tcp: tcp_sendpages() should call tcp_push() once
authorEric Dumazet <eric.dumazet@gmail.com>
Thu, 5 Apr 2012 03:05:35 +0000 (03:05 +0000)
committerDavid S. Miller <davem@davemloft.net>
Thu, 5 Apr 2012 23:04:27 +0000 (19:04 -0400)
commit35f9c09fe9c72eb8ca2b8e89a593e1c151f28fc2
tree6e57508d6633c47b8846cfdbaefc33e6e0128a47
parent78d50217baf36093ab320f95bae0d6452daec85c
tcp: tcp_sendpages() should call tcp_push() once

commit 2f533844242 (tcp: allow splice() to build full TSO packets) added
a regression for splice() calls using SPLICE_F_MORE.

We need to call tcp_flush() at the end of the last page processed in
tcp_sendpages(), or else transmits can be deferred and future sends
stall.

Add a new internal flag, MSG_SENDPAGE_NOTLAST, acting like MSG_MORE, but
with different semantic.

For all sendpage() providers, its a transparent change. Only
sock_sendpage() and tcp_sendpages() can differentiate the two different
flags provided by pipe_to_sendpage()

Reported-by: Tom Herbert <therbert@google.com>
Cc: Nandita Dukkipati <nanditad@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: H.K. Jerry Chu <hkchu@google.com>
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Mahesh Bandewar <maheshb@google.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail>com>
Signed-off-by: David S. Miller <davem@davemloft.net>
fs/splice.c
include/linux/socket.h
net/ipv4/tcp.c
net/socket.c