From: Menglong Dong <menglong8.dong@gmail.com>
To: Mat Martineau <mathew.j.martineau@linux.intel.com>
Cc: pabeni@redhat.com, mptcp@lists.linux.dev,
Menglong Dong <imagedong@tencent.com>
Subject: Re: [PATCH mptcp] net: mptcp: add statistics for mptcp socket in use
Date: Wed, 21 Sep 2022 11:25:44 +0800 [thread overview]
Message-ID: <CADxym3Ydc36U-Z2he0t-v8USoPEALWcO5d=TCL-hYsNvvrC1oQ@mail.gmail.com> (raw)
In-Reply-To: <44a260ec-6a0c-15f8-516b-6304ae9610d6@linux.intel.com>
On Wed, Sep 21, 2022 at 6:36 AM Mat Martineau
<mathew.j.martineau@linux.intel.com> wrote:
>
> On Tue, 20 Sep 2022, Menglong Dong wrote:
>
> > On Tue, Sep 20, 2022 at 6:37 AM Mat Martineau
> > <mathew.j.martineau@linux.intel.com> wrote:
> >>
> >> On Mon, 19 Sep 2022, menglong8.dong@gmail.com wrote:
> >>
> >>> From: Menglong Dong <imagedong@tencent.com>
> >>>
> >>> Do the statistics of mptcp socket in use with sock_prot_inuse_add().
> >>> Therefore, we can get the count of used mptcp socket from
> >>> /proc/net/protocols:
> >>>
> >>> & cat /proc/net/protocols
> >>> protocol size sockets memory press maxhdr slab module cl co di ac io in de sh ss gs se re sp bi br ha uh gp em
> >>> MPTCPv6 2048 0 0 no 0 yes kernel y n y y y y y y y y y y n n n y y y n
> >>> MPTCP 1896 1 0 no 0 yes kernel y n y y y y y y y y y y n n n y y y n
> >>>
> >>> Signed-off-by: Menglong Dong <imagedong@tencent.com>
> >>
> >> Hello Menglong -
> >>
> >> Thanks for your patch.
> >>
> >> One minor thing: please use the subject line tags listed in
> >> https://github.com/multipath-tcp/mptcp_net-next/wiki/Patch-prefixes :
> >> either "[PATCH mptcp-net]" or "[PATCH mptcp-next]", so it's clear which
> >> tree the patch is intended for.
> >>
> >
> > Thanks for your remind, I'll send the next version with
> > mptcp-next tag.
> >
> >>> ---
> >>> net/mptcp/protocol.c | 19 +++++++++++++------
> >>> net/mptcp/subflow.c | 3 +++
> >>> 2 files changed, 16 insertions(+), 6 deletions(-)
> >>>
> >>> diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
> >>> index 45ed50e9aec9..4da77aa8b070 100644
> >>> --- a/net/mptcp/protocol.c
> >>> +++ b/net/mptcp/protocol.c
> >>> @@ -2311,6 +2311,7 @@ static void __mptcp_close_ssk(struct sock *sk, struct sock *ssk,
> >>> tcp_set_state(ssk, TCP_CLOSE);
> >>> mptcp_subflow_queue_clean(ssk);
> >>> inet_csk_listen_stop(ssk);
> >>> + sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
> >>
> >> The code in this function is for a closing subflow, not a closing MPTCP
> >> socket. I don't think this call belongs here.
> >>
> >
> > I think this is the path of the listening mptcp close.
> > The mptcp socket in listening status is not hashed
> > to the token hash table, or has 'MPTCP_FALLBACK_DONE'
> > flags. Therefore, the use statistics of it is not freed in
> > mptcp_destroy_common().
> >
> > Hmm...maybe there is a better code path for
> > this part?
>
> Ah, right: the above block of code only runs once for the single listener
> subflow.
>
> mptcp_destroy_common() still seems like the better place to adjust the
> counter for the TCP_LISTEN case. Does it work to check the sk_state of
> __mptcp_nmpc_socket(msk) for TCP_LISTEN before the loop in
> mptcp_destroy_common() that calls __mptcp_close_ssk(), like this?
>
I'm afraid that this can't work, as msk->subflow will be disposed
in mptcp_destroy() before mptcp_destroy_common().
Does it work if we move mptcp_dispose_initial_subflow()
behind mptcp_destroy_common()?
Otherwise, how about we pass a MPTCP_CF_LISTEN flag
to mptcp_destroy_common(), like this:
ssock = __mptcp_nmpc_socket(msk);
if (ssock && inet_sk_state_load(ssock->sk) == TCP_LISTEN)
listener = true;
mptcp_dispose_initial_subflow(msk);
mptcp_destroy_common(msk, listener ? MPTCP_CF_LISTEN : 0);
Thanks!
Menglong Dong
> ssock = __mptcp_nmpc_socket(msk);
> if (ssock && inet_sk_state_load(ssock->sk) == TCP_LISTEN)
> listener = true;
>
> >
> >>> }
> >>> __tcp_close(ssk, 0);
> >>>
> >>> @@ -3067,6 +3068,9 @@ void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags)
> >>> skb_rbtree_purge(&msk->out_of_order_queue);
> >>> mptcp_data_unlock(sk);
> >>>
> >>> + if (!sk_unhashed(sk) || __mptcp_check_fallback(msk))
>
> then you could change this to
>
> if (!sk_unhashed(sk) || __mptcp_check_fallback(msk) || listener)
>
> ?
>
>
> Thanks,
>
> Mat
>
>
> >>> + sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
> >>> +
> >>> /* move all the rx fwd alloc into the sk_mem_reclaim_final in
> >>> * inet_sock_destruct() will dispose it
> >>> */
> >>> @@ -3513,6 +3517,7 @@ static int mptcp_stream_connect(struct socket *sock, struct sockaddr *uaddr,
> >>> mptcp_token_destroy(msk);
> >>> inet_sk_state_store(sock->sk, TCP_SYN_SENT);
> >>> subflow = mptcp_subflow_ctx(ssock->sk);
> >>> + sock_prot_inuse_add(sock_net(sock->sk), sock->sk->sk_prot, 1);
> >>> #ifdef CONFIG_TCP_MD5SIG
> >>> /* no MPTCP if MD5SIG is enabled on this socket or we may run out of
> >>> * TCP option space.
> >>> @@ -3547,12 +3552,13 @@ static int mptcp_stream_connect(struct socket *sock, struct sockaddr *uaddr,
> >>> static int mptcp_listen(struct socket *sock, int backlog)
> >>> {
> >>> struct mptcp_sock *msk = mptcp_sk(sock->sk);
> >>> + struct sock *sk = sock->sk;
> >>
> >> Changing all of the "sock->sk" text in this function to "sk" creates a lot
> >> of diffs that aren't related to maintaining the inuse statistics. If you'd
> >> like to do that refactoring change, please split that into a separate
> >> patch for mptcp-next.
> >>
> >
> > Ok, I'll split it into a separate patch.
> >
> > Thanks!
> > Menglong Dong
> >
> >> -Mat
> >>
> >>> struct socket *ssock;
> >>> int err;
> >>>
> >>> pr_debug("msk=%p", msk);
> >>>
> >>> - lock_sock(sock->sk);
> >>> + lock_sock(sk);
> >>> ssock = __mptcp_nmpc_socket(msk);
> >>> if (!ssock) {
> >>> err = -EINVAL;
> >>> @@ -3560,16 +3566,17 @@ static int mptcp_listen(struct socket *sock, int backlog)
> >>> }
> >>>
> >>> mptcp_token_destroy(msk);
> >>> - inet_sk_state_store(sock->sk, TCP_LISTEN);
> >>> - sock_set_flag(sock->sk, SOCK_RCU_FREE);
> >>> + inet_sk_state_store(sk, TCP_LISTEN);
> >>> + sock_set_flag(sk, SOCK_RCU_FREE);
> >>> + sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
> >>>
> >>> err = ssock->ops->listen(ssock, backlog);
> >>> - inet_sk_state_store(sock->sk, inet_sk_state_load(ssock->sk));
> >>> + inet_sk_state_store(sk, inet_sk_state_load(ssock->sk));
> >>> if (!err)
> >>> - mptcp_copy_inaddrs(sock->sk, ssock->sk);
> >>> + mptcp_copy_inaddrs(sk, ssock->sk);
> >>>
> >>> unlock:
> >>> - release_sock(sock->sk);
> >>> + release_sock(sk);
> >>> return err;
> >>> }
> >>>
> >>> diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
> >>> index 07dd23d0fe04..da6cfa73a3bd 100644
> >>> --- a/net/mptcp/subflow.c
> >>> +++ b/net/mptcp/subflow.c
> >>> @@ -747,6 +747,9 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
> >>> 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));
> >>> + sock_prot_inuse_add(sock_net(new_msk),
> >>> + new_msk->sk_prot,
> >>> + 1);
> >>> ctx->conn = new_msk;
> >>> new_msk = NULL;
> >>>
> >>> --
> >>> 2.37.2
> >>>
> >>>
> >>>
> >>
> >> --
> >> Mat Martineau
> >> Intel
> >
>
> --
> Mat Martineau
> Intel
next prev parent reply other threads:[~2022-09-21 3:25 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-19 13:21 [PATCH mptcp] net: mptcp: add statistics for mptcp socket in use menglong8.dong
2022-09-19 15:14 ` net: mptcp: add statistics for mptcp socket in use: Tests Results MPTCP CI
2022-09-19 22:37 ` [PATCH mptcp] net: mptcp: add statistics for mptcp socket in use Mat Martineau
2022-09-20 2:10 ` Menglong Dong
2022-09-20 11:44 ` Matthieu Baerts
2022-09-20 13:47 ` Menglong Dong
2022-09-20 22:36 ` Mat Martineau
2022-09-21 3:25 ` Menglong Dong [this message]
2022-09-20 10:52 ` net: mptcp: add statistics for mptcp socket in use: Tests Results MPTCP CI
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='CADxym3Ydc36U-Z2he0t-v8USoPEALWcO5d=TCL-hYsNvvrC1oQ@mail.gmail.com' \
--to=menglong8.dong@gmail.com \
--cc=imagedong@tencent.com \
--cc=mathew.j.martineau@linux.intel.com \
--cc=mptcp@lists.linux.dev \
--cc=pabeni@redhat.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).