Content-Type: multipart/mixed; boundary="===============6907801556688099847==" MIME-Version: 1.0 From: Mat Martineau To: mptcp at lists.01.org Subject: [MPTCP] [RFC PATCH v4 17/17] mptcp: Make connection_list a real list of subflows Date: Fri, 30 Nov 2018 12:11:12 -0800 Message-ID: <20181130201112.21806-18-mathew.j.martineau@linux.intel.com> In-Reply-To: 20181130201112.21806-1-mathew.j.martineau@linux.intel.com X-Status: X-Keywords: X-UID: 870 --===============6907801556688099847== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable From: Peter Krystad Signed-off-by: Peter Krystad --- include/net/mptcp.h | 8 +- net/mptcp/protocol.c | 195 ++++++++++++++++++++++++++++++------------- 2 files changed, 142 insertions(+), 61 deletions(-) diff --git a/include/net/mptcp.h b/include/net/mptcp.h index d3b7083859b7..8c0b21652b70 100644 --- a/include/net/mptcp.h +++ b/include/net/mptcp.h @@ -17,6 +17,7 @@ #define __NET_MPTCP_H = #include +#include = /* MPTCP option subtypes */ = @@ -44,10 +45,14 @@ struct mptcp_sock { u64 write_seq; atomic64_t ack_seq; u32 token; - struct socket *connection_list; /* @@ needs to be a list */ + spinlock_t conn_list_lock; + struct hlist_head conn_list; struct socket *subflow; /* outgoing connect, listener or !mp_capable */ }; = +#define mptcp_for_each_subflow(__msk, __subflow) \ + hlist_for_each_entry_rcu(__subflow, &((__msk)->conn_list), node) + static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) { return (struct mptcp_sock *)sk; @@ -80,6 +85,7 @@ static inline struct mptcp_skb_cb *mptcp_skb_priv_cb(stru= ct sk_buff *skb) struct subflow_sock { /* tcp_sock must be the first member */ struct tcp_sock sk; + struct hlist_node node;// conn_list of subflows u64 local_key; u64 map_seq; u32 map_subflow_seq; diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 117c83eab584..32213102790b 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -71,39 +71,53 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr= *msg, size_t len) struct mptcp_sock *msk =3D mptcp_sk(sk); int mss_now, size_goal, poffset, ret; struct mptcp_skb_cb *mcb =3D NULL; + struct subflow_sock *subflow; struct page *page =3D NULL; + struct hlist_node *node; struct sk_buff *skb; struct sock *ssk; size_t psize; = pr_debug("msk=3D%p", msk); - if (!msk->connection_list && msk->subflow) { + if (msk->subflow) { pr_debug("fallback passthrough"); return sock_sendmsg(msk->subflow, msg); } = + rcu_read_lock(); + node =3D rcu_dereference(hlist_first_rcu(&msk->conn_list)); + subflow =3D hlist_entry(node, struct subflow_sock, node); + ssk =3D sock_sk(subflow); + sock_hold(ssk); + rcu_read_unlock(); + if (!msg_data_left(msg)) { pr_debug("empty send"); - return sock_sendmsg(msk->connection_list, msg); + ret =3D sock_sendmsg(sock_sk(subflow)->sk_socket, msg); + goto put_out; } = - ssk =3D msk->connection_list->sk; + pr_debug("conn_list->subflow=3D%p", subflow); = - if (msg->msg_flags & ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL)) - return -ENOTSUPP; + if (msg->msg_flags & ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL)) { + ret =3D -ENOTSUPP; + goto put_out; + } = /* Initial experiment: new page per send. Real code will * maintain list of active pages and DSS mappings, append to the * end and honor zerocopy */ page =3D alloc_page(GFP_KERNEL); - if (!page) - return -ENOMEM; + if (!page) { + ret =3D -ENOMEM; + goto put_out; + } = mcb =3D kzalloc(sizeof(*mcb), GFP_KERNEL); if (!mcb) { - put_page(page); - return -ENOMEM; + ret =3D -ENOMEM; + goto put_out; } = /* Copy to page */ @@ -115,9 +129,8 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr= *msg, size_t len) pr_debug("left=3D%zu", msg_data_left(msg)); = if (!psize) { - put_page(page); - kfree(mcb); - return -EINVAL; + ret =3D -EINVAL; + goto put_out; } = lock_sock(sk); @@ -134,9 +147,8 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr= *msg, size_t len) = ret =3D do_tcp_sendpages(ssk, page, poffset, min_t(int, size_goal, psize), msg->msg_flags | MSG_SENDPAGE_NOTLAST); - put_page(page); if (ret <=3D 0) - goto error_out; + goto release_out; = if (skb =3D=3D tcp_write_queue_tail(ssk)) pr_err("no new skb %p/%p", sk, ssk); @@ -169,10 +181,15 @@ static int mptcp_sendmsg(struct sock *sk, struct msgh= dr *msg, size_t len) = tcp_push(ssk, msg->msg_flags, mss_now, tcp_sk(ssk)->nonagle, size_goal); = -error_out: +release_out: release_sock(ssk); release_sock(sk); = +put_out: + if (page) + put_page(page); + + sock_put(ssk); kfree(mcb); return ret; } @@ -306,19 +323,25 @@ static int mptcp_recvmsg(struct sock *sk, struct msgh= dr *msg, size_t len, struct mptcp_sock *msk =3D mptcp_sk(sk); struct subflow_sock *subflow; struct mptcp_read_arg arg; + struct hlist_node *node; read_descriptor_t desc; struct tcp_sock *tp; struct sock *ssk; int copied =3D 0; long timeo; = - if (!msk->connection_list) { + if (msk->subflow) { pr_debug("fallback-read subflow=3D%p", msk->subflow->sk); return sock_recvmsg(msk->subflow, msg, flags); } = - ssk =3D msk->connection_list->sk; - subflow =3D subflow_sk(ssk); + rcu_read_lock(); + node =3D rcu_dereference(hlist_first_rcu(&msk->conn_list)); + subflow =3D hlist_entry(node, struct subflow_sock, node); + ssk =3D sock_sk(subflow); + sock_hold(ssk); + rcu_read_unlock(); + tp =3D tcp_sk(ssk); = desc.arg.data =3D &arg; @@ -474,6 +497,8 @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr= *msg, size_t len, release_sock(ssk); release_sock(sk); = + sock_put(ssk); + return copied; } = @@ -483,22 +508,50 @@ static int mptcp_init_sock(struct sock *sk) = pr_debug("msk=3D%p", msk); = + INIT_HLIST_HEAD(&msk->conn_list); + spin_lock_init(&msk->conn_list_lock); + return 0; } = static void mptcp_close(struct sock *sk, long timeout) { struct mptcp_sock *msk =3D mptcp_sk(sk); + struct hlist_node *node; + struct socket *ssk =3D NULL; + struct subflow_sock *subflow; = + spin_lock_bh(&msk->conn_list_lock); if (msk->subflow) { - pr_debug("subflow=3D%p", msk->subflow->sk); - sock_release(msk->subflow); + ssk =3D msk->subflow; + msk->subflow =3D NULL; } - - if (msk->connection_list) { - pr_debug("conn_list->subflow=3D%p", msk->connection_list->sk); - sock_release(msk->connection_list); + spin_unlock_bh(&msk->conn_list_lock); + if (ssk !=3D NULL) { + pr_debug("subflow=3D%p", ssk->sk); + sock_release(ssk); } + + do { + spin_lock_bh(&msk->conn_list_lock); + /* The spin lock was just acquired, so tell + * rcu_dereference_check() this access to the list is properly + * protected even though the rcu_read_lock is not held. + */ + node =3D rcu_dereference_check( + hlist_first_rcu(&msk->conn_list), + spin_is_locked(&msk->conn_list_lock) || + !IS_ENABLED(CONFIG_SMP)); + if (node =3D=3D NULL) + break; + subflow =3D hlist_entry(node, struct subflow_sock, node); + hlist_del_rcu(node); + spin_unlock_bh(&msk->conn_list_lock); + synchronize_rcu(); + pr_debug("conn_list->subflow=3D%p", subflow); + sock_release(sock_sk(subflow)->sk_socket); + } while (1); + spin_unlock_bh(&msk->conn_list_lock); } = static struct sock *mptcp_accept(struct sock *sk, int flags, int *err, @@ -537,7 +590,10 @@ static struct sock *mptcp_accept(struct sock *sk, int = flags, int *err, msk->token =3D subflow->token; pr_debug("token=3D%u", msk->token); token_update_accept(new_sock->sk, mp->sk); - msk->connection_list =3D new_sock; + spin_lock_bh(&msk->conn_list_lock); + hlist_add_head_rcu(&subflow->node, &msk->conn_list); + msk->subflow =3D NULL; + spin_unlock_bh(&msk->conn_list_lock); = crypto_key_sha1(msk->remote_key, NULL, &ack_seq); msk->write_seq =3D subflow->idsn + 1; @@ -568,46 +624,44 @@ static int mptcp_setsockopt(struct sock *sk, int leve= l, int optname, char __user *uoptval, unsigned int optlen) { struct mptcp_sock *msk =3D mptcp_sk(sk); - struct socket *subflow; char __kernel *optval; = - pr_debug("msk=3D%p", msk); - if (msk->connection_list) { - subflow =3D msk->connection_list; - pr_debug("conn_list->subflow=3D%p", subflow->sk); - } else { - subflow =3D msk->subflow; - pr_debug("subflow=3D%p", subflow->sk); - } - /* will be treated as __user in subflow_setsockopt */ optval =3D (char __kernel __force *) uoptval; = - return kernel_setsockopt(subflow, level, optname, optval, optlen); + pr_debug("msk=3D%p", msk); + if (msk->subflow) { + pr_debug("subflow=3D%p", msk->subflow->sk); + return kernel_setsockopt(msk->subflow, level, optname, optval, optlen); + } + + /* @@ the meaning of setsockopt() when the socket is connected and + * there are multiple subflows is not defined. + */ + return 0; } = static int mptcp_getsockopt(struct sock *sk, int level, int optname, char __user *uoptval, int __user *uoption) { struct mptcp_sock *msk =3D mptcp_sk(sk); - struct socket *subflow; char __kernel *optval; int __kernel *option; = - pr_debug("msk=3D%p", msk); - if (msk->connection_list) { - subflow =3D msk->connection_list; - pr_debug("conn_list->subflow=3D%p", subflow->sk); - } else { - subflow =3D msk->subflow; - pr_debug("subflow=3D%p", subflow->sk); - } - /* will be treated as __user in subflow_getsockopt */ optval =3D (char __kernel __force *) uoptval; option =3D (int __kernel __force *) uoption; = - return kernel_getsockopt(subflow, level, optname, optval, option); + pr_debug("msk=3D%p", msk); + if (msk->subflow) { + pr_debug("subflow=3D%p", msk->subflow->sk); + return kernel_getsockopt(msk->subflow, level, optname, optval, option); + } + + /* @@ the meaning of setsockopt() when the socket is connected and + * there are multiple subflows is not defined. + */ + return 0; } = static int mptcp_get_port(struct sock *sk, unsigned short snum) @@ -634,8 +688,10 @@ void mptcp_finish_connect(struct sock *sk, int mp_capa= ble) msk->local_key =3D subflow->local_key; msk->token =3D subflow->token; pr_debug("token=3D%u", msk->token); - msk->connection_list =3D msk->subflow; + spin_lock_bh(&msk->conn_list_lock); + hlist_add_head_rcu(&subflow->node, &msk->conn_list); msk->subflow =3D NULL; + spin_unlock_bh(&msk->conn_list_lock); = crypto_key_sha1(msk->remote_key, NULL, &ack_seq); msk->write_seq =3D subflow->idsn + 1; @@ -703,17 +759,32 @@ int mptcp_stream_connect(struct socket *sock, struct = sockaddr *uaddr, int mptcp_stream_getname(struct socket *sock, struct sockaddr *uaddr, int = peer) { struct mptcp_sock *msk =3D mptcp_sk(sock->sk); - struct socket *subflow; - int err =3D -EPERM; + struct subflow_sock *subflow; + struct hlist_node *node; + struct sock *ssk; + int ret; = - if (msk->connection_list) - subflow =3D msk->connection_list; - else - subflow =3D msk->subflow; + pr_debug("msk=3D%p", msk); = - err =3D inet_getname(subflow, uaddr, peer); + if (msk->subflow) { + pr_debug("subflow=3D%p", msk->subflow->sk); + return inet_getname(msk->subflow, uaddr, peer); + } = - return err; + /* @@ the meaning of getname() for the remote peer when the socket + * is connected and there are multiple subflows is not defined. + * For now just use the first subflow on the list. + */ + rcu_read_lock(); + node =3D rcu_dereference(hlist_first_rcu(&msk->conn_list)); + subflow =3D hlist_entry(node, struct subflow_sock, node); + ssk =3D sock_sk(subflow); + sock_hold(ssk); + rcu_read_unlock(); + + ret =3D inet_getname(sock_sk(subflow)->sk_socket, uaddr, peer); + sock_put(ssk); + return ret; } = int mptcp_stream_listen(struct socket *sock, int backlog) @@ -747,18 +818,22 @@ int mptcp_stream_accept(struct socket *sock, struct s= ocket *newsock, int flags, int mptcp_stream_shutdown(struct socket *sock, int how) { struct mptcp_sock *msk =3D mptcp_sk(sock->sk); + struct subflow_sock *subflow; int ret =3D 0; = pr_debug("sk=3D%p, how=3D%d", msk, how); = if (msk->subflow) { pr_debug("subflow=3D%p", msk->subflow->sk); - ret =3D kernel_sock_shutdown(msk->subflow, how); + return kernel_sock_shutdown(msk->subflow, how); } - if (msk->connection_list) { - pr_debug("conn_list->subflow=3D%p", msk->connection_list->sk); - ret =3D kernel_sock_shutdown(msk->connection_list, how); + + rcu_read_lock(); + mptcp_for_each_subflow(msk, subflow) { + pr_debug("conn_list->subflow=3D%p", subflow); + ret =3D kernel_sock_shutdown(sock_sk(subflow)->sk_socket, how); } + rcu_read_unlock(); = return ret; } -- = 2.19.1 --===============6907801556688099847==--