From 01f83d69844d307be2aa6fea88b0e8fe5cbdb2f4 Mon Sep 17 00:00:00 2001 From: Alexey Kuznetsov Date: Wed, 15 Sep 2010 10:27:52 -0700 Subject: tcp: Prevent overzealous packetization by SWS logic. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If peer uses tiny MSS (say, 75 bytes) and similarly tiny advertised window, the SWS logic will packetize to half the MSS unnecessarily. This causes problems with some embedded devices. However for large MSS devices we do want to half-MSS packetize otherwise we never get enough packets into the pipe for things like fast retransmit and recovery to work. Be careful also to handle the case where MSS > window, otherwise we'll never send until the probe timer. Reported-by: ツ Leandro Melo de Sales Signed-off-by: David S. Miller --- include/net/tcp.h | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/include/net/tcp.h b/include/net/tcp.h index eaa9582779d0..3e4b33e36602 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -475,8 +475,22 @@ extern unsigned int tcp_current_mss(struct sock *sk); /* Bound MSS / TSO packet size with the half of the window */ static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize) { - if (tp->max_window && pktsize > (tp->max_window >> 1)) - return max(tp->max_window >> 1, 68U - tp->tcp_header_len); + int cutoff; + + /* When peer uses tiny windows, there is no use in packetizing + * to sub-MSS pieces for the sake of SWS or making sure there + * are enough packets in the pipe for fast recovery. + * + * On the other hand, for extremely large MSS devices, handling + * smaller than MSS windows in this way does make sense. + */ + if (tp->max_window >= 512) + cutoff = (tp->max_window >> 1); + else + cutoff = tp->max_window; + + if (cutoff && pktsize > cutoff) + return max_t(int, cutoff, 68U - tp->tcp_header_len); else return pktsize; } -- cgit v1.2.3 From f0f9deae9e7c421fa0c1c627beb8e174325e1ba7 Mon Sep 17 00:00:00 2001 From: Herbert Xu Date: Fri, 17 Sep 2010 16:55:03 -0700 Subject: netpoll: Disable IRQ around RCU dereference in netpoll_rx We cannot use rcu_dereference_bh safely in netpoll_rx as we may be called with IRQs disabled. We could however simply disable IRQs as that too causes BH to be disabled and is safe in either case. Thanks to John Linville for discovering this bug and providing a patch. Signed-off-by: Herbert Xu Signed-off-by: David S. Miller --- include/linux/netpoll.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'include') diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h index 791d5109f34c..50d8009be86c 100644 --- a/include/linux/netpoll.h +++ b/include/linux/netpoll.h @@ -63,20 +63,20 @@ static inline bool netpoll_rx(struct sk_buff *skb) unsigned long flags; bool ret = false; - rcu_read_lock_bh(); + local_irq_save(flags); npinfo = rcu_dereference_bh(skb->dev->npinfo); if (!npinfo || (list_empty(&npinfo->rx_np) && !npinfo->rx_flags)) goto out; - spin_lock_irqsave(&npinfo->rx_lock, flags); + spin_lock(&npinfo->rx_lock); /* check rx_flags again with the lock held */ if (npinfo->rx_flags && __netpoll_rx(skb)) ret = true; - spin_unlock_irqrestore(&npinfo->rx_lock, flags); + spin_unlock(&npinfo->rx_lock); out: - rcu_read_unlock_bh(); + local_irq_restore(flags); return ret; } -- cgit v1.2.3