tcp: tsq: relax tcp_small_queue_check() when rtx queue contains a single skb
authorEric Dumazet <edumazet@google.com>
Tue, 17 Oct 2023 12:45:26 +0000 (12:45 +0000)
committerJakub Kicinski <kuba@kernel.org>
Thu, 19 Oct 2023 01:06:36 +0000 (18:06 -0700)
commitf921a4a5bffa8a0005b190fb9421a7fc1fd716b6
tree59b25955bceeeb2be0a4d8671e94829a6fb07a7d
parenta0ca6b9dfef0b3cc83aa8bb485ed61a018f84982
tcp: tsq: relax tcp_small_queue_check() when rtx queue contains a single skb

In commit 75eefc6c59fd ("tcp: tsq: add a shortcut in tcp_small_queue_check()")
we allowed to send an skb regardless of TSQ limits being hit if rtx queue
was empty or had a single skb, in order to better fill the pipe
when/if TX completions were slow.

Then later, commit 75c119afe14f ("tcp: implement rb-tree based
retransmit queue") accidentally removed the special case for
one skb in rtx queue.

Stefan Wahren reported a regression in single TCP flow throughput
using a 100Mbit fec link, starting from commit 65466904b015 ("tcp: adjust
TSO packet sizes based on min_rtt"). This last commit only made the
regression more visible, because it locked the TCP flow on a particular
behavior where TSQ prevented two skbs being pushed downstream,
adding silences on the wire between each TSO packet.

Many thanks to Stefan for his invaluable help !

Fixes: 75c119afe14f ("tcp: implement rb-tree based retransmit queue")
Link: https://lore.kernel.org/netdev/7f31ddc8-9971-495e-a1f6-819df542e0af@gmx.net/
Reported-by: Stefan Wahren <wahrenst@gmx.net>
Tested-by: Stefan Wahren <wahrenst@gmx.net>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Link: https://lore.kernel.org/r/20231017124526.4060202-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net/ipv4/tcp_output.c