tcp: Ignore TCP_NOPUSH socketopt by default
authorSepherosa Ziehau <sephe@dragonflybsd.org>
Thu, 17 May 2012 09:58:41 +0000 (17:58 +0800)
committerSepherosa Ziehau <sephe@dragonflybsd.org>
Fri, 18 May 2012 01:41:57 +0000 (09:41 +0800)
commit5cf02bc43f01bf69cff57b4d534939b6efb2176f
tree0b576c2cd77e33fe02e20a9c89e8a1b0540ff3e2
parent8af81f5b7b0b9db385dc852541744202d48b5f93
tcp: Ignore TCP_NOPUSH socketopt by default

For ill optimized programs which misuses this sockopt will cause
unpredicted length of network stalling, if the total sending size
is not TCP sending segment size aligned.

sysctl node net.inet.tcp.disable_nopush controls whether TCP_NOPUSH
will take effect or not

I am not going to fight agaist the stupid programs in the wild.

DragonFly-bug: http://bugs.dragonflybsd.org/issues/2368

This is actually _not_ a bug on our side.
sys/netinet/tcp_usrreq.c