From: Mat Martineau <mathew.j.martineau@linux.intel.com>
To: edumazet@google.com, netdev@vger.kernel.org
Cc: Florian Westphal <fw@strlen.de>,
cpaasch@apple.com, pabeni@redhat.com,
peter.krystad@linux.intel.com, dcaratti@redhat.com,
matthieu.baerts@tessares.net
Subject: [RFC PATCH net-next 21/33] mptcp: add and use mptcp_subflow_hold
Date: Mon, 17 Jun 2019 15:57:56 -0700 [thread overview]
Message-ID: <20190617225808.665-22-mathew.j.martineau@linux.intel.com> (raw)
In-Reply-To: <20190617225808.665-1-mathew.j.martineau@linux.intel.com>
From: Florian Westphal <fw@strlen.de>
subflow sockets already have lifetime managed by RCU, so we can
switch to atomic_inc_not_zero and skip/pretend we did not find
such socket in the mptcp subflow list.
This is required to get rid of synchronize_rcu() from mptcp_close().
Signed-off-by: Florian Westphal <fw@strlen.de>
---
net/mptcp/protocol.c | 104 +++++++++++++++++++++++++++----------------
1 file changed, 66 insertions(+), 38 deletions(-)
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index c00e837a1766..0db4099d9c13 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -24,14 +24,35 @@ static inline bool before64(__u64 seq1, __u64 seq2)
#define after64(seq2, seq1) before64(seq1, seq2)
+static bool mptcp_subflow_hold(struct subflow_context *subflow)
+{
+ struct sock *sk = mptcp_subflow_tcp_socket(subflow)->sk;
+
+ return refcount_inc_not_zero(&sk->sk_refcnt);
+}
+
+static struct sock *mptcp_subflow_get_ref(const struct mptcp_sock *msk)
+{
+ struct subflow_context *subflow;
+
+ rcu_read_lock();
+ mptcp_for_each_subflow(msk, subflow) {
+ if (mptcp_subflow_hold(subflow)) {
+ rcu_read_unlock();
+ return mptcp_subflow_tcp_socket(subflow)->sk;
+ }
+ }
+
+ rcu_read_unlock();
+ return NULL;
+}
+
static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
{
struct mptcp_sock *msk = mptcp_sk(sk);
int mss_now, size_goal, poffset, ret;
struct mptcp_ext *mpext = NULL;
- struct subflow_context *subflow;
struct page *page = NULL;
- struct hlist_node *node;
struct sk_buff *skb;
struct sock *ssk;
size_t psize;
@@ -42,20 +63,17 @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
return sock_sendmsg(msk->subflow, msg);
}
- rcu_read_lock();
- node = rcu_dereference(hlist_first_rcu(&msk->conn_list));
- subflow = hlist_entry(node, struct subflow_context, node);
- ssk = mptcp_subflow_tcp_socket(subflow)->sk;
- sock_hold(ssk);
- rcu_read_unlock();
+ ssk = mptcp_subflow_get_ref(msk);
+ if (!ssk)
+ return -ENOTCONN;
if (!msg_data_left(msg)) {
pr_debug("empty send");
- ret = sock_sendmsg(mptcp_subflow_tcp_socket(subflow), msg);
+ ret = sock_sendmsg(ssk->sk_socket, msg);
goto put_out;
}
- pr_debug("conn_list->subflow=%p", subflow);
+ pr_debug("conn_list->subflow=%p", ssk);
if (msg->msg_flags & ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL)) {
ret = -ENOTSUPP;
@@ -293,7 +311,6 @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
struct mptcp_sock *msk = mptcp_sk(sk);
struct subflow_context *subflow;
struct mptcp_read_arg arg;
- struct hlist_node *node;
read_descriptor_t desc;
struct tcp_sock *tp;
struct sock *ssk;
@@ -306,13 +323,11 @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
return sock_recvmsg(msk->subflow, msg, flags);
}
- rcu_read_lock();
- node = rcu_dereference(hlist_first_rcu(&msk->conn_list));
- subflow = hlist_entry(node, struct subflow_context, node);
- ssk = mptcp_subflow_tcp_socket(subflow)->sk;
- sock_hold(ssk);
- rcu_read_unlock();
+ ssk = mptcp_subflow_get_ref(msk);
+ if (!ssk)
+ return -ENOTCONN;
+ subflow = subflow_ctx(ssk);
tp = tcp_sk(ssk);
lock_sock(sk);
@@ -778,8 +793,6 @@ static int mptcp_getname(struct socket *sock, struct sockaddr *uaddr,
int peer)
{
struct mptcp_sock *msk = mptcp_sk(sock->sk);
- struct subflow_context *subflow;
- struct hlist_node *node;
struct sock *ssk;
int ret;
@@ -794,14 +807,11 @@ static int mptcp_getname(struct socket *sock, struct sockaddr *uaddr,
* is connected and there are multiple subflows is not defined.
* For now just use the first subflow on the list.
*/
- rcu_read_lock();
- node = rcu_dereference(hlist_first_rcu(&msk->conn_list));
- subflow = hlist_entry(node, struct subflow_context, node);
- ssk = mptcp_subflow_tcp_socket(subflow)->sk;
- sock_hold(ssk);
- rcu_read_unlock();
+ ssk = mptcp_subflow_get_ref(msk);
+ if (!ssk)
+ return -ENOTCONN;
- ret = inet_getname(mptcp_subflow_tcp_socket(subflow), uaddr, peer);
+ ret = inet_getname(ssk->sk_socket, uaddr, peer);
sock_put(ssk);
return ret;
}
@@ -837,26 +847,44 @@ static int mptcp_stream_accept(struct socket *sock, struct socket *newsock,
static __poll_t mptcp_poll(struct file *file, struct socket *sock,
struct poll_table_struct *wait)
{
- const struct mptcp_sock *msk;
struct subflow_context *subflow;
+ const struct mptcp_sock *msk;
struct sock *sk = sock->sk;
- struct hlist_node *node;
- struct sock *ssk;
- __poll_t ret;
+ __poll_t ret = 0;
+ unsigned int i;
msk = mptcp_sk(sk);
if (msk->subflow)
return tcp_poll(file, msk->subflow, wait);
- rcu_read_lock();
- node = rcu_dereference(hlist_first_rcu(&msk->conn_list));
- subflow = hlist_entry(node, struct subflow_context, node);
- ssk = mptcp_subflow_tcp_socket(subflow)->sk;
- sock_hold(ssk);
- rcu_read_unlock();
+ i = 0;
+ for (;;) {
+ struct subflow_context *tmp = NULL;
+ int j = 0;
+
+ rcu_read_lock();
+ mptcp_for_each_subflow(msk, subflow) {
+ if (j < i) {
+ j++;
+ continue;
+ }
+
+ if (!mptcp_subflow_hold(subflow))
+ continue;
+
+ tmp = subflow;
+ i++;
+ break;
+ }
+ rcu_read_unlock();
+
+ if (!tmp)
+ break;
+
+ ret |= tcp_poll(file, mptcp_subflow_tcp_socket(tmp), wait);
+ sock_put(mptcp_subflow_tcp_socket(tmp)->sk);
+ }
- ret = tcp_poll(file, ssk->sk_socket, wait);
- sock_put(ssk);
return ret;
}
--
2.22.0
next prev parent reply other threads:[~2019-06-17 22:59 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-06-17 22:57 [RFC PATCH net-next 00/33] Multipath TCP Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 01/33] tcp: Add MPTCP option number Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 02/33] tcp: Define IPPROTO_MPTCP Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 03/33] mptcp: Add MPTCP socket stubs Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 04/33] mptcp: Handle MPTCP TCP options Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 05/33] mptcp: Associate MPTCP context with TCP socket Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 06/33] tcp: Expose tcp struct and routine for MPTCP Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 07/33] mptcp: Handle MP_CAPABLE options for outgoing connections Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 08/33] mptcp: add mptcp_poll Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 09/33] tcp, ulp: Add clone operation to tcp_ulp_ops Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 10/33] mptcp: Create SUBFLOW socket for incoming connections Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 11/33] mptcp: Add key generation and token tree Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 12/33] mptcp: Add shutdown() socket operation Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 13/33] mptcp: Add setsockopt()/getsockopt() socket operations Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 14/33] tcp: clean ext on tx recycle Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 15/33] mptcp: Add MPTCP to skb extensions Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 16/33] tcp: Prevent coalesce/collapse when skb has MPTCP extensions Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 17/33] tcp: Export low-level TCP functions Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 18/33] mptcp: Write MPTCP DSS headers to outgoing data packets Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 19/33] mptcp: Implement MPTCP receive path Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 20/33] mptcp: Make connection_list a real list of subflows Mat Martineau
2019-06-17 22:57 ` Mat Martineau [this message]
2019-06-17 22:57 ` [RFC PATCH net-next 22/33] mptcp: add basic kselftest program Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 23/33] mptcp: selftests: switch to netns+veth based tests Mat Martineau
2019-06-17 22:57 ` [RFC PATCH net-next 24/33] mptcp: selftests: Add capture option Mat Martineau
2019-06-17 22:58 ` [RFC PATCH net-next 25/33] mptcp: use sk_page_frag() in sendmsg Mat Martineau
2019-06-17 22:58 ` [RFC PATCH net-next 26/33] mptcp: sendmsg() do spool all the provided data Mat Martineau
2019-06-17 22:58 ` [RFC PATCH net-next 27/33] mptcp: allow collapsing consecutive sendpages on the same substream Mat Martineau
2019-06-17 22:58 ` [RFC PATCH net-next 28/33] tcp: Check for filled TCP option space before SACK Mat Martineau
2019-06-17 22:58 ` [RFC PATCH net-next 29/33] mptcp: accept: don't leak mptcp socket structure Mat Martineau
2019-06-17 22:58 ` [RFC PATCH net-next 30/33] mptcp: switch sublist to mptcp socket lock protection Mat Martineau
2019-06-17 22:58 ` [RFC PATCH net-next 31/33] mptcp: Add path manager interface Mat Martineau
2019-06-17 22:58 ` [RFC PATCH net-next 32/33] mptcp: Add ADD_ADDR handling Mat Martineau
2019-06-17 22:58 ` [RFC PATCH net-next 33/33] mptcp: Add handling of incoming MP_JOIN requests Mat Martineau
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20190617225808.665-22-mathew.j.martineau@linux.intel.com \
--to=mathew.j.martineau@linux.intel.com \
--cc=cpaasch@apple.com \
--cc=dcaratti@redhat.com \
--cc=edumazet@google.com \
--cc=fw@strlen.de \
--cc=matthieu.baerts@tessares.net \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=peter.krystad@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).