From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: [PATCH net-next 4/6] tcp: fix behavior for epoll edge trigger Date: Fri, 15 May 2015 07:53:56 -0700 Message-ID: <1431701638-24451-5-git-send-email-edumazet@google.com> References: <1431701638-24451-1-git-send-email-edumazet@google.com> Cc: netdev , Jason Baron , Neal Cardwell , Yuchung Cheng , Eric Dumazet , Eric Dumazet To: "David S. Miller" Return-path: Received: from mail-ig0-f201.google.com ([209.85.213.201]:32923 "EHLO mail-ig0-f201.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754802AbbEOOy0 (ORCPT ); Fri, 15 May 2015 10:54:26 -0400 Received: by igdh15 with SMTP id h15so4968302igd.0 for ; Fri, 15 May 2015 07:54:26 -0700 (PDT) In-Reply-To: <1431701638-24451-1-git-send-email-edumazet@google.com> Sender: netdev-owner@vger.kernel.org List-ID: Under memory pressure, tcp_sendmsg() can fail to queue a packet while no packet is present in write queue. If we return -EAGAIN with no packet in write queue, no ACK packet will ever come to raise EPOLLOUT. We need to allow one skb per TCP socket, and make sure that tcp sockets can release their forward allocations under pressure. This is a followup to commit 790ba4566c1a ("tcp: set SOCK_NOSPACE under memory pressure") Signed-off-by: Eric Dumazet --- net/ipv4/tcp.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-) diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index ecccfdc50d76..9eabfd3e0925 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -815,9 +815,20 @@ struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp) /* The TCP header must be at least 32-bit aligned. */ size = ALIGN(size, 4); + if (unlikely(tcp_under_memory_pressure(sk))) + sk_mem_reclaim_partial(sk); + skb = alloc_skb_fclone(size + sk->sk_prot->max_header, gfp); - if (skb) { - if (sk_wmem_schedule(sk, skb->truesize)) { + if (likely(skb)) { + bool mem_schedule; + + if (skb_queue_len(&sk->sk_write_queue) == 0) { + mem_schedule = true; + sk_forced_mem_schedule(sk, skb->truesize); + } else { + mem_schedule = sk_wmem_schedule(sk, skb->truesize); + } + if (likely(mem_schedule)) { skb_reserve(skb, sk->sk_prot->max_header); /* * Make sure that we have exactly size bytes -- 2.2.0.rc0.207.ga3a616c