linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* Ability to change MSS using TCP_MAXSEG
@ 2006-03-03 22:31 Jerry Fox
  0 siblings, 0 replies; 2+ messages in thread
From: Jerry Fox @ 2006-03-03 22:31 UTC (permalink / raw)
  To: linux-kernel

Hello-

I'll post this time with mentioning "advertised" MSS
in the subject  :-)

I have a question about the advertised MSS and the
TCP_MAXSEG setsock option.   Note: I am using the
Stevens sock code (http://www.unpbook.com/src.html) to
do testing.  

I set the TCP_MAXSEG via –X to 256 (yes, I
intentionally want poor performance), and used gdb to
verify that setsockopt gets called before the
connect() call, as described in the tcp (7) manpage. 
However, Ethereal shows the advmss is still 1460.  
Looking in the kernel
(net/ipv4/tcp_output.c::tcp_connect_init), I can see
the mss_clamp gets the user_mss, but the advmss is
still calculated based on dst_metric.  Host route MTU
settings and interface MTU settings work as expected,
but is there a way to advertise a different MSS per
session/socket? 

For ref, I see the same behavior on:
IA64+Debian/HPCGL+2.4.20
X86+Fedora+2.6.15
PPC+MontaVista+2.6.14

Please cc: me personally on any replies.

Thanks!
Jerry Fox
 



^ permalink raw reply	[flat|nested] 2+ messages in thread

* Ability to change MSS using TCP_MAXSEG
@ 2006-03-03 22:43 Jerry Fox
  0 siblings, 0 replies; 2+ messages in thread
From: Jerry Fox @ 2006-03-03 22:43 UTC (permalink / raw)
  To: linux-kernel

Hello-

I'll post this time with mentioning "advertised" MSS
in the subject  :-)

I have a question about the advertised MSS and the
TCP_MAXSEG setsock option.   Note: I am using the
Stevens sock code (http://www.unpbook.com/src.html) to
do testing.  

I set the TCP_MAXSEG via –X to 256 (yes, I
intentionally want poor performance), and used gdb to
verify that setsockopt gets called before the
connect() call, as described in the tcp (7) manpage. 
However, Ethereal shows the advmss is still 1460.  
Looking in the kernel
(net/ipv4/tcp_output.c::tcp_connect_init), I can see
the mss_clamp gets the user_mss, but the advmss is
still calculated based on dst_metric.  Host route MTU
settings and interface MTU settings work as expected,
but is there a way to advertise a different MSS per
session/socket? 

For ref, I see the same behavior on:
IA64+Debian/HPCGL+2.4.20
X86+Fedora+2.6.15
PPC+MontaVista+2.6.14

Please cc: me personally on any replies.

Thanks!
Jerry Fox
 



^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2006-03-03 22:43 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2006-03-03 22:31 Ability to change MSS using TCP_MAXSEG Jerry Fox
2006-03-03 22:43 Jerry Fox

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).