mptcp_sk(sock->sk);
 	struct socket *ssock;
+	struct sock *newsk;
 	int err;
 
 	pr_debug("msk=%p", msk);
@@ -3759,16 +3760,26 @@ static int mptcp_stream_accept(struct socket *sock, struct socket *newsock,
 	if (!ssock)
 		return -EINVAL;
 
-	err = ssock->ops->accept(sock, newsock, flags, kern);
-	if (err == 0 && !mptcp_is_tcpsk(newsock->sk)) {
-		struct mptcp_sock *msk = mptcp_sk(newsock->sk);
+	newsk = mptcp_accept(sock->sk, flags, &err, kern);
+	if (!newsk)
+		return err;
+
+	lock_sock(newsk);
+
+	sock_rps_record_flow(newsk);
+	WARN_ON(!((1 << newsk->sk_state) &
+		  (TCPF_ESTABLISHED | TCPF_SYN_RECV |
+		  TCPF_CLOSE_WAIT | TCPF_CLOSE)));
+
+	sock_graft(newsk, newsock);
+
+	newsock->state = SS_CONNECTED;
+	if (!mptcp_is_tcpsk(newsock->sk)) {
+		struct mptcp_sock *msk = mptcp_sk(newsk);
 		struct mptcp_subflow_context *subflow;
-		struct sock *newsk = newsock->sk;
 
 		set_bit(SOCK_CUSTOM_SOCKOPT, &newsock->flags);
 
-		lock_sock(newsk);
-
 		/* set ssk->sk_socket of accept()ed flows to mptcp socket.
 		 * This is needed so NOSPACE flag can be set from tcp stack.
 		 */
@@ -3778,10 +3789,10 @@ static int mptcp_stream_accept(struct socket *sock, struct socket *newsock,
 			if (!ssk->sk_socket)
 				mptcp_sock_graft(ssk, newsock);
 		}
-		release_sock(newsk);
 	}
+	release_sock(newsk);
 
-	return err;
+	return 0;
 }
 
 static __poll_t mptcp_check_writeable(struct mptcp_sock *msk)
Signature · 1 line
2.39.0
author maintainer
#21 lore raw

PATCH v2 12/13 security, lsm: Introduce security_mptcp_add_subflow()

In reply to Paolo Abeni

MPTCP can create subflows in kernel context, and later indirectly expose them to user-space, via the owning mptcp socket.

As discussed in the reported link, the above causes unexpected failures for server, MPTCP-enabled applications.

Let's introduce a new LSM hook to allow the security module to relabel the subflow according to the owing process.

Note that the new hook requires both the mptcp socket and the new subflow. This could allow future extensions, e.g. explicitly validating the mptcp <-> subflow linkage.

---
v1 -> v2:
- fix build issue with !CONFIG_SECURITY_NETWORK

---
 include/linux/lsm_hook_defs.h | 1 +
 include/linux/lsm_hooks.h     | 9 +++++++++
 include/linux/security.h      | 6 ++++++
 net/mptcp/subflow.c           | 6 ++++++
 security/security.c           | 5 +++++
 5 files changed, 27 insertions(+)
include/linux/lsm_hook_defs.h +1 -0
index ed6cb2ac55fa..860e11e3a26b 100644
@@ -343,6 +343,7 @@ LSM_HOOK(void, LSM_RET_VOID, sctp_sk_clone, struct sctp_association *asoc,
 	 struct sock *sk, struct sock *newsk)
 LSM_HOOK(int, 0, sctp_assoc_established, struct sctp_association *asoc,
 	 struct sk_buff *skb)
+LSM_HOOK(int, 0, mptcp_add_subflow, struct sock *sk, struct sock *ssk)
 #endif /* CONFIG_SECURITY_NETWORK */
 
 #ifdef CONFIG_SECURITY_INFINIBAND
include/linux/lsm_hooks.h +9 -0
index 0a5ba81f7367..84c9c4d4341e 100644
@@ -1096,6 +1096,15 @@
  *	@skb pointer to skbuff of association packet.
  *	Return 0 if permission is granted.
  *
+ * Security hooks for MPTCP
+ *
+ * @mptcp_add_subflow
+ * 	Update the labeling for the given MPTCP subflow, to match to
+ * 	owning MPTCP socket.
+ * 	@sk: the owning MPTCP socket
+ * 	@ssk: the new subflow
+ * 	Return 0 if successful, otherwise < 0 error code.
+ *
  * Security hooks for Infiniband
  *
  * @ib_pkey_access:
include/linux/security.h +6 -0
index 5b67f208f7de..82d50b2ba683 100644
@@ -1479,6 +1479,7 @@ void security_sctp_sk_clone(struct sctp_association *asoc, struct sock *sk,
 			    struct sock *newsk);
 int security_sctp_assoc_established(struct sctp_association *asoc,
 				    struct sk_buff *skb);
+int security_mptcp_add_subflow(struct sock *sk, struct sock *ssk);
 
 #else	/* CONFIG_SECURITY_NETWORK */
 static inline int security_unix_stream_connect(struct sock *sock,
@@ -1706,6 +1707,11 @@ static inline int security_sctp_assoc_established(struct sctp_association *asoc,
 {
 	return 0;
 }
+
+static inline int security_mptcp_add_subflow(struct sock *sk, struct sock *ssk)
+{
+	return 0;
+}
 #endif	/* CONFIG_SECURITY_NETWORK */
 
 #ifdef CONFIG_SECURITY_INFINIBAND
net/mptcp/subflow.c +6 -0
index a1f8bb745c1b..c9a7460d469a 100644
@@ -1690,6 +1690,10 @@ int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
 
 	lock_sock_nested(sf->sk, SINGLE_DEPTH_NESTING);
 
+	err = security_mptcp_add_subflow(sk, sf->sk);
+	if (err)
+		goto release_ssk;
+
 	/* the newly created socket has to be in the same cgroup as its parent */
 	mptcp_attach_cgroup(sk, sf->sk);
 
@@ -1702,6 +1706,8 @@ int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
 	get_net_track(net, &sf->sk->ns_tracker, GFP_KERNEL);
 	sock_inuse_add(net, 1);
 	err = tcp_set_ulp(sf->sk, "mptcp");
+
+release_ssk:
 	release_sock(sf->sk);
 
 	if (err) {
security/security.c +5 -0
index d1571900a8c7..3491a4fc2b1f 100644
@@ -2493,6 +2493,11 @@ int security_sctp_assoc_established(struct sctp_association *asoc,
 }
 EXPORT_SYMBOL(security_sctp_assoc_established);
 
+int security_mptcp_add_subflow(struct sock *sk, struct sock *ssk)
+{
+	return call_int_hook(mptcp_add_subflow, 0, sk, ssk);
+}
+
 #endif	/* CONFIG_SECURITY_NETWORK */
 
 #ifdef CONFIG_SECURITY_INFINIBAND
Signature · 1 line
2.39.0
author maintainer
#22 lore raw

PATCH v2 13/13 selinux: Implement mptcp_add_subflow hook

In reply to Paolo Abeni

Newly added subflows should inherit the LSM label from the associated msk socket regarless current context.

This patch implements the above copying sid and class from the msk context, deleting the existing subflow label, if any, and then re-creating a new one.

The new helper reuses the selinux_netlbl_sk_security_free() function, and the latter can end-up being called multiple times with the same argument; we additionally need to make it idempotent.

---
v1 -> v2:
- cleanup selinux_netlbl_sk_security_free() (Paul)
- inherit label from msk (Paul)

v1 -> v2:
- fix build issue with !CONFIG_NETLABEL

---
 security/selinux/hooks.c    | 16 ++++++++++++++++
 security/selinux/netlabel.c |  8 ++++++--
 2 files changed, 22 insertions(+), 2 deletions(-)
security/selinux/hooks.c +16 -0
index 3c5be76a9199..1e0ca10a6c02 100644
@@ -5476,6 +5476,21 @@ static void selinux_sctp_sk_clone(struct sctp_association *asoc, struct sock *sk
 	selinux_netlbl_sctp_sk_clone(sk, newsk);
 }
 
+static int selinux_mptcp_add_subflow(struct sock *sk, struct sock *ssk)
+{
+	struct sk_security_struct *ssksec = ssk->sk_security;
+	struct sk_security_struct *sksec = sk->sk_security;
+
+	ssksec->sclass = sksec->sclass;
+	ssksec->sid = sksec->sid;
+
+	/* replace the existing subflow label deleting the existing one
+	 * and re-recrating a new label using the current context
+	 */
+	selinux_netlbl_sk_security_free(ssksec);
+	return selinux_netlbl_socket_post_create(ssk, ssk->sk_family);
+}
+
 static int selinux_inet_conn_request(const struct sock *sk, struct sk_buff *skb,
 				     struct request_sock *req)
 {
@@ -7216,6 +7231,7 @@ static struct security_hook_list selinux_hooks[] __lsm_ro_after_init = {
 	LSM_HOOK_INIT(sctp_sk_clone, selinux_sctp_sk_clone),
 	LSM_HOOK_INIT(sctp_bind_connect, selinux_sctp_bind_connect),
 	LSM_HOOK_INIT(sctp_assoc_established, selinux_sctp_assoc_established),
+	LSM_HOOK_INIT(mptcp_add_subflow, selinux_mptcp_add_subflow),
 	LSM_HOOK_INIT(inet_conn_request, selinux_inet_conn_request),
 	LSM_HOOK_INIT(inet_csk_clone, selinux_inet_csk_clone),
 	LSM_HOOK_INIT(inet_conn_established, selinux_inet_conn_established),
security/selinux/netlabel.c +6 -2
index 1321f15799e2..33187e38def7 100644
@@ -155,8 +155,12 @@ void selinux_netlbl_err(struct sk_buff *skb, u16 family, int error, int gateway)
  */
 void selinux_netlbl_sk_security_free(struct sk_security_struct *sksec)
 {
-	if (sksec->nlbl_secattr != NULL)
-		netlbl_secattr_free(sksec->nlbl_secattr);
+	if (!sksec->nlbl_secattr)
+		return;
+
+	netlbl_secattr_free(sksec->nlbl_secattr);
+	sksec->nlbl_secattr = NULL;
+	sksec->nlbl_state = NLBL_UNSET;
 }
 
 /**
Signature · 1 line
2.39.0
#23 lore raw

Re: selinux: Implement mptcp_add_subflow hook: Tests Results

Hi Paolo,

Thank you for your modifications, that's great!

Our CI did some validations and here is its report:

- KVM Validation: normal (except selftest_mptcp_join):
- Unstable: 2 failed test(s): packetdrill_syscalls selftest_simult_flows 🔴:
- Task: https://cirrus-ci.com/task/4799310072643584
- Summary: https://api.cirrus-ci.com/v1/artifact/task/4799310072643584/summary/summary.txt

- KVM Validation: normal (only selftest_mptcp_join):
- Success! ✅:
- Task: https://cirrus-ci.com/task/5925209979486208
- Summary: https://api.cirrus-ci.com/v1/artifact/task/5925209979486208/summary/summary.txt

- KVM Validation: debug (except selftest_mptcp_join):
- Unstable: 2 failed test(s): packetdrill_syscalls selftest_diag 🔴:
- Task: https://cirrus-ci.com/task/5362260026064896
- Summary: https://api.cirrus-ci.com/v1/artifact/task/5362260026064896/summary/summary.txt

- KVM Validation: debug (only selftest_mptcp_join):
- Success! ✅:
- Task: https://cirrus-ci.com/task/6488159932907520
- Summary: https://api.cirrus-ci.com/v1/artifact/task/6488159932907520/summary/summary.txt

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/ee85063baf78

If there are some issues, you can reproduce them using the same environment as the one used by the CI thanks to a docker image, e.g.:

    $ cd [kernel source code]
    $ docker run -v "${PWD}:${PWD}:rw" -w "${PWD}" --privileged --rm -it \
        --pull always mptcp/mptcp-upstream-virtme-docker:latest \
        auto-debug

For more details:

https://github.com/multipath-tcp/mptcp-upstream-virtme-docker

Please note that despite all the efforts that have been already done to have a stable tests suite when executed on a public CI like here, it is possible some reported issues are not due to your modifications. Still, do not hesitate to help us improve that ;-)

Cheers,
MPTCP GH Action bot
Bot operated by Matthieu Baerts (Tessares)

author maintainer
#24 lore raw

Re: selinux: Implement mptcp_add_subflow hook: Tests Results

Quoted 16 lines · MPTCP CI

Hi Paolo,

Thank you for your modifications, that's great!

Our CI did some validations and here is its report:

- KVM Validation: normal (except selftest_mptcp_join):
- Unstable: 2 failed test(s): packetdrill_syscalls selftest_simult_flows 🔴:
- Task: https://cirrus-ci.com/task/4799310072643584
- Summary: https://api.cirrus-ci.com/v1/artifact/task/4799310072643584/summary/summary.txt

- KVM Validation: normal (only selftest_mptcp_join):
- Success! ✅:
- Task: https://cirrus-ci.com/task/5925209979486208
- Summary: https://api.cirrus-ci.com/v1/artifact/task/5925209979486208/summary/summary.txt

Quoted 4 lines

- KVM Validation: debug (except selftest_mptcp_join):
- Unstable: 2 failed test(s): packetdrill_syscalls selftest_diag 🔴:
- Task: https://cirrus-ci.com/task/5362260026064896
- Summary: https://api.cirrus-ci.com/v1/artifact/task/5362260026064896/summary/summary.txt

Note that packetdrill failures are expected: the close_before_accept.pkt script should be updated accordingly

I can reproduce - to some extent, say once in many runs - the diag failure. I'll investigate it.

I can't reproduce the simult_flows fail, which is quite suspect to me. I think this series should not impact on such test in any way.

/P

#9 lore raw

Re: [PATCH mptcp-next v2 08/13] mptcp: move first subflow allocation at mpc access time

In reply to Paolo Abeni

Hi Paolo,

Quoted 8 lines · Paolo Abeni

In the long run this will simplify the mptcp code and will allow for more consistent behavior. Move the first subflow allocation out of the sock->init ops into the __mptcp_nmpc_socket() helper.

Since the first subflow creation can now happen after the first setsockopt() we additionally need to invoke mptcp_sockopt_sync() on it.

Thank you for this improvement! I have some questions below if you don't mind :)

(...)

Quoted 14 lines
net/mptcp/protocol.c +10 -0
@@ -116,6 +104,31 @@ static int __mptcp_socket_create(struct mptcp_sock *msk)
 	return 0;
 }
 
+/* Returns the first subflow socket if available and the MPC
+ * handshake is not started yet.
+ */
+struct socket *__mptcp_nmpc_socket(struct mptcp_sock *msk)
+{
+	struct sock *sk = (struct sock *)msk;
+	int ret;
+
+	if (!((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN)))
+		return ERR_PTR(-EINVAL);

net/mptcp/protocol.c static int __mptcp_socket_create(struct mptcp_sock *msk) · line 126

Quoted 10 lines
net/mptcp/protocol.c +10 -0
+
+	if (!msk->subflow) {
+		if (msk->first)
+			return ERR_PTR(-EINVAL);
+
+		ret = __mptcp_socket_create(msk);
+		if (ret)
+			return ERR_PTR(ret);
+
+		mptcp_sockopt_sync(msk, msk->first);

Do you think the function should be renamed? Maybe not but it looks strange to me to create the socket and do the sync here with this name. Maybe it is just me but with this name, I would expect it just to returns the first subflow or NULL, but not create a socket if not, no?

Mainly to make it clear a "[gs]etsockopt()" before the MPC will create a new socket.

Quoted 17 lines
net/mptcp/protocol.c +6 -0
+	}
+
+	return msk->subflow;
+}
+
 static void mptcp_drop(struct sock *sk, struct sk_buff *skb)
 {
 	sk_drops_add(sk, skb);
@@ -1699,6 +1712,7 @@ static int mptcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg,
 {
 	unsigned int saved_flags = msg->msg_flags;
 	struct mptcp_sock *msk = mptcp_sk(sk);
+	struct socket *ssock;
 	struct sock *ssk;
 	int ret;
 

net/mptcp/protocol.c static int mptcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg, · line 1728

Quoted 9 lines
net/mptcp/protocol.c +4 -2
@@ -1708,8 +1722,11 @@ static int mptcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg,
 	 * Since the defer_connect flag is cleared after the first succsful
 	 * fastopen attempt, no need to check for additional subflow status.
 	 */
-	if (msg->msg_flags & MSG_FASTOPEN && !__mptcp_nmpc_socket(msk))
-		return -EINVAL;
+	if (msg->msg_flags & MSG_FASTOPEN) {
+		ssock = __mptcp_nmpc_socket(msk);
+		if (IS_ERR(ssock))
+			return PTR_ERR(ssock);> +	}

Here 'ssock' could be declared in this 'if' section: I guess it doesn't really matter but out of curiosity, what's the code style to apply here? Fine where it is or moved only where needed?

Quoted 105 lines
net/mptcp/protocol.c +6 -10
 	if (!msk->first)
 		return -EINVAL;
 
@@ -2768,10 +2785,6 @@ static int mptcp_init_sock(struct sock *sk)
 	if (unlikely(!net->mib.mptcp_statistics) && !mptcp_mib_alloc(net))
 		return -ENOMEM;
 
-	ret = __mptcp_socket_create(mptcp_sk(sk));
-	if (ret)
-		return ret;
-
 	ret = mptcp_init_sched(mptcp_sk(sk),
 			       mptcp_sched_find(mptcp_get_scheduler(net)));
 	if (ret)
@@ -3592,8 +3605,8 @@ static int mptcp_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 	int err = -EINVAL;
 
 	ssock = __mptcp_nmpc_socket(msk);
-	if (!ssock)
-		return -EINVAL;
+	if (IS_ERR(ssock))
+		return PTR_ERR(ssock);
 
 	mptcp_token_destroy(msk);
 	inet_sk_state_store(sk, TCP_SYN_SENT);
@@ -3681,8 +3694,8 @@ static int mptcp_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 
 	lock_sock(sock->sk);
 	ssock = __mptcp_nmpc_socket(msk);
-	if (!ssock) {
-		err = -EINVAL;
+	if (IS_ERR(ssock)) {
+		err = PTR_ERR(ssock);
 		goto unlock;
 	}
 
@@ -3718,8 +3731,8 @@ static int mptcp_listen(struct socket *sock, int backlog)
 
 	lock_sock(sk);
 	ssock = __mptcp_nmpc_socket(msk);
-	if (!ssock) {
-		err = -EINVAL;
+	if (IS_ERR(ssock)) {
+		err = PTR_ERR(ssock);
 		goto unlock;
 	}
 
net/mptcp/protocol.h +1 -1
index 5f1a30959b5c..3a055438c65e 100644
@@ -632,7 +632,7 @@ void __mptcp_subflow_send_ack(struct sock *ssk);
 void mptcp_subflow_reset(struct sock *ssk);
 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
-struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
+struct socket *__mptcp_nmpc_socket(struct mptcp_sock *msk);
 bool __mptcp_close(struct sock *sk, long timeout);
 void mptcp_cancel_work(struct sock *sk);
 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
net/mptcp/sockopt.c +6 -7
index 8a9656248b0f..9bddae9c5c58 100644
@@ -301,9 +301,9 @@ static int mptcp_setsockopt_sol_socket(struct mptcp_sock *msk, int optname,
 	case SO_BINDTOIFINDEX:
 		lock_sock(sk);
 		ssock = __mptcp_nmpc_socket(msk);
-		if (!ssock) {
+		if (IS_ERR(ssock)) {
 			release_sock(sk);
-			return -EINVAL;
+			return PTR_ERR(ssock);
 		}
 
 		ret = sock_setsockopt(ssock, SOL_SOCKET, optname, optval, optlen);
@@ -396,9 +396,9 @@ static int mptcp_setsockopt_v6(struct mptcp_sock *msk, int optname,
 	case IPV6_FREEBIND:
 		lock_sock(sk);
 		ssock = __mptcp_nmpc_socket(msk);
-		if (!ssock) {
+		if (IS_ERR(ssock)) {
 			release_sock(sk);
-			return -EINVAL;
+			return PTR_ERR(ssock);
 		}
 
 		ret = tcp_setsockopt(ssock->sk, SOL_IPV6, optname, optval, optlen);
@@ -693,9 +693,9 @@ static int mptcp_setsockopt_sol_ip_set_transparent(struct mptcp_sock *msk, int o
 	lock_sock(sk);
 
 	ssock = __mptcp_nmpc_socket(msk);
-	if (!ssock) {
+	if (IS_ERR(ssock)) {
 		release_sock(sk);
-		return -EINVAL;
+		return PTR_ERR(ssock);
 	}
 
 	issk = inet_sk(ssock->sk);
@@ -762,13 +762,15 @@ static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int
 {
 	struct sock *sk = (struct sock *)msk;
 	struct socket *sock;
-	int ret = -EINVAL;

net/mptcp/sockopt.c static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int · line 773

Quoted 10 lines
net/mptcp/sockopt.c +4 -1
+	int ret;
 
 	/* Limit to first subflow, before the connection establishment */
 	lock_sock(sk);
 	sock = __mptcp_nmpc_socket(msk);
-	if (!sock)
+	if (IS_ERR(sock)) {
+		ret = PTR_ERR(sock);
 		goto unlock;
+	}

Just to be sure: here and above the error with __mptcp_nmpc_socket() should not happen most of the time: only when we reach some limits, e.g. mem, etc., right?

So it is fine to potentially return a different error than before (and potentially different than what TCP what doing), is it correct?

net/mptcp/sockopt.c static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int · line 880

Quoted 12 lines
net/mptcp/sockopt.c +1 -1
 	ret = tcp_setsockopt(sock->sk, level, optname, optval, optlen);
 
@@ -872,7 +874,7 @@ static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int
 	}
 
 	ssock = __mptcp_nmpc_socket(msk);
-	if (!ssock)
+	if (IS_ERR(ssock))
 		goto out;
 
 	ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);

Why here you don't assign the error that is returned? (Maybe a comment should be added?)

Cheers,
Matt

Signature · 2 lines
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net
#16 lore raw

Re: [PATCH mptcp-next v2 11/13] mptcp: refactor mptcp_stream_accept()

In reply to Paolo Abeni

Hi Paolo,

Quoted 5 lines · Paolo Abeni

Rewrite the mptcp socket accept op, partially open-codying inet_accept(), instead of indirectly calling it.

This way we can avoid acquiring the new socket lock twice and we can avoid a couple of indirect calls.

Maybe a stupid idea but could we eventually split inet_accept() to have a __inet_accept() version with what has to be done while holding the lock and reusable in MPTCP?

From MPTCP, we would not need to duplicate the code but call __inet_accept() (no more indirection) and be ready for Zero Copy stuff or other modifications of __inet_accept() later if needed.

Cheers,
Matt

Signature · 2 lines
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net
#20 lore raw

Re: [PATCH mptcp-next v2 11/13] mptcp: refactor mptcp_stream_accept()

In reply to Paolo Abeni

Hi Paolo,

Quoted 5 lines · Paolo Abeni

Rewrite the mptcp socket accept op, partially open-codying inet_accept(), instead of indirectly calling it.

This way we can avoid acquiring the new socket lock twice and we can avoid a couple of indirect calls.

Maybe a stupid idea but could we eventually split inet_accept() to have a __inet_accept() version with what has to be done while holding the lock and reusable in MPTCP?

From MPTCP, we would not need to duplicate the code but call __inet_accept() (no more indirection) and be ready for Zero Copy stuff or other modifications of __inet_accept() later if needed.

Cheers,
Matt

Signature · 2 lines
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net
#14 lore raw

Re: [PATCH mptcp-next v2 10/13] mptcp: fastclose msk when cleaning unaccepted sockets

In reply to Paolo Abeni
Quoted 19 lines · Paolo Abeni

When cleaning up unaccepted mptcp socket still laying inside the listener queue at listener close time, such sockets will go through a regular close, waiting for a timeout before shutting down the subflows.

There is no need to keep the kernel resources in use for such a possibly long time: short-circuit to fast-close.

---
 net/mptcp/protocol.c | 7 +++++--
 net/mptcp/subflow.c  | 2 +-
 2 files changed, 6 insertions(+), 3 deletions(-)
net/mptcp/protocol.c +0 -0
index b867e3eec5b9..f742d558a1b8 100644

net/mptcp/protocol.c bool __mptcp_close(struct sock *sk, long timeout) · line 2982

Quoted 8 lines
net/mptcp/protocol.c +3 -2
@@ -2977,10 +2977,13 @@ bool __mptcp_close(struct sock *sk, long timeout)
 		goto cleanup;
 	}
 
-	if (mptcp_check_readable(msk)) {
-		/* the msk has read data, do the MPTCP equivalent of TCP reset */
+	if (mptcp_check_readable(msk) || timeout < 0) {
+		/* If the msk has read data, or the caller explicitly ask it,
+		 * do the MPTCP equivalent of TCP reset, aka MTCP fastclose

(small typo here: s/MTCP/MPTCP/ -- I can fix it when applying the patch)

Cheers,
Matt

Signature · 2 lines
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net
author maintainer
#10 lore raw

Re: [PATCH mptcp-next v2 08/13] mptcp: move first subflow allocation at mpc access time

In reply to Matthieu Baerts
Quoted 32 lines · Matthieu Baerts
Quoted 14 lines · Paolo Abeni
net/mptcp/protocol.c +10 -0
@@ -116,6 +104,31 @@ static int __mptcp_socket_create(struct mptcp_sock *msk)
 	return 0;
 }
 
+/* Returns the first subflow socket if available and the MPC
+ * handshake is not started yet.
+ */
+struct socket *__mptcp_nmpc_socket(struct mptcp_sock *msk)
+{
+	struct sock *sk = (struct sock *)msk;
+	int ret;
+
+	if (!((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN)))
+		return ERR_PTR(-EINVAL);

net/mptcp/protocol.c static int __mptcp_socket_create(struct mptcp_sock *msk) · line 126

Quoted 10 lines
net/mptcp/protocol.c +10 -0
+
+	if (!msk->subflow) {
+		if (msk->first)
+			return ERR_PTR(-EINVAL);
+
+		ret = __mptcp_socket_create(msk);
+		if (ret)
+			return ERR_PTR(ret);
+
+		mptcp_sockopt_sync(msk, msk->first);

Do you think the function should be renamed? Maybe not but it looks strange to me to create the socket and do the sync here with this name. Maybe it is just me but with this name, I would expect it just to returns the first subflow or NULL, but not create a socket if not, no?

Quoted 2 lines

Mainly to make it clear a "[gs]etsockopt()" before the MPC will create a new socket.

I don't have a good alternative name handy. You have good suggestions, they are more then welcome!

IMHO the function name is not bad: it does not imply allocation, nor lack of such thing.

Quoted 6 lines
Quoted 6 lines
net/mptcp/protocol.c +0 -2
@@ -1708,8 +1722,11 @@ static int mptcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg,
 	 * Since the defer_connect flag is cleared after the first succsful
 	 * fastopen attempt, no need to check for additional subflow status.
 	 */
-	if (msg->msg_flags & MSG_FASTOPEN && !__mptcp_nmpc_socket(msk))
-		return -EINVAL;
Quoted 8 lines

net/mptcp/protocol.c static int mptcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg,

Quoted 4 lines
net/mptcp/protocol.c +4 -0
+	if (msg->msg_flags & MSG_FASTOPEN) {
+		ssock = __mptcp_nmpc_socket(msk);
+		if (IS_ERR(ssock))
+			return PTR_ERR(ssock);> +	}

Here 'ssock' could be declared in this 'if' section: I guess it doesn't really matter but out of curiosity, what's the code style to apply here? Fine where it is or moved only where needed?

AFAIK there is no strict codying style rule (even if the preference should go for the smaller scope). I opted for this way just to reduce the number of added line - otherwise an additional empty line would have been mandatory.

Quoted 20 lines
Quoted 5 lines
net/mptcp/sockopt.c +0 -1
@@ -762,13 +762,15 @@ static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int
 {
 	struct sock *sk = (struct sock *)msk;
 	struct socket *sock;
-	int ret = -EINVAL;

net/mptcp/sockopt.c static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int · line 773

Quoted 10 lines
net/mptcp/sockopt.c +4 -1
+	int ret;
 
 	/* Limit to first subflow, before the connection establishment */
 	lock_sock(sk);
 	sock = __mptcp_nmpc_socket(msk);
-	if (!sock)
+	if (IS_ERR(sock)) {
+		ret = PTR_ERR(sock);
 		goto unlock;
+	}

Just to be sure: here and above the error with __mptcp_nmpc_socket() should not happen most of the time: only when we reach some limits, e.g. mem, etc., right?

Quoted 2 lines

So it is fine to potentially return a different error than before (and potentially different than what TCP what doing), is it correct?

Well the situation is a bit fuzzy. __mptcp_nmpc_socket() can now return an additional error code (ENOMEM) compared to the code prior to this patch. But plain tcp could already return such error for TCP_FASTOPEN_KEY. For the other sockopt addressed with mptcp_setsockopt_first_sf_only(), that will introduce a new possible error code. I think that is fine, the user-space application must not restrict the set of accepted error code for a given syscall.

Quoted 10 lines
Quoted 10 lines
net/mptcp/sockopt.c +1 -1
 	ret = tcp_setsockopt(sock->sk, level, optname, optval, optlen);
 
@@ -872,7 +874,7 @@ static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int
 	}
 
 	ssock = __mptcp_nmpc_socket(msk);
-	if (!ssock)
+	if (IS_ERR(ssock))
 		goto out;
 
Quoted 4 lines

net/mptcp/sockopt.c static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int · line 880

Quoted 1 line
net/mptcp/sockopt.c +0 -0
 	ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);

Why here you don't assign the error that is returned? (Maybe a comment should be added?)

I guess it felt strange returning ENOMEM for a get operation. I can switch to full error reporting if you prefer.

Thanks!

Paolo

author maintainer
#17 lore raw

Re: [PATCH mptcp-next v2 11/13] mptcp: refactor mptcp_stream_accept()

In reply to Matthieu Baerts
Quoted 6 lines · Matthieu Baerts

Hi Paolo,

Quoted 3 lines · Paolo Abeni

Rewrite the mptcp socket accept op, partially open-codying inet_accept(), instead of indirectly calling it.

Quoted 6 lines
Quoted 2 lines

This way we can avoid acquiring the new socket lock twice and we can avoid a couple of indirect calls.

Maybe a stupid idea but could we eventually split inet_accept() to have a __inet_accept() version with what has to be done while holding the lock and reusable in MPTCP?

I thought about the above but in the end that would be a very small helper:

        sock_rps_record_flow(sk2);
        WARN_ON(!((1 << sk2->sk_state) &
                  (TCPF_ESTABLISHED | TCPF_SYN_RECV |
                  TCPF_CLOSE_WAIT | TCPF_CLOSE)));

        if (test_bit(SOCK_SUPPORT_ZC, &sock->flags))
                set_bit(SOCK_SUPPORT_ZC, &newsock->flags);
        sock_graft(sk2, newsock);

        newsock->state = SS_CONNECTED;

with the larger part - the WARN_ON - being likely obsoleted.

I can do that, if you prefer.

Thanks,

Paolo

#11 lore raw

Re: [PATCH mptcp-next v2 08/13] mptcp: move first subflow allocation at mpc access time

In reply to Paolo Abeni

Hi Paolo,

Thank you for your reply!

Quoted 33 lines · Paolo Abeni
Quoted 32 lines · Matthieu Baerts
Quoted 14 lines · Paolo Abeni
net/mptcp/protocol.c +10 -0
@@ -116,6 +104,31 @@ static int __mptcp_socket_create(struct mptcp_sock *msk)
 	return 0;
 }
 
+/* Returns the first subflow socket if available and the MPC
+ * handshake is not started yet.
+ */
+struct socket *__mptcp_nmpc_socket(struct mptcp_sock *msk)
+{
+	struct sock *sk = (struct sock *)msk;
+	int ret;
+
+	if (!((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN)))
+		return ERR_PTR(-EINVAL);

net/mptcp/protocol.c static int __mptcp_socket_create(struct mptcp_sock *msk) · line 126

Quoted 10 lines
net/mptcp/protocol.c +10 -0
+
+	if (!msk->subflow) {
+		if (msk->first)
+			return ERR_PTR(-EINVAL);
+
+		ret = __mptcp_socket_create(msk);
+		if (ret)
+			return ERR_PTR(ret);
+
+		mptcp_sockopt_sync(msk, msk->first);

Do you think the function should be renamed? Maybe not but it looks strange to me to create the socket and do the sync here with this name. Maybe it is just me but with this name, I would expect it just to returns the first subflow or NULL, but not create a socket if not, no?

Quoted 5 lines
Quoted 2 lines

Mainly to make it clear a "[gs]etsockopt()" before the MPC will create a new socket.

I don't have a good alternative name handy. You have good suggestions, they are more then welcome!

Indeed, not easy to find an alternative.

__mptcp_initial_or_new_sf_socket_before_estab()

might be too long :þ

(but quite explicit :) )

Quoted 2 lines

IMHO the function name is not bad: it does not imply allocation, nor lack of such thing.

You are right. If there is no better names, we can keep it like that.

If you want, I can add this line in the comment when applying the patch (if there is no v3):

Be careful that a new socket might be created if needed.

Quoted 15 lines
Quoted 6 lines
Quoted 6 lines
net/mptcp/protocol.c +0 -2
@@ -1708,8 +1722,11 @@ static int mptcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg,
 	 * Since the defer_connect flag is cleared after the first succsful
 	 * fastopen attempt, no need to check for additional subflow status.
 	 */
-	if (msg->msg_flags & MSG_FASTOPEN && !__mptcp_nmpc_socket(msk))
-		return -EINVAL;
Quoted 8 lines

net/mptcp/protocol.c static int mptcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg,

Quoted 4 lines
net/mptcp/protocol.c +4 -0
+	if (msg->msg_flags & MSG_FASTOPEN) {
+		ssock = __mptcp_nmpc_socket(msk);
+		if (IS_ERR(ssock))
+			return PTR_ERR(ssock);> +	}

Here 'ssock' could be declared in this 'if' section: I guess it doesn't really matter but out of curiosity, what's the code style to apply here? Fine where it is or moved only where needed?

Quoted 4 lines

AFAIK there is no strict codying style rule (even if the preference should go for the smaller scope). I opted for this way just to reduce the number of added line - otherwise an additional empty line would have been mandatory.

Thank you for the explanation. OK to keep it like that!

Quoted 23 lines
Quoted 20 lines
Quoted 5 lines
net/mptcp/sockopt.c +0 -1
@@ -762,13 +762,15 @@ static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int
 {
 	struct sock *sk = (struct sock *)msk;
 	struct socket *sock;
-	int ret = -EINVAL;

net/mptcp/sockopt.c static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int · line 773

Quoted 10 lines
net/mptcp/sockopt.c +4 -1
+	int ret;
 
 	/* Limit to first subflow, before the connection establishment */
 	lock_sock(sk);
 	sock = __mptcp_nmpc_socket(msk);
-	if (!sock)
+	if (IS_ERR(sock)) {
+		ret = PTR_ERR(sock);
 		goto unlock;
+	}

Just to be sure: here and above the error with __mptcp_nmpc_socket() should not happen most of the time: only when we reach some limits, e.g. mem, etc., right?

Quoted 2 lines

So it is fine to potentially return a different error than before (and potentially different than what TCP what doing), is it correct?

Quoted 7 lines

Well the situation is a bit fuzzy. __mptcp_nmpc_socket() can now return an additional error code (ENOMEM) compared to the code prior to this patch. But plain tcp could already return such error for TCP_FASTOPEN_KEY. For the other sockopt addressed with mptcp_setsockopt_first_sf_only(), that will introduce a new possible error code. I think that is fine, the user-space application must not restrict the set of accepted error code for a given syscall.

OK so additional error codes should then be rare. Then it is fine. If they get the ENOMEM now, they would have got it later. If they ignore it, that's fine. If they consider it as a big issue resulting on not doing the connect/bind/listen, that's fine as well.

Thank you for having looked!

Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
net/mptcp/sockopt.c +1 -1
 	ret = tcp_setsockopt(sock->sk, level, optname, optval, optlen);
 
@@ -872,7 +874,7 @@ static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int
 	}
 
 	ssock = __mptcp_nmpc_socket(msk);
-	if (!ssock)
+	if (IS_ERR(ssock))
 		goto out;
 
Quoted 7 lines
Quoted 4 lines

net/mptcp/sockopt.c static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int · line 880

Quoted 1 line
net/mptcp/sockopt.c +0 -0
 	ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);

Why here you don't assign the error that is returned? (Maybe a comment should be added?)

I guess it felt strange returning ENOMEM for a get operation. I can switch to full error reporting if you prefer.

Indeed, you are right.

Do you mind if I add this when applying the patch (if there is no v3):

/* don't return err linked to the socket creation for a getsockopt */

Cheers,
Matt

Signature · 2 lines
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net
#18 lore raw

Re: [PATCH mptcp-next v2 11/13] mptcp: refactor mptcp_stream_accept()

In reply to Paolo Abeni

Hi Paolo,

Quoted 29 lines · Paolo Abeni
Quoted 6 lines · Matthieu Baerts

Hi Paolo,

Quoted 3 lines · Paolo Abeni

Rewrite the mptcp socket accept op, partially open-codying inet_accept(), instead of indirectly calling it.

Quoted 6 lines
Quoted 2 lines

This way we can avoid acquiring the new socket lock twice and we can avoid a couple of indirect calls.

Maybe a stupid idea but could we eventually split inet_accept() to have a __inet_accept() version with what has to be done while holding the lock and reusable in MPTCP?

I thought about the above but in the end that would be a very small helper:

        sock_rps_record_flow(sk2);
        WARN_ON(!((1 << sk2->sk_state) &
                  (TCPF_ESTABLISHED | TCPF_SYN_RECV |
                  TCPF_CLOSE_WAIT | TCPF_CLOSE)));

        if (test_bit(SOCK_SUPPORT_ZC, &sock->flags))
                set_bit(SOCK_SUPPORT_ZC, &newsock->flags);
        sock_graft(sk2, newsock);

        newsock->state = SS_CONNECTED;
Quoted 3 lines

with the larger part - the WARN_ON - being likely obsoleted.

I can do that, if you prefer.

I would prefer but please tell me if you think we should not.

I'm maybe a bit too paranoiac because we had some nasty bugs in the past with mptcp.org due to modifications on TCP side. (But I know the situation is different here: MPTCP is upstream, people might now think about it when changing the core and a great net maintainer is super aware of MPTCP :-D )

If the chunk is "duplicated", it would be nice to at least add a comment in the code :)

Cheers,
Matt

Signature · 2 lines
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net
author maintainer
#19 lore raw

Re: [PATCH mptcp-next v2 11/13] mptcp: refactor mptcp_stream_accept()

Quoted 34 lines · Matthieu Baerts
Quoted 27 lines · Paolo Abeni
Quoted 10 lines · Matthieu Baerts
Quoted 5 lines · Paolo Abeni

Rewrite the mptcp socket accept op, partially open-codying inet_accept(), instead of indirectly calling it.

This way we can avoid acquiring the new socket lock twice and we can avoid a couple of indirect calls.

Maybe a stupid idea but could we eventually split inet_accept() to have a __inet_accept() version with what has to be done while holding the lock and reusable in MPTCP?

I thought about the above but in the end that would be a very small helper:

        sock_rps_record_flow(sk2);
        WARN_ON(!((1 << sk2->sk_state) &
                  (TCPF_ESTABLISHED | TCPF_SYN_RECV |
                  TCPF_CLOSE_WAIT | TCPF_CLOSE)));

        if (test_bit(SOCK_SUPPORT_ZC, &sock->flags))
                set_bit(SOCK_SUPPORT_ZC, &newsock->flags);
        sock_graft(sk2, newsock);

        newsock->state = SS_CONNECTED;
Quoted 3 lines

with the larger part - the WARN_ON - being likely obsoleted.

I can do that, if you prefer.

I would prefer but please tell me if you think we should not.

Quoted 8 lines

I'm maybe a bit too paranoiac because we had some nasty bugs in the past with mptcp.org due to modifications on TCP side. (But I know the situation is different here: MPTCP is upstream, people might now think about it when changing the core and a great net maintainer is super aware of MPTCP :-D )

If the chunk is "duplicated", it would be nice to at least add a comment in the code :)

Ok I can send out a v3 with the bunch of additional comments discussed in the other patches and the new helper (will add one more patch).

Cheers,

Paolo