Series comparison

-[PATCH net-next 0/7] mptcp: rx path refactor
+[PATCH mptcp-next v2 0/7] mptcp: rx path refactor
-Paolo worked on this RX path refactor for these two main reasons:
+This is a batch of changes I had sitting in my local tree for a while.
 Why another refactor you may ask? Two main resons:
-- Currently, the MPTCP RX path introduces quite a bit of 'exceptional'
+- currently the mptcp RX path introduces quite a bit of 'exceptional'
-  accounting/locking processing WRT to plain TCP, adding up to the
+ accounting/locking processing WRT to plain TCP, adding up to the
-  implementation complexity in a miserable way.
+ implementation complexity in a misurable way
+- the performance gap WRT plain TCP for single subflow connections is
-- The performance gap WRT plain TCP for single subflow connections is
+ quite measurable.
   quite measurable.
 The present refactor addresses both the above items: most of the
 additional complexity is dropped, and single stream performances
-increase measurably, from 55Gbps to 71Gbps in Paolo's loopback test. As
+increase measurably - from 55Gbps to 71Gbps in my loopback test. As a
-a reference, plain TCP was around 84Gbps on the same host.
+reference, plain TCP is around 84Gps on the same host.
-The above comes to a price: the patch are invasive, even in subtle ways.
+The above comes to a price: the patch are invasive, even in subtle ways:
 the chance of destabilizing the implementation is real (ence the
 additional, intentional '-next' into the subj).
-Note: patch 5/7 removes the sk_forward_alloc_get() helper, which caused
+In any case keeping the patch hidden for longer was not going to do any
-some trivial modifications in different places in the net tree: sockets,
+good, so here we are.
 IPv4, sched. That's why a few more people have been Cc here. Feel free
 to only look at this patch 5/7.
-Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
+Changes from v1:
----
+  - fixed several data stream corruption and wake-up misses due
     to multi subflows races
   - added patches 1-3 mainly to address the above
   - added an additional follow-up patch (patch 7) with more cleanup
 Paolo Abeni (7):
-      mptcp: consolidate subflow cleanup
+  mptcp: prevent excessive coalescing on receive
-      mptcp: drop __mptcp_fastopen_gen_msk_ackseq()
+  tcp: fix recvbuffer adjust on sleeping rcvmsg
-      mptcp: move the whole rx path under msk socket lock protection
+  mptcp: don't always assume copied data in mptcp_cleanup_rbuf()
-      mptcp: cleanup mem accounting
+  mptcp: consolidate subflow cleanup
-      net: dismiss sk_forward_alloc_get()
+  mptcp: move the whole rx path under msk socket lock protection
-      mptcp: dismiss __mptcp_rmem()
+  mptcp: cleanup mem accounting.
-      mptcp: micro-optimize __mptcp_move_skb()
+  net: dismiss sk_forward_alloc_get()
  include/net/sock.h   |  13 ---
  net/core/sock.c      |   2 +-
  net/ipv4/af_inet.c   |   2 +-
  net/ipv4/inet_diag.c |   2 +-
- net/mptcp/fastopen.c |  27 +----
+ net/mptcp/fastopen.c |   4 +-
- net/mptcp/protocol.c | 317 ++++++++++++++++-----------------------------------
+ net/mptcp/protocol.c | 259 +++++++++++++------------------------------
- net/mptcp/protocol.h |  22 ++--
+ net/mptcp/protocol.h |   6 +-
- net/mptcp/subflow.c  |  36 +++---
+ net/mptcp/subflow.c  |  33 +++---
  net/sched/em_meta.c  |   2 +-
-files changed, 134 insertions(+), 289 deletions(-)
+files changed, 104 insertions(+), 219 deletions(-)
 ---
 base-commit: b4cb730862cf4f59ac3dcb83b9ac4eeb29dbfb0e
 change-id: 20250106-net-next-mptcp-rx-path-refactor-f44579efb57c
-Best regards,
 --
-Matthieu Baerts (NGI0) <matttbe@kernel.org>
+.45.2

-[PATCH net-next 2/7] mptcp: drop __mptcp_fastopen_gen_msk_ackseq()
+[PATCH mptcp-next v2 1/7] mptcp: prevent excessive coalescing on receive
-From: Paolo Abeni <pabeni@redhat.com>
+Currently the skb size after coalescing is only limited by the skb
+layout (the skb must not carry frag_list). A single coalesced skb
-When we will move the whole RX path under the msk socket lock, updating
+covering several MSS can potentially fill completely the receive
-the already queued skb for passive fastopen socket at 3rd ack time will
+buffer. In such a case, the snd win will zero until the receive buffer
-be extremely painful and race prone
+will be empty again, affecting tput badly.
 The map_seq for already enqueued skbs is used only to allow correct
 coalescing with later data; preventing collapsing to the first skb of
 a fastopen connect we can completely remove the
 __mptcp_fastopen_gen_msk_ackseq() helper.
 Before dropping this helper, a new item had to be added to the
 mptcp_skb_cb structure. Because this item will be frequently tested in
 the fast path -- almost on every packet -- and because there is free
 space there, a single byte is used instead of a bitfield. This micro
 optimisation slightly reduces the number of CPU operations to do the
 associated check.
 Signed-off-by: Paolo Abeni <pabeni@redhat.com>
-Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
-Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
 ---
- net/mptcp/fastopen.c | 24 ++----------------------
+No fixes tag because the problem is not very visible in practice
- net/mptcp/protocol.c |  4 +++-
+currently, but will be apparent after the rx refactor.
- net/mptcp/protocol.h |  5 ++---
+Still I hope this could affect positively the simlut flows self-tests
- net/mptcp/subflow.c  |  3 ---
+---
-files changed, 7 insertions(+), 29 deletions(-)
+ net/mptcp/protocol.c | 1 +
 file changed, 1 insertion(+)
-diff --git a/net/mptcp/fastopen.c b/net/mptcp/fastopen.c
-index XXXXXXX..XXXXXXX 100644
---- a/net/mptcp/fastopen.c
-+++ b/net/mptcp/fastopen.c
-@@ -XXX,XX +XXX,XX @@ void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subf
-     tp->copied_seq += skb->len;
-     subflow->ssn_offset += skb->len;
--    /* initialize a dummy sequence number, we will update it at MPC
--     * completion, if needed
--     */
-+    /* Only the sequence delta is relevant */
-     MPTCP_SKB_CB(skb)->map_seq = -skb->len;
-     MPTCP_SKB_CB(skb)->end_seq = 0;
-     MPTCP_SKB_CB(skb)->offset = 0;
-     MPTCP_SKB_CB(skb)->has_rxtstamp = TCP_SKB_CB(skb)->has_rxtstamp;
-+    MPTCP_SKB_CB(skb)->cant_coalesce = 1;
-     mptcp_data_lock(sk);
-@@ -XXX,XX +XXX,XX @@ void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subf
-     mptcp_data_unlock(sk);
- }
--
--void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
--                     const struct mptcp_options_received *mp_opt)
--{
--    struct sock *sk = (struct sock *)msk;
--    struct sk_buff *skb;
--
--    skb = skb_peek_tail(&sk->sk_receive_queue);
--    if (skb) {
--        WARN_ON_ONCE(MPTCP_SKB_CB(skb)->end_seq);
--        pr_debug("msk %p moving seq %llx -> %llx end_seq %llx -> %llx\n", sk,
--             MPTCP_SKB_CB(skb)->map_seq, MPTCP_SKB_CB(skb)->map_seq + msk->ack_seq,
--             MPTCP_SKB_CB(skb)->end_seq, MPTCP_SKB_CB(skb)->end_seq + msk->ack_seq);
--        MPTCP_SKB_CB(skb)->map_seq += msk->ack_seq;
--        MPTCP_SKB_CB(skb)->end_seq += msk->ack_seq;
--    }
--
--    pr_debug("msk=%p ack_seq=%llx\n", msk, msk->ack_seq);
--}
 diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
 index XXXXXXX..XXXXXXX 100644
 --- a/net/mptcp/protocol.c
 +++ b/net/mptcp/protocol.c
 @@ -XXX,XX +XXX,XX @@ static bool mptcp_try_coalesce(struct sock *sk, struct sk_buff *to,
-     bool fragstolen;
      int delta;
--    if (MPTCP_SKB_CB(from)->offset ||
+     if (MPTCP_SKB_CB(from)->offset ||
-+    if (unlikely(MPTCP_SKB_CB(to)->cant_coalesce) ||
++        ((to->len + from->len) > (sk->sk_rcvbuf >> 3)) ||
 +        MPTCP_SKB_CB(from)->offset ||
          ((to->len + from->len) > (sk->sk_rcvbuf >> 3)) ||
          !skb_try_coalesce(to, from, &fragstolen, &delta))
          return false;
-@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skb(struct mptcp_sock *msk, struct sock *ssk,
      MPTCP_SKB_CB(skb)->end_seq = MPTCP_SKB_CB(skb)->map_seq + copy_len;
      MPTCP_SKB_CB(skb)->offset = offset;
      MPTCP_SKB_CB(skb)->has_rxtstamp = has_rxtstamp;
 +    MPTCP_SKB_CB(skb)->cant_coalesce = 0;
      if (MPTCP_SKB_CB(skb)->map_seq == msk->ack_seq) {
          /* in sequence */
 diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
 index XXXXXXX..XXXXXXX 100644
 --- a/net/mptcp/protocol.h
 +++ b/net/mptcp/protocol.h
@@ -XXX,XX +XXX,XX @@ struct mptcp_skb_cb {
      u64 map_seq;
      u64 end_seq;
      u32 offset;
 -    u8  has_rxtstamp:1;
 +    u8  has_rxtstamp;
 +    u8  cant_coalesce;
  };
  #define MPTCP_SKB_CB(__skb)    ((struct mptcp_skb_cb *)&((__skb)->cb[0]))
@@ -XXX,XX +XXX,XX @@ void mptcp_event_pm_listener(const struct sock *ssk,
                   enum mptcp_event_type event);
  bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
 -void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
 -                     const struct mptcp_options_received *mp_opt);
  void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
                            struct request_sock *req);
  int mptcp_nl_fill_addr(struct sk_buff *skb,
 diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
 index XXXXXXX..XXXXXXX 100644
 --- a/net/mptcp/subflow.c
 +++ b/net/mptcp/subflow.c
@@ -XXX,XX +XXX,XX @@ void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
      subflow_set_remote_key(msk, subflow, mp_opt);
      WRITE_ONCE(subflow->fully_established, true);
      WRITE_ONCE(msk->fully_established, true);
 -
 -    if (subflow->is_mptfo)
 -        __mptcp_fastopen_gen_msk_ackseq(msk, subflow, mp_opt);
  }
  static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 --
-.47.1
+.45.2

-[PATCH net-next 7/7] mptcp: micro-optimize __mptcp_move_skb()
+[PATCH mptcp-next v2 2/7] tcp: fix recvbuffer adjust on sleeping rcvmsg
-From: Paolo Abeni <pabeni@redhat.com>
+If the recvmsg() blocks after receiving some data - i.e. due to
 SO_RCVLOWAT - the MPTCP code will attempt multiple times to
 adjust the receive buffer size, wrongly accounting every time the
 cumulative of received data - instead of accounting only for the
 delta.
-After the RX path refactor the mentioned function is expected to run
+Address the issue moving mptcp_rcv_space_adjust just after the
-frequently, let's optimize it a bit.
+data reception and passing it only the just received bytes.
-Scan for ready subflow from the last processed one, and stop after
+This also remove an unneeded difference between the TCP and MPTCP
-traversing the list once or reaching the msk memory limit - instead of
+RX code path implementation.
 looking for dubious per-subflow conditions.
 Also re-order the memory limit checks, to avoid duplicate tests.
+Fixes: 581302298524 ("mptcp: error out earlier on disconnect")
 Signed-off-by: Paolo Abeni <pabeni@redhat.com>
-Reviewed-by: Mat Martineau <martineau@kernel.org>
-Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
 ---
- net/mptcp/protocol.c | 111 +++++++++++++++++++++++----------------------------
+Not strictly related to the refactor, found while investigating
- net/mptcp/protocol.h |   2 +
+the CI failure
-files changed, 52 insertions(+), 61 deletions(-)
+---
  net/mptcp/protocol.c | 6 +++---
 file changed, 3 insertions(+), 3 deletions(-)
 diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
 index XXXXXXX..XXXXXXX 100644
 --- a/net/mptcp/protocol.c
 +++ b/net/mptcp/protocol.c
-@@ -XXX,XX +XXX,XX @@ static void mptcp_dss_corruption(struct mptcp_sock *msk, struct sock *ssk)
+@@ -XXX,XX +XXX,XX @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
      goto out;
  }
- static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
++static void mptcp_rcv_space_adjust(struct mptcp_sock *msk, int copied);
 -                       struct sock *ssk,
 -                       unsigned int *bytes)
 +                       struct sock *ssk)
  {
      struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
      struct sock *sk = (struct sock *)msk;
 -    unsigned int moved = 0;
      bool more_data_avail;
      struct tcp_sock *tp;
 -    bool done = false;
 +    bool ret = false;
      pr_debug("msk=%p ssk=%p\n", msk, ssk);
      tp = tcp_sk(ssk);
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
          struct sk_buff *skb;
          bool fin;
 +        if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
 +            break;
 +
-         /* try to move as much data as available */
+ static int __mptcp_recvmsg_mskq(struct mptcp_sock *msk,
-         map_remaining = subflow->map_data_len -
+                 struct msghdr *msg,
-                 mptcp_subflow_get_map_offset(subflow);
+                 size_t len, int flags,
+@@ -XXX,XX +XXX,XX @@ static int __mptcp_recvmsg_mskq(struct mptcp_sock *msk,
          skb = skb_peek(&ssk->sk_receive_queue);
 -        if (!skb) {
 -            /* With racing move_skbs_to_msk() and __mptcp_move_skbs(),
 -             * a different CPU can have already processed the pending
 -             * data, stop here or we can enter an infinite loop
 -             */
 -            if (!moved)
 -                done = true;
 +        if (unlikely(!skb))
              break;
--        }
+     }
-         if (__mptcp_check_fallback(msk)) {
++    mptcp_rcv_space_adjust(msk, copied);
-             /* Under fallback skbs have no MPTCP extension and TCP could
+     return copied;
-@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
+ }
-         offset = seq - TCP_SKB_CB(skb)->seq;
+@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
-         fin = TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN;
+         }
--        if (fin) {
--            done = true;
+         pr_debug("block timeout %ld\n", timeo);
-+        if (fin)
+-        mptcp_rcv_space_adjust(msk, copied);
-             seq++;
+         err = sk_wait_data(sk, &timeo, NULL);
--        }
+         if (err < 0) {
+             err = copied ? : err;
-         if (offset < skb->len) {
+@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
-             size_t len = skb->len - offset;
+         }
+     }
--            if (tp->urg_data)
--                done = true;
+-    mptcp_rcv_space_adjust(msk, copied);
 -
--            if (__mptcp_move_skb(msk, ssk, skb, offset, len))
+ out_err:
--                moved += len;
+     if (cmsg_flags && copied >= 0) {
-+            ret = __mptcp_move_skb(msk, ssk, skb, offset, len) || ret;
+         if (cmsg_flags & MPTCP_CMSG_TS)
              seq += len;
              if (unlikely(map_remaining < len)) {
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
              }
              sk_eat_skb(ssk, skb);
 -            done = true;
          }
          WRITE_ONCE(tp->copied_seq, seq);
          more_data_avail = mptcp_subflow_data_available(ssk);
 -        if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf) {
 -            done = true;
 -            break;
 -        }
      } while (more_data_avail);
 -    if (moved > 0)
 +    if (ret)
          msk->last_data_recv = tcp_jiffies32;
 -    *bytes += moved;
 -    return done;
 +    return ret;
  }
  static bool __mptcp_ofo_queue(struct mptcp_sock *msk)
@@ -XXX,XX +XXX,XX @@ void __mptcp_error_report(struct sock *sk)
  static bool move_skbs_to_msk(struct mptcp_sock *msk, struct sock *ssk)
  {
      struct sock *sk = (struct sock *)msk;
 -    unsigned int moved = 0;
 +    bool moved;
 -    __mptcp_move_skbs_from_subflow(msk, ssk, &moved);
 +    moved = __mptcp_move_skbs_from_subflow(msk, ssk);
      __mptcp_ofo_queue(msk);
      if (unlikely(ssk->sk_err)) {
          if (!sock_owned_by_user(sk))
@@ -XXX,XX +XXX,XX @@ static bool move_skbs_to_msk(struct mptcp_sock *msk, struct sock *ssk)
       */
      if (mptcp_pending_data_fin(sk, NULL))
          mptcp_schedule_work(sk);
 -    return moved > 0;
 +    return moved;
  }
  static void __mptcp_rcvbuf_update(struct sock *sk, struct sock *ssk)
@@ -XXX,XX +XXX,XX @@ static void __mptcp_data_ready(struct sock *sk, struct sock *ssk)
      __mptcp_rcvbuf_update(sk, ssk);
 -    /* over limit? can't append more skbs to msk, Also, no need to wake-up*/
 -    if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
 -        return;
 -
      /* Wake-up the reader only for in-sequence data */
      if (move_skbs_to_msk(msk, ssk) && mptcp_epollin_ready(sk))
          sk->sk_data_ready(sk);
@@ -XXX,XX +XXX,XX @@ bool mptcp_schedule_work(struct sock *sk)
      return false;
  }
 -static struct sock *mptcp_subflow_recv_lookup(const struct mptcp_sock *msk)
 -{
 -    struct mptcp_subflow_context *subflow;
 -
 -    msk_owned_by_me(msk);
 -
 -    mptcp_for_each_subflow(msk, subflow) {
 -        if (READ_ONCE(subflow->data_avail))
 -            return mptcp_subflow_tcp_sock(subflow);
 -    }
 -
 -    return NULL;
 -}
 -
  static bool mptcp_skb_can_collapse_to(u64 write_seq,
                        const struct sk_buff *skb,
                        const struct mptcp_ext *mpext)
@@ -XXX,XX +XXX,XX @@ static void mptcp_rcv_space_adjust(struct mptcp_sock *msk, int copied)
      msk->rcvq_space.time = mstamp;
  }
 +static struct mptcp_subflow_context *
 +__mptcp_first_ready_from(struct mptcp_sock *msk,
 +             struct mptcp_subflow_context *subflow)
 +{
 +    struct mptcp_subflow_context *start_subflow = subflow;
 +
 +    while (!READ_ONCE(subflow->data_avail)) {
 +        subflow = mptcp_next_subflow(msk, subflow);
 +        if (subflow == start_subflow)
 +            return NULL;
 +    }
 +    return subflow;
 +}
 +
  static bool __mptcp_move_skbs(struct sock *sk)
  {
      struct mptcp_subflow_context *subflow;
      struct mptcp_sock *msk = mptcp_sk(sk);
 -    unsigned int moved = 0;
 -    bool ret, done;
 +    bool ret = false;
 +
 +    if (list_empty(&msk->conn_list))
 +        return false;
      /* verify we can move any data from the subflow, eventually updating */
      if (!(sk->sk_userlocks & SOCK_RCVBUF_LOCK))
          mptcp_for_each_subflow(msk, subflow)
              __mptcp_rcvbuf_update(sk, subflow->tcp_sock);
 -    if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
 -        return false;
 -
 -    do {
 -        struct sock *ssk = mptcp_subflow_recv_lookup(msk);
 +    subflow = list_first_entry(&msk->conn_list,
 +                   struct mptcp_subflow_context, node);
 +    for (;;) {
 +        struct sock *ssk;
          bool slowpath;
 -        if (unlikely(!ssk))
 +        /*
 +         * As an optimization avoid traversing the subflows list
 +         * and ev. acquiring the subflow socket lock before baling out
 +         */
 +        if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
              break;
 -        slowpath = lock_sock_fast(ssk);
 -        done = __mptcp_move_skbs_from_subflow(msk, ssk, &moved);
 +        subflow = __mptcp_first_ready_from(msk, subflow);
 +        if (!subflow)
 +            break;
 +        ssk = mptcp_subflow_tcp_sock(subflow);
 +        slowpath = lock_sock_fast(ssk);
 +        ret = __mptcp_move_skbs_from_subflow(msk, ssk) || ret;
          if (unlikely(ssk->sk_err))
              __mptcp_error_report(sk);
          unlock_sock_fast(ssk, slowpath);
 -    } while (!done);
 -    ret = moved > 0 || __mptcp_ofo_queue(msk);
 +        subflow = mptcp_next_subflow(msk, subflow);
 +    }
 +
 +    __mptcp_ofo_queue(msk);
      if (ret)
          mptcp_check_data_fin((struct sock *)msk);
      return ret;
 diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
 index XXXXXXX..XXXXXXX 100644
 --- a/net/mptcp/protocol.h
 +++ b/net/mptcp/protocol.h
@@ -XXX,XX +XXX,XX @@ struct mptcp_sock {
      list_for_each_entry(__subflow, &((__msk)->conn_list), node)
  #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)            \
      list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
 +#define mptcp_next_subflow(__msk, __subflow)                \
 +    list_next_entry_circular(__subflow, &((__msk)->conn_list), node)
  extern struct genl_family mptcp_genl_family;
 --
-.47.1
+.45.2

-[PATCH net-next 6/7] mptcp: dismiss __mptcp_rmem()
+[PATCH mptcp-next v2 3/7] mptcp: don't always assume copied data in mptcp_cleanup_rbuf()
-From: Paolo Abeni <pabeni@redhat.com>
+Under some corner cases the MPTCP protocol can end-up invoking
 mptcp_cleanup_rbuf() when no data has been copied, but such helper
 assumes the opposite condition.
-After the RX path refactor, it become a wrapper for sk_rmem_alloc
+Explicitly drop such assumption and performs the costly call only
-access, with a slightly misleading name. Just drop it.
+when strictly needed - before releasing the msk socket lock.
+Fixes: fd8976790a6c ("mptcp: be careful on MPTCP-level ack.")
 Signed-off-by: Paolo Abeni <pabeni@redhat.com>
-Reviewed-by: Mat Martineau <martineau@kernel.org>
-Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
 ---
- net/mptcp/protocol.c |  8 ++++----
+ net/mptcp/protocol.c | 18 +++++++++---------
- net/mptcp/protocol.h | 11 ++---------
+file changed, 9 insertions(+), 9 deletions(-)
 files changed, 6 insertions(+), 13 deletions(-)
 diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
 index XXXXXXX..XXXXXXX 100644
 --- a/net/mptcp/protocol.c
 +++ b/net/mptcp/protocol.c
-@@ -XXX,XX +XXX,XX @@ static void mptcp_cleanup_rbuf(struct mptcp_sock *msk, int copied)
+@@ -XXX,XX +XXX,XX @@ static void mptcp_send_ack(struct mptcp_sock *msk)
          mptcp_subflow_send_ack(mptcp_subflow_tcp_sock(subflow));
  }
 -static void mptcp_subflow_cleanup_rbuf(struct sock *ssk)
 +static void mptcp_subflow_cleanup_rbuf(struct sock *ssk, int copied)
  {
      bool slow;
      slow = lock_sock_fast(ssk);
      if (tcp_can_send_ack(ssk))
 -        tcp_cleanup_rbuf(ssk, 1);
 +        tcp_cleanup_rbuf(ssk, copied);
      unlock_sock_fast(ssk, slow);
  }
@@ -XXX,XX +XXX,XX @@ static bool mptcp_subflow_could_cleanup(const struct sock *ssk, bool rx_empty)
                    (ICSK_ACK_PUSHED2 | ICSK_ACK_PUSHED)));
  }
 -static void mptcp_cleanup_rbuf(struct mptcp_sock *msk)
 +static void mptcp_cleanup_rbuf(struct mptcp_sock *msk, int copied)
  {
      int old_space = READ_ONCE(msk->old_wspace);
      struct mptcp_subflow_context *subflow;
@@ -XXX,XX +XXX,XX @@ static void mptcp_cleanup_rbuf(struct mptcp_sock *msk)
      int space =  __mptcp_space(sk);
      bool cleanup, rx_empty;
-     cleanup = (space > 0) && (space >= (old_space << 1)) && copied;
+-    cleanup = (space > 0) && (space >= (old_space << 1));
--    rx_empty = !__mptcp_rmem(sk) && copied;
+-    rx_empty = !__mptcp_rmem(sk);
-+    rx_empty = !sk_rmem_alloc_get(sk) && copied;
++    cleanup = (space > 0) && (space >= (old_space << 1)) && copied;
 +    rx_empty = !__mptcp_rmem(sk) && copied;
      mptcp_for_each_subflow(msk, subflow) {
          struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
-@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
-         WRITE_ONCE(tp->copied_seq, seq);
+         if (cleanup || mptcp_subflow_could_cleanup(ssk, rx_empty))
-         more_data_avail = mptcp_subflow_data_available(ssk);
+-            mptcp_subflow_cleanup_rbuf(ssk);
++            mptcp_subflow_cleanup_rbuf(ssk, copied);
--        if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) {
+     }
-+        if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf) {
+ }
-             done = true;
-             break;
+@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
          copied += bytes_read;
 -        /* be sure to advertise window change */
 -        mptcp_cleanup_rbuf(msk);
 -
          if (skb_queue_empty(&msk->receive_queue) && __mptcp_move_skbs(msk))
              continue;
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
          }
-@@ -XXX,XX +XXX,XX @@ static void __mptcp_data_ready(struct sock *sk, struct sock *ssk)
-     __mptcp_rcvbuf_update(sk, ssk);
+         pr_debug("block timeout %ld\n", timeo);
++        mptcp_cleanup_rbuf(msk, copied);
-     /* over limit? can't append more skbs to msk, Also, no need to wake-up*/
+         err = sk_wait_data(sk, &timeo, NULL);
--    if (__mptcp_rmem(sk) > sk->sk_rcvbuf)
+         if (err < 0) {
-+    if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
+             err = copied ? : err;
-         return;
+@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
+         }
-     /* Wake-up the reader only for in-sequence data */
+     }
-@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs(struct sock *sk)
-         mptcp_for_each_subflow(msk, subflow)
++    mptcp_cleanup_rbuf(msk, copied);
-             __mptcp_rcvbuf_update(sk, subflow->tcp_sock);
++
+ out_err:
--    if (__mptcp_rmem(sk) > sk->sk_rcvbuf)
+     if (cmsg_flags && copied >= 0) {
-+    if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
+         if (cmsg_flags & MPTCP_CMSG_TS)
          return false;
      do {
 diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
 index XXXXXXX..XXXXXXX 100644
 --- a/net/mptcp/protocol.h
 +++ b/net/mptcp/protocol.h
@@ -XXX,XX +XXX,XX @@ static inline void msk_owned_by_me(const struct mptcp_sock *msk)
  #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
  #endif
 -/* the msk socket don't use the backlog, also account for the bulk
 - * free memory
 - */
 -static inline int __mptcp_rmem(const struct sock *sk)
 -{
 -    return atomic_read(&sk->sk_rmem_alloc);
 -}
 -
  static inline int mptcp_win_from_space(const struct sock *sk, int space)
  {
      return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
@@ -XXX,XX +XXX,XX @@ static inline int mptcp_space_from_win(const struct sock *sk, int win)
  static inline int __mptcp_space(const struct sock *sk)
  {
 -    return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
 +    return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) -
 +                    sk_rmem_alloc_get(sk));
  }
  static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
 --
-.47.1
+.45.2

-[PATCH net-next 1/7] mptcp: consolidate subflow cleanup
+[PATCH mptcp-next v2 4/7] mptcp: consolidate subflow cleanup
-From: Paolo Abeni <pabeni@redhat.com>
 Consolidate all the cleanup actions requiring the worker in a single
 helper and ensure the dummy data fin creation for fallback socket is
 performed only when the tcp rx queue is empty.
 There are no functional changes intended, but this will simplify the
 next patch, when the tcp rx queue spooling could be delayed at release_cb
 time.
 Signed-off-by: Paolo Abeni <pabeni@redhat.com>
-Reviewed-by: Mat Martineau <martineau@kernel.org>
-Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
 ---
  net/mptcp/subflow.c | 33 ++++++++++++++++++---------------
 file changed, 18 insertions(+), 15 deletions(-)
 diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
 ...
 -        mptcp_update_rcv_data_fin(msk, READ_ONCE(msk->ack_seq), true))
 -        mptcp_schedule_work(parent);
  }
  void mptcp_subflow_queue_clean(struct sock *listener_sk, struct sock *listener_ssk)
 --
-.47.1
+.45.2

-[PATCH net-next 3/7] mptcp: move the whole rx path under msk socket lock protection
+[PATCH mptcp-next v2 5/7] mptcp: move the whole rx path under msk socket lock protection
-From: Paolo Abeni <pabeni@redhat.com>
 After commit c2e6048fa1cf ("mptcp: fix race in release_cb") we can
 move the whole MPTCP rx path under the socket lock leveraging the
 release_cb.
 We can drop a bunch of spin_lock pairs in the receive functions, use
 ...
 When the skbs move is delayed at msk release callback time, even the
 msk rcvbuf update is delayed; additionally take care of such action in
 __mptcp_move_skbs().
 Signed-off-by: Paolo Abeni <pabeni@redhat.com>
-Reviewed-by: Mat Martineau <martineau@kernel.org>
-Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
 ---
- net/mptcp/fastopen.c |   1 +
+v1 -> v2:
- net/mptcp/protocol.c | 123 ++++++++++++++++++++++++---------------------------
+ - cleanup && fixup msk sk_rcvbuf update
  - fix missed wakeup due to bad placed __mptcp_move_skbs(); move it
    form __mptcp_recvmsg_mskq into mptcp_recvmsg()
  - added missing '\n' in debug message format string
  - keep 'snd_una' && friends update under the mptcp data lock
 ---
  net/mptcp/fastopen.c |   2 +
  net/mptcp/protocol.c | 123 ++++++++++++++++++++-----------------------
  net/mptcp/protocol.h |   2 +-
-files changed, 60 insertions(+), 66 deletions(-)
+files changed, 61 insertions(+), 66 deletions(-)
 diff --git a/net/mptcp/fastopen.c b/net/mptcp/fastopen.c
 index XXXXXXX..XXXXXXX 100644
 --- a/net/mptcp/fastopen.c
 +++ b/net/mptcp/fastopen.c
 @@ -XXX,XX +XXX,XX @@ void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subf
-     MPTCP_SKB_CB(skb)->cant_coalesce = 1;
+     MPTCP_SKB_CB(skb)->has_rxtstamp = TCP_SKB_CB(skb)->has_rxtstamp;
      mptcp_data_lock(sk);
 +    DEBUG_NET_WARN_ON_ONCE(sock_owned_by_user_nocheck(sk));
      mptcp_set_owner_r(skb, sk);
      __skb_queue_tail(&sk->sk_receive_queue, skb);
+@@ -XXX,XX +XXX,XX @@ void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflo
+     struct sock *sk = (struct sock *)msk;
+     struct sk_buff *skb;
++    DEBUG_NET_WARN_ON_ONCE(sock_owned_by_user_nocheck(sk));
+     skb = skb_peek_tail(&sk->sk_receive_queue);
+     if (skb) {
+         WARN_ON_ONCE(MPTCP_SKB_CB(skb)->end_seq);
 diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
 index XXXXXXX..XXXXXXX 100644
 --- a/net/mptcp/protocol.c
 +++ b/net/mptcp/protocol.c
 @@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 ...
      struct rb_root  out_of_order_queue;
 -    struct sk_buff_head receive_queue;
      struct list_head conn_list;
      struct list_head rtx_queue;
      struct mptcp_data_frag *first_pending;
 --
-.47.1
+.45.2

-[PATCH net-next 4/7] mptcp: cleanup mem accounting
+[PATCH mptcp-next v2 6/7] mptcp: cleanup mem accounting.
-From: Paolo Abeni <pabeni@redhat.com>
 After the previous patch, updating sk_forward_memory is cheap and
-we can drop a lot of complexity from the MPTCP memory accounting,
+we can drop a lot of complexity from the MPTCP memory acconting,
 removing the custom fwd mem allocations for rmem.
 Signed-off-by: Paolo Abeni <pabeni@redhat.com>
-Reviewed-by: Mat Martineau <martineau@kernel.org>
+---
-Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
+v1 -> v2:
  - keep 'snd_una' and recovery-related fields under the msk
    data lock
  - dropped unneeded code in __mptcp_move_skbs()
 ---
  net/mptcp/fastopen.c |   2 +-
- net/mptcp/protocol.c | 115 ++++-----------------------------------------------
+ net/mptcp/protocol.c | 115 +++----------------------------------------
  net/mptcp/protocol.h |   4 +-
 files changed, 10 insertions(+), 111 deletions(-)
 diff --git a/net/mptcp/fastopen.c b/net/mptcp/fastopen.c
 index XXXXXXX..XXXXXXX 100644
 ...
 -    return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
 +    return atomic_read(&sk->sk_rmem_alloc);
  }
  static inline int mptcp_win_from_space(const struct sock *sk, int space)
 --
-.47.1
+.45.2

-[PATCH net-next 5/7] net: dismiss sk_forward_alloc_get()
+[PATCH mptcp-next v2 7/7] net: dismiss sk_forward_alloc_get()
-From: Paolo Abeni <pabeni@redhat.com>
 After the previous patch we can remove the forward_alloc_get
 proto callback, basically reverting commit 292e6077b040 ("net: introduce
 sk_forward_alloc_get()") and commit 66d58f046c9d ("net: use
 sk_forward_alloc_get() in sk_get_meminfo()").
 Signed-off-by: Paolo Abeni <pabeni@redhat.com>
-Acked-by: Mat Martineau <martineau@kernel.org>
-Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
 ---
  include/net/sock.h   | 13 -------------
  net/core/sock.c      |  2 +-
  net/ipv4/af_inet.c   |  2 +-
  net/ipv4/inet_diag.c |  2 +-
 ...
 -    dst->value = sk_forward_alloc_get(sk);
 +    dst->value = READ_ONCE(sk->sk_forward_alloc);
  }
  META_COLLECTOR(int_sk_sndbuf)
 --
-.47.1
+.45.2

Paolo worked on this RX path refactor for these two main reasons:

- Currently, the MPTCP RX path introduces quite a bit of 'exceptional'
  accounting/locking processing WRT to plain TCP, adding up to the
  implementation complexity in a miserable way.

- The performance gap WRT plain TCP for single subflow connections is
  quite measurable.

The present refactor addresses both the above items: most of the
additional complexity is dropped, and single stream performances
increase measurably, from 55Gbps to 71Gbps in Paolo's loopback test. As
a reference, plain TCP was around 84Gbps on the same host.

The above comes to a price: the patch are invasive, even in subtle ways.

Note: patch 5/7 removes the sk_forward_alloc_get() helper, which caused
some trivial modifications in different places in the net tree: sockets,
IPv4, sched. That's why a few more people have been Cc here. Feel free
to only look at this patch 5/7.

Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
Paolo Abeni (7):
      mptcp: consolidate subflow cleanup
      mptcp: drop __mptcp_fastopen_gen_msk_ackseq()
      mptcp: move the whole rx path under msk socket lock protection
      mptcp: cleanup mem accounting
      net: dismiss sk_forward_alloc_get()
      mptcp: dismiss __mptcp_rmem()
      mptcp: micro-optimize __mptcp_move_skb()

include/net/sock.h   |  13 ---
 net/core/sock.c      |   2 +-
 net/ipv4/af_inet.c   |   2 +-
 net/ipv4/inet_diag.c |   2 +-
 net/mptcp/fastopen.c |  27 +----
 net/mptcp/protocol.c | 317 ++++++++++++++++-----------------------------------
 net/mptcp/protocol.h |  22 ++--
 net/mptcp/subflow.c  |  36 +++---
 net/sched/em_meta.c  |   2 +-
 9 files changed, 134 insertions(+), 289 deletions(-)
---
base-commit: b4cb730862cf4f59ac3dcb83b9ac4eeb29dbfb0e
change-id: 20250106-net-next-mptcp-rx-path-refactor-f44579efb57c

Best regards,
-- 
Matthieu Baerts (NGI0) <matttbe@kernel.org>

From: Paolo Abeni <pabeni@redhat.com>

Consolidate all the cleanup actions requiring the worker in a single
helper and ensure the dummy data fin creation for fallback socket is
performed only when the tcp rx queue is empty.

There are no functional changes intended, but this will simplify the
next patch, when the tcp rx queue spooling could be delayed at release_cb
time.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
 net/mptcp/subflow.c | 33 ++++++++++++++++++---------------
 1 file changed, 18 insertions(+), 15 deletions(-)

diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -XXX,XX +XXX,XX @@ static void mptcp_subflow_discard_data(struct sock *ssk, struct sk_buff *skb,
 		subflow->map_valid = 0;
 }
 
-/* sched mptcp worker to remove the subflow if no more data is pending */
+static bool subflow_is_done(const struct sock *sk)
+{
+	return sk->sk_shutdown & RCV_SHUTDOWN || sk->sk_state == TCP_CLOSE;
+}
+
+/* sched mptcp worker for subflow cleanup if no more data is pending */
 static void subflow_sched_work_if_closed(struct mptcp_sock *msk, struct sock *ssk)
 {
 	struct sock *sk = (struct sock *)msk;
@@ -XXX,XX +XXX,XX @@ static void subflow_sched_work_if_closed(struct mptcp_sock *msk, struct sock *ss
 		    inet_sk_state_load(sk) != TCP_ESTABLISHED)))
 		return;
 
-	if (skb_queue_empty(&ssk->sk_receive_queue) &&
-	    !test_and_set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &msk->flags))
+	if (!skb_queue_empty(&ssk->sk_receive_queue))
+		return;
+
+	if (!test_and_set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &msk->flags))
+		mptcp_schedule_work(sk);
+
+	/* when the fallback subflow closes the rx side, trigger a 'dummy'
+	 * ingress data fin, so that the msk state will follow along
+	 */
+	if (__mptcp_check_fallback(msk) && subflow_is_done(ssk) &&
+	    msk->first == ssk &&
+	    mptcp_update_rcv_data_fin(msk, READ_ONCE(msk->ack_seq), true))
 		mptcp_schedule_work(sk);
 }
 
@@ -XXX,XX +XXX,XX @@ static void __subflow_state_change(struct sock *sk)
 	rcu_read_unlock();
 }
 
-static bool subflow_is_done(const struct sock *sk)
-{
-	return sk->sk_shutdown & RCV_SHUTDOWN || sk->sk_state == TCP_CLOSE;
-}
-
 static void subflow_state_change(struct sock *sk)
 {
 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
@@ -XXX,XX +XXX,XX @@ static void subflow_state_change(struct sock *sk)
 		subflow_error_report(sk);
 
 	subflow_sched_work_if_closed(mptcp_sk(parent), sk);
-
-	/* when the fallback subflow closes the rx side, trigger a 'dummy'
-	 * ingress data fin, so that the msk state will follow along
-	 */
-	if (__mptcp_check_fallback(msk) && subflow_is_done(sk) && msk->first == sk &&
-	    mptcp_update_rcv_data_fin(msk, READ_ONCE(msk->ack_seq), true))
-		mptcp_schedule_work(parent);
 }
 
 void mptcp_subflow_queue_clean(struct sock *listener_sk, struct sock *listener_ssk)

-- 
2.47.1

From: Paolo Abeni <pabeni@redhat.com>

When we will move the whole RX path under the msk socket lock, updating
the already queued skb for passive fastopen socket at 3rd ack time will
be extremely painful and race prone

The map_seq for already enqueued skbs is used only to allow correct
coalescing with later data; preventing collapsing to the first skb of
a fastopen connect we can completely remove the
__mptcp_fastopen_gen_msk_ackseq() helper.

Before dropping this helper, a new item had to be added to the
mptcp_skb_cb structure. Because this item will be frequently tested in
the fast path -- almost on every packet -- and because there is free
space there, a single byte is used instead of a bitfield. This micro
optimisation slightly reduces the number of CPU operations to do the
associated check.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
 net/mptcp/fastopen.c | 24 ++----------------------
 net/mptcp/protocol.c |  4 +++-
 net/mptcp/protocol.h |  5 ++---
 net/mptcp/subflow.c  |  3 ---
 4 files changed, 7 insertions(+), 29 deletions(-)

-- 
2.47.1

From: Paolo Abeni <pabeni@redhat.com>

After commit c2e6048fa1cf ("mptcp: fix race in release_cb") we can
move the whole MPTCP rx path under the socket lock leveraging the
release_cb.

We can drop a bunch of spin_lock pairs in the receive functions, use
a single receive queue and invoke __mptcp_move_skbs only when subflows
ask for it.

This will allow more cleanup in the next patch.

Some changes are worth specific mention:

The msk rcvbuf update now always happens under both the msk and the
subflow socket lock: we can drop a bunch of ONCE annotation and
consolidate the checks.

When the skbs move is delayed at msk release callback time, even the
msk rcvbuf update is delayed; additionally take care of such action in
__mptcp_move_skbs().

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
 net/mptcp/fastopen.c |   1 +
 net/mptcp/protocol.c | 123 ++++++++++++++++++++++++---------------------------
 net/mptcp/protocol.h |   2 +-
 3 files changed, 60 insertions(+), 66 deletions(-)

diff --git a/net/mptcp/fastopen.c b/net/mptcp/fastopen.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/fastopen.c
+++ b/net/mptcp/fastopen.c
@@ -XXX,XX +XXX,XX @@ void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subf
 	MPTCP_SKB_CB(skb)->cant_coalesce = 1;
 
 	mptcp_data_lock(sk);
+	DEBUG_NET_WARN_ON_ONCE(sock_owned_by_user_nocheck(sk));
 
 	mptcp_set_owner_r(skb, sk);
 	__skb_queue_tail(&sk->sk_receive_queue, skb);
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 	bool more_data_avail;
 	struct tcp_sock *tp;
 	bool done = false;
-	int sk_rbuf;
-
-	sk_rbuf = READ_ONCE(sk->sk_rcvbuf);
-
-	if (!(sk->sk_userlocks & SOCK_RCVBUF_LOCK)) {
-		int ssk_rbuf = READ_ONCE(ssk->sk_rcvbuf);
-
-		if (unlikely(ssk_rbuf > sk_rbuf)) {
-			WRITE_ONCE(sk->sk_rcvbuf, ssk_rbuf);
-			sk_rbuf = ssk_rbuf;
-		}
-	}
 
 	pr_debug("msk=%p ssk=%p\n", msk, ssk);
 	tp = tcp_sk(ssk);
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 		WRITE_ONCE(tp->copied_seq, seq);
 		more_data_avail = mptcp_subflow_data_available(ssk);
 
-		if (atomic_read(&sk->sk_rmem_alloc) > sk_rbuf) {
+		if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) {
 			done = true;
 			break;
 		}
@@ -XXX,XX +XXX,XX @@ static bool move_skbs_to_msk(struct mptcp_sock *msk, struct sock *ssk)
 	return moved > 0;
 }
 
+static void __mptcp_rcvbuf_update(struct sock *sk, struct sock *ssk)
+{
+	if (unlikely(ssk->sk_rcvbuf > sk->sk_rcvbuf))
+		WRITE_ONCE(sk->sk_rcvbuf, ssk->sk_rcvbuf);
+}
+
+static void __mptcp_data_ready(struct sock *sk, struct sock *ssk)
+{
+	struct mptcp_sock *msk = mptcp_sk(sk);
+
+	__mptcp_rcvbuf_update(sk, ssk);
+
+	/* over limit? can't append more skbs to msk, Also, no need to wake-up*/
+	if (__mptcp_rmem(sk) > sk->sk_rcvbuf)
+		return;
+
+	/* Wake-up the reader only for in-sequence data */
+	if (move_skbs_to_msk(msk, ssk) && mptcp_epollin_ready(sk))
+		sk->sk_data_ready(sk);
+}
+
 void mptcp_data_ready(struct sock *sk, struct sock *ssk)
 {
 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
-	struct mptcp_sock *msk = mptcp_sk(sk);
-	int sk_rbuf, ssk_rbuf;
 
 	/* The peer can send data while we are shutting down this
 	 * subflow at msk destruction time, but we must avoid enqueuing
@@ -XXX,XX +XXX,XX @@ void mptcp_data_ready(struct sock *sk, struct sock *ssk)
 	if (unlikely(subflow->disposable))
 		return;
 
-	ssk_rbuf = READ_ONCE(ssk->sk_rcvbuf);
-	sk_rbuf = READ_ONCE(sk->sk_rcvbuf);
-	if (unlikely(ssk_rbuf > sk_rbuf))
-		sk_rbuf = ssk_rbuf;
-
-	/* over limit? can't append more skbs to msk, Also, no need to wake-up*/
-	if (__mptcp_rmem(sk) > sk_rbuf)
-		return;
-
-	/* Wake-up the reader only for in-sequence data */
 	mptcp_data_lock(sk);
-	if (move_skbs_to_msk(msk, ssk) && mptcp_epollin_ready(sk))
-		sk->sk_data_ready(sk);
+	if (!sock_owned_by_user(sk))
+		__mptcp_data_ready(sk, ssk);
+	else
+		__set_bit(MPTCP_DEQUEUE, &mptcp_sk(sk)->cb_flags);
 	mptcp_data_unlock(sk);
 }
 
@@ -XXX,XX +XXX,XX @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 
 static void mptcp_rcv_space_adjust(struct mptcp_sock *msk, int copied);
 
-static int __mptcp_recvmsg_mskq(struct mptcp_sock *msk,
+static int __mptcp_recvmsg_mskq(struct sock *sk,
 				struct msghdr *msg,
 				size_t len, int flags,
 				struct scm_timestamping_internal *tss,
 				int *cmsg_flags)
 {
+	struct mptcp_sock *msk = mptcp_sk(sk);
 	struct sk_buff *skb, *tmp;
 	int copied = 0;
 
-	skb_queue_walk_safe(&msk->receive_queue, skb, tmp) {
+	skb_queue_walk_safe(&sk->sk_receive_queue, skb, tmp) {
 		u32 offset = MPTCP_SKB_CB(skb)->offset;
 		u32 data_len = skb->len - offset;
 		u32 count = min_t(size_t, len - copied, data_len);
@@ -XXX,XX +XXX,XX @@ static int __mptcp_recvmsg_mskq(struct mptcp_sock *msk,
 			/* we will bulk release the skb memory later */
 			skb->destructor = NULL;
 			WRITE_ONCE(msk->rmem_released, msk->rmem_released + skb->truesize);
-			__skb_unlink(skb, &msk->receive_queue);
+			__skb_unlink(skb, &sk->sk_receive_queue);
 			__kfree_skb(skb);
 			msk->bytes_consumed += count;
 		}
@@ -XXX,XX +XXX,XX @@ static void __mptcp_update_rmem(struct sock *sk)
 	WRITE_ONCE(msk->rmem_released, 0);
 }
 
-static void __mptcp_splice_receive_queue(struct sock *sk)
+static bool __mptcp_move_skbs(struct sock *sk)
 {
+	struct mptcp_subflow_context *subflow;
 	struct mptcp_sock *msk = mptcp_sk(sk);
-
-	skb_queue_splice_tail_init(&sk->sk_receive_queue, &msk->receive_queue);
-}
-
-static bool __mptcp_move_skbs(struct mptcp_sock *msk)
-{
-	struct sock *sk = (struct sock *)msk;
 	unsigned int moved = 0;
 	bool ret, done;
 
+	/* verify we can move any data from the subflow, eventually updating */
+	if (!(sk->sk_userlocks & SOCK_RCVBUF_LOCK))
+		mptcp_for_each_subflow(msk, subflow)
+			__mptcp_rcvbuf_update(sk, subflow->tcp_sock);
+
+	if (__mptcp_rmem(sk) > sk->sk_rcvbuf)
+		return false;
+
 	do {
 		struct sock *ssk = mptcp_subflow_recv_lookup(msk);
 		bool slowpath;
 
-		/* we can have data pending in the subflows only if the msk
-		 * receive buffer was full at subflow_data_ready() time,
-		 * that is an unlikely slow path.
-		 */
-		if (likely(!ssk))
+		if (unlikely(!ssk))
 			break;
 
 		slowpath = lock_sock_fast(ssk);
-		mptcp_data_lock(sk);
 		__mptcp_update_rmem(sk);
 		done = __mptcp_move_skbs_from_subflow(msk, ssk, &moved);
-		mptcp_data_unlock(sk);
 
 		if (unlikely(ssk->sk_err))
 			__mptcp_error_report(sk);
 		unlock_sock_fast(ssk, slowpath);
 	} while (!done);
 
-	/* acquire the data lock only if some input data is pending */
 	ret = moved > 0;
 	if (!RB_EMPTY_ROOT(&msk->out_of_order_queue) ||
-	    !skb_queue_empty_lockless(&sk->sk_receive_queue)) {
-		mptcp_data_lock(sk);
+	    !skb_queue_empty(&sk->sk_receive_queue)) {
 		__mptcp_update_rmem(sk);
 		ret |= __mptcp_ofo_queue(msk);
-		__mptcp_splice_receive_queue(sk);
-		mptcp_data_unlock(sk);
 	}
 	if (ret)
 		mptcp_check_data_fin((struct sock *)msk);
-	return !skb_queue_empty(&msk->receive_queue);
+	return ret;
 }
 
 static unsigned int mptcp_inq_hint(const struct sock *sk)
@@ -XXX,XX +XXX,XX @@ static unsigned int mptcp_inq_hint(const struct sock *sk)
 	const struct mptcp_sock *msk = mptcp_sk(sk);
 	const struct sk_buff *skb;
 
-	skb = skb_peek(&msk->receive_queue);
+	skb = skb_peek(&sk->sk_receive_queue);
 	if (skb) {
 		u64 hint_val = READ_ONCE(msk->ack_seq) - MPTCP_SKB_CB(skb)->map_seq;
 
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 	while (copied < len) {
 		int err, bytes_read;
 
-		bytes_read = __mptcp_recvmsg_mskq(msk, msg, len - copied, flags, &tss, &cmsg_flags);
+		bytes_read = __mptcp_recvmsg_mskq(sk, msg, len - copied, flags, &tss, &cmsg_flags);
 		if (unlikely(bytes_read < 0)) {
 			if (!copied)
 				copied = bytes_read;
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 
 		copied += bytes_read;
 
-		if (skb_queue_empty(&msk->receive_queue) && __mptcp_move_skbs(msk))
+		if (skb_queue_empty(&sk->sk_receive_queue) && __mptcp_move_skbs(sk))
 			continue;
 
 		/* only the MPTCP socket status is relevant here. The exit
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 				/* race breaker: the shutdown could be after the
 				 * previous receive queue check
 				 */
-				if (__mptcp_move_skbs(msk))
+				if (__mptcp_move_skbs(sk))
 					continue;
 				break;
 			}
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 		}
 	}
 
-	pr_debug("msk=%p rx queue empty=%d:%d copied=%d\n",
-		 msk, skb_queue_empty_lockless(&sk->sk_receive_queue),
-		 skb_queue_empty(&msk->receive_queue), copied);
+	pr_debug("msk=%p rx queue empty=%d copied=%d\n",
+		 msk, skb_queue_empty(&sk->sk_receive_queue), copied);
 
 	release_sock(sk);
 	return copied;
@@ -XXX,XX +XXX,XX @@ static void __mptcp_init_sock(struct sock *sk)
 	INIT_LIST_HEAD(&msk->join_list);
 	INIT_LIST_HEAD(&msk->rtx_queue);
 	INIT_WORK(&msk->work, mptcp_worker);
-	__skb_queue_head_init(&msk->receive_queue);
 	msk->out_of_order_queue = RB_ROOT;
 	msk->first_pending = NULL;
 	WRITE_ONCE(msk->rmem_fwd_alloc, 0);
@@ -XXX,XX +XXX,XX @@ void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags)
 	mptcp_for_each_subflow_safe(msk, subflow, tmp)
 		__mptcp_close_ssk(sk, mptcp_subflow_tcp_sock(subflow), subflow, flags);
 
-	/* move to sk_receive_queue, sk_stream_kill_queues will purge it */
-	mptcp_data_lock(sk);
-	skb_queue_splice_tail_init(&msk->receive_queue, &sk->sk_receive_queue);
 	__skb_queue_purge(&sk->sk_receive_queue);
 	skb_rbtree_purge(&msk->out_of_order_queue);
-	mptcp_data_unlock(sk);
 
 	/* move all the rx fwd alloc into the sk_mem_reclaim_final in
 	 * inet_sock_destruct() will dispose it
@@ -XXX,XX +XXX,XX @@ void __mptcp_check_push(struct sock *sk, struct sock *ssk)
 
 #define MPTCP_FLAGS_PROCESS_CTX_NEED (BIT(MPTCP_PUSH_PENDING) | \
 				      BIT(MPTCP_RETRANSMIT) | \
-				      BIT(MPTCP_FLUSH_JOIN_LIST))
+				      BIT(MPTCP_FLUSH_JOIN_LIST) | \
+				      BIT(MPTCP_DEQUEUE))
 
 /* processes deferred events and flush wmem */
 static void mptcp_release_cb(struct sock *sk)
@@ -XXX,XX +XXX,XX @@ static void mptcp_release_cb(struct sock *sk)
 			__mptcp_push_pending(sk, 0);
 		if (flags & BIT(MPTCP_RETRANSMIT))
 			__mptcp_retrans(sk);
+		if ((flags & BIT(MPTCP_DEQUEUE)) && __mptcp_move_skbs(sk)) {
+			/* notify ack seq update */
+			mptcp_cleanup_rbuf(msk, 0);
+			sk->sk_data_ready(sk);
+		}
 
 		cond_resched();
 		spin_lock_bh(&sk->sk_lock.slock);
@@ -XXX,XX +XXX,XX @@ static int mptcp_ioctl(struct sock *sk, int cmd, int *karg)
 			return -EINVAL;
 
 		lock_sock(sk);
-		__mptcp_move_skbs(msk);
+		if (__mptcp_move_skbs(sk))
+			mptcp_cleanup_rbuf(msk, 0);
 		*karg = mptcp_inq_hint(sk);
 		release_sock(sk);
 		break;
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -XXX,XX +XXX,XX @@
 #define MPTCP_FLUSH_JOIN_LIST	5
 #define MPTCP_SYNC_STATE	6
 #define MPTCP_SYNC_SNDBUF	7
+#define MPTCP_DEQUEUE		8
 
 struct mptcp_skb_cb {
 	u64 map_seq;
@@ -XXX,XX +XXX,XX @@ struct mptcp_sock {
 	struct work_struct work;
 	struct sk_buff  *ooo_last_skb;
 	struct rb_root  out_of_order_queue;
-	struct sk_buff_head receive_queue;
 	struct list_head conn_list;
 	struct list_head rtx_queue;
 	struct mptcp_data_frag *first_pending;

-- 
2.47.1

From: Paolo Abeni <pabeni@redhat.com>

After the previous patch, updating sk_forward_memory is cheap and
we can drop a lot of complexity from the MPTCP memory accounting,
removing the custom fwd mem allocations for rmem.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
 net/mptcp/fastopen.c |   2 +-
 net/mptcp/protocol.c | 115 ++++-----------------------------------------------
 net/mptcp/protocol.h |   4 +-
 3 files changed, 10 insertions(+), 111 deletions(-)

diff --git a/net/mptcp/fastopen.c b/net/mptcp/fastopen.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/fastopen.c
+++ b/net/mptcp/fastopen.c
@@ -XXX,XX +XXX,XX @@ void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subf
 	mptcp_data_lock(sk);
 	DEBUG_NET_WARN_ON_ONCE(sock_owned_by_user_nocheck(sk));
 
-	mptcp_set_owner_r(skb, sk);
+	skb_set_owner_r(skb, sk);
 	__skb_queue_tail(&sk->sk_receive_queue, skb);
 	mptcp_sk(sk)->bytes_received += skb->len;
 
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -XXX,XX +XXX,XX @@ static void mptcp_drop(struct sock *sk, struct sk_buff *skb)
 	__kfree_skb(skb);
 }
 
-static void mptcp_rmem_fwd_alloc_add(struct sock *sk, int size)
-{
-	WRITE_ONCE(mptcp_sk(sk)->rmem_fwd_alloc,
-		   mptcp_sk(sk)->rmem_fwd_alloc + size);
-}
-
-static void mptcp_rmem_charge(struct sock *sk, int size)
-{
-	mptcp_rmem_fwd_alloc_add(sk, -size);
-}
-
 static bool mptcp_try_coalesce(struct sock *sk, struct sk_buff *to,
 			       struct sk_buff *from)
 {
@@ -XXX,XX +XXX,XX @@ static bool mptcp_try_coalesce(struct sock *sk, struct sk_buff *to,
 	 * negative one
 	 */
 	atomic_add(delta, &sk->sk_rmem_alloc);
-	mptcp_rmem_charge(sk, delta);
+	sk_mem_charge(sk, delta);
 	kfree_skb_partial(from, fragstolen);
 
 	return true;
@@ -XXX,XX +XXX,XX @@ static bool mptcp_ooo_try_coalesce(struct mptcp_sock *msk, struct sk_buff *to,
 	return mptcp_try_coalesce((struct sock *)msk, to, from);
 }
 
-static void __mptcp_rmem_reclaim(struct sock *sk, int amount)
-{
-	amount >>= PAGE_SHIFT;
-	mptcp_rmem_charge(sk, amount << PAGE_SHIFT);
-	__sk_mem_reduce_allocated(sk, amount);
-}
-
-static void mptcp_rmem_uncharge(struct sock *sk, int size)
-{
-	struct mptcp_sock *msk = mptcp_sk(sk);
-	int reclaimable;
-
-	mptcp_rmem_fwd_alloc_add(sk, size);
-	reclaimable = msk->rmem_fwd_alloc - sk_unused_reserved_mem(sk);
-
-	/* see sk_mem_uncharge() for the rationale behind the following schema */
-	if (unlikely(reclaimable >= PAGE_SIZE))
-		__mptcp_rmem_reclaim(sk, reclaimable);
-}
-
-static void mptcp_rfree(struct sk_buff *skb)
-{
-	unsigned int len = skb->truesize;
-	struct sock *sk = skb->sk;
-
-	atomic_sub(len, &sk->sk_rmem_alloc);
-	mptcp_rmem_uncharge(sk, len);
-}
-
-void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk)
-{
-	skb_orphan(skb);
-	skb->sk = sk;
-	skb->destructor = mptcp_rfree;
-	atomic_add(skb->truesize, &sk->sk_rmem_alloc);
-	mptcp_rmem_charge(sk, skb->truesize);
-}
-
 /* "inspired" by tcp_data_queue_ofo(), main differences:
  * - use mptcp seqs
  * - don't cope with sacks
@@ -XXX,XX +XXX,XX @@ static void mptcp_data_queue_ofo(struct mptcp_sock *msk, struct sk_buff *skb)
 
 end:
 	skb_condense(skb);
-	mptcp_set_owner_r(skb, sk);
-}
-
-static bool mptcp_rmem_schedule(struct sock *sk, struct sock *ssk, int size)
-{
-	struct mptcp_sock *msk = mptcp_sk(sk);
-	int amt, amount;
-
-	if (size <= msk->rmem_fwd_alloc)
-		return true;
-
-	size -= msk->rmem_fwd_alloc;
-	amt = sk_mem_pages(size);
-	amount = amt << PAGE_SHIFT;
-	if (!__sk_mem_raise_allocated(sk, size, amt, SK_MEM_RECV))
-		return false;
-
-	mptcp_rmem_fwd_alloc_add(sk, amount);
-	return true;
+	skb_set_owner_r(skb, sk);
 }
 
 static bool __mptcp_move_skb(struct mptcp_sock *msk, struct sock *ssk,
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skb(struct mptcp_sock *msk, struct sock *ssk,
 	skb_orphan(skb);
 
 	/* try to fetch required memory from subflow */
-	if (!mptcp_rmem_schedule(sk, ssk, skb->truesize)) {
+	if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
 		MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_RCVPRUNED);
 		goto drop;
 	}
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skb(struct mptcp_sock *msk, struct sock *ssk,
 		if (tail && mptcp_try_coalesce(sk, tail, skb))
 			return true;
 
-		mptcp_set_owner_r(skb, sk);
+		skb_set_owner_r(skb, sk);
 		__skb_queue_tail(&sk->sk_receive_queue, skb);
 		return true;
 	} else if (after64(MPTCP_SKB_CB(skb)->map_seq, msk->ack_seq)) {
@@ -XXX,XX +XXX,XX @@ static int __mptcp_recvmsg_mskq(struct sock *sk,
 		}
 
 		if (!(flags & MSG_PEEK)) {
-			/* we will bulk release the skb memory later */
+			/* avoid the indirect call, we know the destructor is sock_wfree */
 			skb->destructor = NULL;
-			WRITE_ONCE(msk->rmem_released, msk->rmem_released + skb->truesize);
+			atomic_sub(skb->truesize, &sk->sk_rmem_alloc);
+			sk_mem_uncharge(sk, skb->truesize);
 			__skb_unlink(skb, &sk->sk_receive_queue);
 			__kfree_skb(skb);
 			msk->bytes_consumed += count;
@@ -XXX,XX +XXX,XX @@ static void mptcp_rcv_space_adjust(struct mptcp_sock *msk, int copied)
 	msk->rcvq_space.time = mstamp;
 }
 
-static void __mptcp_update_rmem(struct sock *sk)
-{
-	struct mptcp_sock *msk = mptcp_sk(sk);
-
-	if (!msk->rmem_released)
-		return;
-
-	atomic_sub(msk->rmem_released, &sk->sk_rmem_alloc);
-	mptcp_rmem_uncharge(sk, msk->rmem_released);
-	WRITE_ONCE(msk->rmem_released, 0);
-}
-
 static bool __mptcp_move_skbs(struct sock *sk)
 {
 	struct mptcp_subflow_context *subflow;
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs(struct sock *sk)
 			break;
 
 		slowpath = lock_sock_fast(ssk);
-		__mptcp_update_rmem(sk);
 		done = __mptcp_move_skbs_from_subflow(msk, ssk, &moved);
 
 		if (unlikely(ssk->sk_err))
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs(struct sock *sk)
 		unlock_sock_fast(ssk, slowpath);
 	} while (!done);
 
-	ret = moved > 0;
-	if (!RB_EMPTY_ROOT(&msk->out_of_order_queue) ||
-	    !skb_queue_empty(&sk->sk_receive_queue)) {
-		__mptcp_update_rmem(sk);
-		ret |= __mptcp_ofo_queue(msk);
-	}
+	ret = moved > 0 || __mptcp_ofo_queue(msk);
 	if (ret)
 		mptcp_check_data_fin((struct sock *)msk);
 	return ret;
@@ -XXX,XX +XXX,XX @@ static void __mptcp_init_sock(struct sock *sk)
 	INIT_WORK(&msk->work, mptcp_worker);
 	msk->out_of_order_queue = RB_ROOT;
 	msk->first_pending = NULL;
-	WRITE_ONCE(msk->rmem_fwd_alloc, 0);
-	WRITE_ONCE(msk->rmem_released, 0);
 	msk->timer_ival = TCP_RTO_MIN;
 	msk->scaling_ratio = TCP_DEFAULT_SCALING_RATIO;
 
@@ -XXX,XX +XXX,XX @@ static void __mptcp_destroy_sock(struct sock *sk)
 
 	sk->sk_prot->destroy(sk);
 
-	WARN_ON_ONCE(READ_ONCE(msk->rmem_fwd_alloc));
-	WARN_ON_ONCE(msk->rmem_released);
 	sk_stream_kill_queues(sk);
 	xfrm_sk_free_policy(sk);
 
@@ -XXX,XX +XXX,XX @@ void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags)
 	/* move all the rx fwd alloc into the sk_mem_reclaim_final in
 	 * inet_sock_destruct() will dispose it
 	 */
-	sk_forward_alloc_add(sk, msk->rmem_fwd_alloc);
-	WRITE_ONCE(msk->rmem_fwd_alloc, 0);
 	mptcp_token_destroy(msk);
 	mptcp_pm_free_anno_list(msk);
 	mptcp_free_local_addr_list(msk);
@@ -XXX,XX +XXX,XX @@ static void mptcp_release_cb(struct sock *sk)
 		if (__test_and_clear_bit(MPTCP_SYNC_SNDBUF, &msk->cb_flags))
 			__mptcp_sync_sndbuf(sk);
 	}
-
-	__mptcp_update_rmem(sk);
 }
 
 /* MP_JOIN client subflow must wait for 4th ack before sending any data:
@@ -XXX,XX +XXX,XX @@ static void mptcp_shutdown(struct sock *sk, int how)
 		__mptcp_wr_shutdown(sk);
 }
 
-static int mptcp_forward_alloc_get(const struct sock *sk)
-{
-	return READ_ONCE(sk->sk_forward_alloc) +
-	       READ_ONCE(mptcp_sk(sk)->rmem_fwd_alloc);
-}
-
 static int mptcp_ioctl_outq(const struct mptcp_sock *msk, u64 v)
 {
 	const struct sock *sk = (void *)msk;
@@ -XXX,XX +XXX,XX @@ static struct proto mptcp_prot = {
 	.hash		= mptcp_hash,
 	.unhash		= mptcp_unhash,
 	.get_port	= mptcp_get_port,
-	.forward_alloc_get	= mptcp_forward_alloc_get,
 	.stream_memory_free	= mptcp_stream_memory_free,
 	.sockets_allocated	= &mptcp_sockets_allocated,
 
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -XXX,XX +XXX,XX @@ struct mptcp_sock {
 	u64		rcv_data_fin_seq;
 	u64		bytes_retrans;
 	u64		bytes_consumed;
-	int		rmem_fwd_alloc;
 	int		snd_burst;
 	int		old_wspace;
 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
@@ -XXX,XX +XXX,XX @@ struct mptcp_sock {
 	u32		last_ack_recv;
 	unsigned long	timer_ival;
 	u32		token;
-	int		rmem_released;
 	unsigned long	flags;
 	unsigned long	cb_flags;
 	bool		recovery;		/* closing subflow write queue reinjected */
@@ -XXX,XX +XXX,XX @@ static inline void msk_owned_by_me(const struct mptcp_sock *msk)
  */
 static inline int __mptcp_rmem(const struct sock *sk)
 {
-	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
+	return atomic_read(&sk->sk_rmem_alloc);
 }
 
 static inline int mptcp_win_from_space(const struct sock *sk, int space)

-- 
2.47.1

From: Paolo Abeni <pabeni@redhat.com>

After the previous patch we can remove the forward_alloc_get
proto callback, basically reverting commit 292e6077b040 ("net: introduce
sk_forward_alloc_get()") and commit 66d58f046c9d ("net: use
sk_forward_alloc_get() in sk_get_meminfo()").

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
 include/net/sock.h   | 13 -------------
 net/core/sock.c      |  2 +-
 net/ipv4/af_inet.c   |  2 +-
 net/ipv4/inet_diag.c |  2 +-
 net/sched/em_meta.c  |  2 +-
 5 files changed, 4 insertions(+), 17 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index XXXXXXX..XXXXXXX 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -XXX,XX +XXX,XX @@ struct proto {
 	unsigned int		inuse_idx;
 #endif
 
-#if IS_ENABLED(CONFIG_MPTCP)
-	int			(*forward_alloc_get)(const struct sock *sk);
-#endif
-
 	bool			(*stream_memory_free)(const struct sock *sk, int wake);
 	bool			(*sock_is_readable)(struct sock *sk);
 	/* Memory pressure */
@@ -XXX,XX +XXX,XX @@ int sock_load_diag_module(int family, int protocol);
 
 INDIRECT_CALLABLE_DECLARE(bool tcp_stream_memory_free(const struct sock *sk, int wake));
 
-static inline int sk_forward_alloc_get(const struct sock *sk)
-{
-#if IS_ENABLED(CONFIG_MPTCP)
-	if (sk->sk_prot->forward_alloc_get)
-		return sk->sk_prot->forward_alloc_get(sk);
-#endif
-	return READ_ONCE(sk->sk_forward_alloc);
-}
-
 static inline bool __sk_stream_memory_free(const struct sock *sk, int wake)
 {
 	if (READ_ONCE(sk->sk_wmem_queued) >= READ_ONCE(sk->sk_sndbuf))
diff --git a/net/core/sock.c b/net/core/sock.c
index XXXXXXX..XXXXXXX 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -XXX,XX +XXX,XX @@ void sk_get_meminfo(const struct sock *sk, u32 *mem)
 	mem[SK_MEMINFO_RCVBUF] = READ_ONCE(sk->sk_rcvbuf);
 	mem[SK_MEMINFO_WMEM_ALLOC] = sk_wmem_alloc_get(sk);
 	mem[SK_MEMINFO_SNDBUF] = READ_ONCE(sk->sk_sndbuf);
-	mem[SK_MEMINFO_FWD_ALLOC] = sk_forward_alloc_get(sk);
+	mem[SK_MEMINFO_FWD_ALLOC] = READ_ONCE(sk->sk_forward_alloc);
 	mem[SK_MEMINFO_WMEM_QUEUED] = READ_ONCE(sk->sk_wmem_queued);
 	mem[SK_MEMINFO_OPTMEM] = atomic_read(&sk->sk_omem_alloc);
 	mem[SK_MEMINFO_BACKLOG] = READ_ONCE(sk->sk_backlog.len);
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index XXXXXXX..XXXXXXX 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -XXX,XX +XXX,XX @@ void inet_sock_destruct(struct sock *sk)
 	WARN_ON_ONCE(atomic_read(&sk->sk_rmem_alloc));
 	WARN_ON_ONCE(refcount_read(&sk->sk_wmem_alloc));
 	WARN_ON_ONCE(sk->sk_wmem_queued);
-	WARN_ON_ONCE(sk_forward_alloc_get(sk));
+	WARN_ON_ONCE(sk->sk_forward_alloc);
 
 	kfree(rcu_dereference_protected(inet->inet_opt, 1));
 	dst_release(rcu_dereference_protected(sk->sk_dst_cache, 1));
diff --git a/net/ipv4/inet_diag.c b/net/ipv4/inet_diag.c
index XXXXXXX..XXXXXXX 100644
--- a/net/ipv4/inet_diag.c
+++ b/net/ipv4/inet_diag.c
@@ -XXX,XX +XXX,XX @@ int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
 		struct inet_diag_meminfo minfo = {
 			.idiag_rmem = sk_rmem_alloc_get(sk),
 			.idiag_wmem = READ_ONCE(sk->sk_wmem_queued),
-			.idiag_fmem = sk_forward_alloc_get(sk),
+			.idiag_fmem = READ_ONCE(sk->sk_forward_alloc),
 			.idiag_tmem = sk_wmem_alloc_get(sk),
 		};
 
diff --git a/net/sched/em_meta.c b/net/sched/em_meta.c
index XXXXXXX..XXXXXXX 100644
--- a/net/sched/em_meta.c
+++ b/net/sched/em_meta.c
@@ -XXX,XX +XXX,XX @@ META_COLLECTOR(int_sk_fwd_alloc)
 		*err = -1;
 		return;
 	}
-	dst->value = sk_forward_alloc_get(sk);
+	dst->value = READ_ONCE(sk->sk_forward_alloc);
 }
 
 META_COLLECTOR(int_sk_sndbuf)

-- 
2.47.1

From: Paolo Abeni <pabeni@redhat.com>

After the RX path refactor, it become a wrapper for sk_rmem_alloc
access, with a slightly misleading name. Just drop it.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
 net/mptcp/protocol.c |  8 ++++----
 net/mptcp/protocol.h | 11 ++---------
 2 files changed, 6 insertions(+), 13 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -XXX,XX +XXX,XX @@ static void mptcp_cleanup_rbuf(struct mptcp_sock *msk, int copied)
 	bool cleanup, rx_empty;
 
 	cleanup = (space > 0) && (space >= (old_space << 1)) && copied;
-	rx_empty = !__mptcp_rmem(sk) && copied;
+	rx_empty = !sk_rmem_alloc_get(sk) && copied;
 
 	mptcp_for_each_subflow(msk, subflow) {
 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 		WRITE_ONCE(tp->copied_seq, seq);
 		more_data_avail = mptcp_subflow_data_available(ssk);
 
-		if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) {
+		if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf) {
 			done = true;
 			break;
 		}
@@ -XXX,XX +XXX,XX @@ static void __mptcp_data_ready(struct sock *sk, struct sock *ssk)
 	__mptcp_rcvbuf_update(sk, ssk);
 
 	/* over limit? can't append more skbs to msk, Also, no need to wake-up*/
-	if (__mptcp_rmem(sk) > sk->sk_rcvbuf)
+	if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
 		return;
 
 	/* Wake-up the reader only for in-sequence data */
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs(struct sock *sk)
 		mptcp_for_each_subflow(msk, subflow)
 			__mptcp_rcvbuf_update(sk, subflow->tcp_sock);
 
-	if (__mptcp_rmem(sk) > sk->sk_rcvbuf)
+	if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
 		return false;
 
 	do {
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -XXX,XX +XXX,XX @@ static inline void msk_owned_by_me(const struct mptcp_sock *msk)
 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
 #endif
 
-/* the msk socket don't use the backlog, also account for the bulk
- * free memory
- */
-static inline int __mptcp_rmem(const struct sock *sk)
-{
-	return atomic_read(&sk->sk_rmem_alloc);
-}
-
 static inline int mptcp_win_from_space(const struct sock *sk, int space)
 {
 	return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
@@ -XXX,XX +XXX,XX @@ static inline int mptcp_space_from_win(const struct sock *sk, int win)
 
 static inline int __mptcp_space(const struct sock *sk)
 {
-	return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
+	return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) -
+				    sk_rmem_alloc_get(sk));
 }
 
 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)

-- 
2.47.1

From: Paolo Abeni <pabeni@redhat.com>

After the RX path refactor the mentioned function is expected to run
frequently, let's optimize it a bit.

Scan for ready subflow from the last processed one, and stop after
traversing the list once or reaching the msk memory limit - instead of
looking for dubious per-subflow conditions.
Also re-order the memory limit checks, to avoid duplicate tests.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
 net/mptcp/protocol.c | 111 +++++++++++++++++++++++----------------------------
 net/mptcp/protocol.h |   2 +
 2 files changed, 52 insertions(+), 61 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -XXX,XX +XXX,XX @@ static void mptcp_dss_corruption(struct mptcp_sock *msk, struct sock *ssk)
 }
 
 static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
-					   struct sock *ssk,
-					   unsigned int *bytes)
+					   struct sock *ssk)
 {
 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
 	struct sock *sk = (struct sock *)msk;
-	unsigned int moved = 0;
 	bool more_data_avail;
 	struct tcp_sock *tp;
-	bool done = false;
+	bool ret = false;
 
 	pr_debug("msk=%p ssk=%p\n", msk, ssk);
 	tp = tcp_sk(ssk);
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 		struct sk_buff *skb;
 		bool fin;
 
+		if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
+			break;
+
 		/* try to move as much data as available */
 		map_remaining = subflow->map_data_len -
 				mptcp_subflow_get_map_offset(subflow);
 
 		skb = skb_peek(&ssk->sk_receive_queue);
-		if (!skb) {
-			/* With racing move_skbs_to_msk() and __mptcp_move_skbs(),
-			 * a different CPU can have already processed the pending
-			 * data, stop here or we can enter an infinite loop
-			 */
-			if (!moved)
-				done = true;
+		if (unlikely(!skb))
 			break;
-		}
 
 		if (__mptcp_check_fallback(msk)) {
 			/* Under fallback skbs have no MPTCP extension and TCP could
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 
 		offset = seq - TCP_SKB_CB(skb)->seq;
 		fin = TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN;
-		if (fin) {
-			done = true;
+		if (fin)
 			seq++;
-		}
 
 		if (offset < skb->len) {
 			size_t len = skb->len - offset;
 
-			if (tp->urg_data)
-				done = true;
-
-			if (__mptcp_move_skb(msk, ssk, skb, offset, len))
-				moved += len;
+			ret = __mptcp_move_skb(msk, ssk, skb, offset, len) || ret;
 			seq += len;
 
 			if (unlikely(map_remaining < len)) {
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 			}
 
 			sk_eat_skb(ssk, skb);
-			done = true;
 		}
 
 		WRITE_ONCE(tp->copied_seq, seq);
 		more_data_avail = mptcp_subflow_data_available(ssk);
 
-		if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf) {
-			done = true;
-			break;
-		}
 	} while (more_data_avail);
 
-	if (moved > 0)
+	if (ret)
 		msk->last_data_recv = tcp_jiffies32;
-	*bytes += moved;
-	return done;
+	return ret;
 }
 
 static bool __mptcp_ofo_queue(struct mptcp_sock *msk)
@@ -XXX,XX +XXX,XX @@ void __mptcp_error_report(struct sock *sk)
 static bool move_skbs_to_msk(struct mptcp_sock *msk, struct sock *ssk)
 {
 	struct sock *sk = (struct sock *)msk;
-	unsigned int moved = 0;
+	bool moved;
 
-	__mptcp_move_skbs_from_subflow(msk, ssk, &moved);
+	moved = __mptcp_move_skbs_from_subflow(msk, ssk);
 	__mptcp_ofo_queue(msk);
 	if (unlikely(ssk->sk_err)) {
 		if (!sock_owned_by_user(sk))
@@ -XXX,XX +XXX,XX @@ static bool move_skbs_to_msk(struct mptcp_sock *msk, struct sock *ssk)
 	 */
 	if (mptcp_pending_data_fin(sk, NULL))
 		mptcp_schedule_work(sk);
-	return moved > 0;
+	return moved;
 }
 
 static void __mptcp_rcvbuf_update(struct sock *sk, struct sock *ssk)
@@ -XXX,XX +XXX,XX @@ static void __mptcp_data_ready(struct sock *sk, struct sock *ssk)
 
 	__mptcp_rcvbuf_update(sk, ssk);
 
-	/* over limit? can't append more skbs to msk, Also, no need to wake-up*/
-	if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
-		return;
-
 	/* Wake-up the reader only for in-sequence data */
 	if (move_skbs_to_msk(msk, ssk) && mptcp_epollin_ready(sk))
 		sk->sk_data_ready(sk);
@@ -XXX,XX +XXX,XX @@ bool mptcp_schedule_work(struct sock *sk)
 	return false;
 }
 
-static struct sock *mptcp_subflow_recv_lookup(const struct mptcp_sock *msk)
-{
-	struct mptcp_subflow_context *subflow;
-
-	msk_owned_by_me(msk);
-
-	mptcp_for_each_subflow(msk, subflow) {
-		if (READ_ONCE(subflow->data_avail))
-			return mptcp_subflow_tcp_sock(subflow);
-	}
-
-	return NULL;
-}
-
 static bool mptcp_skb_can_collapse_to(u64 write_seq,
 				      const struct sk_buff *skb,
 				      const struct mptcp_ext *mpext)
@@ -XXX,XX +XXX,XX @@ static void mptcp_rcv_space_adjust(struct mptcp_sock *msk, int copied)
 	msk->rcvq_space.time = mstamp;
 }
 
+static struct mptcp_subflow_context *
+__mptcp_first_ready_from(struct mptcp_sock *msk,
+			 struct mptcp_subflow_context *subflow)
+{
+	struct mptcp_subflow_context *start_subflow = subflow;
+
+	while (!READ_ONCE(subflow->data_avail)) {
+		subflow = mptcp_next_subflow(msk, subflow);
+		if (subflow == start_subflow)
+			return NULL;
+	}
+	return subflow;
+}
+
 static bool __mptcp_move_skbs(struct sock *sk)
 {
 	struct mptcp_subflow_context *subflow;
 	struct mptcp_sock *msk = mptcp_sk(sk);
-	unsigned int moved = 0;
-	bool ret, done;
+	bool ret = false;
+
+	if (list_empty(&msk->conn_list))
+		return false;
 
 	/* verify we can move any data from the subflow, eventually updating */
 	if (!(sk->sk_userlocks & SOCK_RCVBUF_LOCK))
 		mptcp_for_each_subflow(msk, subflow)
 			__mptcp_rcvbuf_update(sk, subflow->tcp_sock);
 
-	if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
-		return false;
-
-	do {
-		struct sock *ssk = mptcp_subflow_recv_lookup(msk);
+	subflow = list_first_entry(&msk->conn_list,
+				   struct mptcp_subflow_context, node);
+	for (;;) {
+		struct sock *ssk;
 		bool slowpath;
 
-		if (unlikely(!ssk))
+		/*
+		 * As an optimization avoid traversing the subflows list
+		 * and ev. acquiring the subflow socket lock before baling out
+		 */
+		if (sk_rmem_alloc_get(sk) > sk->sk_rcvbuf)
 			break;
 
-		slowpath = lock_sock_fast(ssk);
-		done = __mptcp_move_skbs_from_subflow(msk, ssk, &moved);
+		subflow = __mptcp_first_ready_from(msk, subflow);
+		if (!subflow)
+			break;
 
+		ssk = mptcp_subflow_tcp_sock(subflow);
+		slowpath = lock_sock_fast(ssk);
+		ret = __mptcp_move_skbs_from_subflow(msk, ssk) || ret;
 		if (unlikely(ssk->sk_err))
 			__mptcp_error_report(sk);
 		unlock_sock_fast(ssk, slowpath);
-	} while (!done);
 
-	ret = moved > 0 || __mptcp_ofo_queue(msk);
+		subflow = mptcp_next_subflow(msk, subflow);
+	}
+
+	__mptcp_ofo_queue(msk);
 	if (ret)
 		mptcp_check_data_fin((struct sock *)msk);
 	return ret;
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -XXX,XX +XXX,XX @@ struct mptcp_sock {
 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)			\
 	list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
+#define mptcp_next_subflow(__msk, __subflow)				\
+	list_next_entry_circular(__subflow, &((__msk)->conn_list), node)
 
 extern struct genl_family mptcp_genl_family;

-- 
2.47.1

This is a batch of changes I had sitting in my local tree for a while.
Why another refactor you may ask? Two main resons:

- currently the mptcp RX path introduces quite a bit of 'exceptional'
 accounting/locking processing WRT to plain TCP, adding up to the
 implementation complexity in a misurable way
- the performance gap WRT plain TCP for single subflow connections is
 quite measurable.

The present refactor addresses both the above items: most of the
additional complexity is dropped, and single stream performances
increase measurably - from 55Gbps to 71Gbps in my loopback test. As a
reference, plain TCP is around 84Gps on the same host.

The above comes to a price: the patch are invasive, even in subtle ways:
the chance of destabilizing the implementation is real (ence the
additional, intentional '-next' into the subj).

In any case keeping the patch hidden for longer was not going to do any
good, so here we are.

Changes from v1:
  - fixed several data stream corruption and wake-up misses due
    to multi subflows races
  - added patches 1-3 mainly to address the above
  - added an additional follow-up patch (patch 7) with more cleanup

Paolo Abeni (7):
  mptcp: prevent excessive coalescing on receive
  tcp: fix recvbuffer adjust on sleeping rcvmsg
  mptcp: don't always assume copied data in mptcp_cleanup_rbuf()
  mptcp: consolidate subflow cleanup
  mptcp: move the whole rx path under msk socket lock protection
  mptcp: cleanup mem accounting.
  net: dismiss sk_forward_alloc_get()

-- 
2.45.2

Currently the skb size after coalescing is only limited by the skb
layout (the skb must not carry frag_list). A single coalesced skb
covering several MSS can potentially fill completely the receive
buffer. In such a case, the snd win will zero until the receive buffer
will be empty again, affecting tput badly.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
No fixes tag because the problem is not very visible in practice
currently, but will be apparent after the rx refactor.
Still I hope this could affect positively the simlut flows self-tests
---
 net/mptcp/protocol.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -XXX,XX +XXX,XX @@ static bool mptcp_try_coalesce(struct sock *sk, struct sk_buff *to,
 	int delta;
 
 	if (MPTCP_SKB_CB(from)->offset ||
+	    ((to->len + from->len) > (sk->sk_rcvbuf >> 3)) ||
 	    !skb_try_coalesce(to, from, &fragstolen, &delta))
 		return false;
 
-- 
2.45.2

If the recvmsg() blocks after receiving some data - i.e. due to
SO_RCVLOWAT - the MPTCP code will attempt multiple times to
adjust the receive buffer size, wrongly accounting every time the
cumulative of received data - instead of accounting only for the
delta.

Address the issue moving mptcp_rcv_space_adjust just after the
data reception and passing it only the just received bytes.

This also remove an unneeded difference between the TCP and MPTCP
RX code path implementation.

Fixes: 581302298524 ("mptcp: error out earlier on disconnect")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
Not strictly related to the refactor, found while investigating
the CI failure
---
 net/mptcp/protocol.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -XXX,XX +XXX,XX @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 	goto out;
 }
 
+static void mptcp_rcv_space_adjust(struct mptcp_sock *msk, int copied);
+
 static int __mptcp_recvmsg_mskq(struct mptcp_sock *msk,
 				struct msghdr *msg,
 				size_t len, int flags,
@@ -XXX,XX +XXX,XX @@ static int __mptcp_recvmsg_mskq(struct mptcp_sock *msk,
 			break;
 	}
 
+	mptcp_rcv_space_adjust(msk, copied);
 	return copied;
 }
 
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 		}
 
 		pr_debug("block timeout %ld\n", timeo);
-		mptcp_rcv_space_adjust(msk, copied);
 		err = sk_wait_data(sk, &timeo, NULL);
 		if (err < 0) {
 			err = copied ? : err;
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 		}
 	}
 
-	mptcp_rcv_space_adjust(msk, copied);
-
 out_err:
 	if (cmsg_flags && copied >= 0) {
 		if (cmsg_flags & MPTCP_CMSG_TS)
-- 
2.45.2

Under some corner cases the MPTCP protocol can end-up invoking
mptcp_cleanup_rbuf() when no data has been copied, but such helper
assumes the opposite condition.

Explicitly drop such assumption and performs the costly call only
when strictly needed - before releasing the msk socket lock.

Fixes: fd8976790a6c ("mptcp: be careful on MPTCP-level ack.")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
 net/mptcp/protocol.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -XXX,XX +XXX,XX @@ static void mptcp_send_ack(struct mptcp_sock *msk)
 		mptcp_subflow_send_ack(mptcp_subflow_tcp_sock(subflow));
 }
 
-static void mptcp_subflow_cleanup_rbuf(struct sock *ssk)
+static void mptcp_subflow_cleanup_rbuf(struct sock *ssk, int copied)
 {
 	bool slow;
 
 	slow = lock_sock_fast(ssk);
 	if (tcp_can_send_ack(ssk))
-		tcp_cleanup_rbuf(ssk, 1);
+		tcp_cleanup_rbuf(ssk, copied);
 	unlock_sock_fast(ssk, slow);
 }
 
@@ -XXX,XX +XXX,XX @@ static bool mptcp_subflow_could_cleanup(const struct sock *ssk, bool rx_empty)
 			      (ICSK_ACK_PUSHED2 | ICSK_ACK_PUSHED)));
 }
 
-static void mptcp_cleanup_rbuf(struct mptcp_sock *msk)
+static void mptcp_cleanup_rbuf(struct mptcp_sock *msk, int copied)
 {
 	int old_space = READ_ONCE(msk->old_wspace);
 	struct mptcp_subflow_context *subflow;
@@ -XXX,XX +XXX,XX @@ static void mptcp_cleanup_rbuf(struct mptcp_sock *msk)
 	int space =  __mptcp_space(sk);
 	bool cleanup, rx_empty;
 
-	cleanup = (space > 0) && (space >= (old_space << 1));
-	rx_empty = !__mptcp_rmem(sk);
+	cleanup = (space > 0) && (space >= (old_space << 1)) && copied;
+	rx_empty = !__mptcp_rmem(sk) && copied;
 
 	mptcp_for_each_subflow(msk, subflow) {
 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
 
 		if (cleanup || mptcp_subflow_could_cleanup(ssk, rx_empty))
-			mptcp_subflow_cleanup_rbuf(ssk);
+			mptcp_subflow_cleanup_rbuf(ssk, copied);
 	}
 }
 
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 
 		copied += bytes_read;
 
-		/* be sure to advertise window change */
-		mptcp_cleanup_rbuf(msk);
-
 		if (skb_queue_empty(&msk->receive_queue) && __mptcp_move_skbs(msk))
 			continue;
 
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 		}
 
 		pr_debug("block timeout %ld\n", timeo);
+		mptcp_cleanup_rbuf(msk, copied);
 		err = sk_wait_data(sk, &timeo, NULL);
 		if (err < 0) {
 			err = copied ? : err;
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 		}
 	}
 
+	mptcp_cleanup_rbuf(msk, copied);
+
 out_err:
 	if (cmsg_flags && copied >= 0) {
 		if (cmsg_flags & MPTCP_CMSG_TS)
-- 
2.45.2

Consolidate all the cleanup actions requiring the worker in a single
helper and ensure the dummy data fin creation for fallback socket is
performed only when the tcp rx queue is empty.

There are no functional changes intended, but this will simplify the
next patch, when the tcp rx queue spooling could be delayed at release_cb
time.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
 net/mptcp/subflow.c | 33 ++++++++++++++++++---------------
 1 file changed, 18 insertions(+), 15 deletions(-)

After commit c2e6048fa1cf ("mptcp: fix race in release_cb") we can
move the whole MPTCP rx path under the socket lock leveraging the
release_cb.

We can drop a bunch of spin_lock pairs in the receive functions, use
a single receive queue and invoke __mptcp_move_skbs only when subflows
ask for it.

This will allow more cleanup in the next patch.

Some changes are worth specific mention:

The msk rcvbuf update now always happens under both the msk and the
subflow socket lock: we can drop a bunch of ONCE annotation and
consolidate the checks.

When the skbs move is delayed at msk release callback time, even the
msk rcvbuf update is delayed; additionally take care of such action in
__mptcp_move_skbs().

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
v1 -> v2:
 - cleanup && fixup msk sk_rcvbuf update
 - fix missed wakeup due to bad placed __mptcp_move_skbs(); move it
   form __mptcp_recvmsg_mskq into mptcp_recvmsg()
 - added missing '\n' in debug message format string
 - keep 'snd_una' && friends update under the mptcp data lock
---
 net/mptcp/fastopen.c |   2 +
 net/mptcp/protocol.c | 123 ++++++++++++++++++++-----------------------
 net/mptcp/protocol.h |   2 +-
 3 files changed, 61 insertions(+), 66 deletions(-)

diff --git a/net/mptcp/fastopen.c b/net/mptcp/fastopen.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/fastopen.c
+++ b/net/mptcp/fastopen.c
@@ -XXX,XX +XXX,XX @@ void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subf
 	MPTCP_SKB_CB(skb)->has_rxtstamp = TCP_SKB_CB(skb)->has_rxtstamp;
 
 	mptcp_data_lock(sk);
+	DEBUG_NET_WARN_ON_ONCE(sock_owned_by_user_nocheck(sk));
 
 	mptcp_set_owner_r(skb, sk);
 	__skb_queue_tail(&sk->sk_receive_queue, skb);
@@ -XXX,XX +XXX,XX @@ void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflo
 	struct sock *sk = (struct sock *)msk;
 	struct sk_buff *skb;
 
+	DEBUG_NET_WARN_ON_ONCE(sock_owned_by_user_nocheck(sk));
 	skb = skb_peek_tail(&sk->sk_receive_queue);
 	if (skb) {
 		WARN_ON_ONCE(MPTCP_SKB_CB(skb)->end_seq);
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 	bool more_data_avail;
 	struct tcp_sock *tp;
 	bool done = false;
-	int sk_rbuf;
-
-	sk_rbuf = READ_ONCE(sk->sk_rcvbuf);
-
-	if (!(sk->sk_userlocks & SOCK_RCVBUF_LOCK)) {
-		int ssk_rbuf = READ_ONCE(ssk->sk_rcvbuf);
-
-		if (unlikely(ssk_rbuf > sk_rbuf)) {
-			WRITE_ONCE(sk->sk_rcvbuf, ssk_rbuf);
-			sk_rbuf = ssk_rbuf;
-		}
-	}
 
 	pr_debug("msk=%p ssk=%p\n", msk, ssk);
 	tp = tcp_sk(ssk);
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 		WRITE_ONCE(tp->copied_seq, seq);
 		more_data_avail = mptcp_subflow_data_available(ssk);
 
-		if (atomic_read(&sk->sk_rmem_alloc) > sk_rbuf) {
+		if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) {
 			done = true;
 			break;
 		}
@@ -XXX,XX +XXX,XX @@ static bool move_skbs_to_msk(struct mptcp_sock *msk, struct sock *ssk)
 	return moved > 0;
 }
 
+static void __mptcp_rcvbuf_update(struct sock *sk, struct sock *ssk)
+{
+	if (unlikely(ssk->sk_rcvbuf > sk->sk_rcvbuf))
+		WRITE_ONCE(sk->sk_rcvbuf, ssk->sk_rcvbuf);
+}
+
+static void __mptcp_data_ready(struct sock *sk, struct sock *ssk)
+{
+	struct mptcp_sock *msk = mptcp_sk(sk);
+
+	__mptcp_rcvbuf_update(sk, ssk);
+
+	/* over limit? can't append more skbs to msk, Also, no need to wake-up*/
+	if (__mptcp_rmem(sk) > sk->sk_rcvbuf)
+		return;
+
+	/* Wake-up the reader only for in-sequence data */
+	if (move_skbs_to_msk(msk, ssk) && mptcp_epollin_ready(sk))
+		sk->sk_data_ready(sk);
+}
+
 void mptcp_data_ready(struct sock *sk, struct sock *ssk)
 {
 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
-	struct mptcp_sock *msk = mptcp_sk(sk);
-	int sk_rbuf, ssk_rbuf;
 
 	/* The peer can send data while we are shutting down this
 	 * subflow at msk destruction time, but we must avoid enqueuing
@@ -XXX,XX +XXX,XX @@ void mptcp_data_ready(struct sock *sk, struct sock *ssk)
 	if (unlikely(subflow->disposable))
 		return;
 
-	ssk_rbuf = READ_ONCE(ssk->sk_rcvbuf);
-	sk_rbuf = READ_ONCE(sk->sk_rcvbuf);
-	if (unlikely(ssk_rbuf > sk_rbuf))
-		sk_rbuf = ssk_rbuf;
-
-	/* over limit? can't append more skbs to msk, Also, no need to wake-up*/
-	if (__mptcp_rmem(sk) > sk_rbuf)
-		return;
-
-	/* Wake-up the reader only for in-sequence data */
 	mptcp_data_lock(sk);
-	if (move_skbs_to_msk(msk, ssk) && mptcp_epollin_ready(sk))
-		sk->sk_data_ready(sk);
+	if (!sock_owned_by_user(sk))
+		__mptcp_data_ready(sk, ssk);
+	else
+		__set_bit(MPTCP_DEQUEUE, &mptcp_sk(sk)->cb_flags);
 	mptcp_data_unlock(sk);
 }
 
@@ -XXX,XX +XXX,XX @@ static int mptcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 
 static void mptcp_rcv_space_adjust(struct mptcp_sock *msk, int copied);
 
-static int __mptcp_recvmsg_mskq(struct mptcp_sock *msk,
+static int __mptcp_recvmsg_mskq(struct sock *sk,
 				struct msghdr *msg,
 				size_t len, int flags,
 				struct scm_timestamping_internal *tss,
 				int *cmsg_flags)
 {
+	struct mptcp_sock *msk = mptcp_sk(sk);
 	struct sk_buff *skb, *tmp;
 	int copied = 0;
 
-	skb_queue_walk_safe(&msk->receive_queue, skb, tmp) {
+	skb_queue_walk_safe(&sk->sk_receive_queue, skb, tmp) {
 		u32 offset = MPTCP_SKB_CB(skb)->offset;
 		u32 data_len = skb->len - offset;
 		u32 count = min_t(size_t, len - copied, data_len);
@@ -XXX,XX +XXX,XX @@ static int __mptcp_recvmsg_mskq(struct mptcp_sock *msk,
 			/* we will bulk release the skb memory later */
 			skb->destructor = NULL;
 			WRITE_ONCE(msk->rmem_released, msk->rmem_released + skb->truesize);
-			__skb_unlink(skb, &msk->receive_queue);
+			__skb_unlink(skb, &sk->sk_receive_queue);
 			__kfree_skb(skb);
 			msk->bytes_consumed += count;
 		}
@@ -XXX,XX +XXX,XX @@ static void __mptcp_update_rmem(struct sock *sk)
 	WRITE_ONCE(msk->rmem_released, 0);
 }
 
-static void __mptcp_splice_receive_queue(struct sock *sk)
+static bool __mptcp_move_skbs(struct sock *sk)
 {
+	struct mptcp_subflow_context *subflow;
 	struct mptcp_sock *msk = mptcp_sk(sk);
-
-	skb_queue_splice_tail_init(&sk->sk_receive_queue, &msk->receive_queue);
-}
-
-static bool __mptcp_move_skbs(struct mptcp_sock *msk)
-{
-	struct sock *sk = (struct sock *)msk;
 	unsigned int moved = 0;
 	bool ret, done;
 
+	/* verify we can move any data from the subflow, eventually updating */
+	if (!(sk->sk_userlocks & SOCK_RCVBUF_LOCK))
+		mptcp_for_each_subflow(msk, subflow)
+			__mptcp_rcvbuf_update(sk, subflow->tcp_sock);
+
+	if (__mptcp_rmem(sk) > sk->sk_rcvbuf)
+		return false;
+
 	do {
 		struct sock *ssk = mptcp_subflow_recv_lookup(msk);
 		bool slowpath;
 
-		/* we can have data pending in the subflows only if the msk
-		 * receive buffer was full at subflow_data_ready() time,
-		 * that is an unlikely slow path.
-		 */
-		if (likely(!ssk))
+		if (unlikely(!ssk))
 			break;
 
 		slowpath = lock_sock_fast(ssk);
-		mptcp_data_lock(sk);
 		__mptcp_update_rmem(sk);
 		done = __mptcp_move_skbs_from_subflow(msk, ssk, &moved);
-		mptcp_data_unlock(sk);
 
 		if (unlikely(ssk->sk_err))
 			__mptcp_error_report(sk);
 		unlock_sock_fast(ssk, slowpath);
 	} while (!done);
 
-	/* acquire the data lock only if some input data is pending */
 	ret = moved > 0;
 	if (!RB_EMPTY_ROOT(&msk->out_of_order_queue) ||
-	    !skb_queue_empty_lockless(&sk->sk_receive_queue)) {
-		mptcp_data_lock(sk);
+	    !skb_queue_empty(&sk->sk_receive_queue)) {
 		__mptcp_update_rmem(sk);
 		ret |= __mptcp_ofo_queue(msk);
-		__mptcp_splice_receive_queue(sk);
-		mptcp_data_unlock(sk);
 	}
 	if (ret)
 		mptcp_check_data_fin((struct sock *)msk);
-	return !skb_queue_empty(&msk->receive_queue);
+	return ret;
 }
 
 static unsigned int mptcp_inq_hint(const struct sock *sk)
@@ -XXX,XX +XXX,XX @@ static unsigned int mptcp_inq_hint(const struct sock *sk)
 	const struct mptcp_sock *msk = mptcp_sk(sk);
 	const struct sk_buff *skb;
 
-	skb = skb_peek(&msk->receive_queue);
+	skb = skb_peek(&sk->sk_receive_queue);
 	if (skb) {
 		u64 hint_val = READ_ONCE(msk->ack_seq) - MPTCP_SKB_CB(skb)->map_seq;
 
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 	while (copied < len) {
 		int err, bytes_read;
 
-		bytes_read = __mptcp_recvmsg_mskq(msk, msg, len - copied, flags, &tss, &cmsg_flags);
+		bytes_read = __mptcp_recvmsg_mskq(sk, msg, len - copied, flags, &tss, &cmsg_flags);
 		if (unlikely(bytes_read < 0)) {
 			if (!copied)
 				copied = bytes_read;
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 
 		copied += bytes_read;
 
-		if (skb_queue_empty(&msk->receive_queue) && __mptcp_move_skbs(msk))
+		if (skb_queue_empty(&sk->sk_receive_queue) && __mptcp_move_skbs(sk))
 			continue;
 
 		/* only the MPTCP socket status is relevant here. The exit
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 				/* race breaker: the shutdown could be after the
 				 * previous receive queue check
 				 */
-				if (__mptcp_move_skbs(msk))
+				if (__mptcp_move_skbs(sk))
 					continue;
 				break;
 			}
@@ -XXX,XX +XXX,XX @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 		}
 	}
 
-	pr_debug("msk=%p rx queue empty=%d:%d copied=%d\n",
-		 msk, skb_queue_empty_lockless(&sk->sk_receive_queue),
-		 skb_queue_empty(&msk->receive_queue), copied);
+	pr_debug("msk=%p rx queue empty=%d copied=%d\n",
+		 msk, skb_queue_empty(&sk->sk_receive_queue), copied);
 
 	release_sock(sk);
 	return copied;
@@ -XXX,XX +XXX,XX @@ static void __mptcp_init_sock(struct sock *sk)
 	INIT_LIST_HEAD(&msk->join_list);
 	INIT_LIST_HEAD(&msk->rtx_queue);
 	INIT_WORK(&msk->work, mptcp_worker);
-	__skb_queue_head_init(&msk->receive_queue);
 	msk->out_of_order_queue = RB_ROOT;
 	msk->first_pending = NULL;
 	WRITE_ONCE(msk->rmem_fwd_alloc, 0);
@@ -XXX,XX +XXX,XX @@ void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags)
 	mptcp_for_each_subflow_safe(msk, subflow, tmp)
 		__mptcp_close_ssk(sk, mptcp_subflow_tcp_sock(subflow), subflow, flags);
 
-	/* move to sk_receive_queue, sk_stream_kill_queues will purge it */
-	mptcp_data_lock(sk);
-	skb_queue_splice_tail_init(&msk->receive_queue, &sk->sk_receive_queue);
 	__skb_queue_purge(&sk->sk_receive_queue);
 	skb_rbtree_purge(&msk->out_of_order_queue);
-	mptcp_data_unlock(sk);
 
 	/* move all the rx fwd alloc into the sk_mem_reclaim_final in
 	 * inet_sock_destruct() will dispose it
@@ -XXX,XX +XXX,XX @@ void __mptcp_check_push(struct sock *sk, struct sock *ssk)
 
 #define MPTCP_FLAGS_PROCESS_CTX_NEED (BIT(MPTCP_PUSH_PENDING) | \
 				      BIT(MPTCP_RETRANSMIT) | \
-				      BIT(MPTCP_FLUSH_JOIN_LIST))
+				      BIT(MPTCP_FLUSH_JOIN_LIST) | \
+				      BIT(MPTCP_DEQUEUE))
 
 /* processes deferred events and flush wmem */
 static void mptcp_release_cb(struct sock *sk)
@@ -XXX,XX +XXX,XX @@ static void mptcp_release_cb(struct sock *sk)
 			__mptcp_push_pending(sk, 0);
 		if (flags & BIT(MPTCP_RETRANSMIT))
 			__mptcp_retrans(sk);
+		if ((flags & BIT(MPTCP_DEQUEUE)) && __mptcp_move_skbs(sk)) {
+			/* notify ack seq update */
+			mptcp_cleanup_rbuf(msk, 0);
+			sk->sk_data_ready(sk);
+		}
 
 		cond_resched();
 		spin_lock_bh(&sk->sk_lock.slock);
@@ -XXX,XX +XXX,XX @@ static int mptcp_ioctl(struct sock *sk, int cmd, int *karg)
 			return -EINVAL;
 
 		lock_sock(sk);
-		__mptcp_move_skbs(msk);
+		if (__mptcp_move_skbs(sk))
+			mptcp_cleanup_rbuf(msk, 0);
 		*karg = mptcp_inq_hint(sk);
 		release_sock(sk);
 		break;
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -XXX,XX +XXX,XX @@
 #define MPTCP_FLUSH_JOIN_LIST	5
 #define MPTCP_SYNC_STATE	6
 #define MPTCP_SYNC_SNDBUF	7
+#define MPTCP_DEQUEUE		8
 
 struct mptcp_skb_cb {
 	u64 map_seq;
@@ -XXX,XX +XXX,XX @@ struct mptcp_sock {
 	struct work_struct work;
 	struct sk_buff  *ooo_last_skb;
 	struct rb_root  out_of_order_queue;
-	struct sk_buff_head receive_queue;
 	struct list_head conn_list;
 	struct list_head rtx_queue;
 	struct mptcp_data_frag *first_pending;
-- 
2.45.2

After the previous patch, updating sk_forward_memory is cheap and
we can drop a lot of complexity from the MPTCP memory acconting,
removing the custom fwd mem allocations for rmem.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
v1 -> v2:
 - keep 'snd_una' and recovery-related fields under the msk
   data lock
 - dropped unneeded code in __mptcp_move_skbs()
---
 net/mptcp/fastopen.c |   2 +-
 net/mptcp/protocol.c | 115 +++----------------------------------------
 net/mptcp/protocol.h |   4 +-
 3 files changed, 10 insertions(+), 111 deletions(-)

diff --git a/net/mptcp/fastopen.c b/net/mptcp/fastopen.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/fastopen.c
+++ b/net/mptcp/fastopen.c
@@ -XXX,XX +XXX,XX @@ void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subf
 	mptcp_data_lock(sk);
 	DEBUG_NET_WARN_ON_ONCE(sock_owned_by_user_nocheck(sk));
 
-	mptcp_set_owner_r(skb, sk);
+	skb_set_owner_r(skb, sk);
 	__skb_queue_tail(&sk->sk_receive_queue, skb);
 	mptcp_sk(sk)->bytes_received += skb->len;
 
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -XXX,XX +XXX,XX @@ static void mptcp_drop(struct sock *sk, struct sk_buff *skb)
 	__kfree_skb(skb);
 }
 
-static void mptcp_rmem_fwd_alloc_add(struct sock *sk, int size)
-{
-	WRITE_ONCE(mptcp_sk(sk)->rmem_fwd_alloc,
-		   mptcp_sk(sk)->rmem_fwd_alloc + size);
-}
-
-static void mptcp_rmem_charge(struct sock *sk, int size)
-{
-	mptcp_rmem_fwd_alloc_add(sk, -size);
-}
-
 static bool mptcp_try_coalesce(struct sock *sk, struct sk_buff *to,
 			       struct sk_buff *from)
 {
@@ -XXX,XX +XXX,XX @@ static bool mptcp_try_coalesce(struct sock *sk, struct sk_buff *to,
 	 * negative one
 	 */
 	atomic_add(delta, &sk->sk_rmem_alloc);
-	mptcp_rmem_charge(sk, delta);
+	sk_mem_charge(sk, delta);
 	kfree_skb_partial(from, fragstolen);
 
 	return true;
@@ -XXX,XX +XXX,XX @@ static bool mptcp_ooo_try_coalesce(struct mptcp_sock *msk, struct sk_buff *to,
 	return mptcp_try_coalesce((struct sock *)msk, to, from);
 }
 
-static void __mptcp_rmem_reclaim(struct sock *sk, int amount)
-{
-	amount >>= PAGE_SHIFT;
-	mptcp_rmem_charge(sk, amount << PAGE_SHIFT);
-	__sk_mem_reduce_allocated(sk, amount);
-}
-
-static void mptcp_rmem_uncharge(struct sock *sk, int size)
-{
-	struct mptcp_sock *msk = mptcp_sk(sk);
-	int reclaimable;
-
-	mptcp_rmem_fwd_alloc_add(sk, size);
-	reclaimable = msk->rmem_fwd_alloc - sk_unused_reserved_mem(sk);
-
-	/* see sk_mem_uncharge() for the rationale behind the following schema */
-	if (unlikely(reclaimable >= PAGE_SIZE))
-		__mptcp_rmem_reclaim(sk, reclaimable);
-}
-
-static void mptcp_rfree(struct sk_buff *skb)
-{
-	unsigned int len = skb->truesize;
-	struct sock *sk = skb->sk;
-
-	atomic_sub(len, &sk->sk_rmem_alloc);
-	mptcp_rmem_uncharge(sk, len);
-}
-
-void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk)
-{
-	skb_orphan(skb);
-	skb->sk = sk;
-	skb->destructor = mptcp_rfree;
-	atomic_add(skb->truesize, &sk->sk_rmem_alloc);
-	mptcp_rmem_charge(sk, skb->truesize);
-}
-
 /* "inspired" by tcp_data_queue_ofo(), main differences:
  * - use mptcp seqs
  * - don't cope with sacks
@@ -XXX,XX +XXX,XX @@ static void mptcp_data_queue_ofo(struct mptcp_sock *msk, struct sk_buff *skb)
 
 end:
 	skb_condense(skb);
-	mptcp_set_owner_r(skb, sk);
-}
-
-static bool mptcp_rmem_schedule(struct sock *sk, struct sock *ssk, int size)
-{
-	struct mptcp_sock *msk = mptcp_sk(sk);
-	int amt, amount;
-
-	if (size <= msk->rmem_fwd_alloc)
-		return true;
-
-	size -= msk->rmem_fwd_alloc;
-	amt = sk_mem_pages(size);
-	amount = amt << PAGE_SHIFT;
-	if (!__sk_mem_raise_allocated(sk, size, amt, SK_MEM_RECV))
-		return false;
-
-	mptcp_rmem_fwd_alloc_add(sk, amount);
-	return true;
+	skb_set_owner_r(skb, sk);
 }
 
 static bool __mptcp_move_skb(struct mptcp_sock *msk, struct sock *ssk,
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skb(struct mptcp_sock *msk, struct sock *ssk,
 	skb_orphan(skb);
 
 	/* try to fetch required memory from subflow */
-	if (!mptcp_rmem_schedule(sk, ssk, skb->truesize)) {
+	if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
 		MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_RCVPRUNED);
 		goto drop;
 	}
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skb(struct mptcp_sock *msk, struct sock *ssk,
 		if (tail && mptcp_try_coalesce(sk, tail, skb))
 			return true;
 
-		mptcp_set_owner_r(skb, sk);
+		skb_set_owner_r(skb, sk);
 		__skb_queue_tail(&sk->sk_receive_queue, skb);
 		return true;
 	} else if (after64(MPTCP_SKB_CB(skb)->map_seq, msk->ack_seq)) {
@@ -XXX,XX +XXX,XX @@ static int __mptcp_recvmsg_mskq(struct sock *sk,
 		}
 
 		if (!(flags & MSG_PEEK)) {
-			/* we will bulk release the skb memory later */
+			/* avoid the indirect call, we know the destructor is sock_wfree */
 			skb->destructor = NULL;
-			WRITE_ONCE(msk->rmem_released, msk->rmem_released + skb->truesize);
+			atomic_sub(skb->truesize, &sk->sk_rmem_alloc);
+			sk_mem_uncharge(sk, skb->truesize);
 			__skb_unlink(skb, &sk->sk_receive_queue);
 			__kfree_skb(skb);
 			msk->bytes_consumed += count;
@@ -XXX,XX +XXX,XX @@ static void mptcp_rcv_space_adjust(struct mptcp_sock *msk, int copied)
 	msk->rcvq_space.time = mstamp;
 }
 
-static void __mptcp_update_rmem(struct sock *sk)
-{
-	struct mptcp_sock *msk = mptcp_sk(sk);
-
-	if (!msk->rmem_released)
-		return;
-
-	atomic_sub(msk->rmem_released, &sk->sk_rmem_alloc);
-	mptcp_rmem_uncharge(sk, msk->rmem_released);
-	WRITE_ONCE(msk->rmem_released, 0);
-}
-
 static bool __mptcp_move_skbs(struct sock *sk)
 {
 	struct mptcp_subflow_context *subflow;
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs(struct sock *sk)
 			break;
 
 		slowpath = lock_sock_fast(ssk);
-		__mptcp_update_rmem(sk);
 		done = __mptcp_move_skbs_from_subflow(msk, ssk, &moved);
 
 		if (unlikely(ssk->sk_err))
@@ -XXX,XX +XXX,XX @@ static bool __mptcp_move_skbs(struct sock *sk)
 		unlock_sock_fast(ssk, slowpath);
 	} while (!done);
 
-	ret = moved > 0;
-	if (!RB_EMPTY_ROOT(&msk->out_of_order_queue) ||
-	    !skb_queue_empty(&sk->sk_receive_queue)) {
-		__mptcp_update_rmem(sk);
-		ret |= __mptcp_ofo_queue(msk);
-	}
+	ret = moved > 0 || __mptcp_ofo_queue(msk);
 	if (ret)
 		mptcp_check_data_fin((struct sock *)msk);
 	return ret;
@@ -XXX,XX +XXX,XX @@ static void __mptcp_init_sock(struct sock *sk)
 	INIT_WORK(&msk->work, mptcp_worker);
 	msk->out_of_order_queue = RB_ROOT;
 	msk->first_pending = NULL;
-	WRITE_ONCE(msk->rmem_fwd_alloc, 0);
-	WRITE_ONCE(msk->rmem_released, 0);
 	msk->timer_ival = TCP_RTO_MIN;
 	msk->scaling_ratio = TCP_DEFAULT_SCALING_RATIO;
 
@@ -XXX,XX +XXX,XX @@ static void __mptcp_destroy_sock(struct sock *sk)
 
 	sk->sk_prot->destroy(sk);
 
-	WARN_ON_ONCE(READ_ONCE(msk->rmem_fwd_alloc));
-	WARN_ON_ONCE(msk->rmem_released);
 	sk_stream_kill_queues(sk);
 	xfrm_sk_free_policy(sk);
 
@@ -XXX,XX +XXX,XX @@ void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags)
 	/* move all the rx fwd alloc into the sk_mem_reclaim_final in
 	 * inet_sock_destruct() will dispose it
 	 */
-	sk_forward_alloc_add(sk, msk->rmem_fwd_alloc);
-	WRITE_ONCE(msk->rmem_fwd_alloc, 0);
 	mptcp_token_destroy(msk);
 	mptcp_pm_free_anno_list(msk);
 	mptcp_free_local_addr_list(msk);
@@ -XXX,XX +XXX,XX @@ static void mptcp_release_cb(struct sock *sk)
 		if (__test_and_clear_bit(MPTCP_SYNC_SNDBUF, &msk->cb_flags))
 			__mptcp_sync_sndbuf(sk);
 	}
-
-	__mptcp_update_rmem(sk);
 }
 
 /* MP_JOIN client subflow must wait for 4th ack before sending any data:
@@ -XXX,XX +XXX,XX @@ static void mptcp_shutdown(struct sock *sk, int how)
 		__mptcp_wr_shutdown(sk);
 }
 
-static int mptcp_forward_alloc_get(const struct sock *sk)
-{
-	return READ_ONCE(sk->sk_forward_alloc) +
-	       READ_ONCE(mptcp_sk(sk)->rmem_fwd_alloc);
-}
-
 static int mptcp_ioctl_outq(const struct mptcp_sock *msk, u64 v)
 {
 	const struct sock *sk = (void *)msk;
@@ -XXX,XX +XXX,XX @@ static struct proto mptcp_prot = {
 	.hash		= mptcp_hash,
 	.unhash		= mptcp_unhash,
 	.get_port	= mptcp_get_port,
-	.forward_alloc_get	= mptcp_forward_alloc_get,
 	.stream_memory_free	= mptcp_stream_memory_free,
 	.sockets_allocated	= &mptcp_sockets_allocated,
 
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index XXXXXXX..XXXXXXX 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -XXX,XX +XXX,XX @@ struct mptcp_sock {
 	u64		rcv_data_fin_seq;
 	u64		bytes_retrans;
 	u64		bytes_consumed;
-	int		rmem_fwd_alloc;
 	int		snd_burst;
 	int		old_wspace;
 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
@@ -XXX,XX +XXX,XX @@ struct mptcp_sock {
 	u32		last_ack_recv;
 	unsigned long	timer_ival;
 	u32		token;
-	int		rmem_released;
 	unsigned long	flags;
 	unsigned long	cb_flags;
 	bool		recovery;		/* closing subflow write queue reinjected */
@@ -XXX,XX +XXX,XX @@ static inline void msk_owned_by_me(const struct mptcp_sock *msk)
  */
 static inline int __mptcp_rmem(const struct sock *sk)
 {
-	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
+	return atomic_read(&sk->sk_rmem_alloc);
 }
 
 static inline int mptcp_win_from_space(const struct sock *sk, int space)
-- 
2.45.2